rust/tantivy
1
0
Fork 0
mirror of https://github.com/quickwit-oss/tantivy.git synced 2026-01-04 16:22:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

Compare commits

base: rust:softcommits
Branches Tags
rust:main rust:paul.masurel/seek_into_the_danger_zone_result rust:paul.masurel/skip-first-doc-on-intersection rust:paul.masurel/search_into_danger_zone_fastfield rust:bucket_id_agg rust:paradedb/fix-agg-validation rust:flatheadmill-geo rust:paradedb-paradedb/fix-overflow-issue rust:low_card_optimization rust:paul.masurel/clippy rust:qw-airmail-20250916 rust:unit-test-aggregation-sum-other rust:clippy-and-cleanups rust:trinity.pointard/opti-sstable-sorted-ords-to-term rust:release_0.24 rust:dependabot/cargo/rand-0.9.1 rust:dependabot/cargo/lru-0.14.0 rust:dependabot/cargo/fs4-0.13.1 rust:tagged-user-input-ast rust:1686a/parse-agg-res rust:dependabot/cargo/rand_distr-0.5.1 rust:issue/2411-adding-panic-handler-to-rayon-merge-thread-pool rust:dependabot/cargo/nom-8 rust:upgrade_deps rust:test_parser rust:dependabot/github_actions/codecov/codecov-action-5 rust:binggan-0.12.0 rust:agg_format rust:path_collector rust:test_order_bug rust:test_columnar_multi rust:columnar_optionalmask_in_multivalues rust:prefix-phrase-query-optim rust:tantivy-with-oneshot rust:update_example rust:store-reader-enumerate rust:document-deserialize-seed rust:check_doc_mem rust:experimental-do-not-merge rust:expdonotmerge rust:cherry-pick-vec-instead-of-btreemap rust:validate_length rust:raphael_bench_avx_filter rust:raphael_optim_memset2 rust:query-field-name-avoid-alloc rust:check_bug rust:missing_test rust:list_fields rust:block-cache-control rust:trinity--multilayer-sstable rust:tracing-tantivy2 rust:tracing-tantivy rust:index_writer_err rust:default_fast_field_tokenizer rust:fmassot/add-preserve-original-ascii-folding rust:numeric_with_str_col rust:straightforward-simplify rust:fmassot/fake-test-branch rust:fmassot/dynamic-tokenizer-poc-follow-up rust:fmassot/add-box-token-filter-and-refactor rust:dynamic-tokenizer-proof-of-concept rust:fmassot/add-box-tokenizer-and-token-filter rust:tokenizer-api-0.1.1 rust:addconversion rust:removed-extra-enum-alternative-for-datetime-precision rust:issue/1981-reproduce-bm25-bug rust:tokenizer_docs rust:merge-proptests rust:columnar-todo rust:quickwit-0.5 rust:remove-byteorder rust:remove_dyn rust:pr/alexcole/1855-inject-bm25-statistics rust:evance-expose-build-query-from-user-input-ast rust:missing-specialized-method-arcdyncolumn rust:fast-u64-range-query rust:0.19.2 rust:columnar-cli2 rust:trinity--pub-term-as-slice rust:typed-column3 rust:columnar-merge3 rust:columnar-merge-exploring-fn rust:use_columnar rust:typed-column2 rust:term_agg_bucket_limit rust:range rust:typed-column rust:columnar-main rust:forced-types rust:u128-columnar rust:use_stats rust:truncation_comment rust:0.19.1 rust:trinity/yoke-documents rust:column-handle rust:columnar-cli rust:columnar rust:revert-1711-sparse_dense_index rust:commit-changes rust:sparse_dense_index_headroom rust:optional_column rust:no-column-for-multivalued rust:fmassot--bench-hdfs-with-array rust:bugfix-position-broken rust:float rust:simplify rust:debug_time rust:sparse_codecs rust:fix_estimate rust:column-reader rust:column-reader-b rust:columnreader-simple rust:col-trait-refact rust:ip_fastfield rust:column-trait-2 rust:refact-dyn-column rust:column-trait rust:refact-codec-trait rust:pr/PSeitz/1485-1 rust:fastfieldcodec rust:termmap_per_field rust:fastfield-codecs-tests rust:rayon-indexing-writers rust:issue/1251 rust:index-bench rust:fix_open_bytes_index rust:warming rust:revert-1222-issue/1195 rust:add-random-bench-for-fast-field rust:wasm-friendly-exploratory rust:revert-1159-patch-1 rust:debug-position-panic-0.15.3 rust:fix-issue-1111 rust:remove_rand rust:unused-deps rust:hfb/phrase_match_slop rust:0.15.3 rust:expose-min-max-on-multi-fast-field-reader rust:revert-1094-expose-min-max-on-multi-fast-field-reader rust:hotfix-1088 rust:owned-bytes-instead-of-rawdoc rust:rihardsk-date-ranges rust:issue/997 rust:0.13.3 rust:fail-fast-store-checkpoint rust:debugging-store rust:issue/reproduce-897 rust:issue/969-reproduce rust:troublescooter-static rust:intermediate-mergeable-troublescooter rust:barrotsteindev-filter-collector-tpredicatevalue rust:issue/938b rust:issue/938 rust:barrotsteindev-filter-collector rust:issue/922 rust:issue/896 rust:quickfix-explain rust:slog rust:limit-random-seeks rust:0.13.1 rust:issue/866b rust:bugfix-unittest-macos rust:githubactions rust:fieldnorm_readers rust:blockwand rust:nodeffeatfails rust:removedaliveiterator rust:robyoung-introduce-offset-to-top-docs rust:specialization-of-foreach-for-union-scorer rust:elshize-bmw rust:cutt rust:nrt rust:storewriter-out rust:segment_writer_in_ram2 rust:polling rust:audunhalland-query-boost rust:poll-meta rust:segment_writer_in_ram rust:bundle rust:petr-tik-n662_footer_is_incompatible_impl rust:race-condition-in-tests rust:hotfix-0.10.3-issue681 rust:remove-tru rust:kkoziara-pre-tokenized-text rust:issue/681 rust:issue/680 rust:refactoring/field rust:updating/uuid rust:fix-bench rust:hotfix-656 rust:readd-macos-ci rust:incremental-search rust:crate-division rust:streamer-with-state rust:criterion rust:python-binding-changes rust:bump-version rust:gh-pages rust:issue/554 rust:perf/experimental-union-for_each rust:issue/526b rust:softcommits rust:0.9.1-hotfix rust:revert-521-fix-non-english-stemmers rust:issue/500-hotfix rust:hotfix/0.8.2 rust:issue/483 rust:bug/merge-deted rust:issue/407 rust:issue/457 rust:segment_fruits rust:doc rust:dds/lenient rust:issue/webasm rust:issue/396 rust:broken_collector_with_result rust:dss/automaton-weight-suggestion rust:wasm rust:common-crawl rust:tantivy-imhotep
rust:0.25.0 rust:0.22.1 rust:0.24.2 rust:0.24.1 rust:tantivy-query-grammar-v0.24.0 rust:tantivy-columnar-v0.5.0 rust:tantivy-bitpacker-v0.8.0 rust:ownedbytes-v0.9.0 rust:tantivy-stacker-v0.5.0 rust:tantivy-sstable-v0.5.0 rust:tantivy-common-v0.9.0 rust:tantivy-tokenizer-api-v0.5.0 rust:0.24 rust:qw-airmail-20250224-hotfix-merge-panic rust:qw-airmail-20250219-hotfix-merge-panic rust:qw-airmail-202406 rust:qw-airmail-20240611 rust:qw-airmail-20240606 rust:quickwit-rev rust:0.22.0 rust:tantivy-stacker-v0.3.0 rust:ownedbytes-v0.7.0 rust:tantivy-tokenizer-api-v0.3.0 rust:tantivy-columnar-v0.3.0 rust:tantivy-common-v0.7.0 rust:tantivy-query-grammar-v0.22.0 rust:tantivy-sstable-v0.3.0 rust:tantivy-bitpacker-v0.6.0 rust:0.21.1 rust:0.21 rust:0.20.2 rust:0.20.1 rust:0.20 rust:quickwit-0.5-rev rust:0.19.2 rust:0.19.1 rust:0.19 rust:0.18.1 rust:0.18 rust:0.17 rust:0.16.1 rust:0.15.3 rust:0.15.2 rust:0.15.1 rust:0.15 rust:0.14 rust:0.13.3 rust:0.13.2 rust:0.13.1 rust:0.13 rust:0.12 rust:0.11.3 rust:0.11.1 rust:0.11 rust:0.10.3 rust:0.1 rust:0.10.2 rust:0.10.1 rust:0.10.0 rust:0.9.1 rust:0.9 rust:0.8.2 rust:0.8.2b rust:0.8.1 rust:0.8.0 rust:0.7.2 rust:0.7.1 rust:0.7.0 rust:0.6.1 rust:0.6.0 rust:0.5.2 rust:0.5.1 rust:0.5.0 rust:0.4.3 rust:0.4.2 rust:0.4.0 rust:0.3.1 rust:0.3.0 rust:0.2.0
..
compare: rust:issue/webasm
Branches Tags
rust:paul.masurel/seek_into_the_danger_zone_result rust:paul.masurel/skip-first-doc-on-intersection rust:main rust:paul.masurel/search_into_danger_zone_fastfield rust:bucket_id_agg rust:paradedb/fix-agg-validation rust:flatheadmill-geo rust:paradedb-paradedb/fix-overflow-issue rust:low_card_optimization rust:paul.masurel/clippy rust:qw-airmail-20250916 rust:unit-test-aggregation-sum-other rust:clippy-and-cleanups rust:trinity.pointard/opti-sstable-sorted-ords-to-term rust:release_0.24 rust:dependabot/cargo/rand-0.9.1 rust:dependabot/cargo/lru-0.14.0 rust:dependabot/cargo/fs4-0.13.1 rust:tagged-user-input-ast rust:1686a/parse-agg-res rust:dependabot/cargo/rand_distr-0.5.1 rust:issue/2411-adding-panic-handler-to-rayon-merge-thread-pool rust:dependabot/cargo/nom-8 rust:upgrade_deps rust:test_parser rust:dependabot/github_actions/codecov/codecov-action-5 rust:binggan-0.12.0 rust:agg_format rust:path_collector rust:test_order_bug rust:test_columnar_multi rust:columnar_optionalmask_in_multivalues rust:prefix-phrase-query-optim rust:tantivy-with-oneshot rust:update_example rust:store-reader-enumerate rust:document-deserialize-seed rust:check_doc_mem rust:experimental-do-not-merge rust:expdonotmerge rust:cherry-pick-vec-instead-of-btreemap rust:validate_length rust:raphael_bench_avx_filter rust:raphael_optim_memset2 rust:query-field-name-avoid-alloc rust:check_bug rust:missing_test rust:list_fields rust:block-cache-control rust:trinity--multilayer-sstable rust:tracing-tantivy2 rust:tracing-tantivy rust:index_writer_err rust:default_fast_field_tokenizer rust:fmassot/add-preserve-original-ascii-folding rust:numeric_with_str_col rust:straightforward-simplify rust:fmassot/fake-test-branch rust:fmassot/dynamic-tokenizer-poc-follow-up rust:fmassot/add-box-token-filter-and-refactor rust:dynamic-tokenizer-proof-of-concept rust:fmassot/add-box-tokenizer-and-token-filter rust:tokenizer-api-0.1.1 rust:addconversion rust:removed-extra-enum-alternative-for-datetime-precision rust:issue/1981-reproduce-bm25-bug rust:tokenizer_docs rust:merge-proptests rust:columnar-todo rust:quickwit-0.5 rust:remove-byteorder rust:remove_dyn rust:pr/alexcole/1855-inject-bm25-statistics rust:evance-expose-build-query-from-user-input-ast rust:missing-specialized-method-arcdyncolumn rust:fast-u64-range-query rust:0.19.2 rust:columnar-cli2 rust:trinity--pub-term-as-slice rust:typed-column3 rust:columnar-merge3 rust:columnar-merge-exploring-fn rust:use_columnar rust:typed-column2 rust:term_agg_bucket_limit rust:range rust:typed-column rust:columnar-main rust:forced-types rust:u128-columnar rust:use_stats rust:truncation_comment rust:0.19.1 rust:trinity/yoke-documents rust:column-handle rust:columnar-cli rust:columnar rust:revert-1711-sparse_dense_index rust:commit-changes rust:sparse_dense_index_headroom rust:optional_column rust:no-column-for-multivalued rust:fmassot--bench-hdfs-with-array rust:bugfix-position-broken rust:float rust:simplify rust:debug_time rust:sparse_codecs rust:fix_estimate rust:column-reader rust:column-reader-b rust:columnreader-simple rust:col-trait-refact rust:ip_fastfield rust:column-trait-2 rust:refact-dyn-column rust:column-trait rust:refact-codec-trait rust:pr/PSeitz/1485-1 rust:fastfieldcodec rust:termmap_per_field rust:fastfield-codecs-tests rust:rayon-indexing-writers rust:issue/1251 rust:index-bench rust:fix_open_bytes_index rust:warming rust:revert-1222-issue/1195 rust:add-random-bench-for-fast-field rust:wasm-friendly-exploratory rust:revert-1159-patch-1 rust:debug-position-panic-0.15.3 rust:fix-issue-1111 rust:remove_rand rust:unused-deps rust:hfb/phrase_match_slop rust:0.15.3 rust:expose-min-max-on-multi-fast-field-reader rust:revert-1094-expose-min-max-on-multi-fast-field-reader rust:hotfix-1088 rust:owned-bytes-instead-of-rawdoc rust:rihardsk-date-ranges rust:issue/997 rust:0.13.3 rust:fail-fast-store-checkpoint rust:debugging-store rust:issue/reproduce-897 rust:issue/969-reproduce rust:troublescooter-static rust:intermediate-mergeable-troublescooter rust:barrotsteindev-filter-collector-tpredicatevalue rust:issue/938b rust:issue/938 rust:barrotsteindev-filter-collector rust:issue/922 rust:issue/896 rust:quickfix-explain rust:slog rust:limit-random-seeks rust:0.13.1 rust:issue/866b rust:bugfix-unittest-macos rust:githubactions rust:fieldnorm_readers rust:blockwand rust:nodeffeatfails rust:removedaliveiterator rust:robyoung-introduce-offset-to-top-docs rust:specialization-of-foreach-for-union-scorer rust:elshize-bmw rust:cutt rust:nrt rust:storewriter-out rust:segment_writer_in_ram2 rust:polling rust:audunhalland-query-boost rust:poll-meta rust:segment_writer_in_ram rust:bundle rust:petr-tik-n662_footer_is_incompatible_impl rust:race-condition-in-tests rust:hotfix-0.10.3-issue681 rust:remove-tru rust:kkoziara-pre-tokenized-text rust:issue/681 rust:issue/680 rust:refactoring/field rust:updating/uuid rust:fix-bench rust:hotfix-656 rust:readd-macos-ci rust:incremental-search rust:crate-division rust:streamer-with-state rust:criterion rust:python-binding-changes rust:bump-version rust:gh-pages rust:issue/554 rust:perf/experimental-union-for_each rust:issue/526b rust:softcommits rust:0.9.1-hotfix rust:revert-521-fix-non-english-stemmers rust:issue/500-hotfix rust:hotfix/0.8.2 rust:issue/483 rust:bug/merge-deted rust:issue/407 rust:issue/457 rust:segment_fruits rust:doc rust:dds/lenient rust:issue/webasm rust:issue/396 rust:broken_collector_with_result rust:dss/automaton-weight-suggestion rust:wasm rust:common-crawl rust:tantivy-imhotep
rust:0.25.0 rust:0.22.1 rust:0.24.2 rust:0.24.1 rust:tantivy-query-grammar-v0.24.0 rust:tantivy-columnar-v0.5.0 rust:tantivy-bitpacker-v0.8.0 rust:ownedbytes-v0.9.0 rust:tantivy-stacker-v0.5.0 rust:tantivy-sstable-v0.5.0 rust:tantivy-common-v0.9.0 rust:tantivy-tokenizer-api-v0.5.0 rust:0.24 rust:qw-airmail-20250224-hotfix-merge-panic rust:qw-airmail-20250219-hotfix-merge-panic rust:qw-airmail-202406 rust:qw-airmail-20240611 rust:qw-airmail-20240606 rust:quickwit-rev rust:0.22.0 rust:tantivy-stacker-v0.3.0 rust:ownedbytes-v0.7.0 rust:tantivy-tokenizer-api-v0.3.0 rust:tantivy-columnar-v0.3.0 rust:tantivy-common-v0.7.0 rust:tantivy-query-grammar-v0.22.0 rust:tantivy-sstable-v0.3.0 rust:tantivy-bitpacker-v0.6.0 rust:0.21.1 rust:0.21 rust:0.20.2 rust:0.20.1 rust:0.20 rust:quickwit-0.5-rev rust:0.19.2 rust:0.19.1 rust:0.19 rust:0.18.1 rust:0.18 rust:0.17 rust:0.16.1 rust:0.15.3 rust:0.15.2 rust:0.15.1 rust:0.15 rust:0.14 rust:0.13.3 rust:0.13.2 rust:0.13.1 rust:0.13 rust:0.12 rust:0.11.3 rust:0.11.1 rust:0.11 rust:0.10.3 rust:0.1 rust:0.10.2 rust:0.10.1 rust:0.10.0 rust:0.9.1 rust:0.9 rust:0.8.2 rust:0.8.2b rust:0.8.1 rust:0.8.0 rust:0.7.2 rust:0.7.1 rust:0.7.0 rust:0.6.1 rust:0.6.0 rust:0.5.2 rust:0.5.1 rust:0.5.0 rust:0.4.3 rust:0.4.2 rust:0.4.0 rust:0.3.1 rust:0.3.0 rust:0.2.0

2 Commits
softcommit ... issue/weba

Author SHA1 Message Date
Paul Masurel
507e46f814 Added static directory 2018-10-04 23:28:44 +09:00
Paul Masurel
3d3da2d66f Compiling in WebAssembly 2018-10-04 08:45:04 +09:00
144 changed files with 4589 additions and 8843 deletions
Expand all files Collapse all files

4
.travis.yml
View File

@@ -29,7 +29,7 @@ addons:
matrix: matrix:
include: include:
# Android # Android
- env: TARGET=aarch64-linux-android DISABLE_TESTS - env: TARGET=aarch64-linux-android DISABLE_TESTS=1
#- env: TARGET=arm-linux-androideabi DISABLE_TESTS=1 #- env: TARGET=arm-linux-androideabi DISABLE_TESTS=1
#- env: TARGET=armv7-linux-androideabi DISABLE_TESTS=1 #- env: TARGET=armv7-linux-androideabi DISABLE_TESTS=1
#- env: TARGET=i686-linux-android DISABLE_TESTS=1 #- env: TARGET=i686-linux-android DISABLE_TESTS=1
@@ -77,4 +77,4 @@ before_cache:
notifications: notifications:
email: email:
on_success: never on_success: never

46
CHANGELOG.md
View File

@@ -1,49 +1,3 @@
Tantivy 0.9.0
=====================
*0.9.0 index format is not compatible with the
previous index format.*
- MAJOR BUGFIX :
Some `Mmap` objects were being leaked, and would never get released. (@fulmicoton)
- Removed most unsafe (@fulmicoton)
- Indexer memory footprint improved. (VInt comp, inlining the first block. (@fulmicoton)
- Stemming in other language possible (@pentlander)
- Segments with no docs are deleted earlier (@barrotsteindev)
- Added grouped add and delete operations.
They are guaranteed to happen together (i.e. they cannot be split by a commit).
In addition, adds are guaranteed to happen on the same segment. (@elbow-jason)
- Removed `INT_STORED` and `INT_INDEXED`. It is now possible to use `STORED` and `INDEXED`
for int fields. (@fulmicoton)
- Added DateTime field (@barrotsteindev)
- Added IndexReader. By default, index is reloaded automatically upon new commits (@fulmicoton)
- SIMD linear search within blocks (@fulmicoton)
Tantivy 0.8.2
=====================
Fixing build for x86_64 platforms. (#496)
No need to update from 0.8.1 if tantivy
is building on your platform.
Tantivy 0.8.1
=====================
Hotfix of #476.
Merge was reflecting deletes before commit was passed.
Thanks @barrotsteindev for reporting the bug.
Tantivy 0.8.0
=====================
*No change in the index format*
- API Breaking change in the collector API. (@jwolfe, @fulmicoton)
- Multithreaded search (@jwolfe, @fulmicoton)
Tantivy 0.7.1
=====================
*No change in the index format*
- Bugfix: NGramTokenizer panics on non ascii chars
- Added a space usage API
Tantivy 0.7 Tantivy 0.7
===================== =====================

42
Cargo.toml
View File

@@ -1,6 +1,6 @@
[package] [package]
name = "tantivy" name = "tantivy"
version = "0.9.0" version = "0.7.0"
authors = ["Paul Masurel <paul.masurel@gmail.com>"] authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT" license = "MIT"
categories = ["database-implementations", "data-structures"] categories = ["database-implementations", "data-structures"]
@@ -12,12 +12,12 @@ readme = "README.md"
keywords = ["search", "information", "retrieval"] keywords = ["search", "information", "retrieval"]
[dependencies] [dependencies]
base64 = "0.10.0" base64 = "0.9.1"
byteorder = "1.0" byteorder = "1.0"
lazy_static = "1" lazy_static = "1"
regex = "1.0" regex = "1.0"
tantivy-fst = "0.1" fst = {version="0.3", default-features=false}
memmap = {version = "0.7", optional=true} fst-regex = { version="0.2", optional=true}
lz4 = {version="1.20", optional=true} lz4 = {version="1.20", optional=true}
snap = {version="0.2"} snap = {version="0.2"}
atomicwrites = {version="0.2.2", optional=true} atomicwrites = {version="0.2.2", optional=true}
@@ -29,38 +29,33 @@ serde = "1.0"
serde_derive = "1.0" serde_derive = "1.0"
serde_json = "1.0" serde_json = "1.0"
num_cpus = "1.2" num_cpus = "1.2"
fs2={version="0.4", optional=true} itertools = "0.7"
itertools = "0.8"
levenshtein_automata = {version="0.1", features=["fst_automaton"]} levenshtein_automata = {version="0.1", features=["fst_automaton"]}
notify = {version="4", optional=true}
bit-set = "0.5" bit-set = "0.5"
uuid = { version = "0.7.2", features = ["v4", "serde"] } uuid = { version = "0.7", features = ["v4", "serde"] }
crossbeam = "0.5" crossbeam = "0.4"
crossbeam-channel = "0.2"
futures = "0.1" futures = "0.1"
futures-cpupool = "0.1" futures-cpupool = "0.1"
owning_ref = "0.4" owning_ref = "0.4"
stable_deref_trait = "1.0.0" stable_deref_trait = "1.0.0"
rust-stemmers = "1.1" rust-stemmers = "1"
downcast-rs = { version="1.0" } downcast = { version="0.9" }
bitpacking = "0.6" matches = "0.1"
census = "0.2" bitpacking = "0.5"
census = "0.1"
fnv = "1.0.6" fnv = "1.0.6"
owned-read = "0.4" owned-read = "0.4"
failure = "0.1" failure = "0.1"
htmlescape = "0.3.1" htmlescape = "0.3.1"
fail = "0.2" fail = "0.2"
scoped-pool = "1.0"
murmurhash32 = "0.2"
chrono = "0.4"
[target.'cfg(windows)'.dependencies] [target.'cfg(windows)'.dependencies]
winapi = "0.2" winapi = "0.2"
[dev-dependencies] [dev-dependencies]
rand = "0.6" rand = "0.5"
maplit = "1" maplit = "1"
matches = "0.1.8"
time = "0.1.42"
[profile.release] [profile.release]
opt-level = 3 opt-level = 3
@@ -73,12 +68,13 @@ overflow-checks = true
[features] [features]
# by default no-fail is disabled. We manually enable it when running test. # by default no-fail is disabled. We manually enable it when running test.
default = ["mmap", "no_fail"] default = ["mmap", "no_fail", "regex_query"]
mmap = ["atomicwrites", "fs2", "memmap", "notify"] mmap = ["fst/mmap", "atomicwrites"]
regex_query = ["fst-regex"]
lz4-compression = ["lz4"] lz4-compression = ["lz4"]
no_fail = ["fail/no_fail"] no_fail = ["fail/no_fail"]
unstable = [] # useful for benches.
wasm-bindgen = ["uuid/wasm-bindgen"]
[badges] [badges]
travis-ci = { repository = "tantivy-search/tantivy" } travis-ci = { repository = "tantivy-search/tantivy" }

27
README.md
View File

@@ -17,29 +17,19 @@
[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/6)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/6) [![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/6)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/6)
[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/7)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/7) [![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/7)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/7)
[![Become a patron](https://c5.patreon.com/external/logo/become_a_patron_button.png)](https://www.patreon.com/fulmicoton)
**Tantivy** is a **full text search engine library** written in rust. **Tantivy** is a **full text search engine library** written in rust.
It is closer to [Apache Lucene](https://lucene.apache.org/) than to [Elasticsearch](https://www.elastic.co/products/elasticsearch) and [Apache Solr](https://lucene.apache.org/solr/) in the sense it is not It is closer to Lucene than to Elastic Search and Solr in the sense it is not
an off-the-shelf search engine server, but rather a crate that can be used an off-the-shelf search engine server, but rather a crate that can be used
to build such a search engine. to build such a search engine.
Tantivy is, in fact, strongly inspired by Lucene's design. Tantivy is, in fact, strongly inspired by Lucene's design.
# Benchmark
Tantivy is typically faster than Lucene, but the results will depend on
the nature of the queries in your workload.
The following [benchmark](https://tantivy-search.github.io/bench/) break downs
performance for different type of queries / collection.
# Features # Features
- Full-text search - Full-text search
- Configurable tokenizer. (stemming available for 17 latin languages. Third party support for Chinese ([tantivy-jieba](https://crates.io/crates/tantivy-jieba) and [cang-jie](https://crates.io/crates/cang-jie)) and [Japanese](https://crates.io/crates/tantivy-tokenizer-tiny-segmenter)
- Fast (check out the :racehorse: :sparkles: [benchmark](https://tantivy-search.github.io/bench/) :sparkles: :racehorse:) - Fast (check out the :racehorse: :sparkles: [benchmark](https://tantivy-search.github.io/bench/) :sparkles: :racehorse:)
- Tiny startup time (<10ms), perfect for command line tools - Tiny startup time (<10ms), perfect for command line tools
- BM25 scoring (the same as lucene) - BM25 scoring (the same as lucene)
@@ -51,7 +41,6 @@ performance for different type of queries / collection.
- SIMD integer compression when the platform/CPU includes the SSE2 instruction set. - SIMD integer compression when the platform/CPU includes the SSE2 instruction set.
- Single valued and multivalued u64 and i64 fast fields (equivalent of doc values in Lucene) - Single valued and multivalued u64 and i64 fast fields (equivalent of doc values in Lucene)
- `&[u8]` fast fields - `&[u8]` fast fields
- Text, i64, u64, dates and hierarchical facet fields
- LZ4 compressed document store - LZ4 compressed document store
- Range queries - Range queries
- Faceted search - Faceted search
@@ -87,7 +76,7 @@ It will walk you through getting a wikipedia search engine up and running in a f
Tantivy compiles on stable rust but requires `Rust >= 1.27`. Tantivy compiles on stable rust but requires `Rust >= 1.27`.
To check out and run tests, you can simply run : To check out and run tests, you can simply run :
git clone https://github.com/tantivy-search/tantivy.git git clone git@github.com:tantivy-search/tantivy.git
cd tantivy cd tantivy
cargo build cargo build
@@ -96,14 +85,6 @@ To check out and run tests, you can simply run :
Some tests will not run with just `cargo test` because of `fail-rs`. Some tests will not run with just `cargo test` because of `fail-rs`.
To run the tests exhaustively, run `./run-tests.sh`. To run the tests exhaustively, run `./run-tests.sh`.
# How can I support this project ? # Contribute
There are many ways to support this project. Send me an email (paul.masurel at gmail.com) if you want to contribute to tantivy.
- If you use tantivy, tell us about your experience on [gitter](https://gitter.im/tantivy-search/tantivy) or by email (paul.masurel@gmail.com)
- Report bugs
- Write a blog post
- Complete documentation
- Contribute code (you can join [our gitter](https://gitter.im/tantivy-search/tantivy) )
- Talk about tantivy around you
- Drop a word on on [![Say Thanks!](https://img.shields.io/badge/Say%20Thanks-!-1EAEDB.svg)](https://saythanks.io/to/fulmicoton) or even [![Become a patron](https://c5.patreon.com/external/logo/become_a_patron_button.png)](https://www.patreon.com/fulmicoton)

99
examples/basic_search.rs
View File

@@ -16,12 +16,10 @@ extern crate tempdir;
// Importing tantivy... // Importing tantivy...
#[macro_use] #[macro_use]
extern crate tantivy; extern crate tantivy;
use tantivy::collector::TopDocs; use tantivy::collector::TopCollector;
use tantivy::query::QueryParser; use tantivy::query::QueryParser;
use tantivy::schema::*; use tantivy::schema::*;
use tantivy::Index; use tantivy::Index;
use tantivy::ReloadPolicy;
use tempdir::TempDir;
fn main() -> tantivy::Result<()> { fn main() -> tantivy::Result<()> {
// Let's create a temporary directory for the // Let's create a temporary directory for the
@@ -36,7 +34,7 @@ fn main() -> tantivy::Result<()> {
// be indexed". // be indexed".
// first we need to define a schema ... // first we need to define a schema ...
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
// Our first field is title. // Our first field is title.
// We want full-text search for it, and we also want // We want full-text search for it, and we also want
@@ -107,37 +105,37 @@ fn main() -> tantivy::Result<()> {
// For convenience, tantivy also comes with a macro to // For convenience, tantivy also comes with a macro to
// reduce the boilerplate above. // reduce the boilerplate above.
index_writer.add_document(doc!( index_writer.add_document(doc!(
title => "Of Mice and Men", title => "Of Mice and Men",
body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \ body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
bank and runs deep and green. The water is warm too, for it has slipped twinkling \ bank and runs deep and green. The water is warm too, for it has slipped twinkling \
over the yellow sands in the sunlight before reaching the narrow pool. On one \ over the yellow sands in the sunlight before reaching the narrow pool. On one \
side of the river the golden foothill slopes curve up to the strong and rocky \ side of the river the golden foothill slopes curve up to the strong and rocky \
Gabilan Mountains, but on the valley side the water is lined with trees—willows \ Gabilan Mountains, but on the valley side the water is lined with trees—willows \
fresh and green with every spring, carrying in their lower leaf junctures the \ fresh and green with every spring, carrying in their lower leaf junctures the \
debris of the winter’s flooding; and sycamores with mottled, white, recumbent \ debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
limbs and branches that arch over the pool" limbs and branches that arch over the pool"
)); ));
index_writer.add_document(doc!( index_writer.add_document(doc!(
title => "Of Mice and Men", title => "Of Mice and Men",
body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \ body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
bank and runs deep and green. The water is warm too, for it has slipped twinkling \ bank and runs deep and green. The water is warm too, for it has slipped twinkling \
over the yellow sands in the sunlight before reaching the narrow pool. On one \ over the yellow sands in the sunlight before reaching the narrow pool. On one \
side of the river the golden foothill slopes curve up to the strong and rocky \ side of the river the golden foothill slopes curve up to the strong and rocky \
Gabilan Mountains, but on the valley side the water is lined with trees—willows \ Gabilan Mountains, but on the valley side the water is lined with trees—willows \
fresh and green with every spring, carrying in their lower leaf junctures the \ fresh and green with every spring, carrying in their lower leaf junctures the \
debris of the winter’s flooding; and sycamores with mottled, white, recumbent \ debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
limbs and branches that arch over the pool" limbs and branches that arch over the pool"
)); ));
// Multivalued field just need to be repeated. // Multivalued field just need to be repeated.
index_writer.add_document(doc!( index_writer.add_document(doc!(
title => "Frankenstein", title => "Frankenstein",
title => "The Modern Prometheus", title => "The Modern Prometheus",
body => "You will rejoice to hear that no disaster has accompanied the commencement of an \ body => "You will rejoice to hear that no disaster has accompanied the commencement of an \
enterprise which you have regarded with such evil forebodings. I arrived here \ enterprise which you have regarded with such evil forebodings. I arrived here \
yesterday, and my first task is to assure my dear sister of my welfare and \ yesterday, and my first task is to assure my dear sister of my welfare and \
increasing confidence in the success of my undertaking." increasing confidence in the success of my undertaking."
)); ));
// This is an example, so we will only index 3 documents // This is an example, so we will only index 3 documents
@@ -171,33 +169,24 @@ fn main() -> tantivy::Result<()> {
// //
// ### Searcher // ### Searcher
// //
// A reader is required to get search the index. // Let's search our index. Start by reloading
// It acts as a `Searcher` pool that reloads itself, // searchers in the index. This should be done
// depending on a `ReloadPolicy`. // after every `commit()`.
// index.load_searchers()?;
// For a search server you will typically create one reader for the entire lifetime of your
// program, and acquire a new searcher for every single request.
//
// In the code below, we rely on the 'ON_COMMIT' policy: the reader
// will reload the index automatically after each commit.
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::OnCommit)
.try_into()?;
// We now need to acquire a searcher. // We now need to acquire a searcher.
//
// A searcher points to snapshotted, immutable version of the index.
//
// Some search experience might require more than // Some search experience might require more than
// one query. Using the same searcher ensures that all of these queries will run on the // one query.
// same version of the index. //
// The searcher ensure that we get to work
// with a consistent version of the index.
// //
// Acquiring a `searcher` is very cheap. // Acquiring a `searcher` is very cheap.
// //
// You should acquire a searcher every time you start processing a request and // You should acquire a searcher every time you
// start processing a request and
// and release it right after your query is finished. // and release it right after your query is finished.
let searcher = reader.searcher(); let searcher = index.searcher();
// ### Query // ### Query
@@ -223,10 +212,15 @@ fn main() -> tantivy::Result<()> {
// //
// We are not interested in all of the documents but // We are not interested in all of the documents but
// only in the top 10. Keeping track of our top 10 best documents // only in the top 10. Keeping track of our top 10 best documents
// is the role of the TopDocs. // is the role of the TopCollector.
let mut top_collector = TopCollector::with_limit(10);
// We can now perform our query. // We can now perform our query.
let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?; searcher.search(&*query, &mut top_collector)?;
// Our top collector now contains the 10
// most relevant doc ids...
let doc_addresses = top_collector.docs();
// The actual documents still need to be // The actual documents still need to be
// retrieved from Tantivy's store. // retrieved from Tantivy's store.
@@ -234,10 +228,13 @@ fn main() -> tantivy::Result<()> {
// Since the body field was not configured as stored, // Since the body field was not configured as stored,
// the document returned will only contain // the document returned will only contain
// a title. // a title.
for (_score, doc_address) in top_docs {
for doc_address in doc_addresses {
let retrieved_doc = searcher.doc(doc_address)?; let retrieved_doc = searcher.doc(doc_address)?;
println!("{}", schema.to_json(&retrieved_doc)); println!("{}", schema.to_json(&retrieved_doc));
} }
Ok(()) Ok(())
} }
use tempdir::TempDir;

187
examples/custom_collector.rs
View File

@@ -1,187 +0,0 @@
// # Custom collector example
//
// This example shows how you can implement your own
// collector. As an example, we will compute a collector
// that computes the standard deviation of a given fast field.
//
// Of course, you can have a look at the tantivy's built-in collectors
// such as the `CountCollector` for more examples.
extern crate tempdir;
// ---
// Importing tantivy...
#[macro_use]
extern crate tantivy;
use tantivy::collector::{Collector, SegmentCollector};
use tantivy::fastfield::FastFieldReader;
use tantivy::query::QueryParser;
use tantivy::schema::Field;
use tantivy::schema::{Schema, FAST, INDEXED, TEXT};
use tantivy::Index;
use tantivy::SegmentReader;
#[derive(Default)]
struct Stats {
count: usize,
sum: f64,
squared_sum: f64,
}
impl Stats {
pub fn count(&self) -> usize {
self.count
}
pub fn mean(&self) -> f64 {
self.sum / (self.count as f64)
}
fn square_mean(&self) -> f64 {
self.squared_sum / (self.count as f64)
}
pub fn standard_deviation(&self) -> f64 {
let mean = self.mean();
(self.square_mean() - mean * mean).sqrt()
}
fn non_zero_count(self) -> Option<Stats> {
if self.count == 0 {
None
} else {
Some(self)
}
}
}
struct StatsCollector {
field: Field,
}
impl StatsCollector {
fn with_field(field: Field) -> StatsCollector {
StatsCollector { field }
}
}
impl Collector for StatsCollector {
// That's the type of our result.
// Our standard deviation will be a float.
type Fruit = Option<Stats>;
type Child = StatsSegmentCollector;
fn for_segment(
&self,
_segment_local_id: u32,
segment: &SegmentReader,
) -> tantivy::Result<StatsSegmentCollector> {
let fast_field_reader = segment.fast_field_reader(self.field)?;
Ok(StatsSegmentCollector {
fast_field_reader,
stats: Stats::default(),
})
}
fn requires_scoring(&self) -> bool {
// this collector does not care about score.
false
}
fn merge_fruits(&self, segment_stats: Vec<Option<Stats>>) -> tantivy::Result<Option<Stats>> {
let mut stats = Stats::default();
for segment_stats_opt in segment_stats {
if let Some(segment_stats) = segment_stats_opt {
stats.count += segment_stats.count;
stats.sum += segment_stats.sum;
stats.squared_sum += segment_stats.squared_sum;
}
}
Ok(stats.non_zero_count())
}
}
struct StatsSegmentCollector {
fast_field_reader: FastFieldReader<u64>,
stats: Stats,
}
impl SegmentCollector for StatsSegmentCollector {
type Fruit = Option<Stats>;
fn collect(&mut self, doc: u32, _score: f32) {
let value = self.fast_field_reader.get(doc) as f64;
self.stats.count += 1;
self.stats.sum += value;
self.stats.squared_sum += value * value;
}
fn harvest(self) -> <Self as SegmentCollector>::Fruit {
self.stats.non_zero_count()
}
}
fn main() -> tantivy::Result<()> {
// # Defining the schema
//
// The Tantivy index requires a very strict schema.
// The schema declares which fields are in the index,
// and for each field, its type and "the way it should
// be indexed".
// first we need to define a schema ...
let mut schema_builder = Schema::builder();
// We'll assume a fictional index containing
// products, and with a name, a description, and a price.
let product_name = schema_builder.add_text_field("name", TEXT);
let product_description = schema_builder.add_text_field("description", TEXT);
let price = schema_builder.add_u64_field("price", INDEXED | FAST);
let schema = schema_builder.build();
// # Indexing documents
//
// Lets index a bunch of fake documents for the sake of
// this example.
let index = Index::create_in_ram(schema.clone());
let mut index_writer = index.writer(50_000_000)?;
index_writer.add_document(doc!(
product_name => "Super Broom 2000",
product_description => "While it is ok for short distance travel, this broom \
was designed quiditch. It will up your game.",
price => 30_200u64
));
index_writer.add_document(doc!(
product_name => "Turbulobroom",
product_description => "You might have heard of this broom before : it is the sponsor of the Wales team.\
You'll enjoy its sharp turns, and rapid acceleration",
price => 29_240u64
));
index_writer.add_document(doc!(
product_name => "Broomio",
product_description => "Great value for the price. This broom is a market favorite",
price => 21_240u64
));
index_writer.add_document(doc!(
product_name => "Whack a Mole",
product_description => "Prime quality bat.",
price => 5_200u64
));
index_writer.commit()?;
let reader = index.reader()?;
let searcher = reader.searcher();
let query_parser = QueryParser::for_index(&index, vec![product_name, product_description]);
// here we want to get a hit on the 'ken' in Frankenstein
let query = query_parser.parse_query("broom")?;
if let Some(stats) = searcher.search(&query, &StatsCollector::with_field(price))? {
println!("count: {}", stats.count());
println!("mean: {}", stats.mean());
println!("standard deviation: {}", stats.standard_deviation());
}
Ok(())
}

24
examples/custom_tokenizer.rs
View File

@@ -5,7 +5,7 @@
#[macro_use] #[macro_use]
extern crate tantivy; extern crate tantivy;
use tantivy::collector::TopDocs; use tantivy::collector::TopCollector;
use tantivy::query::QueryParser; use tantivy::query::QueryParser;
use tantivy::schema::*; use tantivy::schema::*;
use tantivy::tokenizer::NgramTokenizer; use tantivy::tokenizer::NgramTokenizer;
@@ -20,7 +20,7 @@ fn main() -> tantivy::Result<()> {
// be indexed". // be indexed".
// first we need to define a schema ... // first we need to define a schema ...
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
// Our first field is title. // Our first field is title.
// In this example we want to use NGram searching // In this example we want to use NGram searching
@@ -68,12 +68,12 @@ fn main() -> tantivy::Result<()> {
// heap for the indexer can increase its throughput. // heap for the indexer can increase its throughput.
let mut index_writer = index.writer(50_000_000)?; let mut index_writer = index.writer(50_000_000)?;
index_writer.add_document(doc!( index_writer.add_document(doc!(
title => "The Old Man and the Sea", title => "The Old Man and the Sea",
body => "He was an old man who fished alone in a skiff in the Gulf Stream and \ body => "He was an old man who fished alone in a skiff in the Gulf Stream and \
he had gone eighty-four days now without taking a fish." he had gone eighty-four days now without taking a fish."
)); ));
index_writer.add_document(doc!( index_writer.add_document(doc!(
title => "Of Mice and Men", title => "Of Mice and Men",
body => r#"A few miles south of Soledad, the Salinas River drops in close to the hillside body => r#"A few miles south of Soledad, the Salinas River drops in close to the hillside
bank and runs deep and green. The water is warm too, for it has slipped twinkling bank and runs deep and green. The water is warm too, for it has slipped twinkling
over the yellow sands in the sunlight before reaching the narrow pool. On one over the yellow sands in the sunlight before reaching the narrow pool. On one
@@ -84,16 +84,16 @@ fn main() -> tantivy::Result<()> {
limbs and branches that arch over the pool"# limbs and branches that arch over the pool"#
)); ));
index_writer.add_document(doc!( index_writer.add_document(doc!(
title => "Frankenstein", title => "Frankenstein",
body => r#"You will rejoice to hear that no disaster has accompanied the commencement of an body => r#"You will rejoice to hear that no disaster has accompanied the commencement of an
enterprise which you have regarded with such evil forebodings. I arrived here enterprise which you have regarded with such evil forebodings. I arrived here
yesterday, and my first task is to assure my dear sister of my welfare and yesterday, and my first task is to assure my dear sister of my welfare and
increasing confidence in the success of my undertaking."# increasing confidence in the success of my undertaking."#
)); ));
index_writer.commit()?; index_writer.commit()?;
index.load_searchers()?;
let reader = index.reader()?; let searcher = index.searcher();
let searcher = reader.searcher();
// The query parser can interpret human queries. // The query parser can interpret human queries.
// Here, if the user does not specify which // Here, if the user does not specify which
@@ -104,9 +104,11 @@ fn main() -> tantivy::Result<()> {
// here we want to get a hit on the 'ken' in Frankenstein // here we want to get a hit on the 'ken' in Frankenstein
let query = query_parser.parse_query("ken")?; let query = query_parser.parse_query("ken")?;
let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?; let mut top_collector = TopCollector::with_limit(10);
searcher.search(&*query, &mut top_collector)?;
for (_, doc_address) in top_docs { let doc_addresses = top_collector.docs();
for doc_address in doc_addresses {
let retrieved_doc = searcher.doc(doc_address)?; let retrieved_doc = searcher.doc(doc_address)?;
println!("{}", schema.to_json(&retrieved_doc)); println!("{}", schema.to_json(&retrieved_doc));
} }

25
examples/deleting_updating_documents.rs
View File

@@ -10,20 +10,16 @@
// Importing tantivy... // Importing tantivy...
#[macro_use] #[macro_use]
extern crate tantivy; extern crate tantivy;
use tantivy::collector::TopDocs; use tantivy::collector::TopCollector;
use tantivy::query::TermQuery; use tantivy::query::TermQuery;
use tantivy::schema::*; use tantivy::schema::*;
use tantivy::Index; use tantivy::Index;
use tantivy::IndexReader;
// A simple helper function to fetch a single document // A simple helper function to fetch a single document
// given its id from our index. // given its id from our index.
// It will be helpful to check our work. // It will be helpful to check our work.
fn extract_doc_given_isbn( fn extract_doc_given_isbn(index: &Index, isbn_term: &Term) -> tantivy::Result<Option<Document>> {
reader: &IndexReader, let searcher = index.searcher();
isbn_term: &Term,
) -> tantivy::Result<Option<Document>> {
let searcher = reader.searcher();
// This is the simplest query you can think of. // This is the simplest query you can think of.
// It matches all of the documents containing a specific term. // It matches all of the documents containing a specific term.
@@ -31,9 +27,10 @@ fn extract_doc_given_isbn(
// The second argument is here to tell we don't care about decoding positions, // The second argument is here to tell we don't care about decoding positions,
// or term frequencies. // or term frequencies.
let term_query = TermQuery::new(isbn_term.clone(), IndexRecordOption::Basic); let term_query = TermQuery::new(isbn_term.clone(), IndexRecordOption::Basic);
let top_docs = searcher.search(&term_query, &TopDocs::with_limit(1))?; let mut top_collector = TopCollector::with_limit(1);
searcher.search(&term_query, &mut top_collector)?;
if let Some((_score, doc_address)) = top_docs.first() { if let Some(doc_address) = top_collector.docs().first() {
let doc = searcher.doc(*doc_address)?; let doc = searcher.doc(*doc_address)?;
Ok(Some(doc)) Ok(Some(doc))
} else { } else {
@@ -47,7 +44,7 @@ fn main() -> tantivy::Result<()> {
// //
// Check out the *basic_search* example if this makes // Check out the *basic_search* example if this makes
// small sense to you. // small sense to you.
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
// Tantivy does not really have a notion of primary id. // Tantivy does not really have a notion of primary id.
// This may change in the future. // This may change in the future.
@@ -89,12 +86,12 @@ fn main() -> tantivy::Result<()> {
isbn => "978-9176370711", isbn => "978-9176370711",
)); ));
index_writer.commit()?; index_writer.commit()?;
let reader = index.reader()?; index.load_searchers()?;
let frankenstein_isbn = Term::from_field_text(isbn, "978-9176370711"); let frankenstein_isbn = Term::from_field_text(isbn, "978-9176370711");
// Oops our frankenstein doc seems mispelled // Oops our frankenstein doc seems mispelled
let frankenstein_doc_misspelled = extract_doc_given_isbn(&reader, &frankenstein_isbn)?.unwrap(); let frankenstein_doc_misspelled = extract_doc_given_isbn(&index, &frankenstein_isbn)?.unwrap();
assert_eq!( assert_eq!(
schema.to_json(&frankenstein_doc_misspelled), schema.to_json(&frankenstein_doc_misspelled),
r#"{"isbn":["978-9176370711"],"title":["Frankentein"]}"#, r#"{"isbn":["978-9176370711"],"title":["Frankentein"]}"#,
@@ -133,10 +130,10 @@ fn main() -> tantivy::Result<()> {
// Everything happened as if the document was updated. // Everything happened as if the document was updated.
index_writer.commit()?; index_writer.commit()?;
// We reload our searcher to make our change available to clients. // We reload our searcher to make our change available to clients.
reader.reload()?; index.load_searchers()?;
// No more typo! // No more typo!
let frankenstein_new_doc = extract_doc_given_isbn(&reader, &frankenstein_isbn)?.unwrap(); let frankenstein_new_doc = extract_doc_given_isbn(&index, &frankenstein_isbn)?.unwrap();
assert_eq!( assert_eq!(
schema.to_json(&frankenstein_new_doc), schema.to_json(&frankenstein_new_doc),
r#"{"isbn":["978-9176370711"],"title":["Frankenstein"]}"#, r#"{"isbn":["978-9176370711"],"title":["Frankenstein"]}"#,

11
examples/faceted_search.rs
View File

@@ -25,7 +25,7 @@ fn main() -> tantivy::Result<()> {
// Let's create a temporary directory for the // Let's create a temporary directory for the
// sake of this example // sake of this example
let index_path = TempDir::new("tantivy_facet_example_dir")?; let index_path = TempDir::new("tantivy_facet_example_dir")?;
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
schema_builder.add_text_field("name", TEXT | STORED); schema_builder.add_text_field("name", TEXT | STORED);
@@ -55,17 +55,18 @@ fn main() -> tantivy::Result<()> {
index_writer.commit()?; index_writer.commit()?;
let reader = index.reader()?; index.load_searchers()?;
let searcher = reader.searcher(); let searcher = index.searcher();
let mut facet_collector = FacetCollector::for_field(tags); let mut facet_collector = FacetCollector::for_field(tags);
facet_collector.add_facet("/pools"); facet_collector.add_facet("/pools");
let facet_counts = searcher.search(&AllQuery, &facet_collector).unwrap(); searcher.search(&AllQuery, &mut facet_collector).unwrap();
let counts = facet_collector.harvest();
// This lists all of the facet counts // This lists all of the facet counts
let facets: Vec<(&Facet, u64)> = facet_counts.get("/pools").collect(); let facets: Vec<(&Facet, u64)> = counts.get("/pools").collect();
assert_eq!( assert_eq!(
facets, facets,
vec![ vec![

43
examples/integer_range_search.rs
View File

@@ -1,43 +0,0 @@
// # Searching a range on an indexed int field.
//
// Below is an example of creating an indexed integer field in your schema
// You can use RangeQuery to get a Count of all occurrences in a given range.
#[macro_use]
extern crate tantivy;
use tantivy::collector::Count;
use tantivy::query::RangeQuery;
use tantivy::schema::{Schema, INDEXED};
use tantivy::Index;
use tantivy::Result;
fn run() -> Result<()> {
// For the sake of simplicity, this schema will only have 1 field
let mut schema_builder = Schema::builder();
// `INDEXED` is a short-hand to indicate that our field should be "searchable".
let year_field = schema_builder.add_u64_field("year", INDEXED);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let reader = index.reader()?;
{
let mut index_writer = index.writer_with_num_threads(1, 6_000_000)?;
for year in 1950u64..2019u64 {
index_writer.add_document(doc!(year_field => year));
}
index_writer.commit()?;
// The index will be a range of years
}
reader.reload()?;
let searcher = reader.searcher();
// The end is excluded i.e. here we are searching up to 1969
let docs_in_the_sixties = RangeQuery::new_u64(year_field, 1960..1970);
// Uses a Count collector to sum the total number of docs in the range
let num_60s_books = searcher.search(&docs_in_the_sixties, &Count)?;
assert_eq!(num_60s_books, 10);
Ok(())
}
fn main() {
run().unwrap()
}

6
examples/iterating_docs_and_positions.rs
View File

@@ -18,7 +18,7 @@ use tantivy::{DocId, DocSet, Postings};
fn main() -> tantivy::Result<()> { fn main() -> tantivy::Result<()> {
// We first create a schema for the sake of the // We first create a schema for the sake of the
// example. Check the `basic_search` example for more information. // example. Check the `basic_search` example for more information.
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
// For this example, we need to make sure to index positions for our title // For this example, we need to make sure to index positions for our title
// field. `TEXT` precisely does this. // field. `TEXT` precisely does this.
@@ -33,9 +33,9 @@ fn main() -> tantivy::Result<()> {
index_writer.add_document(doc!(title => "The modern Promotheus")); index_writer.add_document(doc!(title => "The modern Promotheus"));
index_writer.commit()?; index_writer.commit()?;
let reader = index.reader()?; index.load_searchers()?;
let searcher = reader.searcher(); let searcher = index.searcher();
// A tantivy index is actually a collection of segments. // A tantivy index is actually a collection of segments.
// Similarly, a searcher just wraps a list `segment_reader`. // Similarly, a searcher just wraps a list `segment_reader`.

55
examples/snippet.rs
View File

@@ -10,11 +10,11 @@ extern crate tempdir;
// Importing tantivy... // Importing tantivy...
#[macro_use] #[macro_use]
extern crate tantivy; extern crate tantivy;
use tantivy::collector::TopDocs; use tantivy::collector::TopCollector;
use tantivy::query::QueryParser; use tantivy::query::QueryParser;
use tantivy::schema::*; use tantivy::schema::*;
use tantivy::Index; use tantivy::Index;
use tantivy::{Snippet, SnippetGenerator}; use tantivy::SnippetGenerator;
use tempdir::TempDir; use tempdir::TempDir;
fn main() -> tantivy::Result<()> { fn main() -> tantivy::Result<()> {
@@ -23,7 +23,7 @@ fn main() -> tantivy::Result<()> {
let index_path = TempDir::new("tantivy_example_dir")?; let index_path = TempDir::new("tantivy_example_dir")?;
// # Defining the schema // # Defining the schema
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let title = schema_builder.add_text_field("title", TEXT | STORED); let title = schema_builder.add_text_field("title", TEXT | STORED);
let body = schema_builder.add_text_field("body", TEXT | STORED); let body = schema_builder.add_text_field("body", TEXT | STORED);
let schema = schema_builder.build(); let schema = schema_builder.build();
@@ -35,52 +35,37 @@ fn main() -> tantivy::Result<()> {
// we'll only need one doc for this example. // we'll only need one doc for this example.
index_writer.add_document(doc!( index_writer.add_document(doc!(
title => "Of Mice and Men", title => "Of Mice and Men",
body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \ body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
bank and runs deep and green. The water is warm too, for it has slipped twinkling \ bank and runs deep and green. The water is warm too, for it has slipped twinkling \
over the yellow sands in the sunlight before reaching the narrow pool. On one \ over the yellow sands in the sunlight before reaching the narrow pool. On one \
side of the river the golden foothill slopes curve up to the strong and rocky \ side of the river the golden foothill slopes curve up to the strong and rocky \
Gabilan Mountains, but on the valley side the water is lined with trees—willows \ Gabilan Mountains, but on the valley side the water is lined with trees—willows \
fresh and green with every spring, carrying in their lower leaf junctures the \ fresh and green with every spring, carrying in their lower leaf junctures the \
debris of the winter’s flooding; and sycamores with mottled, white, recumbent \ debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
limbs and branches that arch over the pool" limbs and branches that arch over the pool"
)); ));
// ... // ...
index_writer.commit()?; index_writer.commit()?;
let reader = index.reader()?; index.load_searchers()?;
let searcher = reader.searcher();
let searcher = index.searcher();
let query_parser = QueryParser::for_index(&index, vec![title, body]); let query_parser = QueryParser::for_index(&index, vec![title, body]);
let query = query_parser.parse_query("sycamore spring")?; let query = query_parser.parse_query("sycamore spring")?;
let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?; let mut top_collector = TopCollector::with_limit(10);
searcher.search(&*query, &mut top_collector)?;
let snippet_generator = SnippetGenerator::create(&searcher, &*query, body)?; let snippet_generator = SnippetGenerator::new(&searcher, &*query, body)?;
for (score, doc_address) in top_docs { let doc_addresses = top_collector.docs();
for doc_address in doc_addresses {
let doc = searcher.doc(doc_address)?; let doc = searcher.doc(doc_address)?;
let snippet = snippet_generator.snippet_from_doc(&doc); let snippet = snippet_generator.snippet_from_doc(&doc);
println!("Document score {}:", score);
println!("title: {}", doc.get_first(title).unwrap().text().unwrap()); println!("title: {}", doc.get_first(title).unwrap().text().unwrap());
println!("snippet: {}", snippet.to_html()); println!("snippet: {}", snippet.to_html());
println!("custom highlighting: {}", highlight(snippet));
} }
Ok(()) Ok(())
} }
fn highlight(snippet: Snippet) -> String {
let mut result = String::new();
let mut start_from = 0;
for (start, end) in snippet.highlighted().iter().map(|h| h.bounds()) {
result.push_str(&snippet.fragments()[start_from..start]);
result.push_str(" --> ");
result.push_str(&snippet.fragments()[start..end]);
result.push_str(" <-- ");
start_from = end;
}
result.push_str(&snippet.fragments()[start_from..]);
result
}

50
examples/stop_words.rs
View File

@@ -15,7 +15,7 @@ extern crate tempdir;
// Importing tantivy... // Importing tantivy...
#[macro_use] #[macro_use]
extern crate tantivy; extern crate tantivy;
use tantivy::collector::TopDocs; use tantivy::collector::TopCollector;
use tantivy::query::QueryParser; use tantivy::query::QueryParser;
use tantivy::schema::*; use tantivy::schema::*;
use tantivy::tokenizer::*; use tantivy::tokenizer::*;
@@ -23,7 +23,7 @@ use tantivy::Index;
fn main() -> tantivy::Result<()> { fn main() -> tantivy::Result<()> {
// this example assumes you understand the content in `basic_search` // this example assumes you understand the content in `basic_search`
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
// This configures your custom options for how tantivy will // This configures your custom options for how tantivy will
// store and process your content in the index; The key // store and process your content in the index; The key
@@ -72,44 +72,48 @@ fn main() -> tantivy::Result<()> {
title => "The Old Man and the Sea", title => "The Old Man and the Sea",
body => "He was an old man who fished alone in a skiff in the Gulf Stream and \ body => "He was an old man who fished alone in a skiff in the Gulf Stream and \
he had gone eighty-four days now without taking a fish." he had gone eighty-four days now without taking a fish."
)); ));
index_writer.add_document(doc!( index_writer.add_document(doc!(
title => "Of Mice and Men", title => "Of Mice and Men",
body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \ body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
bank and runs deep and green. The water is warm too, for it has slipped twinkling \ bank and runs deep and green. The water is warm too, for it has slipped twinkling \
over the yellow sands in the sunlight before reaching the narrow pool. On one \ over the yellow sands in the sunlight before reaching the narrow pool. On one \
side of the river the golden foothill slopes curve up to the strong and rocky \ side of the river the golden foothill slopes curve up to the strong and rocky \
Gabilan Mountains, but on the valley side the water is lined with trees—willows \ Gabilan Mountains, but on the valley side the water is lined with trees—willows \
fresh and green with every spring, carrying in their lower leaf junctures the \ fresh and green with every spring, carrying in their lower leaf junctures the \
debris of the winter’s flooding; and sycamores with mottled, white, recumbent \ debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
limbs and branches that arch over the pool" limbs and branches that arch over the pool"
)); ));
index_writer.add_document(doc!( index_writer.add_document(doc!(
title => "Frankenstein", title => "Frankenstein",
body => "You will rejoice to hear that no disaster has accompanied the commencement of an \ body => "You will rejoice to hear that no disaster has accompanied the commencement of an \
enterprise which you have regarded with such evil forebodings. I arrived here \ enterprise which you have regarded with such evil forebodings. I arrived here \
yesterday, and my first task is to assure my dear sister of my welfare and \ yesterday, and my first task is to assure my dear sister of my welfare and \
increasing confidence in the success of my undertaking." increasing confidence in the success of my undertaking."
)); ));
index_writer.commit()?; index_writer.commit()?;
let reader = index.reader()?; index.load_searchers()?;
let searcher = reader.searcher(); let searcher = index.searcher();
let query_parser = QueryParser::for_index(&index, vec![title, body]); let query_parser = QueryParser::for_index(&index, vec![title, body]);
// stop words are applied on the query as well. // stop words are applied on the query as well.
// The following will be equivalent to `title:frankenstein` // The following will be equivalent to `title:frankenstein`
let query = query_parser.parse_query("title:\"the Frankenstein\"")?; let query = query_parser.parse_query("title:\"the Frankenstein\"")?;
let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
for (score, doc_address) in top_docs { let mut top_collector = TopCollector::with_limit(10);
searcher.search(&*query, &mut top_collector)?;
let doc_addresses = top_collector.docs();
for doc_address in doc_addresses {
let retrieved_doc = searcher.doc(doc_address)?; let retrieved_doc = searcher.doc(doc_address)?;
println!("\n==\nDocument score {}:", score);
println!("{}", schema.to_json(&retrieved_doc)); println!("{}", schema.to_json(&retrieved_doc));
} }

4
examples/working_with_json.rs
View File

@@ -9,10 +9,10 @@ fn main() -> tantivy::Result<()> {
// Check out the basic example if this is confusing to you. // Check out the basic example if this is confusing to you.
// //
// first we need to define a schema ... // first we need to define a schema ...
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
schema_builder.add_text_field("title", TEXT | STORED); schema_builder.add_text_field("title", TEXT | STORED);
schema_builder.add_text_field("body", TEXT); schema_builder.add_text_field("body", TEXT);
schema_builder.add_u64_field("year", INDEXED); schema_builder.add_u64_field("year", INT_INDEXED);
let schema = schema_builder.build(); let schema = schema_builder.build();
// Let's assume we have a json-serialized document. // Let's assume we have a json-serialized document.

142
src/collector/chained_collector.rs Normal file
View File

@@ -0,0 +1,142 @@
use collector::Collector;
use DocId;
use Result;
use Score;
use SegmentLocalId;
use SegmentReader;
/// Collector that does nothing.
/// This is used in the chain Collector and will hopefully
/// be optimized away by the compiler.
pub struct DoNothingCollector;
impl Collector for DoNothingCollector {
#[inline]
fn set_segment(&mut self, _: SegmentLocalId, _: &SegmentReader) -> Result<()> {
Ok(())
}
#[inline]
fn collect(&mut self, _doc: DocId, _score: Score) {}
#[inline]
fn requires_scoring(&self) -> bool {
false
}
}
/// Zero-cost abstraction used to collect on multiple collectors.
/// This contraption is only usable if the type of your collectors
/// are known at compile time.
///
/// ```rust
/// #[macro_use]
/// extern crate tantivy;
/// use tantivy::schema::{SchemaBuilder, TEXT};
/// use tantivy::{Index, Result};
/// use tantivy::collector::{CountCollector, TopCollector, chain};
/// use tantivy::query::QueryParser;
///
/// # fn main() { example().unwrap(); }
/// fn example() -> Result<()> {
/// let mut schema_builder = SchemaBuilder::new();
/// let title = schema_builder.add_text_field("title", TEXT);
/// let schema = schema_builder.build();
/// let index = Index::create_in_ram(schema);
/// {
/// let mut index_writer = index.writer(3_000_000)?;
/// index_writer.add_document(doc!(
/// title => "The Name of the Wind",
/// ));
/// index_writer.add_document(doc!(
/// title => "The Diary of Muadib",
/// ));
/// index_writer.add_document(doc!(
/// title => "A Dairy Cow",
/// ));
/// index_writer.add_document(doc!(
/// title => "The Diary of a Young Girl",
/// ));
/// index_writer.commit().unwrap();
/// }
///
/// index.load_searchers()?;
/// let searcher = index.searcher();
///
/// {
/// let mut top_collector = TopCollector::with_limit(2);
/// let mut count_collector = CountCollector::default();
/// {
/// let mut collectors = chain().push(&mut top_collector).push(&mut count_collector);
/// let query_parser = QueryParser::for_index(&index, vec![title]);
/// let query = query_parser.parse_query("diary")?;
/// searcher.search(&*query, &mut collectors).unwrap();
/// }
/// assert_eq!(count_collector.count(), 2);
/// assert!(top_collector.at_capacity());
/// }
///
/// Ok(())
/// }
/// ```
pub struct ChainedCollector<Left: Collector, Right: Collector> {
left: Left,
right: Right,
}
impl<Left: Collector, Right: Collector> ChainedCollector<Left, Right> {
/// Adds a collector
pub fn push<C: Collector>(self, new_collector: &mut C) -> ChainedCollector<Self, &mut C> {
ChainedCollector {
left: self,
right: new_collector,
}
}
}
impl<Left: Collector, Right: Collector> Collector for ChainedCollector<Left, Right> {
fn set_segment(
&mut self,
segment_local_id: SegmentLocalId,
segment: &SegmentReader,
) -> Result<()> {
self.left.set_segment(segment_local_id, segment)?;
self.right.set_segment(segment_local_id, segment)?;
Ok(())
}
fn collect(&mut self, doc: DocId, score: Score) {
self.left.collect(doc, score);
self.right.collect(doc, score);
}
fn requires_scoring(&self) -> bool {
self.left.requires_scoring() || self.right.requires_scoring()
}
}
/// Creates a `ChainedCollector`
pub fn chain() -> ChainedCollector<DoNothingCollector, DoNothingCollector> {
ChainedCollector {
left: DoNothingCollector,
right: DoNothingCollector,
}
}
#[cfg(test)]
mod tests {
use super::*;
use collector::{Collector, CountCollector, TopCollector};
#[test]
fn test_chained_collector() {
let mut top_collector = TopCollector::with_limit(2);
let mut count_collector = CountCollector::default();
{
let mut collectors = chain().push(&mut top_collector).push(&mut count_collector);
collectors.collect(1, 0.2);
collectors.collect(2, 0.1);
collectors.collect(3, 0.5);
}
assert_eq!(count_collector.count(), 3);
assert!(top_collector.at_capacity());
}
}

96
src/collector/count_collector.rs
View File

@@ -1,5 +1,4 @@
use super::Collector; use super::Collector;
use collector::SegmentCollector;
use DocId; use DocId;
use Result; use Result;
use Score; use Score;
@@ -12,14 +11,14 @@ use SegmentReader;
/// ```rust /// ```rust
/// #[macro_use] /// #[macro_use]
/// extern crate tantivy; /// extern crate tantivy;
/// use tantivy::schema::{Schema, TEXT}; /// use tantivy::schema::{SchemaBuilder, TEXT};
/// use tantivy::{Index, Result}; /// use tantivy::{Index, Result};
/// use tantivy::collector::Count; /// use tantivy::collector::CountCollector;
/// use tantivy::query::QueryParser; /// use tantivy::query::QueryParser;
/// ///
/// # fn main() { example().unwrap(); } /// # fn main() { example().unwrap(); }
/// fn example() -> Result<()> { /// fn example() -> Result<()> {
/// let mut schema_builder = Schema::builder(); /// let mut schema_builder = SchemaBuilder::new();
/// let title = schema_builder.add_text_field("title", TEXT); /// let title = schema_builder.add_text_field("title", TEXT);
/// let schema = schema_builder.build(); /// let schema = schema_builder.build();
/// let index = Index::create_in_ram(schema); /// let index = Index::create_in_ram(schema);
@@ -40,90 +39,63 @@ use SegmentReader;
/// index_writer.commit().unwrap(); /// index_writer.commit().unwrap();
/// } /// }
/// ///
/// let reader = index.reader()?; /// index.load_searchers()?;
/// let searcher = reader.searcher(); /// let searcher = index.searcher();
/// ///
/// { /// {
/// let mut count_collector = CountCollector::default();
/// let query_parser = QueryParser::for_index(&index, vec![title]); /// let query_parser = QueryParser::for_index(&index, vec![title]);
/// let query = query_parser.parse_query("diary")?; /// let query = query_parser.parse_query("diary")?;
/// let count = searcher.search(&query, &Count).unwrap(); /// searcher.search(&*query, &mut count_collector).unwrap();
/// ///
/// assert_eq!(count, 2); /// assert_eq!(count_collector.count(), 2);
/// } /// }
/// ///
/// Ok(()) /// Ok(())
/// } /// }
/// ``` /// ```
pub struct Count;
impl Collector for Count {
type Fruit = usize;
type Child = SegmentCountCollector;
fn for_segment(&self, _: SegmentLocalId, _: &SegmentReader) -> Result<SegmentCountCollector> {
Ok(SegmentCountCollector::default())
}
fn requires_scoring(&self) -> bool {
false
}
fn merge_fruits(&self, segment_counts: Vec<usize>) -> Result<usize> {
Ok(segment_counts.into_iter().sum())
}
}
#[derive(Default)] #[derive(Default)]
pub struct SegmentCountCollector { pub struct CountCollector {
count: usize, count: usize,
} }
impl SegmentCollector for SegmentCountCollector { impl CountCollector {
type Fruit = usize; /// Returns the count of documents that were
/// collected.
pub fn count(&self) -> usize {
self.count
}
}
impl Collector for CountCollector {
fn set_segment(&mut self, _: SegmentLocalId, _: &SegmentReader) -> Result<()> {
Ok(())
}
fn collect(&mut self, _: DocId, _: Score) { fn collect(&mut self, _: DocId, _: Score) {
self.count += 1; self.count += 1;
} }
fn harvest(self) -> usize { fn requires_scoring(&self) -> bool {
self.count false
} }
} }
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::{Count, SegmentCountCollector};
use collector::Collector; use collector::{Collector, CountCollector};
use collector::SegmentCollector;
#[test] #[test]
fn test_count_collect_does_not_requires_scoring() { fn test_count_collector() {
assert!(!Count.requires_scoring()); let mut count_collector = CountCollector::default();
} assert_eq!(count_collector.count(), 0);
count_collector.collect(0u32, 1f32);
#[test] assert_eq!(count_collector.count(), 1);
fn test_segment_count_collector() { assert_eq!(count_collector.count(), 1);
{ count_collector.collect(1u32, 1f32);
let count_collector = SegmentCountCollector::default(); assert_eq!(count_collector.count(), 2);
assert_eq!(count_collector.harvest(), 0); assert!(!count_collector.requires_scoring());
}
{
let mut count_collector = SegmentCountCollector::default();
count_collector.collect(0u32, 1f32);
assert_eq!(count_collector.harvest(), 1);
}
{
let mut count_collector = SegmentCountCollector::default();
count_collector.collect(0u32, 1f32);
assert_eq!(count_collector.harvest(), 1);
}
{
let mut count_collector = SegmentCountCollector::default();
count_collector.collect(0u32, 1f32);
count_collector.collect(1u32, 1f32);
assert_eq!(count_collector.harvest(), 2);
}
} }
} }

355
src/collector/facet_collector.rs
View File

@@ -1,17 +1,20 @@
use collector::Collector; use collector::Collector;
use collector::SegmentCollector;
use docset::SkipResult; use docset::SkipResult;
use fastfield::FacetReader; use fastfield::FacetReader;
use schema::Facet; use schema::Facet;
use schema::Field; use schema::Field;
use std::cmp::Ordering; use std::cell::UnsafeCell;
use std::collections::btree_map; use std::collections::btree_map;
use std::collections::BTreeMap; use std::collections::BTreeMap;
use std::collections::BTreeSet; use std::collections::BTreeSet;
use std::collections::BinaryHeap; use std::collections::BinaryHeap;
use std::collections::Bound; use std::collections::Bound;
use std::iter::Peekable; use std::iter::Peekable;
use std::mem;
use std::{u64, usize}; use std::{u64, usize};
use termdict::TermMerger;
use std::cmp::Ordering;
use DocId; use DocId;
use Result; use Result;
use Score; use Score;
@@ -43,6 +46,12 @@ impl<'a> Ord for Hit<'a> {
} }
} }
struct SegmentFacetCounter {
pub facet_reader: FacetReader,
pub facet_ords: Vec<u64>,
pub facet_counts: Vec<u64>,
}
fn facet_depth(facet_bytes: &[u8]) -> usize { fn facet_depth(facet_bytes: &[u8]) -> usize {
if facet_bytes.is_empty() { if facet_bytes.is_empty() {
0 0
@@ -82,14 +91,14 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
/// ```rust /// ```rust
/// #[macro_use] /// #[macro_use]
/// extern crate tantivy; /// extern crate tantivy;
/// use tantivy::schema::{Facet, Schema, TEXT}; /// use tantivy::schema::{Facet, SchemaBuilder, TEXT};
/// use tantivy::{Index, Result}; /// use tantivy::{Index, Result};
/// use tantivy::collector::FacetCollector; /// use tantivy::collector::FacetCollector;
/// use tantivy::query::AllQuery; /// use tantivy::query::AllQuery;
/// ///
/// # fn main() { example().unwrap(); } /// # fn main() { example().unwrap(); }
/// fn example() -> Result<()> { /// fn example() -> Result<()> {
/// let mut schema_builder = Schema::builder(); /// let mut schema_builder = SchemaBuilder::new();
/// ///
/// // Facet have their own specific type. /// // Facet have their own specific type.
/// // It is not a bad practise to put all of your /// // It is not a bad practise to put all of your
@@ -122,19 +131,23 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
/// facet => Facet::from("/lang/en"), /// facet => Facet::from("/lang/en"),
/// facet => Facet::from("/category/biography") /// facet => Facet::from("/category/biography")
/// )); /// ));
/// index_writer.commit()?; /// index_writer.commit().unwrap();
/// } /// }
/// let reader = index.reader()?; ///
/// let searcher = reader.searcher(); /// index.load_searchers()?;
/// let searcher = index.searcher();
/// ///
/// { /// {
/// let mut facet_collector = FacetCollector::for_field(facet); /// let mut facet_collector = FacetCollector::for_field(facet);
/// facet_collector.add_facet("/lang"); /// facet_collector.add_facet("/lang");
/// facet_collector.add_facet("/category"); /// facet_collector.add_facet("/category");
/// let facet_counts = searcher.search(&AllQuery, &facet_collector)?; /// searcher.search(&AllQuery, &mut facet_collector).unwrap();
///
/// // this object contains count aggregate for all of the facets.
/// let counts = facet_collector.harvest();
/// ///
/// // This lists all of the facet counts /// // This lists all of the facet counts
/// let facets: Vec<(&Facet, u64)> = facet_counts /// let facets: Vec<(&Facet, u64)> = counts
/// .get("/category") /// .get("/category")
/// .collect(); /// .collect();
/// assert_eq!(facets, vec![ /// assert_eq!(facets, vec![
@@ -146,10 +159,13 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
/// { /// {
/// let mut facet_collector = FacetCollector::for_field(facet); /// let mut facet_collector = FacetCollector::for_field(facet);
/// facet_collector.add_facet("/category/fiction"); /// facet_collector.add_facet("/category/fiction");
/// let facet_counts = searcher.search(&AllQuery, &facet_collector)?; /// searcher.search(&AllQuery, &mut facet_collector).unwrap();
///
/// // this object contains count aggregate for all of the facets.
/// let counts = facet_collector.harvest();
/// ///
/// // This lists all of the facet counts /// // This lists all of the facet counts
/// let facets: Vec<(&Facet, u64)> = facet_counts /// let facets: Vec<(&Facet, u64)> = counts
/// .get("/category/fiction") /// .get("/category/fiction")
/// .collect(); /// .collect();
/// assert_eq!(facets, vec![ /// assert_eq!(facets, vec![
@@ -162,10 +178,13 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
/// { /// {
/// let mut facet_collector = FacetCollector::for_field(facet); /// let mut facet_collector = FacetCollector::for_field(facet);
/// facet_collector.add_facet("/category/fiction"); /// facet_collector.add_facet("/category/fiction");
/// let facet_counts = searcher.search(&AllQuery, &facet_collector)?; /// searcher.search(&AllQuery, &mut facet_collector).unwrap();
///
/// // this object contains count aggregate for all of the facets.
/// let counts = facet_collector.harvest();
/// ///
/// // This lists all of the facet counts /// // This lists all of the facet counts
/// let facets: Vec<(&Facet, u64)> = facet_counts.top_k("/category/fiction", 1); /// let facets: Vec<(&Facet, u64)> = counts.top_k("/category/fiction", 1);
/// assert_eq!(facets, vec![ /// assert_eq!(facets, vec![
/// (&Facet::from("/category/fiction/fantasy"), 2) /// (&Facet::from("/category/fiction/fantasy"), 2)
/// ]); /// ]);
@@ -175,19 +194,19 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
/// } /// }
/// ``` /// ```
pub struct FacetCollector { pub struct FacetCollector {
facet_ords: Vec<u64>,
field: Field, field: Field,
facets: BTreeSet<Facet>, ff_reader: Option<UnsafeCell<FacetReader>>,
} segment_counters: Vec<SegmentFacetCounter>,
pub struct FacetSegmentCollector {
reader: FacetReader,
facet_ords_buf: Vec<u64>,
// facet_ord -> collapse facet_id // facet_ord -> collapse facet_id
collapse_mapping: Vec<usize>, current_segment_collapse_mapping: Vec<usize>,
// collapse facet_id -> count // collapse facet_id -> count
counts: Vec<u64>, current_segment_counts: Vec<u64>,
// collapse facet_id -> facet_ord // collapse facet_id -> facet_ord
collapse_facet_ords: Vec<u64>, current_collapse_facet_ords: Vec<u64>,
facets: BTreeSet<Facet>,
} }
fn skip<'a, I: Iterator<Item = &'a Facet>>( fn skip<'a, I: Iterator<Item = &'a Facet>>(
@@ -196,7 +215,7 @@ fn skip<'a, I: Iterator<Item = &'a Facet>>(
) -> SkipResult { ) -> SkipResult {
loop { loop {
match collapse_it.peek() { match collapse_it.peek() {
Some(facet_bytes) => match facet_bytes.encoded_str().as_bytes().cmp(target) { Some(facet_bytes) => match facet_bytes.encoded_bytes().cmp(target) {
Ordering::Less => {} Ordering::Less => {}
Ordering::Greater => { Ordering::Greater => {
return SkipResult::OverStep; return SkipResult::OverStep;
@@ -221,8 +240,15 @@ impl FacetCollector {
/// is of the proper type. /// is of the proper type.
pub fn for_field(field: Field) -> FacetCollector { pub fn for_field(field: Field) -> FacetCollector {
FacetCollector { FacetCollector {
facet_ords: Vec::with_capacity(255),
segment_counters: Vec::new(),
field, field,
facets: BTreeSet::default(), ff_reader: None,
facets: BTreeSet::new(),
current_segment_collapse_mapping: Vec::new(),
current_collapse_facet_ords: Vec::new(),
current_segment_counts: Vec::new(),
} }
} }
@@ -252,100 +278,143 @@ impl FacetCollector {
} }
self.facets.insert(facet); self.facets.insert(facet);
} }
}
impl Collector for FacetCollector {
type Fruit = FacetCounts;
type Child = FacetSegmentCollector;
fn for_segment(
&self,
_: SegmentLocalId,
reader: &SegmentReader,
) -> Result<FacetSegmentCollector> {
let facet_reader = reader.facet_reader(self.field)?;
let mut collapse_mapping = Vec::new();
let mut counts = Vec::new();
let mut collapse_facet_ords = Vec::new();
fn set_collapse_mapping(&mut self, facet_reader: &FacetReader) {
self.current_segment_collapse_mapping.clear();
self.current_collapse_facet_ords.clear();
self.current_segment_counts.clear();
let mut collapse_facet_it = self.facets.iter().peekable(); let mut collapse_facet_it = self.facets.iter().peekable();
collapse_facet_ords.push(0); self.current_collapse_facet_ords.push(0);
{ let mut facet_streamer = facet_reader.facet_dict().range().into_stream();
let mut facet_streamer = facet_reader.facet_dict().range().into_stream(); if !facet_streamer.advance() {
if facet_streamer.advance() { return;
'outer: loop { }
// at the begining of this loop, facet_streamer 'outer: loop {
// is positionned on a term that has not been processed yet. // at the begining of this loop, facet_streamer
let skip_result = skip(facet_streamer.key(), &mut collapse_facet_it); // is positionned on a term that has not been processed yet.
match skip_result { let skip_result = skip(facet_streamer.key(), &mut collapse_facet_it);
SkipResult::Reached => { match skip_result {
// we reach a facet we decided to collapse. SkipResult::Reached => {
let collapse_depth = facet_depth(facet_streamer.key()); // we reach a facet we decided to collapse.
let mut collapsed_id = 0; let collapse_depth = facet_depth(facet_streamer.key());
collapse_mapping.push(0); let mut collapsed_id = 0;
while facet_streamer.advance() { self.current_segment_collapse_mapping.push(0);
let depth = facet_depth(facet_streamer.key()); while facet_streamer.advance() {
if depth <= collapse_depth { let depth = facet_depth(facet_streamer.key());
continue 'outer; if depth <= collapse_depth {
} continue 'outer;
if depth == collapse_depth + 1 {
collapsed_id = collapse_facet_ords.len();
collapse_facet_ords.push(facet_streamer.term_ord());
collapse_mapping.push(collapsed_id);
} else {
collapse_mapping.push(collapsed_id);
}
}
break;
} }
SkipResult::End | SkipResult::OverStep => { if depth == collapse_depth + 1 {
collapse_mapping.push(0); collapsed_id = self.current_collapse_facet_ords.len();
if !facet_streamer.advance() { self.current_collapse_facet_ords
break; .push(facet_streamer.term_ord());
} self.current_segment_collapse_mapping.push(collapsed_id);
} else {
self.current_segment_collapse_mapping.push(collapsed_id);
} }
} }
break;
}
SkipResult::End | SkipResult::OverStep => {
self.current_segment_collapse_mapping.push(0);
if !facet_streamer.advance() {
break;
}
} }
} }
} }
counts.resize(collapse_facet_ords.len(), 0);
Ok(FacetSegmentCollector {
reader: facet_reader,
facet_ords_buf: Vec::with_capacity(255),
collapse_mapping,
counts,
collapse_facet_ords,
})
} }
fn requires_scoring(&self) -> bool { fn finalize_segment(&mut self) {
false if self.ff_reader.is_some() {
self.segment_counters.push(SegmentFacetCounter {
facet_reader: self.ff_reader.take().unwrap().into_inner(),
facet_ords: mem::replace(&mut self.current_collapse_facet_ords, Vec::new()),
facet_counts: mem::replace(&mut self.current_segment_counts, Vec::new()),
});
}
} }
fn merge_fruits(&self, segments_facet_counts: Vec<FacetCounts>) -> Result<FacetCounts> { /// Returns the results of the collection.
let mut facet_counts: BTreeMap<Facet, u64> = BTreeMap::new(); ///
for segment_facet_counts in segments_facet_counts { /// This method does not just return the counters,
for (facet, count) in segment_facet_counts.facet_counts { /// it also translates the facet ordinals of the last segment.
*(facet_counts.entry(facet).or_insert(0)) += count; pub fn harvest(mut self) -> FacetCounts {
self.finalize_segment();
let collapsed_facet_ords: Vec<&[u64]> = self
.segment_counters
.iter()
.map(|segment_counter| &segment_counter.facet_ords[..])
.collect();
let collapsed_facet_counts: Vec<&[u64]> = self
.segment_counters
.iter()
.map(|segment_counter| &segment_counter.facet_counts[..])
.collect();
let facet_streams = self
.segment_counters
.iter()
.map(|seg_counts| seg_counts.facet_reader.facet_dict().range().into_stream())
.collect::<Vec<_>>();
let mut facet_merger = TermMerger::new(facet_streams);
let mut facet_counts = BTreeMap::new();
while facet_merger.advance() {
let count = facet_merger
.current_kvs()
.iter()
.map(|it| {
let seg_ord = it.segment_ord;
let term_ord = it.streamer.term_ord();
collapsed_facet_ords[seg_ord]
.binary_search(&term_ord)
.map(|collapsed_term_id| {
if collapsed_term_id == 0 {
0
} else {
collapsed_facet_counts[seg_ord][collapsed_term_id]
}
}).unwrap_or(0)
}).sum();
if count > 0u64 {
let bytes: Vec<u8> = facet_merger.key().to_owned();
// may create an corrupted facet if the term dicitonary is corrupted
let facet = unsafe { Facet::from_encoded(bytes) };
facet_counts.insert(facet, count);
} }
} }
Ok(FacetCounts { facet_counts }) FacetCounts { facet_counts }
} }
} }
impl SegmentCollector for FacetSegmentCollector { impl Collector for FacetCollector {
type Fruit = FacetCounts; fn set_segment(&mut self, _: SegmentLocalId, reader: &SegmentReader) -> Result<()> {
self.finalize_segment();
let facet_reader = reader.facet_reader(self.field)?;
self.set_collapse_mapping(&facet_reader);
self.current_segment_counts
.resize(self.current_collapse_facet_ords.len(), 0);
self.ff_reader = Some(UnsafeCell::new(facet_reader));
Ok(())
}
fn collect(&mut self, doc: DocId, _: Score) { fn collect(&mut self, doc: DocId, _: Score) {
self.reader.facet_ords(doc, &mut self.facet_ords_buf); let facet_reader: &mut FacetReader = unsafe {
&mut *self
.ff_reader
.as_ref()
.expect("collect() was called before set_segment. This should never happen.")
.get()
};
facet_reader.facet_ords(doc, &mut self.facet_ords);
let mut previous_collapsed_ord: usize = usize::MAX; let mut previous_collapsed_ord: usize = usize::MAX;
for &facet_ord in &self.facet_ords_buf { for &facet_ord in &self.facet_ords {
let collapsed_ord = self.collapse_mapping[facet_ord as usize]; let collapsed_ord = self.current_segment_collapse_mapping[facet_ord as usize];
self.counts[collapsed_ord] += if collapsed_ord == previous_collapsed_ord { self.current_segment_counts[collapsed_ord] += if collapsed_ord == previous_collapsed_ord
{
0 0
} else { } else {
1 1
@@ -354,24 +423,8 @@ impl SegmentCollector for FacetSegmentCollector {
} }
} }
/// Returns the results of the collection. fn requires_scoring(&self) -> bool {
/// false
/// This method does not just return the counters,
/// it also translates the facet ordinals of the last segment.
fn harvest(self) -> FacetCounts {
let mut facet_counts = BTreeMap::new();
let facet_dict = self.reader.facet_dict();
for (collapsed_facet_ord, count) in self.counts.iter().cloned().enumerate() {
if count == 0 {
continue;
}
let mut facet = vec![];
let facet_ord = self.collapse_facet_ords[collapsed_facet_ord];
facet_dict.ord_to_term(facet_ord as u64, &mut facet);
// TODO
facet_counts.insert(Facet::from_encoded(facet).unwrap(), count);
}
FacetCounts { facet_counts }
} }
} }
@@ -403,9 +456,9 @@ impl FacetCounts {
let right_bound = if facet.is_root() { let right_bound = if facet.is_root() {
Bound::Unbounded Bound::Unbounded
} else { } else {
let mut facet_after_bytes: String = facet.encoded_str().to_owned(); let mut facet_after_bytes: Vec<u8> = facet.encoded_bytes().to_owned();
facet_after_bytes.push('\u{1}'); facet_after_bytes.push(1u8);
let facet_after = Facet::from_encoded_string(facet_after_bytes); let facet_after = unsafe { Facet::from_encoded(facet_after_bytes) }; // ok logic
Bound::Excluded(facet_after) Bound::Excluded(facet_after)
}; };
let underlying: btree_map::Range<_, _> = self.facet_counts.range((left_bound, right_bound)); let underlying: btree_map::Range<_, _> = self.facet_counts.range((left_bound, right_bound));
@@ -452,14 +505,14 @@ mod tests {
use core::Index; use core::Index;
use query::AllQuery; use query::AllQuery;
use rand::distributions::Uniform; use rand::distributions::Uniform;
use rand::prelude::SliceRandom;
use rand::{thread_rng, Rng}; use rand::{thread_rng, Rng};
use schema::{Document, Facet, Field, Schema}; use schema::Field;
use schema::{Document, Facet, SchemaBuilder};
use std::iter; use std::iter;
#[test] #[test]
fn test_facet_collector_drilldown() { fn test_facet_collector_drilldown() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
let facet_field = schema_builder.add_facet_field("facet"); let facet_field = schema_builder.add_facet_field("facet");
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
@@ -474,20 +527,21 @@ mod tests {
n /= 4; n /= 4;
let leaf = n % 5; let leaf = n % 5;
Facet::from(&format!("/top{}/mid{}/leaf{}", top, mid, leaf)) Facet::from(&format!("/top{}/mid{}/leaf{}", top, mid, leaf))
}) }).collect();
.collect();
for i in 0..num_facets * 10 { for i in 0..num_facets * 10 {
let mut doc = Document::new(); let mut doc = Document::new();
doc.add_facet(facet_field, facets[i % num_facets].clone()); doc.add_facet(facet_field, facets[i % num_facets].clone());
index_writer.add_document(doc); index_writer.add_document(doc);
} }
index_writer.commit().unwrap(); index_writer.commit().unwrap();
let reader = index.reader().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
let mut facet_collector = FacetCollector::for_field(facet_field); let mut facet_collector = FacetCollector::for_field(facet_field);
facet_collector.add_facet(Facet::from("/top1")); facet_collector.add_facet(Facet::from("/top1"));
let counts = searcher.search(&AllQuery, &facet_collector).unwrap(); searcher.search(&AllQuery, &mut facet_collector).unwrap();
let counts: FacetCounts = facet_collector.harvest();
{ {
let facets: Vec<(String, u64)> = counts let facets: Vec<(String, u64)> = counts
.get("/top1") .get("/top1")
@@ -501,16 +555,18 @@ mod tests {
("/top1/mid2", 50), ("/top1/mid2", 50),
("/top1/mid3", 50), ("/top1/mid3", 50),
] ]
.iter() .iter()
.map(|&(facet_str, count)| (String::from(facet_str), count)) .map(|&(facet_str, count)| (String::from(facet_str), count))
.collect::<Vec<_>>() .collect::<Vec<_>>()
); );
} }
} }
#[test] #[test]
#[should_panic(expected = "Tried to add a facet which is a descendant of \ #[should_panic(
an already added facet.")] expected = "Tried to add a facet which is a descendant of \
an already added facet."
)]
fn test_misused_facet_collector() { fn test_misused_facet_collector() {
let mut facet_collector = FacetCollector::for_field(Field(0)); let mut facet_collector = FacetCollector::for_field(Field(0));
facet_collector.add_facet(Facet::from("/country")); facet_collector.add_facet(Facet::from("/country"));
@@ -519,7 +575,7 @@ mod tests {
#[test] #[test]
fn test_doc_unsorted_multifacet() { fn test_doc_unsorted_multifacet() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
let facet_field = schema_builder.add_facet_field("facets"); let facet_field = schema_builder.add_facet_field("facets");
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
@@ -531,12 +587,13 @@ mod tests {
facet_field => Facet::from_text(&"/subjects/B/b"), facet_field => Facet::from_text(&"/subjects/B/b"),
)); ));
index_writer.commit().unwrap(); index_writer.commit().unwrap();
let reader = index.reader().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
assert_eq!(searcher.num_docs(), 1); assert_eq!(searcher.num_docs(), 1);
let mut facet_collector = FacetCollector::for_field(facet_field); let mut facet_collector = FacetCollector::for_field(facet_field);
facet_collector.add_facet("/subjects"); facet_collector.add_facet("/subjects");
let counts = searcher.search(&AllQuery, &facet_collector).unwrap(); searcher.search(&AllQuery, &mut facet_collector).unwrap();
let counts = facet_collector.harvest();
let facets: Vec<(&Facet, u64)> = counts.get("/subjects").collect(); let facets: Vec<(&Facet, u64)> = counts.get("/subjects").collect();
assert_eq!(facets[0].1, 1); assert_eq!(facets[0].1, 1);
} }
@@ -550,7 +607,7 @@ mod tests {
#[test] #[test]
fn test_facet_collector_topk() { fn test_facet_collector_topk() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
let facet_field = schema_builder.add_facet_field("facet"); let facet_field = schema_builder.add_facet_field("facet");
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
@@ -562,28 +619,29 @@ mod tests {
let facet = Facet::from(&format!("/facet/{}", c)); let facet = Facet::from(&format!("/facet/{}", c));
let doc = doc!(facet_field => facet); let doc = doc!(facet_field => facet);
iter::repeat(doc).take(count) iter::repeat(doc).take(count)
}) }).map(|mut doc| {
.map(|mut doc| {
doc.add_facet( doc.add_facet(
facet_field, facet_field,
&format!("/facet/{}", thread_rng().sample(&uniform)), &format!("/facet/{}", thread_rng().sample(&uniform)),
); );
doc doc
}) }).collect();
.collect(); thread_rng().shuffle(&mut docs[..]);
docs[..].shuffle(&mut thread_rng());
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
for doc in docs { for doc in docs {
index_writer.add_document(doc); index_writer.add_document(doc);
} }
index_writer.commit().unwrap(); index_writer.commit().unwrap();
let searcher = index.reader().unwrap().searcher(); index.load_searchers().unwrap();
let searcher = index.searcher();
let mut facet_collector = FacetCollector::for_field(facet_field); let mut facet_collector = FacetCollector::for_field(facet_field);
facet_collector.add_facet("/facet"); facet_collector.add_facet("/facet");
let counts: FacetCounts = searcher.search(&AllQuery, &facet_collector).unwrap(); searcher.search(&AllQuery, &mut facet_collector).unwrap();
let counts: FacetCounts = facet_collector.harvest();
{ {
let facets: Vec<(&Facet, u64)> = counts.top_k("/facet", 3); let facets: Vec<(&Facet, u64)> = counts.top_k("/facet", 3);
assert_eq!( assert_eq!(
@@ -606,13 +664,13 @@ mod bench {
use query::AllQuery; use query::AllQuery;
use rand::{thread_rng, Rng}; use rand::{thread_rng, Rng};
use schema::Facet; use schema::Facet;
use schema::Schema; use schema::SchemaBuilder;
use test::Bencher; use test::Bencher;
use Index; use Index;
#[bench] #[bench]
fn bench_facet_collector(b: &mut Bencher) { fn bench_facet_collector(b: &mut Bencher) {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
let facet_field = schema_builder.add_facet_field("facet"); let facet_field = schema_builder.add_facet_field("facet");
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
@@ -632,11 +690,12 @@ mod bench {
index_writer.add_document(doc); index_writer.add_document(doc);
} }
index_writer.commit().unwrap(); index_writer.commit().unwrap();
let reader = index.reader().unwrap(); index.load_searchers().unwrap();
b.iter(|| { b.iter(|| {
let searcher = index.searcher(); let searcher = index.searcher();
let facet_collector = FacetCollector::for_field(facet_field); let mut facet_collector = FacetCollector::for_field(facet_field);
searcher.search(&AllQuery, &facet_collector).unwrap(); searcher.search(&AllQuery, &mut facet_collector).unwrap();
}); });
} }
} }

7
src/collector/int_facet_collector.rs
View File

@@ -79,7 +79,7 @@ mod tests {
// make sure we have facet counters correctly filled // make sure we have facet counters correctly filled
fn test_facet_collector_results() { fn test_facet_collector_results() {
let mut schema_builder = schema::Schema::builder(); let mut schema_builder = schema::SchemaBuilder::new();
let num_field_i64 = schema_builder.add_i64_field("num_i64", FAST); let num_field_i64 = schema_builder.add_i64_field("num_i64", FAST);
let num_field_u64 = schema_builder.add_u64_field("num_u64", FAST); let num_field_u64 = schema_builder.add_u64_field("num_u64", FAST);
let text_field = schema_builder.add_text_field("text", STRING); let text_field = schema_builder.add_text_field("text", STRING);
@@ -88,7 +88,7 @@ mod tests {
let index = Index::create_in_ram(schema.clone()); let index = Index::create_in_ram(schema.clone());
{ {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
{ {
for i in 0u64..10u64 { for i in 0u64..10u64 {
index_writer.add_document(doc!( index_writer.add_document(doc!(
@@ -101,7 +101,8 @@ mod tests {
assert_eq!(index_writer.commit().unwrap(), 10u64); assert_eq!(index_writer.commit().unwrap(), 10u64);
} }
let searcher = index.reader().searcher(); index.load_searchers().unwrap();
let searcher = index.searcher();
let mut ffvf_i64: IntFacetCollector<I64FastFieldReader> = IntFacetCollector::new(num_field_i64); let mut ffvf_i64: IntFacetCollector<I64FastFieldReader> = IntFacetCollector::new(num_field_i64);
let mut ffvf_u64: IntFacetCollector<U64FastFieldReader> = IntFacetCollector::new(num_field_u64); let mut ffvf_u64: IntFacetCollector<U64FastFieldReader> = IntFacetCollector::new(num_field_u64);

513
src/collector/mod.rs
View File

@@ -1,91 +1,7 @@
/*! /*!
Defines how the documents matching a search query should be processed.
# Collectors
Collectors define the information you want to extract from the documents matching the queries.
In tantivy jargon, we call this information your search "fruit".
Your fruit could for instance be :
- [the count of matching documents](./struct.Count.html)
- [the top 10 documents, by relevancy or by a fast field](./struct.TopDocs.html)
- [facet counts](./struct.FacetCollector.html)
At one point in your code, you will trigger the actual search operation by calling
[the `search(...)` method of your `Searcher` object](../struct.Searcher.html#method.search).
This call will look like this.
```verbatim
let fruit = searcher.search(&query, &collector)?;
```
Here the type of fruit is actually determined as an associated type of the collector (`Collector::Fruit`).
# Combining several collectors
A rich search experience often requires to run several collectors on your search query.
For instance,
- selecting the top-K products matching your query
- counting the matching documents
- computing several facets
- computing statistics about the matching product prices
A simple and efficient way to do that is to pass your collectors as one tuple.
The resulting `Fruit` will then be a typed tuple with each collector's original fruits
in their respective position.
```rust
# extern crate tantivy;
# use tantivy::schema::*;
# use tantivy::*;
# use tantivy::query::*;
use tantivy::collector::{Count, TopDocs};
#
# fn main() -> tantivy::Result<()> {
# let mut schema_builder = Schema::builder();
# let title = schema_builder.add_text_field("title", TEXT);
# let schema = schema_builder.build();
# let index = Index::create_in_ram(schema);
# let mut index_writer = index.writer(3_000_000)?;
# index_writer.add_document(doc!(
# title => "The Name of the Wind",
# ));
# index_writer.add_document(doc!(
# title => "The Diary of Muadib",
# ));
# index_writer.commit()?;
# let reader = index.reader()?;
# let searcher = reader.searcher();
# let query_parser = QueryParser::for_index(&index, vec![title]);
# let query = query_parser.parse_query("diary")?;
let (doc_count, top_docs): (usize, Vec<(Score, DocAddress)>) =
searcher.search(&query, &(Count, TopDocs::with_limit(2)))?;
# Ok(())
# }
```
The `Collector` trait is implemented for up to 4 collectors.
If you have more than 4 collectors, you can either group them into
tuples of tuples `(a,(b,(c,d)))`, or rely on `MultiCollector`'s.
# Combining several collectors dynamically
Combining collectors into a tuple is a zero-cost abstraction: everything
happens as if you had manually implemented a single collector
combining all of our features.
Unfortunately it requires you to know at compile time your collector types.
If on the other hand, the collectors depend on some query parameter,
you can rely on `MultiCollector`'s.
# Implementing your own collectors.
See the `custom_collector` example.
*/ */
use downcast_rs;
use DocId; use DocId;
use Result; use Result;
use Score; use Score;
@@ -93,7 +9,7 @@ use SegmentLocalId;
use SegmentReader; use SegmentReader;
mod count_collector; mod count_collector;
pub use self::count_collector::Count; pub use self::count_collector::CountCollector;
mod multi_collector; mod multi_collector;
pub use self::multi_collector::MultiCollector; pub use self::multi_collector::MultiCollector;
@@ -101,264 +17,237 @@ pub use self::multi_collector::MultiCollector;
mod top_collector; mod top_collector;
mod top_score_collector; mod top_score_collector;
pub use self::top_score_collector::TopDocs; pub use self::top_score_collector::TopScoreCollector;
#[deprecated]
pub use self::top_score_collector::TopScoreCollector as TopCollector;
mod top_field_collector; mod top_field_collector;
pub use self::top_field_collector::TopDocsByField; pub use self::top_field_collector::TopFieldCollector;
mod facet_collector; mod facet_collector;
pub use self::facet_collector::FacetCollector; pub use self::facet_collector::FacetCollector;
/// `Fruit` is the type for the result of our collection. mod chained_collector;
/// e.g. `usize` for the `Count` collector. pub use self::chained_collector::{chain, ChainedCollector};
pub trait Fruit: Send + downcast_rs::Downcast {}
impl<T> Fruit for T where T: Send + downcast_rs::Downcast {}
/// Collectors are in charge of collecting and retaining relevant /// Collectors are in charge of collecting and retaining relevant
/// information from the document found and scored by the query. /// information from the document found and scored by the query.
/// ///
///
/// For instance, /// For instance,
/// ///
/// - keeping track of the top 10 best documents /// - keeping track of the top 10 best documents
/// - computing a breakdown over a fast field /// - computing a breakdown over a fast field
/// - computing the number of documents matching the query /// - computing the number of documents matching the query
/// ///
/// Our search index is in fact a collection of segments, so /// Queries are in charge of pushing the `DocSet` to the collector.
/// a `Collector` trait is actually more of a factory to instance
/// `SegmentCollector`s for each segments.
/// ///
/// The collection logic itself is in the `SegmentCollector`. /// As they work on multiple segments, they first inform
/// the collector of a change in a segment and then
/// call the `collect` method to push the document to the collector.
///
/// Temporally, our collector will receive calls
/// - `.set_segment(0, segment_reader_0)`
/// - `.collect(doc0_of_segment_0)`
/// - `.collect(...)`
/// - `.collect(last_doc_of_segment_0)`
/// - `.set_segment(1, segment_reader_1)`
/// - `.collect(doc0_of_segment_1)`
/// - `.collect(...)`
/// - `.collect(last_doc_of_segment_1)`
/// - `...`
/// - `.collect(last_doc_of_last_segment)`
/// ///
/// Segments are not guaranteed to be visited in any specific order. /// Segments are not guaranteed to be visited in any specific order.
pub trait Collector: Sync { pub trait Collector {
/// `Fruit` is the type for the result of our collection.
/// e.g. `usize` for the `Count` collector.
type Fruit: Fruit;
/// Type of the `SegmentCollector` associated to this collector.
type Child: SegmentCollector<Fruit = Self::Fruit>;
/// `set_segment` is called before beginning to enumerate /// `set_segment` is called before beginning to enumerate
/// on this segment. /// on this segment.
fn for_segment( fn set_segment(
&self, &mut self,
segment_local_id: SegmentLocalId, segment_local_id: SegmentLocalId,
segment: &SegmentReader, segment: &SegmentReader,
) -> Result<Self::Child>; ) -> Result<()>;
/// Returns true iff the collector requires to compute scores for documents.
fn requires_scoring(&self) -> bool;
/// Combines the fruit associated to the collection of each segments
/// into one fruit.
fn merge_fruits(&self, segment_fruits: Vec<Self::Fruit>) -> Result<Self::Fruit>;
}
/// The `SegmentCollector` is the trait in charge of defining the
/// collect operation at the scale of the segment.
///
/// `.collect(doc, score)` will be called for every documents
/// matching the query.
pub trait SegmentCollector: 'static {
/// `Fruit` is the type for the result of our collection.
/// e.g. `usize` for the `Count` collector.
type Fruit: Fruit;
/// The query pushes the scored document to the collector via this method. /// The query pushes the scored document to the collector via this method.
fn collect(&mut self, doc: DocId, score: Score); fn collect(&mut self, doc: DocId, score: Score);
/// Extract the fruit of the collection from the `SegmentCollector`. /// Returns true iff the collector requires to compute scores for documents.
fn harvest(self) -> Self::Fruit; fn requires_scoring(&self) -> bool;
} }
// ----------------------------------------------- impl<'a, C: Collector> Collector for &'a mut C {
// Tuple implementations. fn set_segment(
&mut self,
impl<Left, Right> Collector for (Left, Right) segment_local_id: SegmentLocalId,
where segment: &SegmentReader,
Left: Collector, ) -> Result<()> {
Right: Collector, (*self).set_segment(segment_local_id, segment)
{ }
type Fruit = (Left::Fruit, Right::Fruit); /// The query pushes the scored document to the collector via this method.
type Child = (Left::Child, Right::Child); fn collect(&mut self, doc: DocId, score: Score) {
C::collect(self, doc, score)
fn for_segment(&self, segment_local_id: u32, segment: &SegmentReader) -> Result<Self::Child> {
let left = self.0.for_segment(segment_local_id, segment)?;
let right = self.1.for_segment(segment_local_id, segment)?;
Ok((left, right))
} }
fn requires_scoring(&self) -> bool { fn requires_scoring(&self) -> bool {
self.0.requires_scoring() || self.1.requires_scoring() C::requires_scoring(self)
}
fn merge_fruits(
&self,
children: Vec<(Left::Fruit, Right::Fruit)>,
) -> Result<(Left::Fruit, Right::Fruit)> {
let mut left_fruits = vec![];
let mut right_fruits = vec![];
for (left_fruit, right_fruit) in children {
left_fruits.push(left_fruit);
right_fruits.push(right_fruit);
}
Ok((
self.0.merge_fruits(left_fruits)?,
self.1.merge_fruits(right_fruits)?,
))
} }
} }
impl<Left, Right> SegmentCollector for (Left, Right)
where
Left: SegmentCollector,
Right: SegmentCollector,
{
type Fruit = (Left::Fruit, Right::Fruit);
fn collect(&mut self, doc: DocId, score: Score) {
self.0.collect(doc, score);
self.1.collect(doc, score);
}
fn harvest(self) -> <Self as SegmentCollector>::Fruit {
(self.0.harvest(), self.1.harvest())
}
}
// 3-Tuple
impl<One, Two, Three> Collector for (One, Two, Three)
where
One: Collector,
Two: Collector,
Three: Collector,
{
type Fruit = (One::Fruit, Two::Fruit, Three::Fruit);
type Child = (One::Child, Two::Child, Three::Child);
fn for_segment(&self, segment_local_id: u32, segment: &SegmentReader) -> Result<Self::Child> {
let one = self.0.for_segment(segment_local_id, segment)?;
let two = self.1.for_segment(segment_local_id, segment)?;
let three = self.2.for_segment(segment_local_id, segment)?;
Ok((one, two, three))
}
fn requires_scoring(&self) -> bool {
self.0.requires_scoring() || self.1.requires_scoring() || self.2.requires_scoring()
}
fn merge_fruits(&self, children: Vec<Self::Fruit>) -> Result<Self::Fruit> {
let mut one_fruits = vec![];
let mut two_fruits = vec![];
let mut three_fruits = vec![];
for (one_fruit, two_fruit, three_fruit) in children {
one_fruits.push(one_fruit);
two_fruits.push(two_fruit);
three_fruits.push(three_fruit);
}
Ok((
self.0.merge_fruits(one_fruits)?,
self.1.merge_fruits(two_fruits)?,
self.2.merge_fruits(three_fruits)?,
))
}
}
impl<One, Two, Three> SegmentCollector for (One, Two, Three)
where
One: SegmentCollector,
Two: SegmentCollector,
Three: SegmentCollector,
{
type Fruit = (One::Fruit, Two::Fruit, Three::Fruit);
fn collect(&mut self, doc: DocId, score: Score) {
self.0.collect(doc, score);
self.1.collect(doc, score);
self.2.collect(doc, score);
}
fn harvest(self) -> <Self as SegmentCollector>::Fruit {
(self.0.harvest(), self.1.harvest(), self.2.harvest())
}
}
// 4-Tuple
impl<One, Two, Three, Four> Collector for (One, Two, Three, Four)
where
One: Collector,
Two: Collector,
Three: Collector,
Four: Collector,
{
type Fruit = (One::Fruit, Two::Fruit, Three::Fruit, Four::Fruit);
type Child = (One::Child, Two::Child, Three::Child, Four::Child);
fn for_segment(&self, segment_local_id: u32, segment: &SegmentReader) -> Result<Self::Child> {
let one = self.0.for_segment(segment_local_id, segment)?;
let two = self.1.for_segment(segment_local_id, segment)?;
let three = self.2.for_segment(segment_local_id, segment)?;
let four = self.3.for_segment(segment_local_id, segment)?;
Ok((one, two, three, four))
}
fn requires_scoring(&self) -> bool {
self.0.requires_scoring()
|| self.1.requires_scoring()
|| self.2.requires_scoring()
|| self.3.requires_scoring()
}
fn merge_fruits(&self, children: Vec<Self::Fruit>) -> Result<Self::Fruit> {
let mut one_fruits = vec![];
let mut two_fruits = vec![];
let mut three_fruits = vec![];
let mut four_fruits = vec![];
for (one_fruit, two_fruit, three_fruit, four_fruit) in children {
one_fruits.push(one_fruit);
two_fruits.push(two_fruit);
three_fruits.push(three_fruit);
four_fruits.push(four_fruit);
}
Ok((
self.0.merge_fruits(one_fruits)?,
self.1.merge_fruits(two_fruits)?,
self.2.merge_fruits(three_fruits)?,
self.3.merge_fruits(four_fruits)?,
))
}
}
impl<One, Two, Three, Four> SegmentCollector for (One, Two, Three, Four)
where
One: SegmentCollector,
Two: SegmentCollector,
Three: SegmentCollector,
Four: SegmentCollector,
{
type Fruit = (One::Fruit, Two::Fruit, Three::Fruit, Four::Fruit);
fn collect(&mut self, doc: DocId, score: Score) {
self.0.collect(doc, score);
self.1.collect(doc, score);
self.2.collect(doc, score);
self.3.collect(doc, score);
}
fn harvest(self) -> <Self as SegmentCollector>::Fruit {
(
self.0.harvest(),
self.1.harvest(),
self.2.harvest(),
self.3.harvest(),
)
}
}
impl_downcast!(Fruit);
#[cfg(test)] #[cfg(test)]
pub mod tests; pub mod tests {
use super::*;
use core::SegmentReader;
use fastfield::BytesFastFieldReader;
use fastfield::FastFieldReader;
use schema::Field;
use DocId;
use Score;
use SegmentLocalId;
/// Stores all of the doc ids.
/// This collector is only used for tests.
/// It is unusable in practise, as it does not store
/// the segment ordinals
pub struct TestCollector {
offset: DocId,
segment_max_doc: DocId,
docs: Vec<DocId>,
scores: Vec<Score>,
}
impl TestCollector {
/// Return the exhalist of documents.
pub fn docs(self) -> Vec<DocId> {
self.docs
}
pub fn scores(self) -> Vec<Score> {
self.scores
}
}
impl Default for TestCollector {
fn default() -> TestCollector {
TestCollector {
offset: 0,
segment_max_doc: 0,
docs: Vec::new(),
scores: Vec::new(),
}
}
}
impl Collector for TestCollector {
fn set_segment(&mut self, _: SegmentLocalId, reader: &SegmentReader) -> Result<()> {
self.offset += self.segment_max_doc;
self.segment_max_doc = reader.max_doc();
Ok(())
}
fn collect(&mut self, doc: DocId, score: Score) {
self.docs.push(doc + self.offset);
self.scores.push(score);
}
fn requires_scoring(&self) -> bool {
true
}
}
/// Collects in order all of the fast fields for all of the
/// doc in the `DocSet`
///
/// This collector is mainly useful for tests.
pub struct FastFieldTestCollector {
vals: Vec<u64>,
field: Field,
ff_reader: Option<FastFieldReader<u64>>,
}
impl FastFieldTestCollector {
pub fn for_field(field: Field) -> FastFieldTestCollector {
FastFieldTestCollector {
vals: Vec::new(),
field,
ff_reader: None,
}
}
pub fn vals(self) -> Vec<u64> {
self.vals
}
}
impl Collector for FastFieldTestCollector {
fn set_segment(&mut self, _: SegmentLocalId, reader: &SegmentReader) -> Result<()> {
self.ff_reader = Some(reader.fast_field_reader(self.field)?);
Ok(())
}
fn collect(&mut self, doc: DocId, _score: Score) {
let val = self.ff_reader.as_ref().unwrap().get(doc);
self.vals.push(val);
}
fn requires_scoring(&self) -> bool {
false
}
}
/// Collects in order all of the fast field bytes for all of the
/// docs in the `DocSet`
///
/// This collector is mainly useful for tests.
pub struct BytesFastFieldTestCollector {
vals: Vec<u8>,
field: Field,
ff_reader: Option<BytesFastFieldReader>,
}
impl BytesFastFieldTestCollector {
pub fn for_field(field: Field) -> BytesFastFieldTestCollector {
BytesFastFieldTestCollector {
vals: Vec::new(),
field,
ff_reader: None,
}
}
pub fn vals(self) -> Vec<u8> {
self.vals
}
}
impl Collector for BytesFastFieldTestCollector {
fn set_segment(&mut self, _segment_local_id: u32, segment: &SegmentReader) -> Result<()> {
self.ff_reader = Some(segment.bytes_fast_field_reader(self.field)?);
Ok(())
}
fn collect(&mut self, doc: u32, _score: f32) {
let val = self.ff_reader.as_ref().unwrap().get_val(doc);
self.vals.extend(val);
}
fn requires_scoring(&self) -> bool {
false
}
}
}
#[cfg(all(test, feature = "unstable"))]
mod bench {
use collector::{Collector, CountCollector};
use test::Bencher;
#[bench]
fn build_collector(b: &mut Bencher) {
b.iter(|| {
let mut count_collector = CountCollector::default();
let docs: Vec<u32> = (0..1_000_000).collect();
for doc in docs {
count_collector.collect(doc, 1f32);
}
count_collector.count()
});
}
}

264
src/collector/multi_collector.rs
View File

@@ -1,119 +1,26 @@
use super::Collector; use super::Collector;
use super::SegmentCollector;
use collector::Fruit;
use std::marker::PhantomData;
use DocId; use DocId;
use Result; use Result;
use Score; use Score;
use SegmentLocalId; use SegmentLocalId;
use SegmentReader; use SegmentReader;
use TantivyError;
pub struct MultiFruit {
sub_fruits: Vec<Option<Box<Fruit>>>,
}
pub struct CollectorWrapper<TCollector: Collector>(TCollector);
impl<TCollector: Collector> Collector for CollectorWrapper<TCollector> {
type Fruit = Box<Fruit>;
type Child = Box<BoxableSegmentCollector>;
fn for_segment(
&self,
segment_local_id: u32,
reader: &SegmentReader,
) -> Result<Box<BoxableSegmentCollector>> {
let child = self.0.for_segment(segment_local_id, reader)?;
Ok(Box::new(SegmentCollectorWrapper(child)))
}
fn requires_scoring(&self) -> bool {
self.0.requires_scoring()
}
fn merge_fruits(&self, children: Vec<<Self as Collector>::Fruit>) -> Result<Box<Fruit>> {
let typed_fruit: Vec<TCollector::Fruit> = children
.into_iter()
.map(|untyped_fruit| {
untyped_fruit
.downcast::<TCollector::Fruit>()
.map(|boxed_but_typed| *boxed_but_typed)
.map_err(|_| {
TantivyError::InvalidArgument("Failed to cast child fruit.".to_string())
})
})
.collect::<Result<_>>()?;
let merged_fruit = self.0.merge_fruits(typed_fruit)?;
Ok(Box::new(merged_fruit))
}
}
impl SegmentCollector for Box<BoxableSegmentCollector> {
type Fruit = Box<Fruit>;
fn collect(&mut self, doc: u32, score: f32) {
self.as_mut().collect(doc, score);
}
fn harvest(self) -> Box<Fruit> {
BoxableSegmentCollector::harvest_from_box(self)
}
}
pub trait BoxableSegmentCollector {
fn collect(&mut self, doc: u32, score: f32);
fn harvest_from_box(self: Box<Self>) -> Box<Fruit>;
}
pub struct SegmentCollectorWrapper<TSegmentCollector: SegmentCollector>(TSegmentCollector);
impl<TSegmentCollector: SegmentCollector> BoxableSegmentCollector
for SegmentCollectorWrapper<TSegmentCollector>
{
fn collect(&mut self, doc: u32, score: f32) {
self.0.collect(doc, score);
}
fn harvest_from_box(self: Box<Self>) -> Box<Fruit> {
Box::new(self.0.harvest())
}
}
pub struct FruitHandle<TFruit: Fruit> {
pos: usize,
_phantom: PhantomData<TFruit>,
}
impl<TFruit: Fruit> FruitHandle<TFruit> {
pub fn extract(self, fruits: &mut MultiFruit) -> TFruit {
let boxed_fruit = fruits.sub_fruits[self.pos].take().expect("");
*boxed_fruit
.downcast::<TFruit>()
.map_err(|_| ())
.expect("Failed to downcast collector fruit.")
}
}
/// Multicollector makes it possible to collect on more than one collector. /// Multicollector makes it possible to collect on more than one collector.
/// It should only be used for use cases where the Collector types is unknown /// It should only be used for use cases where the Collector types is unknown
/// at compile time. /// at compile time.
/// /// If the type of the collectors is known, you should prefer to use `ChainedCollector`.
/// If the type of the collectors is known, you can just group yours collectors
/// in a tuple. See the
/// [Combining several collectors section of the collector documentation](./index.html#combining-several-collectors).
/// ///
/// ```rust /// ```rust
/// #[macro_use] /// #[macro_use]
/// extern crate tantivy; /// extern crate tantivy;
/// use tantivy::schema::{Schema, TEXT}; /// use tantivy::schema::{SchemaBuilder, TEXT};
/// use tantivy::{Index, Result}; /// use tantivy::{Index, Result};
/// use tantivy::collector::{Count, TopDocs, MultiCollector}; /// use tantivy::collector::{CountCollector, TopCollector, MultiCollector};
/// use tantivy::query::QueryParser; /// use tantivy::query::QueryParser;
/// ///
/// # fn main() { example().unwrap(); } /// # fn main() { example().unwrap(); }
/// fn example() -> Result<()> { /// fn example() -> Result<()> {
/// let mut schema_builder = Schema::builder(); /// let mut schema_builder = SchemaBuilder::new();
/// let title = schema_builder.add_text_field("title", TEXT); /// let title = schema_builder.add_text_field("title", TEXT);
/// let schema = schema_builder.build(); /// let schema = schema_builder.build();
/// let index = Index::create_in_ram(schema); /// let index = Index::create_in_ram(schema);
@@ -134,118 +41,58 @@ impl<TFruit: Fruit> FruitHandle<TFruit> {
/// index_writer.commit().unwrap(); /// index_writer.commit().unwrap();
/// } /// }
/// ///
/// let reader = index.reader()?; /// index.load_searchers()?;
/// let searcher = reader.searcher(); /// let searcher = index.searcher();
/// ///
/// let mut collectors = MultiCollector::new(); /// {
/// let top_docs_handle = collectors.add_collector(TopDocs::with_limit(2)); /// let mut top_collector = TopCollector::with_limit(2);
/// let count_handle = collectors.add_collector(Count); /// let mut count_collector = CountCollector::default();
/// let query_parser = QueryParser::for_index(&index, vec![title]); /// {
/// let query = query_parser.parse_query("diary")?; /// let mut collectors =
/// let mut multi_fruit = searcher.search(&query, &collectors)?; /// MultiCollector::from(vec![&mut top_collector, &mut count_collector]);
/// /// let query_parser = QueryParser::for_index(&index, vec![title]);
/// let count = count_handle.extract(&mut multi_fruit); /// let query = query_parser.parse_query("diary")?;
/// let top_docs = top_docs_handle.extract(&mut multi_fruit); /// searcher.search(&*query, &mut collectors).unwrap();
/// /// }
/// # assert_eq!(count, 2); /// assert_eq!(count_collector.count(), 2);
/// # assert_eq!(top_docs.len(), 2); /// assert!(top_collector.at_capacity());
/// }
/// ///
/// Ok(()) /// Ok(())
/// } /// }
/// ``` /// ```
#[allow(clippy::type_complexity)]
#[derive(Default)]
pub struct MultiCollector<'a> { pub struct MultiCollector<'a> {
collector_wrappers: collectors: Vec<&'a mut Collector>,
Vec<Box<Collector<Child = Box<BoxableSegmentCollector>, Fruit = Box<Fruit>> + 'a>>,
} }
impl<'a> MultiCollector<'a> { impl<'a> MultiCollector<'a> {
/// Create a new `MultiCollector` /// Constructor
pub fn new() -> Self { pub fn from(collectors: Vec<&'a mut Collector>) -> MultiCollector {
Default::default() MultiCollector { collectors }
}
/// Add a new collector to our `MultiCollector`.
pub fn add_collector<'b: 'a, TCollector: Collector + 'b>(
&mut self,
collector: TCollector,
) -> FruitHandle<TCollector::Fruit> {
let pos = self.collector_wrappers.len();
self.collector_wrappers
.push(Box::new(CollectorWrapper(collector)));
FruitHandle {
pos,
_phantom: PhantomData,
}
} }
} }
impl<'a> Collector for MultiCollector<'a> { impl<'a> Collector for MultiCollector<'a> {
type Fruit = MultiFruit; fn set_segment(
type Child = MultiCollectorChild; &mut self,
fn for_segment(
&self,
segment_local_id: SegmentLocalId, segment_local_id: SegmentLocalId,
segment: &SegmentReader, segment: &SegmentReader,
) -> Result<MultiCollectorChild> { ) -> Result<()> {
let children = self for collector in &mut self.collectors {
.collector_wrappers collector.set_segment(segment_local_id, segment)?;
.iter()
.map(|collector_wrapper| collector_wrapper.for_segment(segment_local_id, segment))
.collect::<Result<Vec<_>>>()?;
Ok(MultiCollectorChild { children })
}
fn requires_scoring(&self) -> bool {
self.collector_wrappers.iter().any(|c| c.requires_scoring())
}
fn merge_fruits(&self, segments_multifruits: Vec<MultiFruit>) -> Result<MultiFruit> {
let mut segment_fruits_list: Vec<Vec<Box<Fruit>>> = (0..self.collector_wrappers.len())
.map(|_| Vec::with_capacity(segments_multifruits.len()))
.collect::<Vec<_>>();
for segment_multifruit in segments_multifruits {
for (idx, segment_fruit_opt) in segment_multifruit.sub_fruits.into_iter().enumerate() {
if let Some(segment_fruit) = segment_fruit_opt {
segment_fruits_list[idx].push(segment_fruit);
}
}
} }
let sub_fruits = self Ok(())
.collector_wrappers
.iter()
.zip(segment_fruits_list)
.map(|(child_collector, segment_fruits)| {
Ok(Some(child_collector.merge_fruits(segment_fruits)?))
})
.collect::<Result<_>>()?;
Ok(MultiFruit { sub_fruits })
} }
}
pub struct MultiCollectorChild {
children: Vec<Box<BoxableSegmentCollector>>,
}
impl SegmentCollector for MultiCollectorChild {
type Fruit = MultiFruit;
fn collect(&mut self, doc: DocId, score: Score) { fn collect(&mut self, doc: DocId, score: Score) {
for child in &mut self.children { for collector in &mut self.collectors {
child.collect(doc, score); collector.collect(doc, score);
} }
} }
fn requires_scoring(&self) -> bool {
fn harvest(self) -> MultiFruit { self.collectors
MultiFruit { .iter()
sub_fruits: self .any(|collector| collector.requires_scoring())
.children
.into_iter()
.map(|child| Some(child.harvest()))
.collect(),
}
} }
} }
@@ -253,41 +100,20 @@ impl SegmentCollector for MultiCollectorChild {
mod tests { mod tests {
use super::*; use super::*;
use collector::{Count, TopDocs}; use collector::{Collector, CountCollector, TopScoreCollector};
use query::TermQuery;
use schema::IndexRecordOption;
use schema::{Schema, TEXT};
use Index;
use Term;
#[test] #[test]
fn test_multi_collector() { fn test_multi_collector() {
let mut schema_builder = Schema::builder(); let mut top_collector = TopScoreCollector::with_limit(2);
let text = schema_builder.add_text_field("text", TEXT); let mut count_collector = CountCollector::default();
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{ {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut collectors =
index_writer.add_document(doc!(text=>"abc")); MultiCollector::from(vec![&mut top_collector, &mut count_collector]);
index_writer.add_document(doc!(text=>"abc abc abc")); collectors.collect(1, 0.2);
index_writer.add_document(doc!(text=>"abc abc")); collectors.collect(2, 0.1);
index_writer.commit().unwrap(); collectors.collect(3, 0.5);
index_writer.add_document(doc!(text=>""));
index_writer.add_document(doc!(text=>"abc abc abc abc"));
index_writer.add_document(doc!(text=>"abc"));
index_writer.commit().unwrap();
} }
let searcher = index.reader().unwrap().searcher(); assert_eq!(count_collector.count(), 3);
let term = Term::from_field_text(text, "abc"); assert!(top_collector.at_capacity());
let query = TermQuery::new(term, IndexRecordOption::Basic);
let mut collectors = MultiCollector::new();
let topdocs_handler = collectors.add_collector(TopDocs::with_limit(2));
let count_handler = collectors.add_collector(Count);
let mut multifruits = searcher.search(&query, &mut collectors).unwrap();
assert_eq!(count_handler.extract(&mut multifruits), 5);
assert_eq!(topdocs_handler.extract(&mut multifruits).len(), 2);
} }
} }

201
src/collector/tests.rs
View File

@@ -1,201 +0,0 @@
use super::*;
use core::SegmentReader;
use fastfield::BytesFastFieldReader;
use fastfield::FastFieldReader;
use schema::Field;
use DocAddress;
use DocId;
use Score;
use SegmentLocalId;
/// Stores all of the doc ids.
/// This collector is only used for tests.
/// It is unusable in pr
///
/// actise, as it does not store
/// the segment ordinals
pub struct TestCollector;
pub struct TestSegmentCollector {
segment_id: SegmentLocalId,
fruit: TestFruit,
}
#[derive(Default)]
pub struct TestFruit {
docs: Vec<DocAddress>,
scores: Vec<Score>,
}
impl TestFruit {
/// Return the list of matching documents exhaustively.
pub fn docs(&self) -> &[DocAddress] {
&self.docs[..]
}
pub fn scores(&self) -> &[Score] {
&self.scores[..]
}
}
impl Collector for TestCollector {
type Fruit = TestFruit;
type Child = TestSegmentCollector;
fn for_segment(
&self,
segment_id: SegmentLocalId,
_reader: &SegmentReader,
) -> Result<TestSegmentCollector> {
Ok(TestSegmentCollector {
segment_id,
fruit: TestFruit::default(),
})
}
fn requires_scoring(&self) -> bool {
true
}
fn merge_fruits(&self, mut children: Vec<TestFruit>) -> Result<TestFruit> {
children.sort_by_key(|fruit| {
if fruit.docs().is_empty() {
0
} else {
fruit.docs()[0].segment_ord()
}
});
let mut docs = vec![];
let mut scores = vec![];
for child in children {
docs.extend(child.docs());
scores.extend(child.scores);
}
Ok(TestFruit { docs, scores })
}
}
impl SegmentCollector for TestSegmentCollector {
type Fruit = TestFruit;
fn collect(&mut self, doc: DocId, score: Score) {
self.fruit.docs.push(DocAddress(self.segment_id, doc));
self.fruit.scores.push(score);
}
fn harvest(self) -> <Self as SegmentCollector>::Fruit {
self.fruit
}
}
/// Collects in order all of the fast fields for all of the
/// doc in the `DocSet`
///
/// This collector is mainly useful for tests.
pub struct FastFieldTestCollector {
field: Field,
}
pub struct FastFieldSegmentCollector {
vals: Vec<u64>,
reader: FastFieldReader<u64>,
}
impl FastFieldTestCollector {
pub fn for_field(field: Field) -> FastFieldTestCollector {
FastFieldTestCollector { field }
}
}
impl Collector for FastFieldTestCollector {
type Fruit = Vec<u64>;
type Child = FastFieldSegmentCollector;
fn for_segment(
&self,
_: SegmentLocalId,
reader: &SegmentReader,
) -> Result<FastFieldSegmentCollector> {
Ok(FastFieldSegmentCollector {
vals: Vec::new(),
reader: reader.fast_field_reader(self.field)?,
})
}
fn requires_scoring(&self) -> bool {
false
}
fn merge_fruits(&self, children: Vec<Vec<u64>>) -> Result<Vec<u64>> {
Ok(children.into_iter().flat_map(|v| v.into_iter()).collect())
}
}
impl SegmentCollector for FastFieldSegmentCollector {
type Fruit = Vec<u64>;
fn collect(&mut self, doc: DocId, _score: Score) {
let val = self.reader.get(doc);
self.vals.push(val);
}
fn harvest(self) -> Vec<u64> {
self.vals
}
}
/// Collects in order all of the fast field bytes for all of the
/// docs in the `DocSet`
///
/// This collector is mainly useful for tests.
pub struct BytesFastFieldTestCollector {
field: Field,
}
pub struct BytesFastFieldSegmentCollector {
vals: Vec<u8>,
reader: BytesFastFieldReader,
}
impl BytesFastFieldTestCollector {
pub fn for_field(field: Field) -> BytesFastFieldTestCollector {
BytesFastFieldTestCollector { field }
}
}
impl Collector for BytesFastFieldTestCollector {
type Fruit = Vec<u8>;
type Child = BytesFastFieldSegmentCollector;
fn for_segment(
&self,
_segment_local_id: u32,
segment: &SegmentReader,
) -> Result<BytesFastFieldSegmentCollector> {
Ok(BytesFastFieldSegmentCollector {
vals: Vec::new(),
reader: segment.bytes_fast_field_reader(self.field)?,
})
}
fn requires_scoring(&self) -> bool {
false
}
fn merge_fruits(&self, children: Vec<Vec<u8>>) -> Result<Vec<u8>> {
Ok(children.into_iter().flat_map(|c| c.into_iter()).collect())
}
}
impl SegmentCollector for BytesFastFieldSegmentCollector {
type Fruit = Vec<u8>;
fn collect(&mut self, doc: u32, _score: f32) {
let data = self.reader.get_val(doc);
self.vals.extend(data);
}
fn harvest(self) -> <Self as SegmentCollector>::Fruit {
self.vals
}
}

223
src/collector/top_collector.rs
View File

@@ -1,59 +1,56 @@
use serde::export::PhantomData;
use std::cmp::Ordering; use std::cmp::Ordering;
use std::collections::BinaryHeap; use std::collections::BinaryHeap;
use DocAddress; use DocAddress;
use DocId; use DocId;
use Result;
use SegmentLocalId; use SegmentLocalId;
use SegmentReader;
/// Contains a feature (field, score, etc.) of a document along with the document address. /// Contains a feature (field, score, etc.) of a document along with the document address.
/// ///
/// It has a custom implementation of `PartialOrd` that reverses the order. This is because the /// It has a custom implementation of `PartialOrd` that reverses the order. This is because the
/// default Rust heap is a max heap, whereas a min heap is needed. /// default Rust heap is a max heap, whereas a min heap is needed.
/// #[derive(Clone, Copy)]
/// WARNING: equality is not what you would expect here. pub struct ComparableDoc<T> {
/// Two elements are equal if their feature is equal, and regardless of whether `doc`
/// is equal. This should be perfectly fine for this usage, but let's make sure this
/// struct is never public.
struct ComparableDoc<T, D> {
feature: T, feature: T,
doc: D, doc_address: DocAddress,
} }
impl<T: PartialOrd, D> PartialOrd for ComparableDoc<T, D> { impl<T: PartialOrd> PartialOrd for ComparableDoc<T> {
fn partial_cmp(&self, other: &Self) -> Option<Ordering> { fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
Some(self.cmp(other)) Some(self.cmp(other))
} }
} }
impl<T: PartialOrd, D> Ord for ComparableDoc<T, D> { impl<T: PartialOrd> Ord for ComparableDoc<T> {
#[inline] #[inline]
fn cmp(&self, other: &Self) -> Ordering { fn cmp(&self, other: &Self) -> Ordering {
other other
.feature .feature
.partial_cmp(&self.feature) .partial_cmp(&self.feature)
.unwrap_or_else(|| Ordering::Equal) .unwrap_or_else(|| other.doc_address.cmp(&self.doc_address))
} }
} }
impl<T: PartialOrd, D> PartialEq for ComparableDoc<T, D> { impl<T: PartialOrd> PartialEq for ComparableDoc<T> {
fn eq(&self, other: &Self) -> bool { fn eq(&self, other: &Self) -> bool {
self.cmp(other) == Ordering::Equal self.cmp(other) == Ordering::Equal
} }
} }
impl<T: PartialOrd, D> Eq for ComparableDoc<T, D> {} impl<T: PartialOrd> Eq for ComparableDoc<T> {}
pub(crate) struct TopCollector<T> { /// The Top Collector keeps track of the K documents
/// sorted by type `T`.
///
/// The implementation is based on a `BinaryHeap`.
/// The theorical complexity for collecting the top `K` out of `n` documents
/// is `O(n log K)`.
pub struct TopCollector<T> {
limit: usize, limit: usize,
_marker: PhantomData<T>, heap: BinaryHeap<ComparableDoc<T>>,
segment_id: u32,
} }
impl<T> TopCollector<T> impl<T: PartialOrd + Clone> TopCollector<T> {
where
T: PartialOrd + Clone,
{
/// Creates a top collector, with a number of documents equal to "limit". /// Creates a top collector, with a number of documents equal to "limit".
/// ///
/// # Panics /// # Panics
@@ -63,157 +60,128 @@ where
panic!("Limit must be strictly greater than 0."); panic!("Limit must be strictly greater than 0.");
} }
TopCollector { TopCollector {
limit,
_marker: PhantomData,
}
}
pub fn limit(&self) -> usize {
self.limit
}
pub fn merge_fruits(
&self,
children: Vec<Vec<(T, DocAddress)>>,
) -> Result<Vec<(T, DocAddress)>> {
if self.limit == 0 {
return Ok(Vec::new());
}
let mut top_collector = BinaryHeap::new();
for child_fruit in children {
for (feature, doc) in child_fruit {
if top_collector.len() < self.limit {
top_collector.push(ComparableDoc { feature, doc });
} else if let Some(mut head) = top_collector.peek_mut() {
if head.feature < feature {
*head = ComparableDoc { feature, doc };
}
}
}
}
Ok(top_collector
.into_sorted_vec()
.into_iter()
.map(|cdoc| (cdoc.feature, cdoc.doc))
.collect())
}
pub(crate) fn for_segment(
&self,
segment_id: SegmentLocalId,
_: &SegmentReader,
) -> Result<TopSegmentCollector<T>> {
Ok(TopSegmentCollector::new(segment_id, self.limit))
}
}
/// The Top Collector keeps track of the K documents
/// sorted by type `T`.
///
/// The implementation is based on a `BinaryHeap`.
/// The theorical complexity for collecting the top `K` out of `n` documents
/// is `O(n log K)`.
pub(crate) struct TopSegmentCollector<T> {
limit: usize,
heap: BinaryHeap<ComparableDoc<T, DocId>>,
segment_id: u32,
}
impl<T: PartialOrd> TopSegmentCollector<T> {
fn new(segment_id: SegmentLocalId, limit: usize) -> TopSegmentCollector<T> {
TopSegmentCollector {
limit, limit,
heap: BinaryHeap::with_capacity(limit), heap: BinaryHeap::with_capacity(limit),
segment_id, segment_id: 0,
} }
} }
}
impl<T: PartialOrd + Clone> TopSegmentCollector<T> { /// Returns K best documents sorted in decreasing order.
pub fn harvest(self) -> Vec<(T, DocAddress)> { ///
let segment_id = self.segment_id; /// Calling this method triggers the sort.
self.heap /// The result of the sort is not cached.
.into_sorted_vec() pub fn docs(&self) -> Vec<DocAddress> {
self.top_docs()
.into_iter() .into_iter()
.map(|comparable_doc| { .map(|(_feature, doc)| doc)
(
comparable_doc.feature,
DocAddress(segment_id, comparable_doc.doc),
)
})
.collect() .collect()
} }
/// Returns K best FeatureDocuments sorted in decreasing order.
///
/// Calling this method triggers the sort.
/// The result of the sort is not cached.
pub fn top_docs(&self) -> Vec<(T, DocAddress)> {
let mut feature_docs: Vec<ComparableDoc<T>> = self.heap.iter().cloned().collect();
feature_docs.sort();
feature_docs
.into_iter()
.map(
|ComparableDoc {
feature,
doc_address,
}| (feature, doc_address),
).collect()
}
/// Return true iff at least K documents have gone through /// Return true iff at least K documents have gone through
/// the collector. /// the collector.
#[inline(always)] #[inline]
pub(crate) fn at_capacity(&self) -> bool { pub fn at_capacity(&self) -> bool {
self.heap.len() >= self.limit self.heap.len() >= self.limit
} }
/// Sets the segment local ID for the collector
pub fn set_segment_id(&mut self, segment_id: SegmentLocalId) {
self.segment_id = segment_id;
}
/// Collects a document scored by the given feature /// Collects a document scored by the given feature
/// ///
/// It collects documents until it has reached the max capacity. Once it reaches capacity, it /// It collects documents until it has reached the max capacity. Once it reaches capacity, it
/// will compare the lowest scoring item with the given one and keep whichever is greater. /// will compare the lowest scoring item with the given one and keep whichever is greater.
#[inline(always)]
pub fn collect(&mut self, doc: DocId, feature: T) { pub fn collect(&mut self, doc: DocId, feature: T) {
if self.at_capacity() { if self.at_capacity() {
// It's ok to unwrap as long as a limit of 0 is forbidden. // It's ok to unwrap as long as a limit of 0 is forbidden.
if let Some(limit_feature) = self.heap.peek().map(|head| head.feature.clone()) { let limit_doc: ComparableDoc<T> = self
if limit_feature < feature { .heap
if let Some(mut head) = self.heap.peek_mut() { .peek()
head.feature = feature; .expect("Top collector with size 0 is forbidden")
head.doc = doc; .clone();
} if limit_doc.feature < feature {
} let mut mut_head = self
.heap
.peek_mut()
.expect("Top collector with size 0 is forbidden");
mut_head.feature = feature;
mut_head.doc_address = DocAddress(self.segment_id, doc);
} }
} else { } else {
// we have not reached capacity yet, so we can just push the let wrapped_doc = ComparableDoc {
// element. feature,
self.heap.push(ComparableDoc { feature, doc }); doc_address: DocAddress(self.segment_id, doc),
};
self.heap.push(wrapped_doc);
} }
} }
} }
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::{TopCollector, TopSegmentCollector}; use super::*;
use DocAddress; use DocId;
use Score; use Score;
#[test] #[test]
fn test_top_collector_not_at_capacity() { fn test_top_collector_not_at_capacity() {
let mut top_collector = TopSegmentCollector::new(0, 4); let mut top_collector = TopCollector::with_limit(4);
top_collector.collect(1, 0.8); top_collector.collect(1, 0.8);
top_collector.collect(3, 0.2); top_collector.collect(3, 0.2);
top_collector.collect(5, 0.3); top_collector.collect(5, 0.3);
assert_eq!( assert!(!top_collector.at_capacity());
top_collector.harvest(), let score_docs: Vec<(Score, DocId)> = top_collector
vec![ .top_docs()
(0.8, DocAddress(0, 1)), .into_iter()
(0.3, DocAddress(0, 5)), .map(|(score, doc_address)| (score, doc_address.doc()))
(0.2, DocAddress(0, 3)) .collect();
] assert_eq!(score_docs, vec![(0.8, 1), (0.3, 5), (0.2, 3)]);
);
} }
#[test] #[test]
fn test_top_collector_at_capacity() { fn test_top_collector_at_capacity() {
let mut top_collector = TopSegmentCollector::new(0, 4); let mut top_collector = TopCollector::with_limit(4);
top_collector.collect(1, 0.8); top_collector.collect(1, 0.8);
top_collector.collect(3, 0.2); top_collector.collect(3, 0.2);
top_collector.collect(5, 0.3); top_collector.collect(5, 0.3);
top_collector.collect(7, 0.9); top_collector.collect(7, 0.9);
top_collector.collect(9, -0.2); top_collector.collect(9, -0.2);
assert_eq!( assert!(top_collector.at_capacity());
top_collector.harvest(), {
vec![ let score_docs: Vec<(Score, DocId)> = top_collector
(0.9, DocAddress(0, 7)), .top_docs()
(0.8, DocAddress(0, 1)), .into_iter()
(0.3, DocAddress(0, 5)), .map(|(score, doc_address)| (score, doc_address.doc()))
(0.2, DocAddress(0, 3)) .collect();
] assert_eq!(score_docs, vec![(0.9, 7), (0.8, 1), (0.3, 5), (0.2, 3)]);
); }
{
let docs: Vec<DocId> = top_collector
.docs()
.into_iter()
.map(|doc_address| doc_address.doc())
.collect();
assert_eq!(docs, vec![7, 1, 5, 3]);
}
} }
#[test] #[test]
@@ -221,4 +189,5 @@ mod tests {
fn test_top_0() { fn test_top_0() {
let _collector: TopCollector<Score> = TopCollector::with_limit(0); let _collector: TopCollector<Score> = TopCollector::with_limit(0);
} }
} }

264
src/collector/top_field_collector.rs
View File

@@ -1,13 +1,12 @@
use super::Collector; use super::Collector;
use collector::top_collector::TopCollector; use collector::top_collector::TopCollector;
use collector::top_collector::TopSegmentCollector;
use collector::SegmentCollector;
use fastfield::FastFieldReader; use fastfield::FastFieldReader;
use fastfield::FastValue; use fastfield::FastValue;
use schema::Field; use schema::Field;
use DocAddress; use DocAddress;
use DocId;
use Result; use Result;
use SegmentLocalId; use Score;
use SegmentReader; use SegmentReader;
/// The Top Field Collector keeps track of the K documents /// The Top Field Collector keeps track of the K documents
@@ -20,139 +19,136 @@ use SegmentReader;
/// ```rust /// ```rust
/// #[macro_use] /// #[macro_use]
/// extern crate tantivy; /// extern crate tantivy;
/// # use tantivy::schema::{Schema, Field, FAST, TEXT}; /// use tantivy::schema::{SchemaBuilder, TEXT, FAST};
/// # use tantivy::{Index, Result, DocAddress}; /// use tantivy::{Index, Result, DocId};
/// # use tantivy::query::{Query, QueryParser}; /// use tantivy::collector::TopFieldCollector;
/// use tantivy::Searcher; /// use tantivy::query::QueryParser;
/// use tantivy::collector::TopDocs;
/// ///
/// # fn main() -> tantivy::Result<()> { /// # fn main() { example().unwrap(); }
/// # let mut schema_builder = Schema::builder(); /// fn example() -> Result<()> {
/// # let title = schema_builder.add_text_field("title", TEXT); /// let mut schema_builder = SchemaBuilder::new();
/// # let rating = schema_builder.add_u64_field("rating", FAST); /// let title = schema_builder.add_text_field("title", TEXT);
/// # let schema = schema_builder.build(); /// let rating = schema_builder.add_u64_field("rating", FAST);
/// # let index = Index::create_in_ram(schema); /// let schema = schema_builder.build();
/// # let mut index_writer = index.writer_with_num_threads(1, 3_000_000)?; /// let index = Index::create_in_ram(schema);
/// # index_writer.add_document(doc!( /// {
/// # title => "The Name of the Wind", /// let mut index_writer = index.writer_with_num_threads(1, 3_000_000)?;
/// # rating => 92u64, /// index_writer.add_document(doc!(
/// # )); /// title => "The Name of the Wind",
/// # index_writer.add_document(doc!(title => "The Diary of Muadib", rating => 97u64)); /// rating => 92u64,
/// # index_writer.add_document(doc!(title => "A Dairy Cow", rating => 63u64)); /// ));
/// # index_writer.add_document(doc!(title => "The Diary of a Young Girl", rating => 80u64)); /// index_writer.add_document(doc!(
/// # index_writer.commit()?; /// title => "The Diary of Muadib",
/// # let reader = index.reader()?; /// rating => 97u64,
/// # let query = QueryParser::for_index(&index, vec![title]).parse_query("diary")?; /// ));
/// # let top_docs = docs_sorted_by_rating(&reader.searcher(), &query, rating)?; /// index_writer.add_document(doc!(
/// # assert_eq!(top_docs, /// title => "A Dairy Cow",
/// # vec![(97u64, DocAddress(0u32, 1)), /// rating => 63u64,
/// # (80u64, DocAddress(0u32, 3))]); /// ));
/// # Ok(()) /// index_writer.add_document(doc!(
/// # } /// title => "The Diary of a Young Girl",
/// # /// rating => 80u64,
/// /// Searches the document matching the given query, and /// ));
/// /// collects the top 10 documents, order by the `field` /// index_writer.commit().unwrap();
/// /// given in argument. /// }
/// ///
/// /// `field` is required to be a FAST field.
/// fn docs_sorted_by_rating(searcher: &Searcher,
/// query: &Query,
/// sort_by_field: Field)
/// -> Result<Vec<(u64, DocAddress)>> {
/// ///
/// // This is where we build our collector! /// index.load_searchers()?;
/// let top_docs_by_rating = TopDocs::with_limit(2).order_by_field(sort_by_field); /// let searcher = index.searcher();
/// ///
/// // ... and here is our documents. Not this is a simple vec. /// {
/// // The `u64` in the pair is the value of our fast field for each documents. /// let mut top_collector = TopFieldCollector::with_limit(rating, 2);
/// searcher.search(query, &top_docs_by_rating) /// let query_parser = QueryParser::for_index(&index, vec![title]);
/// let query = query_parser.parse_query("diary")?;
/// searcher.search(&*query, &mut top_collector).unwrap();
///
/// let score_docs: Vec<(u64, DocId)> = top_collector
/// .top_docs()
/// .into_iter()
/// .map(|(field, doc_address)| (field, doc_address.doc()))
/// .collect();
///
/// assert_eq!(score_docs, vec![(97u64, 1), (80, 3)]);
/// }
///
/// Ok(())
/// } /// }
/// ``` /// ```
pub struct TopDocsByField<T> { pub struct TopFieldCollector<T: FastValue> {
collector: TopCollector<T>,
field: Field, field: Field,
collector: TopCollector<T>,
fast_field: Option<FastFieldReader<T>>,
} }
impl<T: FastValue + PartialOrd + Clone> TopDocsByField<T> { impl<T: FastValue + PartialOrd + Clone> TopFieldCollector<T> {
/// Creates a top field collector, with a number of documents equal to "limit". /// Creates a top field collector, with a number of documents equal to "limit".
/// ///
/// The given field name must be a fast field, otherwise the collector have an error while /// The given field name must be a fast field, otherwise the collector have an error while
/// collecting results. /// collecting results.
/// ///
/// This constructor is crate-private. Client are supposed to call
/// build `TopDocsByField` object using the `TopDocs` API.
///
/// e.g.:
/// `TopDocs::with_limit(2).order_by_field(sort_by_field)`
///
/// # Panics /// # Panics
/// The method panics if limit is 0 /// The method panics if limit is 0
pub(crate) fn new(field: Field, limit: usize) -> TopDocsByField<T> { pub fn with_limit(field: Field, limit: usize) -> Self {
TopDocsByField { TopFieldCollector {
collector: TopCollector::with_limit(limit),
field, field,
collector: TopCollector::with_limit(limit),
fast_field: None,
} }
} }
/// Returns K best documents sorted the given field name in decreasing order.
///
/// Calling this method triggers the sort.
/// The result of the sort is not cached.
pub fn docs(&self) -> Vec<DocAddress> {
self.collector.docs()
}
/// Returns K best FieldDocuments sorted in decreasing order.
///
/// Calling this method triggers the sort.
/// The result of the sort is not cached.
pub fn top_docs(&self) -> Vec<(T, DocAddress)> {
self.collector.top_docs()
}
/// Return true iff at least K documents have gone through
/// the collector.
#[inline]
pub fn at_capacity(&self) -> bool {
self.collector.at_capacity()
}
} }
impl<T: FastValue + PartialOrd + Send + Sync + 'static> Collector for TopDocsByField<T> { impl<T: FastValue + PartialOrd + Clone> Collector for TopFieldCollector<T> {
type Fruit = Vec<(T, DocAddress)>; fn set_segment(&mut self, segment_id: u32, segment: &SegmentReader) -> Result<()> {
self.collector.set_segment_id(segment_id);
self.fast_field = Some(segment.fast_field_reader(self.field)?);
Ok(())
}
type Child = TopFieldSegmentCollector<T>; fn collect(&mut self, doc: DocId, _score: Score) {
let field_value = self
fn for_segment( .fast_field
&self, .as_ref()
segment_local_id: SegmentLocalId, .expect("collect() was called before set_segment. This should never happen.")
reader: &SegmentReader, .get(doc);
) -> Result<TopFieldSegmentCollector<T>> { self.collector.collect(doc, field_value);
let collector = self.collector.for_segment(segment_local_id, reader)?;
let reader = reader.fast_field_reader(self.field)?;
Ok(TopFieldSegmentCollector { collector, reader })
} }
fn requires_scoring(&self) -> bool { fn requires_scoring(&self) -> bool {
false false
} }
fn merge_fruits(
&self,
segment_fruits: Vec<Vec<(T, DocAddress)>>,
) -> Result<Vec<(T, DocAddress)>> {
self.collector.merge_fruits(segment_fruits)
}
}
pub struct TopFieldSegmentCollector<T: FastValue + PartialOrd> {
collector: TopSegmentCollector<T>,
reader: FastFieldReader<T>,
}
impl<T: FastValue + PartialOrd + Send + Sync + 'static> SegmentCollector
for TopFieldSegmentCollector<T>
{
type Fruit = Vec<(T, DocAddress)>;
fn collect(&mut self, doc: u32, _score: f32) {
let field_value = self.reader.get(doc);
self.collector.collect(doc, field_value);
}
fn harvest(self) -> Vec<(T, DocAddress)> {
self.collector.harvest()
}
} }
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::TopDocsByField; use super::*;
use collector::Collector;
use collector::TopDocs;
use query::Query; use query::Query;
use query::QueryParser; use query::QueryParser;
use schema::Field; use schema::Field;
use schema::IntOptions; use schema::IntOptions;
use schema::{Schema, FAST, TEXT}; use schema::Schema;
use DocAddress; use schema::{SchemaBuilder, FAST, TEXT};
use Index; use Index;
use IndexWriter; use IndexWriter;
use TantivyError; use TantivyError;
@@ -162,7 +158,7 @@ mod tests {
#[test] #[test]
fn test_top_collector_not_at_capacity() { fn test_top_collector_not_at_capacity() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
let title = schema_builder.add_text_field(TITLE, TEXT); let title = schema_builder.add_text_field(TITLE, TEXT);
let size = schema_builder.add_u64_field(SIZE, FAST); let size = schema_builder.add_u64_field(SIZE, FAST);
let schema = schema_builder.build(); let schema = schema_builder.build();
@@ -180,24 +176,24 @@ mod tests {
size => 16u64, size => 16u64,
)); ));
}); });
let searcher = index.reader().unwrap().searcher(); let searcher = index.searcher();
let top_collector = TopDocs::with_limit(4).order_by_field(size); let mut top_collector = TopFieldCollector::with_limit(size, 4);
let top_docs: Vec<(u64, DocAddress)> = searcher.search(&query, &top_collector).unwrap(); searcher.search(&*query, &mut top_collector).unwrap();
assert_eq!( assert!(!top_collector.at_capacity());
top_docs,
vec![ let score_docs: Vec<(u64, DocId)> = top_collector
(64, DocAddress(0, 1)), .top_docs()
(16, DocAddress(0, 2)), .into_iter()
(12, DocAddress(0, 0)) .map(|(field, doc_address)| (field, doc_address.doc()))
] .collect();
); assert_eq!(score_docs, vec![(64, 1), (16, 2), (12, 0)]);
} }
#[test] #[test]
#[should_panic] #[should_panic]
fn test_field_does_not_exist() { fn test_field_does_not_exist() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
let title = schema_builder.add_text_field(TITLE, TEXT); let title = schema_builder.add_text_field(TITLE, TEXT);
let size = schema_builder.add_u64_field(SIZE, FAST); let size = schema_builder.add_u64_field(SIZE, FAST);
let schema = schema_builder.build(); let schema = schema_builder.build();
@@ -207,17 +203,15 @@ mod tests {
size => 12u64, size => 12u64,
)); ));
}); });
let searcher = index.reader().unwrap().searcher(); let searcher = index.searcher();
let top_collector: TopDocsByField<u64> = TopDocs::with_limit(4).order_by_field(Field(2)); let segment = searcher.segment_reader(0);
let segment_reader = searcher.segment_reader(0u32); let mut top_collector: TopFieldCollector<u64> = TopFieldCollector::with_limit(Field(2), 4);
top_collector let _ = top_collector.set_segment(0, segment);
.for_segment(0, segment_reader)
.expect("should panic");
} }
#[test] #[test]
fn test_field_not_fast_field() { fn test_field_not_fast_field() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
let title = schema_builder.add_text_field(TITLE, TEXT); let title = schema_builder.add_text_field(TITLE, TEXT);
let size = schema_builder.add_u64_field(SIZE, IntOptions::default()); let size = schema_builder.add_u64_field(SIZE, IntOptions::default());
let schema = schema_builder.build(); let schema = schema_builder.build();
@@ -227,18 +221,28 @@ mod tests {
size => 12u64, size => 12u64,
)); ));
}); });
let searcher = index.reader().unwrap().searcher(); let searcher = index.searcher();
let segment = searcher.segment_reader(0); let segment = searcher.segment_reader(0);
let top_collector: TopDocsByField<u64> = TopDocs::with_limit(4).order_by_field(size); let mut top_collector: TopFieldCollector<u64> = TopFieldCollector::with_limit(size, 4);
assert_matches!( assert_matches!(
top_collector top_collector.set_segment(0, segment),
.for_segment(0, segment) Err(TantivyError::FastFieldError(_))
.map(|_| ())
.unwrap_err(),
TantivyError::FastFieldError(_)
); );
} }
#[test]
#[should_panic]
fn test_collect_before_set_segment() {
let mut top_collector: TopFieldCollector<u64> = TopFieldCollector::with_limit(Field(0), 4);
top_collector.collect(0, 0f32);
}
#[test]
#[should_panic]
fn test_top_0() {
let _: TopFieldCollector<u64> = TopFieldCollector::with_limit(Field(0), 0);
}
fn index( fn index(
query: &str, query: &str,
query_field: Field, query_field: Field,
@@ -250,6 +254,8 @@ mod tests {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
doc_adder(&mut index_writer); doc_adder(&mut index_writer);
index_writer.commit().unwrap(); index_writer.commit().unwrap();
index.load_searchers().unwrap();
let query_parser = QueryParser::for_index(&index, vec![query_field]); let query_parser = QueryParser::for_index(&index, vec![query_field]);
let query = query_parser.parse_query(query).unwrap(); let query = query_parser.parse_query(query).unwrap();
(index, query) (index, query)

218
src/collector/top_score_collector.rs
View File

@@ -1,10 +1,5 @@
use super::Collector; use super::Collector;
use collector::top_collector::TopCollector; use collector::top_collector::TopCollector;
use collector::top_collector::TopSegmentCollector;
use collector::SegmentCollector;
use collector::TopDocsByField;
use fastfield::FastValue;
use schema::Field;
use DocAddress; use DocAddress;
use DocId; use DocId;
use Result; use Result;
@@ -22,15 +17,14 @@ use SegmentReader;
/// ```rust /// ```rust
/// #[macro_use] /// #[macro_use]
/// extern crate tantivy; /// extern crate tantivy;
/// use tantivy::DocAddress; /// use tantivy::schema::{SchemaBuilder, TEXT};
/// use tantivy::schema::{Schema, TEXT}; /// use tantivy::{Index, Result, DocId, Score};
/// use tantivy::{Index, Result}; /// use tantivy::collector::TopScoreCollector;
/// use tantivy::collector::TopDocs;
/// use tantivy::query::QueryParser; /// use tantivy::query::QueryParser;
/// ///
/// # fn main() { example().unwrap(); } /// # fn main() { example().unwrap(); }
/// fn example() -> Result<()> { /// fn example() -> Result<()> {
/// let mut schema_builder = Schema::builder(); /// let mut schema_builder = SchemaBuilder::new();
/// let title = schema_builder.add_text_field("title", TEXT); /// let title = schema_builder.add_text_field("title", TEXT);
/// let schema = schema_builder.build(); /// let schema = schema_builder.build();
/// let index = Index::create_in_ram(schema); /// let index = Index::create_in_ram(schema);
@@ -51,153 +45,143 @@ use SegmentReader;
/// index_writer.commit().unwrap(); /// index_writer.commit().unwrap();
/// } /// }
/// ///
/// let reader = index.reader()?; /// index.load_searchers()?;
/// let searcher = reader.searcher(); /// let searcher = index.searcher();
/// ///
/// let query_parser = QueryParser::for_index(&index, vec![title]); /// {
/// let query = query_parser.parse_query("diary")?; /// let mut top_collector = TopScoreCollector::with_limit(2);
/// let top_docs = searcher.search(&query, &TopDocs::with_limit(2))?; /// let query_parser = QueryParser::for_index(&index, vec![title]);
/// let query = query_parser.parse_query("diary")?;
/// searcher.search(&*query, &mut top_collector).unwrap();
/// ///
/// assert_eq!(&top_docs[0], &(0.7261542, DocAddress(0, 1))); /// let score_docs: Vec<(Score, DocId)> = top_collector
/// assert_eq!(&top_docs[1], &(0.6099695, DocAddress(0, 3))); /// .top_docs()
/// .into_iter()
/// .map(|(score, doc_address)| (score, doc_address.doc()))
/// .collect();
///
/// assert_eq!(score_docs, vec![(0.7261542, 1), (0.6099695, 3)]);
/// }
/// ///
/// Ok(()) /// Ok(())
/// } /// }
/// ``` /// ```
pub struct TopDocs(TopCollector<Score>); pub struct TopScoreCollector {
collector: TopCollector<Score>,
}
impl TopDocs { impl TopScoreCollector {
/// Creates a top score collector, with a number of documents equal to "limit". /// Creates a top score collector, with a number of documents equal to "limit".
/// ///
/// # Panics /// # Panics
/// The method panics if limit is 0 /// The method panics if limit is 0
pub fn with_limit(limit: usize) -> TopDocs { pub fn with_limit(limit: usize) -> TopScoreCollector {
TopDocs(TopCollector::with_limit(limit)) TopScoreCollector {
collector: TopCollector::with_limit(limit),
}
} }
/// Set top-K to rank documents by a given fast field. /// Returns K best scored documents sorted in decreasing order.
/// ///
/// (By default, `TopDocs` collects the top-K documents sorted by /// Calling this method triggers the sort.
/// the similarity score.) /// The result of the sort is not cached.
pub fn order_by_field<T: PartialOrd + FastValue + Clone>( pub fn docs(&self) -> Vec<DocAddress> {
self, self.collector.docs()
field: Field, }
) -> TopDocsByField<T> {
TopDocsByField::new(field, self.0.limit()) /// Returns K best ScoredDocuments sorted in decreasing order.
///
/// Calling this method triggers the sort.
/// The result of the sort is not cached.
pub fn top_docs(&self) -> Vec<(Score, DocAddress)> {
self.collector.top_docs()
}
/// Returns K best ScoredDocuments sorted in decreasing order.
///
/// Calling this method triggers the sort.
/// The result of the sort is not cached.
#[deprecated]
pub fn score_docs(&self) -> Vec<(Score, DocAddress)> {
self.collector.top_docs()
}
/// Return true iff at least K documents have gone through
/// the collector.
#[inline]
pub fn at_capacity(&self) -> bool {
self.collector.at_capacity()
} }
} }
impl Collector for TopDocs { impl Collector for TopScoreCollector {
type Fruit = Vec<(Score, DocAddress)>; fn set_segment(&mut self, segment_id: SegmentLocalId, _: &SegmentReader) -> Result<()> {
self.collector.set_segment_id(segment_id);
Ok(())
}
type Child = TopScoreSegmentCollector; fn collect(&mut self, doc: DocId, score: Score) {
self.collector.collect(doc, score);
fn for_segment(
&self,
segment_local_id: SegmentLocalId,
reader: &SegmentReader,
) -> Result<Self::Child> {
let collector = self.0.for_segment(segment_local_id, reader)?;
Ok(TopScoreSegmentCollector(collector))
} }
fn requires_scoring(&self) -> bool { fn requires_scoring(&self) -> bool {
true true
} }
fn merge_fruits(&self, child_fruits: Vec<Vec<(Score, DocAddress)>>) -> Result<Self::Fruit> {
self.0.merge_fruits(child_fruits)
}
}
/// Segment Collector associated to `TopDocs`.
pub struct TopScoreSegmentCollector(TopSegmentCollector<Score>);
impl SegmentCollector for TopScoreSegmentCollector {
type Fruit = Vec<(Score, DocAddress)>;
fn collect(&mut self, doc: DocId, score: Score) {
self.0.collect(doc, score)
}
fn harvest(self) -> Vec<(Score, DocAddress)> {
self.0.harvest()
}
} }
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::TopDocs; use super::*;
use query::QueryParser; use collector::Collector;
use schema::Schema; use DocId;
use schema::TEXT;
use DocAddress;
use Index;
use Score; use Score;
fn make_index() -> Index {
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{
// writing the segment
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"Hello happy tax payer."));
index_writer.add_document(doc!(text_field=>"Droopy says hello happy tax payer"));
index_writer.add_document(doc!(text_field=>"I like Droopy"));
assert!(index_writer.commit().is_ok());
}
index
}
#[test] #[test]
fn test_top_collector_not_at_capacity() { fn test_top_collector_not_at_capacity() {
let index = make_index(); let mut top_collector = TopScoreCollector::with_limit(4);
let field = index.schema().get_field("text").unwrap(); top_collector.collect(1, 0.8);
let query_parser = QueryParser::for_index(&index, vec![field]); top_collector.collect(3, 0.2);
let text_query = query_parser.parse_query("droopy tax").unwrap(); top_collector.collect(5, 0.3);
let score_docs: Vec<(Score, DocAddress)> = index assert!(!top_collector.at_capacity());
.reader() let score_docs: Vec<(Score, DocId)> = top_collector
.unwrap() .top_docs()
.searcher() .into_iter()
.search(&text_query, &TopDocs::with_limit(4)) .map(|(score, doc_address)| (score, doc_address.doc()))
.unwrap(); .collect();
assert_eq!( assert_eq!(score_docs, vec![(0.8, 1), (0.3, 5), (0.2, 3)]);
score_docs,
vec![
(0.81221175, DocAddress(0u32, 1)),
(0.5376842, DocAddress(0u32, 2)),
(0.48527452, DocAddress(0, 0))
]
);
} }
#[test] #[test]
fn test_top_collector_at_capacity() { fn test_top_collector_at_capacity() {
let index = make_index(); let mut top_collector = TopScoreCollector::with_limit(4);
let field = index.schema().get_field("text").unwrap(); top_collector.collect(1, 0.8);
let query_parser = QueryParser::for_index(&index, vec![field]); top_collector.collect(3, 0.2);
let text_query = query_parser.parse_query("droopy tax").unwrap(); top_collector.collect(5, 0.3);
let score_docs: Vec<(Score, DocAddress)> = index top_collector.collect(7, 0.9);
.reader() top_collector.collect(9, -0.2);
.unwrap() assert!(top_collector.at_capacity());
.searcher() {
.search(&text_query, &TopDocs::with_limit(2)) let score_docs: Vec<(Score, DocId)> = top_collector
.unwrap(); .top_docs()
assert_eq!( .into_iter()
score_docs, .map(|(score, doc_address)| (score, doc_address.doc()))
vec![ .collect();
(0.81221175, DocAddress(0u32, 1)), assert_eq!(score_docs, vec![(0.9, 7), (0.8, 1), (0.3, 5), (0.2, 3)]);
(0.5376842, DocAddress(0u32, 2)), }
] {
); let docs: Vec<DocId> = top_collector
.docs()
.into_iter()
.map(|doc_address| doc_address.doc())
.collect();
assert_eq!(docs, vec![7, 1, 5, 3]);
}
} }
#[test] #[test]
#[should_panic] #[should_panic]
fn test_top_0() { fn test_top_0() {
TopDocs::with_limit(0); TopScoreCollector::with_limit(0);
} }
} }

72
src/common/bitpacker.rs
View File

@@ -1,6 +1,9 @@
use byteorder::{ByteOrder, LittleEndian, WriteBytesExt}; use common::serialize::BinarySerializable;
use std::io; use std::io;
use std::io::Write;
use std::mem;
use std::ops::Deref; use std::ops::Deref;
use std::ptr;
pub(crate) struct BitPacker { pub(crate) struct BitPacker {
mini_buffer: u64, mini_buffer: u64,
@@ -15,7 +18,7 @@ impl BitPacker {
} }
} }
pub fn write<TWrite: io::Write>( pub fn write<TWrite: Write>(
&mut self, &mut self,
val: u64, val: u64,
num_bits: u8, num_bits: u8,
@@ -25,14 +28,14 @@ impl BitPacker {
let num_bits = num_bits as usize; let num_bits = num_bits as usize;
if self.mini_buffer_written + num_bits > 64 { if self.mini_buffer_written + num_bits > 64 {
self.mini_buffer |= val_u64.wrapping_shl(self.mini_buffer_written as u32); self.mini_buffer |= val_u64.wrapping_shl(self.mini_buffer_written as u32);
output.write_u64::<LittleEndian>(self.mini_buffer)?; self.mini_buffer.serialize(output)?;
self.mini_buffer = val_u64.wrapping_shr((64 - self.mini_buffer_written) as u32); self.mini_buffer = val_u64.wrapping_shr((64 - self.mini_buffer_written) as u32);
self.mini_buffer_written = self.mini_buffer_written + num_bits - 64; self.mini_buffer_written = self.mini_buffer_written + num_bits - 64;
} else { } else {
self.mini_buffer |= val_u64 << self.mini_buffer_written; self.mini_buffer |= val_u64 << self.mini_buffer_written;
self.mini_buffer_written += num_bits; self.mini_buffer_written += num_bits;
if self.mini_buffer_written == 64 { if self.mini_buffer_written == 64 {
output.write_u64::<LittleEndian>(self.mini_buffer)?; self.mini_buffer.serialize(output)?;
self.mini_buffer_written = 0; self.mini_buffer_written = 0;
self.mini_buffer = 0u64; self.mini_buffer = 0u64;
} }
@@ -40,18 +43,17 @@ impl BitPacker {
Ok(()) Ok(())
} }
pub fn flush<TWrite: io::Write>(&mut self, output: &mut TWrite) -> io::Result<()> { pub fn flush<TWrite: Write>(&mut self, output: &mut TWrite) -> io::Result<()> {
if self.mini_buffer_written > 0 { if self.mini_buffer_written > 0 {
let num_bytes = (self.mini_buffer_written + 7) / 8; let num_bytes = (self.mini_buffer_written + 7) / 8;
let mut arr: [u8; 8] = [0u8; 8]; let arr: [u8; 8] = unsafe { mem::transmute::<u64, [u8; 8]>(self.mini_buffer.to_le()) };
LittleEndian::write_u64(&mut arr, self.mini_buffer);
output.write_all(&arr[..num_bytes])?; output.write_all(&arr[..num_bytes])?;
self.mini_buffer_written = 0; self.mini_buffer_written = 0;
} }
Ok(()) Ok(())
} }
pub fn close<TWrite: io::Write>(&mut self, output: &mut TWrite) -> io::Result<()> { pub fn close<TWrite: Write>(&mut self, output: &mut TWrite) -> io::Result<()> {
self.flush(output)?; self.flush(output)?;
// Padding the write file to simplify reads. // Padding the write file to simplify reads.
output.write_all(&[0u8; 7])?; output.write_all(&[0u8; 7])?;
@@ -64,7 +66,7 @@ pub struct BitUnpacker<Data>
where where
Data: Deref<Target = [u8]>, Data: Deref<Target = [u8]>,
{ {
num_bits: u64, num_bits: usize,
mask: u64, mask: u64,
data: Data, data: Data,
} }
@@ -80,13 +82,13 @@ where
(1u64 << num_bits) - 1u64 (1u64 << num_bits) - 1u64
}; };
BitUnpacker { BitUnpacker {
num_bits: u64::from(num_bits), num_bits: num_bits as usize,
mask, mask,
data, data,
} }
} }
pub fn get(&self, idx: u64) -> u64 { pub fn get(&self, idx: usize) -> u64 {
if self.num_bits == 0 { if self.num_bits == 0 {
return 0u64; return 0u64;
} }
@@ -97,13 +99,42 @@ where
let addr = addr_in_bits >> 3; let addr = addr_in_bits >> 3;
let bit_shift = addr_in_bits & 7; let bit_shift = addr_in_bits & 7;
debug_assert!( debug_assert!(
addr + 8 <= data.len() as u64, addr + 8 <= data.len(),
"The fast field field should have been padded with 7 bytes." "The fast field field should have been padded with 7 bytes."
); );
let val_unshifted_unmasked: u64 = LittleEndian::read_u64(&data[(addr as usize)..]); #[cfg_attr(feature = "cargo-clippy", allow(clippy::cast_ptr_alignment))]
let val_unshifted_unmasked: u64 =
u64::from_le(unsafe { ptr::read_unaligned(data[addr..].as_ptr() as *const u64) });
let val_shifted = (val_unshifted_unmasked >> bit_shift) as u64; let val_shifted = (val_unshifted_unmasked >> bit_shift) as u64;
val_shifted & mask val_shifted & mask
} }
/// Reads a range of values from the fast field.
///
/// The range of values read is from
/// `[start..start + output.len()[`
pub fn get_range(&self, start: u32, output: &mut [u64]) {
if self.num_bits == 0 {
for val in output.iter_mut() {
*val = 0u64;
}
} else {
let data: &[u8] = &*self.data;
let num_bits = self.num_bits;
let mask = self.mask;
let mut addr_in_bits = (start as usize) * num_bits;
for output_val in output.iter_mut() {
let addr = addr_in_bits >> 3;
let bit_shift = addr_in_bits & 7;
#[cfg_attr(feature = "cargo-clippy", allow(clippy::cast_ptr_alignment))]
let val_unshifted_unmasked: u64 =
unsafe { ptr::read_unaligned(data[addr..].as_ptr() as *const u64) };
let val_shifted = (val_unshifted_unmasked >> bit_shift) as u64;
*output_val = val_shifted & mask;
addr_in_bits += num_bits;
}
}
}
} }
#[cfg(test)] #[cfg(test)]
@@ -129,7 +160,7 @@ mod test {
fn test_bitpacker_util(len: usize, num_bits: u8) { fn test_bitpacker_util(len: usize, num_bits: u8) {
let (bitunpacker, vals) = create_fastfield_bitpacker(len, num_bits); let (bitunpacker, vals) = create_fastfield_bitpacker(len, num_bits);
for (i, val) in vals.iter().enumerate() { for (i, val) in vals.iter().enumerate() {
assert_eq!(bitunpacker.get(i as u64), *val); assert_eq!(bitunpacker.get(i), *val);
} }
} }
@@ -141,4 +172,17 @@ mod test {
test_bitpacker_util(6, 14); test_bitpacker_util(6, 14);
test_bitpacker_util(1000, 14); test_bitpacker_util(1000, 14);
} }
#[test]
fn test_bitpacker_range() {
let (bitunpacker, vals) = create_fastfield_bitpacker(100_000, 12);
let buffer_len = 100;
let mut buffer = vec![0u64; buffer_len];
for start in vec![0, 10, 20, 100, 1_000] {
bitunpacker.get_range(start as u32, &mut buffer[..]);
for i in 0..buffer_len {
assert_eq!(buffer[i], vals[start + i]);
}
}
}
} }

15
src/common/composite_file.rs
View File

@@ -4,8 +4,6 @@ use common::VInt;
use directory::ReadOnlySource; use directory::ReadOnlySource;
use directory::WritePtr; use directory::WritePtr;
use schema::Field; use schema::Field;
use space_usage::FieldUsage;
use space_usage::PerFieldSpaceUsage;
use std::collections::HashMap; use std::collections::HashMap;
use std::io::Write; use std::io::Write;
use std::io::{self, Read}; use std::io::{self, Read};
@@ -39,7 +37,7 @@ impl BinarySerializable for FileAddr {
/// A `CompositeWrite` is used to write a `CompositeFile`. /// A `CompositeWrite` is used to write a `CompositeFile`.
pub struct CompositeWrite<W = WritePtr> { pub struct CompositeWrite<W = WritePtr> {
write: CountingWriter<W>, write: CountingWriter<W>,
offsets: HashMap<FileAddr, u64>, offsets: HashMap<FileAddr, usize>,
} }
impl<W: Write> CompositeWrite<W> { impl<W: Write> CompositeWrite<W> {
@@ -168,17 +166,6 @@ impl CompositeFile {
.get(&FileAddr { field, idx }) .get(&FileAddr { field, idx })
.map(|&(from, to)| self.data.slice(from, to)) .map(|&(from, to)| self.data.slice(from, to))
} }
pub fn space_usage(&self) -> PerFieldSpaceUsage {
let mut fields = HashMap::new();
for (&field_addr, &(start, end)) in self.offsets_index.iter() {
fields
.entry(field_addr.field)
.or_insert_with(|| FieldUsage::empty(field_addr.field))
.add_field_idx(field_addr.idx, end - start);
}
PerFieldSpaceUsage::new(fields)
}
} }
#[cfg(test)] #[cfg(test)]

18
src/common/counting_writer.rs
View File

@@ -3,7 +3,7 @@ use std::io::Write;
pub struct CountingWriter<W> { pub struct CountingWriter<W> {
underlying: W, underlying: W,
written_bytes: u64, written_bytes: usize,
} }
impl<W: Write> CountingWriter<W> { impl<W: Write> CountingWriter<W> {
@@ -14,11 +14,11 @@ impl<W: Write> CountingWriter<W> {
} }
} }
pub fn written_bytes(&self) -> u64 { pub fn written_bytes(&self) -> usize {
self.written_bytes self.written_bytes
} }
pub fn finish(mut self) -> io::Result<(W, u64)> { pub fn finish(mut self) -> io::Result<(W, usize)> {
self.flush()?; self.flush()?;
Ok((self.underlying, self.written_bytes)) Ok((self.underlying, self.written_bytes))
} }
@@ -27,16 +27,10 @@ impl<W: Write> CountingWriter<W> {
impl<W: Write> Write for CountingWriter<W> { impl<W: Write> Write for CountingWriter<W> {
fn write(&mut self, buf: &[u8]) -> io::Result<usize> { fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
let written_size = self.underlying.write(buf)?; let written_size = self.underlying.write(buf)?;
self.written_bytes += written_size as u64; self.written_bytes += written_size;
Ok(written_size) Ok(written_size)
} }
fn write_all(&mut self, buf: &[u8]) -> io::Result<()> {
self.underlying.write_all(buf)?;
self.written_bytes += buf.len() as u64;
Ok(())
}
fn flush(&mut self) -> io::Result<()> { fn flush(&mut self) -> io::Result<()> {
self.underlying.flush() self.underlying.flush()
} }
@@ -54,8 +48,8 @@ mod test {
let mut counting_writer = CountingWriter::wrap(buffer); let mut counting_writer = CountingWriter::wrap(buffer);
let bytes = (0u8..10u8).collect::<Vec<u8>>(); let bytes = (0u8..10u8).collect::<Vec<u8>>();
counting_writer.write_all(&bytes).unwrap(); counting_writer.write_all(&bytes).unwrap();
let (w, len): (Vec<u8>, u64) = counting_writer.finish().unwrap(); let (w, len): (Vec<u8>, usize) = counting_writer.finish().unwrap();
assert_eq!(len, 10u64); assert_eq!(len, 10);
assert_eq!(w.len(), 10); assert_eq!(w.len(), 10);
} }
} }

20
src/common/mod.rs
View File

@@ -10,14 +10,10 @@ pub(crate) use self::bitset::TinySet;
pub(crate) use self::composite_file::{CompositeFile, CompositeWrite}; pub(crate) use self::composite_file::{CompositeFile, CompositeWrite};
pub use self::counting_writer::CountingWriter; pub use self::counting_writer::CountingWriter;
pub use self::serialize::{BinarySerializable, FixedSize}; pub use self::serialize::{BinarySerializable, FixedSize};
pub use self::vint::{read_u32_vint, serialize_vint_u32, write_u32_vint, VInt}; pub use self::vint::VInt;
pub use byteorder::LittleEndian as Endianness; pub use byteorder::LittleEndian as Endianness;
use std::io;
/// Segment's max doc must be `< MAX_DOC_LIMIT`.
///
/// We do not allow segments with more than
pub const MAX_DOC_LIMIT: u32 = 1 << 31;
/// Computes the number of bits that will be used for bitpacking. /// Computes the number of bits that will be used for bitpacking.
/// ///
@@ -56,6 +52,11 @@ pub(crate) fn is_power_of_2(n: usize) -> bool {
(n > 0) && (n & (n - 1) == 0) (n > 0) && (n & (n - 1) == 0)
} }
/// Create a default io error given a string.
pub(crate) fn make_io_err(msg: String) -> io::Error {
io::Error::new(io::ErrorKind::Other, msg)
}
/// Has length trait /// Has length trait
pub trait HasLen { pub trait HasLen {
/// Return length /// Return length
@@ -133,11 +134,4 @@ pub(crate) mod test {
assert_eq!(compute_num_bits(256), 9u8); assert_eq!(compute_num_bits(256), 9u8);
assert_eq!(compute_num_bits(5_000_000_000), 33u8); assert_eq!(compute_num_bits(5_000_000_000), 33u8);
} }
#[test]
fn test_max_doc() {
// this is the first time I write a unit test for a constant.
assert!(((super::MAX_DOC_LIMIT - 1) as i32) >= 0);
assert!((super::MAX_DOC_LIMIT as i32) < 0);
}
} }

125
src/common/vint.rs
View File

@@ -1,5 +1,4 @@
use super::BinarySerializable; use super::BinarySerializable;
use byteorder::{ByteOrder, LittleEndian};
use std::io; use std::io;
use std::io::Read; use std::io::Read;
use std::io::Write; use std::io::Write;
@@ -10,100 +9,6 @@ pub struct VInt(pub u64);
const STOP_BIT: u8 = 128; const STOP_BIT: u8 = 128;
pub fn serialize_vint_u32(val: u32) -> (u64, usize) {
const START_2: u64 = 1 << 7;
const START_3: u64 = 1 << 14;
const START_4: u64 = 1 << 21;
const START_5: u64 = 1 << 28;
const STOP_1: u64 = START_2 - 1;
const STOP_2: u64 = START_3 - 1;
const STOP_3: u64 = START_4 - 1;
const STOP_4: u64 = START_5 - 1;
const MASK_1: u64 = 127;
const MASK_2: u64 = MASK_1 << 7;
const MASK_3: u64 = MASK_2 << 7;
const MASK_4: u64 = MASK_3 << 7;
const MASK_5: u64 = MASK_4 << 7;
let val = u64::from(val);
const STOP_BIT: u64 = 128u64;
match val {
0...STOP_1 => (val | STOP_BIT, 1),
START_2...STOP_2 => (
(val & MASK_1) | ((val & MASK_2) << 1) | (STOP_BIT << (8)),
2,
),
START_3...STOP_3 => (
(val & MASK_1) | ((val & MASK_2) << 1) | ((val & MASK_3) << 2) | (STOP_BIT << (8 * 2)),
3,
),
START_4...STOP_4 => (
(val & MASK_1)
| ((val & MASK_2) << 1)
| ((val & MASK_3) << 2)
| ((val & MASK_4) << 3)
| (STOP_BIT << (8 * 3)),
4,
),
_ => (
(val & MASK_1)
| ((val & MASK_2) << 1)
| ((val & MASK_3) << 2)
| ((val & MASK_4) << 3)
| ((val & MASK_5) << 4)
| (STOP_BIT << (8 * 4)),
5,
),
}
}
/// Returns the number of bytes covered by a
/// serialized vint `u32`.
///
/// Expects a buffer data that starts
/// by the serialized `vint`, scans at most 5 bytes ahead until
/// it finds the vint final byte.
///
/// # May Panic
/// If the payload does not start by a valid `vint`
fn vint_len(data: &[u8]) -> usize {
for (i, &val) in data.iter().enumerate().take(5) {
if val >= STOP_BIT {
return i + 1;
}
}
panic!("Corrupted data. Invalid VInt 32");
}
/// Reads a vint `u32` from a buffer, and
/// consumes its payload data.
///
/// # Panics
///
/// If the buffer does not start by a valid
/// vint payload
pub fn read_u32_vint(data: &mut &[u8]) -> u32 {
let vlen = vint_len(*data);
let mut result = 0u32;
let mut shift = 0u64;
for &b in &data[..vlen] {
result |= u32::from(b & 127u8) << shift;
shift += 7;
}
*data = &data[vlen..];
result
}
/// Write a `u32` as a vint payload.
pub fn write_u32_vint<W: io::Write>(val: u32, writer: &mut W) -> io::Result<()> {
let (val, num_bytes) = serialize_vint_u32(val);
let mut buffer = [0u8; 8];
LittleEndian::write_u64(&mut buffer, val);
writer.write_all(&buffer[..num_bytes])
}
impl VInt { impl VInt {
pub fn val(&self) -> u64 { pub fn val(&self) -> u64 {
self.0 self.0
@@ -119,7 +24,7 @@ impl VInt {
output.extend(&buffer[0..num_bytes]); output.extend(&buffer[0..num_bytes]);
} }
pub fn serialize_into(&self, buffer: &mut [u8; 10]) -> usize { fn serialize_into(&self, buffer: &mut [u8; 10]) -> usize {
let mut remaining = self.0; let mut remaining = self.0;
for (i, b) in buffer.iter_mut().enumerate() { for (i, b) in buffer.iter_mut().enumerate() {
let next_byte: u8 = (remaining % 128u64) as u8; let next_byte: u8 = (remaining % 128u64) as u8;
@@ -159,7 +64,7 @@ impl BinarySerializable for VInt {
return Err(io::Error::new( return Err(io::Error::new(
io::ErrorKind::InvalidData, io::ErrorKind::InvalidData,
"Reach end of buffer while reading VInt", "Reach end of buffer while reading VInt",
)); ))
} }
} }
} }
@@ -169,9 +74,7 @@ impl BinarySerializable for VInt {
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::serialize_vint_u32;
use super::VInt; use super::VInt;
use byteorder::{ByteOrder, LittleEndian};
use common::BinarySerializable; use common::BinarySerializable;
fn aux_test_vint(val: u64) { fn aux_test_vint(val: u64) {
@@ -205,28 +108,4 @@ mod tests {
} }
aux_test_vint(10); aux_test_vint(10);
} }
fn aux_test_serialize_vint_u32(val: u32) {
let mut buffer = [0u8; 10];
let mut buffer2 = [0u8; 10];
let len_vint = VInt(val as u64).serialize_into(&mut buffer);
let (vint, len) = serialize_vint_u32(val);
assert_eq!(len, len_vint, "len wrong for val {}", val);
LittleEndian::write_u64(&mut buffer2, vint);
assert_eq!(&buffer[..len], &buffer2[..len], "array wrong for {}", val);
}
#[test]
fn test_vint_u32() {
aux_test_serialize_vint_u32(0);
aux_test_serialize_vint_u32(1);
aux_test_serialize_vint_u32(5);
for i in 1..3 {
let power_of_128 = 1u32 << (7 * i);
aux_test_serialize_vint_u32(power_of_128 - 1u32);
aux_test_serialize_vint_u32(power_of_128);
aux_test_serialize_vint_u32(power_of_128 + 1u32);
}
aux_test_serialize_vint_u32(u32::max_value());
}
} }

136
src/core/executor.rs
View File

@@ -1,136 +0,0 @@
use crossbeam::channel;
use scoped_pool::{Pool, ThreadConfig};
use Result;
/// Search executor whether search request are single thread or multithread.
///
/// We don't expose Rayon thread pool directly here for several reasons.
///
/// First dependency hell. It is not a good idea to expose the
/// API of a dependency, knowing it might conflict with a different version
/// used by the client. Second, we may stop using rayon in the future.
pub enum Executor {
SingleThread,
ThreadPool(Pool),
}
impl Executor {
/// Creates an Executor that performs all task in the caller thread.
pub fn single_thread() -> Executor {
Executor::SingleThread
}
// Creates an Executor that dispatches the tasks in a thread pool.
pub fn multi_thread(num_threads: usize, prefix: &'static str) -> Executor {
let thread_config = ThreadConfig::new().prefix(prefix);
let pool = Pool::with_thread_config(num_threads, thread_config);
Executor::ThreadPool(pool)
}
// Perform a map in the thread pool.
//
// Regardless of the executor (`SingleThread` or `ThreadPool`), panics in the task
// will propagate to the caller.
pub fn map<
A: Send,
R: Send,
AIterator: Iterator<Item = A>,
F: Sized + Sync + Fn(A) -> Result<R>,
>(
&self,
f: F,
args: AIterator,
) -> Result<Vec<R>> {
match self {
Executor::SingleThread => args.map(f).collect::<Result<_>>(),
Executor::ThreadPool(pool) => {
let args_with_indices: Vec<(usize, A)> = args.enumerate().collect();
let num_fruits = args_with_indices.len();
let fruit_receiver = {
let (fruit_sender, fruit_receiver) = channel::unbounded();
pool.scoped(|scope| {
for arg_with_idx in args_with_indices {
scope.execute(|| {
let (idx, arg) = arg_with_idx;
let fruit = f(arg);
if let Err(err) = fruit_sender.send((idx, fruit)) {
error!("Failed to send search task. It probably means all search threads have panicked. {:?}", err);
}
});
}
});
fruit_receiver
// This ends the scope of fruit_sender.
// This is important as it makes it possible for the fruit_receiver iteration to
// terminate.
};
// This is lame, but safe.
let mut results_with_position = Vec::with_capacity(num_fruits);
for (pos, fruit_res) in fruit_receiver {
let fruit = fruit_res?;
results_with_position.push((pos, fruit));
}
results_with_position.sort_by_key(|(pos, _)| *pos);
assert_eq!(results_with_position.len(), num_fruits);
Ok(results_with_position
.into_iter()
.map(|(_, fruit)| fruit)
.collect::<Vec<_>>())
}
}
}
}
#[cfg(test)]
mod tests {
use super::Executor;
#[test]
#[should_panic(expected = "panic should propagate")]
fn test_panic_propagates_single_thread() {
let _result: Vec<usize> = Executor::single_thread()
.map(
|_| {
panic!("panic should propagate");
},
vec![0].into_iter(),
)
.unwrap();
}
#[test]
#[should_panic] //< unfortunately the panic message is not propagated
fn test_panic_propagates_multi_thread() {
let _result: Vec<usize> = Executor::multi_thread(1, "search-test")
.map(
|_| {
panic!("panic should propagate");
},
vec![0].into_iter(),
)
.unwrap();
}
#[test]
fn test_map_singlethread() {
let result: Vec<usize> = Executor::single_thread()
.map(|i| Ok(i * 2), 0..1_000)
.unwrap();
assert_eq!(result.len(), 1_000);
for i in 0..1_000 {
assert_eq!(result[i], i * 2);
}
}
#[test]
fn test_map_multithread() {
let result: Vec<usize> = Executor::multi_thread(3, "search-test")
.map(|i| Ok(i * 2), 0..10)
.unwrap();
assert_eq!(result.len(), 10);
for i in 0..10 {
assert_eq!(result[i], i * 2);
}
}
}

352
src/core/index.rs
View File

@@ -1,31 +1,31 @@
use super::pool::LeasedItem;
use super::pool::Pool;
use super::segment::create_segment; use super::segment::create_segment;
use super::segment::Segment; use super::segment::Segment;
use core::Executor; use core::searcher::Searcher;
use core::IndexMeta; use core::IndexMeta;
use core::SegmentId; use core::SegmentId;
use core::SegmentMeta; use core::SegmentMeta;
use core::SegmentReader;
use core::META_FILEPATH; use core::META_FILEPATH;
use directory::ManagedDirectory; use directory::ManagedDirectory;
#[cfg(feature = "mmap")] #[cfg(feature = "mmap")]
use directory::MmapDirectory; use directory::MmapDirectory;
use directory::INDEX_WRITER_LOCK;
use directory::{Directory, RAMDirectory}; use directory::{Directory, RAMDirectory};
use error::DataCorruption;
use error::TantivyError; use error::TantivyError;
use indexer::index_writer::open_index_writer; use indexer::index_writer::open_index_writer;
use indexer::index_writer::HEAP_SIZE_MIN; use indexer::index_writer::HEAP_SIZE_MIN;
use indexer::segment_updater::save_new_metas; use indexer::segment_updater::save_new_metas;
use indexer::LockType;
use num_cpus; use num_cpus;
use reader::IndexReader;
use reader::IndexReaderBuilder;
use schema::Field; use schema::Field;
use schema::FieldType; use schema::FieldType;
use schema::Schema; use schema::Schema;
use serde_json; use serde_json;
use std::borrow::BorrowMut; use std::borrow::BorrowMut;
use std::fmt; use std::fmt;
#[cfg(feature = "mmap")]
use std::path::Path; use std::path::Path;
use std::sync::atomic::{AtomicUsize, Ordering};
use std::sync::Arc; use std::sync::Arc;
use tokenizer::BoxedTokenizer; use tokenizer::BoxedTokenizer;
use tokenizer::TokenizerManager; use tokenizer::TokenizerManager;
@@ -36,53 +36,19 @@ fn load_metas(directory: &Directory) -> Result<IndexMeta> {
let meta_data = directory.atomic_read(&META_FILEPATH)?; let meta_data = directory.atomic_read(&META_FILEPATH)?;
let meta_string = String::from_utf8_lossy(&meta_data); let meta_string = String::from_utf8_lossy(&meta_data);
serde_json::from_str(&meta_string) serde_json::from_str(&meta_string)
.map_err(|e| { .map_err(|_| TantivyError::CorruptedFile(META_FILEPATH.clone()))
DataCorruption::new(
META_FILEPATH.clone(),
format!("Meta file cannot be deserialized. {:?}.", e),
)
})
.map_err(From::from)
} }
/// Search Index /// Search Index
#[derive(Clone)]
pub struct Index { pub struct Index {
directory: ManagedDirectory, directory: ManagedDirectory,
schema: Schema, schema: Schema,
executor: Arc<Executor>, num_searchers: Arc<AtomicUsize>,
searcher_pool: Arc<Pool<Searcher>>,
tokenizers: TokenizerManager, tokenizers: TokenizerManager,
} }
impl Index { impl Index {
/// Examines the director to see if it contains an index
pub fn exists<Dir: Directory>(dir: &Dir) -> bool {
dir.exists(&META_FILEPATH)
}
/// Accessor to the search executor.
///
/// This pool is used by default when calling `searcher.search(...)`
/// to perform search on the individual segments.
///
/// By default the executor is single thread, and simply runs in the calling thread.
pub fn search_executor(&self) -> &Executor {
self.executor.as_ref()
}
/// Replace the default single thread search executor pool
/// by a thread pool with a given number of threads.
pub fn set_multithread_executor(&mut self, num_threads: usize) {
self.executor = Arc::new(Executor::multi_thread(num_threads, "thrd-tantivy-search-"));
}
/// Replace the default single thread search executor pool
/// by a thread pool with a given number of threads.
pub fn set_default_multithread_executor(&mut self) {
let default_num_threads = num_cpus::get();
self.set_multithread_executor(default_num_threads);
}
/// Creates a new index using the `RAMDirectory`. /// Creates a new index using the `RAMDirectory`.
/// ///
/// The index will be allocated in anonymous memory. /// The index will be allocated in anonymous memory.
@@ -99,29 +65,9 @@ impl Index {
#[cfg(feature = "mmap")] #[cfg(feature = "mmap")]
pub fn create_in_dir<P: AsRef<Path>>(directory_path: P, schema: Schema) -> Result<Index> { pub fn create_in_dir<P: AsRef<Path>>(directory_path: P, schema: Schema) -> Result<Index> {
let mmap_directory = MmapDirectory::open(directory_path)?; let mmap_directory = MmapDirectory::open(directory_path)?;
if Index::exists(&mmap_directory) {
return Err(TantivyError::IndexAlreadyExists);
}
Index::create(mmap_directory, schema) Index::create(mmap_directory, schema)
} }
/// Opens or creates a new index in the provided directory
pub fn open_or_create<Dir: Directory>(dir: Dir, schema: Schema) -> Result<Index> {
if Index::exists(&dir) {
let index = Index::open(dir)?;
if index.schema() == schema {
Ok(index)
} else {
Err(TantivyError::SchemaError(
"An index exists but the schema does not match.".to_string(),
))
}
} else {
Index::create(dir, schema)
}
}
/// Creates a new index in a temp directory. /// Creates a new index in a temp directory.
/// ///
/// The index will use the `MMapDirectory` in a newly created directory. /// The index will use the `MMapDirectory` in a newly created directory.
@@ -138,15 +84,13 @@ impl Index {
/// Creates a new index given an implementation of the trait `Directory` /// Creates a new index given an implementation of the trait `Directory`
pub fn create<Dir: Directory>(dir: Dir, schema: Schema) -> Result<Index> { pub fn create<Dir: Directory>(dir: Dir, schema: Schema) -> Result<Index> {
let directory = ManagedDirectory::wrap(dir)?; let directory = ManagedDirectory::new(dir)?;
Index::from_directory(directory, schema) Index::from_directory(directory, schema)
} }
/// Create a new index from a directory. /// Create a new index from a directory.
///
/// This will overwrite existing meta.json
fn from_directory(mut directory: ManagedDirectory, schema: Schema) -> Result<Index> { fn from_directory(mut directory: ManagedDirectory, schema: Schema) -> Result<Index> {
save_new_metas(schema.clone(), directory.borrow_mut())?; save_new_metas(schema.clone(), 0, directory.borrow_mut())?;
let metas = IndexMeta::with_schema(schema); let metas = IndexMeta::with_schema(schema);
Index::create_from_metas(directory, &metas) Index::create_from_metas(directory, &metas)
} }
@@ -154,12 +98,15 @@ impl Index {
/// Creates a new index given a directory and an `IndexMeta`. /// Creates a new index given a directory and an `IndexMeta`.
fn create_from_metas(directory: ManagedDirectory, metas: &IndexMeta) -> Result<Index> { fn create_from_metas(directory: ManagedDirectory, metas: &IndexMeta) -> Result<Index> {
let schema = metas.schema.clone(); let schema = metas.schema.clone();
let n_cpus = num_cpus::get();
let index = Index { let index = Index {
directory, directory,
schema, schema,
num_searchers: Arc::new(AtomicUsize::new(n_cpus)),
searcher_pool: Arc::new(Pool::new()),
tokenizers: TokenizerManager::default(), tokenizers: TokenizerManager::default(),
executor: Arc::new(Executor::single_thread()),
}; };
index.load_searchers()?;
Ok(index) Ok(index)
} }
@@ -189,22 +136,6 @@ impl Index {
} }
} }
/// Create a default `IndexReader` for the given index.
///
/// See [`Index.reader_builder()`](#method.reader_builder).
pub fn reader(&self) -> Result<IndexReader> {
self.reader_builder().try_into()
}
/// Create a `IndexReader` for the given index.
///
/// Most project should create at most one reader for a given index.
/// This method is typically called only once per `Index` instance,
/// over the lifetime of most problem.
pub fn reader_builder(&self) -> IndexReaderBuilder {
IndexReaderBuilder::new(self.clone())
}
/// Opens a new directory from an index path. /// Opens a new directory from an index path.
#[cfg(feature = "mmap")] #[cfg(feature = "mmap")]
pub fn open_in_dir<P: AsRef<Path>>(directory_path: P) -> Result<Index> { pub fn open_in_dir<P: AsRef<Path>>(directory_path: P) -> Result<Index> {
@@ -214,7 +145,7 @@ impl Index {
/// Open the index using the provided directory /// Open the index using the provided directory
pub fn open<D: Directory>(directory: D) -> Result<Index> { pub fn open<D: Directory>(directory: D) -> Result<Index> {
let directory = ManagedDirectory::wrap(directory)?; let directory = ManagedDirectory::new(directory)?;
let metas = load_metas(&directory)?; let metas = load_metas(&directory)?;
Index::create_from_metas(directory, &metas) Index::create_from_metas(directory, &metas)
} }
@@ -240,8 +171,7 @@ impl Index {
/// Each thread will receive a budget of `overall_heap_size_in_bytes / num_threads`. /// Each thread will receive a budget of `overall_heap_size_in_bytes / num_threads`.
/// ///
/// # Errors /// # Errors
/// If the lockfile already exists, returns `Error::DirectoryLockBusy` or an `Error::IOError`. /// If the lockfile already exists, returns `Error::FileAlreadyExists`.
///
/// # Panics /// # Panics
/// If the heap size per thread is too small, panics. /// If the heap size per thread is too small, panics.
pub fn writer_with_num_threads( pub fn writer_with_num_threads(
@@ -249,21 +179,7 @@ impl Index {
num_threads: usize, num_threads: usize,
overall_heap_size_in_bytes: usize, overall_heap_size_in_bytes: usize,
) -> Result<IndexWriter> { ) -> Result<IndexWriter> {
let directory_lock = self let directory_lock = LockType::IndexWriterLock.acquire_lock(&self.directory)?;
.directory
.acquire_lock(&INDEX_WRITER_LOCK)
.map_err(|err| {
TantivyError::LockFailure(
err,
Some(
"Failed to acquire index lock. If you are using\
a regular directory, this means there is already an \
`IndexWriter` working on this `Directory`, in this process \
or in a different process."
.to_string(),
),
)
})?;
let heap_size_in_bytes_per_thread = overall_heap_size_in_bytes / num_threads; let heap_size_in_bytes_per_thread = overall_heap_size_in_bytes / num_threads;
open_index_writer( open_index_writer(
self, self,
@@ -343,6 +259,53 @@ impl Index {
.map(|segment_meta| segment_meta.id()) .map(|segment_meta| segment_meta.id())
.collect()) .collect())
} }
/// Sets the number of searchers to use
///
/// Only works after the next call to `load_searchers`
pub fn set_num_searchers(&mut self, num_searchers: usize) {
self.num_searchers.store(num_searchers, Ordering::Release);
}
/// Update searchers so that they reflect the state of the last
/// `.commit()`.
///
/// If indexing happens in the same process as searching,
/// you most likely want to call `.load_searchers()` right after each
/// successful call to `.commit()`.
///
/// If indexing and searching happen in different processes, the way to
/// get the freshest `index` at all time, is to watch `meta.json` and
/// call `load_searchers` whenever a changes happen.
pub fn load_searchers(&self) -> Result<()> {
let _meta_lock = LockType::MetaLock.acquire_lock(self.directory())?;
let searchable_segments = self.searchable_segments()?;
let segment_readers: Vec<SegmentReader> = searchable_segments
.iter()
.map(SegmentReader::open)
.collect::<Result<_>>()?;
let schema = self.schema();
let num_searchers: usize = self.num_searchers.load(Ordering::Acquire);
let searchers = (0..num_searchers)
.map(|_| Searcher::new(schema.clone(), self.clone(), segment_readers.clone()))
.collect();
self.searcher_pool.publish_new_generation(searchers);
Ok(())
}
/// Returns a searcher
///
/// This method should be called every single time a search
/// query is performed.
/// The searchers are taken from a pool of `num_searchers` searchers.
/// If no searcher is available
/// this may block.
///
/// The same searcher must be used for a given query, as it ensures
/// the use of a consistent segment set.
pub fn searcher(&self) -> LeasedItem<Searcher> {
self.searcher_pool.acquire()
}
} }
impl fmt::Debug for Index { impl fmt::Debug for Index {
@@ -351,22 +314,27 @@ impl fmt::Debug for Index {
} }
} }
impl Clone for Index {
fn clone(&self) -> Index {
Index {
directory: self.directory.clone(),
schema: self.schema.clone(),
num_searchers: Arc::clone(&self.num_searchers),
searcher_pool: Arc::clone(&self.searcher_pool),
tokenizers: self.tokenizers.clone(),
}
}
}
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use directory::RAMDirectory; use schema::{SchemaBuilder, INT_INDEXED, TEXT};
use schema::Field;
use schema::{Schema, INDEXED, TEXT};
use std::thread;
use std::time::Duration;
use Index; use Index;
use IndexReader;
use IndexWriter;
use ReloadPolicy;
#[test] #[test]
fn test_indexer_for_field() { fn test_indexer_for_field() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let num_likes_field = schema_builder.add_u64_field("num_likes", INDEXED); let num_likes_field = schema_builder.add_u64_field("num_likes", INT_INDEXED);
let body_field = schema_builder.add_text_field("body", TEXT); let body_field = schema_builder.add_text_field("body", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
@@ -377,164 +345,4 @@ mod tests {
); );
} }
#[test]
fn test_index_exists() {
let directory = RAMDirectory::create();
assert!(!Index::exists(&directory));
assert!(Index::create(directory.clone(), throw_away_schema()).is_ok());
assert!(Index::exists(&directory));
}
#[test]
fn open_or_create_should_create() {
let directory = RAMDirectory::create();
assert!(!Index::exists(&directory));
assert!(Index::open_or_create(directory.clone(), throw_away_schema()).is_ok());
assert!(Index::exists(&directory));
}
#[test]
fn open_or_create_should_open() {
let directory = RAMDirectory::create();
assert!(Index::create(directory.clone(), throw_away_schema()).is_ok());
assert!(Index::exists(&directory));
assert!(Index::open_or_create(directory, throw_away_schema()).is_ok());
}
#[test]
fn create_should_wipeoff_existing() {
let directory = RAMDirectory::create();
assert!(Index::create(directory.clone(), throw_away_schema()).is_ok());
assert!(Index::exists(&directory));
assert!(Index::create(directory.clone(), Schema::builder().build()).is_ok());
}
#[test]
fn open_or_create_exists_but_schema_does_not_match() {
let directory = RAMDirectory::create();
assert!(Index::create(directory.clone(), throw_away_schema()).is_ok());
assert!(Index::exists(&directory));
assert!(Index::open_or_create(directory.clone(), throw_away_schema()).is_ok());
let err = Index::open_or_create(directory, Schema::builder().build());
assert_eq!(
format!("{:?}", err.unwrap_err()),
"SchemaError(\"An index exists but the schema does not match.\")"
);
}
fn throw_away_schema() -> Schema {
let mut schema_builder = Schema::builder();
let _ = schema_builder.add_u64_field("num_likes", INDEXED);
schema_builder.build()
}
#[test]
fn test_index_on_commit_reload_policy() {
let schema = throw_away_schema();
let field = schema.get_field("num_likes").unwrap();
let index = Index::create_in_ram(schema);
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::OnCommit)
.try_into()
.unwrap();
assert_eq!(reader.searcher().num_docs(), 0);
let mut writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
test_index_on_commit_reload_policy_aux(field, &mut writer, &reader);
}
#[cfg(feature = "mmap")]
mod mmap_specific {
use super::*;
use std::path::PathBuf;
use tempdir::TempDir;
#[test]
fn test_index_on_commit_reload_policy_mmap() {
let schema = throw_away_schema();
let field = schema.get_field("num_likes").unwrap();
let tempdir = TempDir::new("index").unwrap();
let tempdir_path = PathBuf::from(tempdir.path());
let index = Index::create_in_dir(&tempdir_path, schema).unwrap();
let mut writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
writer.commit().unwrap();
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::OnCommit)
.try_into()
.unwrap();
assert_eq!(reader.searcher().num_docs(), 0);
test_index_on_commit_reload_policy_aux(field, &mut writer, &reader);
}
#[test]
fn test_index_manual_policy_mmap() {
let schema = throw_away_schema();
let field = schema.get_field("num_likes").unwrap();
let index = Index::create_from_tempdir(schema).unwrap();
let mut writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
writer.commit().unwrap();
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::Manual)
.try_into()
.unwrap();
assert_eq!(reader.searcher().num_docs(), 0);
writer.add_document(doc!(field=>1u64));
writer.commit().unwrap();
thread::sleep(Duration::from_millis(500));
assert_eq!(reader.searcher().num_docs(), 0);
reader.reload().unwrap();
assert_eq!(reader.searcher().num_docs(), 1);
}
#[test]
fn test_index_on_commit_reload_policy_different_directories() {
let schema = throw_away_schema();
let field = schema.get_field("num_likes").unwrap();
let tempdir = TempDir::new("index").unwrap();
let tempdir_path = PathBuf::from(tempdir.path());
let write_index = Index::create_in_dir(&tempdir_path, schema).unwrap();
let read_index = Index::open_in_dir(&tempdir_path).unwrap();
let reader = read_index
.reader_builder()
.reload_policy(ReloadPolicy::OnCommit)
.try_into()
.unwrap();
assert_eq!(reader.searcher().num_docs(), 0);
let mut writer = write_index.writer_with_num_threads(1, 3_000_000).unwrap();
test_index_on_commit_reload_policy_aux(field, &mut writer, &reader);
}
}
fn test_index_on_commit_reload_policy_aux(
field: Field,
writer: &mut IndexWriter,
reader: &IndexReader,
) {
assert_eq!(reader.searcher().num_docs(), 0);
writer.add_document(doc!(field=>1u64));
writer.commit().unwrap();
let mut count = 0;
for _ in 0..100 {
count = reader.searcher().num_docs();
if count > 0 {
break;
}
thread::sleep(Duration::from_millis(100));
}
assert_eq!(count, 1);
writer.add_document(doc!(field=>2u64));
writer.commit().unwrap();
let mut count = 0;
for _ in 0..10 {
count = reader.searcher().num_docs();
if count > 1 {
break;
}
thread::sleep(Duration::from_millis(100));
}
assert_eq!(count, 2);
}
} }

4
src/core/index_meta.rs
View File

@@ -46,13 +46,13 @@ impl fmt::Debug for IndexMeta {
mod tests { mod tests {
use super::IndexMeta; use super::IndexMeta;
use schema::{Schema, TEXT}; use schema::{SchemaBuilder, TEXT};
use serde_json; use serde_json;
#[test] #[test]
fn test_serialize_metas() { fn test_serialize_metas() {
let schema = { let schema = {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
schema_builder.add_text_field("text", TEXT); schema_builder.add_text_field("text", TEXT);
schema_builder.build() schema_builder.build()
}; };

5
src/core/inverted_index_reader.rs
View File

@@ -32,7 +32,10 @@ pub struct InvertedIndexReader {
} }
impl InvertedIndexReader { impl InvertedIndexReader {
#[cfg_attr(feature = "cargo-clippy", allow(clippy::needless_pass_by_value))] // for symetry #[cfg_attr(
feature = "cargo-clippy",
allow(clippy::needless_pass_by_value)
)] // for symetry
pub(crate) fn new( pub(crate) fn new(
termdict: TermDictionary, termdict: TermDictionary,
postings_source: ReadOnlySource, postings_source: ReadOnlySource,

4
src/core/mod.rs
View File

@@ -1,7 +1,7 @@
mod executor;
pub mod index; pub mod index;
mod index_meta; mod index_meta;
mod inverted_index_reader; mod inverted_index_reader;
mod pool;
pub mod searcher; pub mod searcher;
mod segment; mod segment;
mod segment_component; mod segment_component;
@@ -9,7 +9,6 @@ mod segment_id;
mod segment_meta; mod segment_meta;
mod segment_reader; mod segment_reader;
pub use self::executor::Executor;
pub use self::index::Index; pub use self::index::Index;
pub use self::index_meta::IndexMeta; pub use self::index_meta::IndexMeta;
pub use self::inverted_index_reader::InvertedIndexReader; pub use self::inverted_index_reader::InvertedIndexReader;
@@ -24,7 +23,6 @@ pub use self::segment_reader::SegmentReader;
use std::path::PathBuf; use std::path::PathBuf;
lazy_static! { lazy_static! {
/// The meta file contains all the information about the list of segments and the schema /// The meta file contains all the information about the list of segments and the schema
/// of the index. /// of the index.
pub static ref META_FILEPATH: PathBuf = PathBuf::from("meta.json"); pub static ref META_FILEPATH: PathBuf = PathBuf::from("meta.json");

107
src/reader/pool.rs → src/core/pool.rs
View File

@@ -1,5 +1,5 @@
use crossbeam::crossbeam_channel::unbounded; use crossbeam::queue::MsQueue;
use crossbeam::{Receiver, RecvError, Sender}; use std::mem;
use std::ops::{Deref, DerefMut}; use std::ops::{Deref, DerefMut};
use std::sync::atomic::AtomicUsize; use std::sync::atomic::AtomicUsize;
use std::sync::atomic::Ordering; use std::sync::atomic::Ordering;
@@ -10,52 +10,15 @@ pub struct GenerationItem<T> {
item: T, item: T,
} }
/// Queue implementation for the Object Pool below
/// Uses the unbounded Linked-List type queue from crossbeam-channel
/// Splits the Queue into sender and receiver
struct Queue<T> {
sender: Sender<T>,
receiver: Receiver<T>,
}
impl<T> Queue<T> {
fn new() -> Self {
let (s, r) = unbounded();
Queue {
sender: s,
receiver: r,
}
}
/// Sender trait returns a Result type, which is ignored.
/// The Result is not handled at the moment
fn push(&self, elem: T) {
self.sender
.send(elem)
.expect("Sending an item to crossbeam-queue shouldn't fail");
}
/// Relies on the underlying crossbeam-channel Receiver
/// to block on empty queue
fn pop(&self) -> Result<T, RecvError> {
self.receiver.recv()
}
}
/// An object pool
///
/// This is used in tantivy to create a pool of `Searcher`.
/// Object are wrapped in a `LeasedItem` wrapper and are
/// released automatically back into the pool on `Drop`.
pub struct Pool<T> { pub struct Pool<T> {
queue: Arc<Queue<GenerationItem<T>>>, queue: Arc<MsQueue<GenerationItem<T>>>,
freshest_generation: AtomicUsize, freshest_generation: AtomicUsize,
next_generation: AtomicUsize, next_generation: AtomicUsize,
} }
impl<T> Pool<T> { impl<T> Pool<T> {
pub fn new() -> Pool<T> { pub fn new() -> Pool<T> {
let queue = Arc::new(Queue::new()); let queue = Arc::new(MsQueue::new());
Pool { Pool {
queue, queue,
freshest_generation: AtomicUsize::default(), freshest_generation: AtomicUsize::default(),
@@ -63,10 +26,6 @@ impl<T> Pool<T> {
} }
} }
/// Publishes a new generation of `Searcher`.
///
/// After publish, all new `Searcher` acquired will be
/// of the new generation.
pub fn publish_new_generation(&self, items: Vec<T>) { pub fn publish_new_generation(&self, items: Vec<T>) {
let next_generation = self.next_generation.fetch_add(1, Ordering::SeqCst) + 1; let next_generation = self.next_generation.fetch_add(1, Ordering::SeqCst) + 1;
for item in items { for item in items {
@@ -102,14 +61,10 @@ impl<T> Pool<T> {
self.freshest_generation.load(Ordering::Acquire) self.freshest_generation.load(Ordering::Acquire)
} }
/// Acquires a new searcher.
///
/// If no searcher is available, this methods block until
/// a searcher is released.
pub fn acquire(&self) -> LeasedItem<T> { pub fn acquire(&self) -> LeasedItem<T> {
let generation = self.generation(); let generation = self.generation();
loop { loop {
let gen_item = self.queue.pop().unwrap(); let gen_item = self.queue.pop();
if gen_item.generation >= generation { if gen_item.generation >= generation {
return LeasedItem { return LeasedItem {
gen_item: Some(gen_item), gen_item: Some(gen_item),
@@ -125,7 +80,7 @@ impl<T> Pool<T> {
pub struct LeasedItem<T> { pub struct LeasedItem<T> {
gen_item: Option<GenerationItem<T>>, gen_item: Option<GenerationItem<T>>,
recycle_queue: Arc<Queue<GenerationItem<T>>>, recycle_queue: Arc<MsQueue<GenerationItem<T>>>,
} }
impl<T> Deref for LeasedItem<T> { impl<T> Deref for LeasedItem<T> {
@@ -152,9 +107,9 @@ impl<T> DerefMut for LeasedItem<T> {
impl<T> Drop for LeasedItem<T> { impl<T> Drop for LeasedItem<T> {
fn drop(&mut self) { fn drop(&mut self) {
if let Some(gen_item) = self.gen_item.take() { let gen_item: GenerationItem<T> = mem::replace(&mut self.gen_item, None)
self.recycle_queue.push(gen_item); .expect("Unwrapping a leased item should never fail");
} self.recycle_queue.push(gen_item);
} }
} }
@@ -162,7 +117,6 @@ impl<T> Drop for LeasedItem<T> {
mod tests { mod tests {
use super::Pool; use super::Pool;
use super::Queue;
use std::iter; use std::iter;
#[test] #[test]
@@ -179,47 +133,4 @@ mod tests {
assert_eq!(*pool.acquire(), 11); assert_eq!(*pool.acquire(), 11);
} }
} }
#[test]
fn test_queue() {
let q = Queue::new();
let elem = 5;
q.push(elem);
let res = q.pop();
assert_eq!(res.unwrap(), elem);
}
#[test]
fn test_pool_dont_panic_on_empty_pop() {
// When the object pool is exhausted, it shouldn't panic on pop()
use std::sync::Arc;
use std::{thread, time};
// Wrap the pool in an Arc, same way as its used in `core/index.rs`
let pool = Arc::new(Pool::new());
// clone pools outside the move scope of each new thread
let pool1 = Arc::clone(&pool);
let pool2 = Arc::clone(&pool);
let elements_for_pool = vec![1, 2];
pool.publish_new_generation(elements_for_pool);
let mut threads = vec![];
let sleep_dur = time::Duration::from_millis(10);
// spawn one more thread than there are elements in the pool
threads.push(thread::spawn(move || {
// leasing to make sure it's not dropped before sleep is called
let _leased_searcher = &pool.acquire();
thread::sleep(sleep_dur);
}));
threads.push(thread::spawn(move || {
// leasing to make sure it's not dropped before sleep is called
let _leased_searcher = &pool1.acquire();
thread::sleep(sleep_dur);
}));
threads.push(thread::spawn(move || {
// leasing to make sure it's not dropped before sleep is called
let _leased_searcher = &pool2.acquire();
thread::sleep(sleep_dur);
}));
}
} }

103
src/core/searcher.rs
View File

@@ -1,43 +1,17 @@
use collector::Collector; use collector::Collector;
use collector::SegmentCollector;
use core::Executor;
use core::InvertedIndexReader; use core::InvertedIndexReader;
use core::SegmentReader; use core::SegmentReader;
use query::Query; use query::Query;
use query::Scorer;
use query::Weight;
use schema::Document; use schema::Document;
use schema::Schema; use schema::Schema;
use schema::{Field, Term}; use schema::{Field, Term};
use space_usage::SearcherSpaceUsage;
use std::fmt; use std::fmt;
use std::sync::Arc; use std::sync::Arc;
use store::StoreReader;
use termdict::TermMerger; use termdict::TermMerger;
use DocAddress; use DocAddress;
use Index; use Index;
use Result; use Result;
fn collect_segment<C: Collector>(
collector: &C,
weight: &Weight,
segment_ord: u32,
segment_reader: &SegmentReader,
) -> Result<C::Fruit> {
let mut scorer = weight.scorer(segment_reader)?;
let mut segment_collector = collector.for_segment(segment_ord as u32, segment_reader)?;
if let Some(delete_bitset) = segment_reader.delete_bitset() {
scorer.for_each(&mut |doc, score| {
if !delete_bitset.is_deleted(doc) {
segment_collector.collect(doc, score);
}
});
} else {
scorer.for_each(&mut |doc, score| segment_collector.collect(doc, score));
}
Ok(segment_collector.harvest())
}
/// Holds a list of `SegmentReader`s ready for search. /// Holds a list of `SegmentReader`s ready for search.
/// ///
/// It guarantees that the `Segment` will not be removed before /// It guarantees that the `Segment` will not be removed before
@@ -47,7 +21,6 @@ pub struct Searcher {
schema: Schema, schema: Schema,
index: Index, index: Index,
segment_readers: Vec<SegmentReader>, segment_readers: Vec<SegmentReader>,
store_readers: Vec<StoreReader>,
} }
impl Searcher { impl Searcher {
@@ -57,15 +30,10 @@ impl Searcher {
index: Index, index: Index,
segment_readers: Vec<SegmentReader>, segment_readers: Vec<SegmentReader>,
) -> Searcher { ) -> Searcher {
let store_readers = segment_readers
.iter()
.map(|segment_reader| segment_reader.get_store_reader())
.collect();
Searcher { Searcher {
schema, schema,
index, index,
segment_readers, segment_readers,
store_readers,
} }
} }
@@ -80,8 +48,8 @@ impl Searcher {
/// the request to the right `Segment`. /// the request to the right `Segment`.
pub fn doc(&self, doc_address: DocAddress) -> Result<Document> { pub fn doc(&self, doc_address: DocAddress) -> Result<Document> {
let DocAddress(segment_local_id, doc_id) = doc_address; let DocAddress(segment_local_id, doc_id) = doc_address;
let store_reader = &self.store_readers[segment_local_id as usize]; let segment_reader = &self.segment_readers[segment_local_id as usize];
store_reader.get(doc_id) segment_reader.doc(doc_id)
} }
/// Access the schema associated to the index of this searcher. /// Access the schema associated to the index of this searcher.
@@ -104,8 +72,7 @@ impl Searcher {
.iter() .iter()
.map(|segment_reader| { .map(|segment_reader| {
u64::from(segment_reader.inverted_index(term.field()).doc_freq(term)) u64::from(segment_reader.inverted_index(term.field()).doc_freq(term))
}) }).sum::<u64>()
.sum::<u64>()
} }
/// Return the list of segment readers /// Return the list of segment readers
@@ -118,58 +85,9 @@ impl Searcher {
&self.segment_readers[segment_ord as usize] &self.segment_readers[segment_ord as usize]
} }
/// Runs a query on the segment readers wrapped by the searcher. /// Runs a query on the segment readers wrapped by the searcher
/// pub fn search<C: Collector>(&self, query: &Query, collector: &mut C) -> Result<()> {
/// Search works as follows : query.search(self, collector)
///
/// First the weight object associated to the query is created.
///
/// Then, the query loops over the segments and for each segment :
/// - setup the collector and informs it that the segment being processed has changed.
/// - creates a SegmentCollector for collecting documents associated to the segment
/// - creates a `Scorer` object associated for this segment
/// - iterate through the matched documents and push them to the segment collector.
///
/// Finally, the Collector merges each of the child collectors into itself for result usability
/// by the caller.
pub fn search<C: Collector>(&self, query: &Query, collector: &C) -> Result<C::Fruit> {
let executor = self.index.search_executor();
self.search_with_executor(query, collector, executor)
}
/// Same as [`search(...)`](#method.search) but multithreaded.
///
/// The current implementation is rather naive :
/// multithreading is by splitting search into as many task
/// as there are segments.
///
/// It is powerless at making search faster if your index consists in
/// one large segment.
///
/// Also, keep in my multithreading a single query on several
/// threads will not improve your throughput. It can actually
/// hurt it. It will however, decrease the average response time.
pub fn search_with_executor<C: Collector>(
&self,
query: &Query,
collector: &C,
executor: &Executor,
) -> Result<C::Fruit> {
let scoring_enabled = collector.requires_scoring();
let weight = query.weight(self, scoring_enabled)?;
let segment_readers = self.segment_readers();
let fruits = executor.map(
|(segment_ord, segment_reader)| {
collect_segment(
collector,
weight.as_ref(),
segment_ord as u32,
segment_reader,
)
},
segment_readers.iter().enumerate(),
)?;
collector.merge_fruits(fruits)
} }
/// Return the field searcher associated to a `Field`. /// Return the field searcher associated to a `Field`.
@@ -181,15 +99,6 @@ impl Searcher {
.collect::<Vec<_>>(); .collect::<Vec<_>>();
FieldSearcher::new(inv_index_readers) FieldSearcher::new(inv_index_readers)
} }
/// Summarize total space usage of this searcher.
pub fn space_usage(&self) -> SearcherSpaceUsage {
let mut space_usage = SearcherSpaceUsage::new();
for segment_reader in self.segment_readers.iter() {
space_usage.add_segment(segment_reader.space_usage());
}
space_usage
}
} }
pub struct FieldSearcher { pub struct FieldSearcher {

2
src/core/segment_component.rs
View File

@@ -41,6 +41,6 @@ impl SegmentComponent {
SegmentComponent::STORE, SegmentComponent::STORE,
SegmentComponent::DELETE, SegmentComponent::DELETE,
]; ];
SEGMENT_COMPONENTS.iter() SEGMENT_COMPONENTS.into_iter()
} }
} }

4
src/core/segment_id.rs
View File

@@ -19,7 +19,7 @@ pub struct SegmentId(Uuid);
#[cfg(test)] #[cfg(test)]
lazy_static! { lazy_static! {
static ref AUTO_INC_COUNTER: atomic::AtomicUsize = atomic::AtomicUsize::default(); static ref AUTO_INC_COUNTER: atomic::AtomicUsize = atomic::AtomicUsize::default();
static ref ZERO_ARRAY: [u8; 8] = [0u8; 8]; static ref EMPTY_ARR: [u8; 8] = [0u8; 8];
} }
// During tests, we generate the segment id in a autoincrement manner // During tests, we generate the segment id in a autoincrement manner
@@ -30,7 +30,7 @@ lazy_static! {
#[cfg(test)] #[cfg(test)]
fn create_uuid() -> Uuid { fn create_uuid() -> Uuid {
let new_auto_inc_id = (*AUTO_INC_COUNTER).fetch_add(1, atomic::Ordering::SeqCst); let new_auto_inc_id = (*AUTO_INC_COUNTER).fetch_add(1, atomic::Ordering::SeqCst);
Uuid::from_fields(new_auto_inc_id as u32, 0, 0, &*ZERO_ARRAY).unwrap() Uuid::from_fields(new_auto_inc_id as u32, 0, 0, &*EMPTY_ARR).unwrap()
} }
#[cfg(not(test))] #[cfg(not(test))]

49
src/core/segment_reader.rs
View File

@@ -4,7 +4,6 @@ use core::InvertedIndexReader;
use core::Segment; use core::Segment;
use core::SegmentComponent; use core::SegmentComponent;
use core::SegmentId; use core::SegmentId;
use directory::ReadOnlySource;
use error::TantivyError; use error::TantivyError;
use fastfield::DeleteBitSet; use fastfield::DeleteBitSet;
use fastfield::FacetReader; use fastfield::FacetReader;
@@ -13,10 +12,10 @@ use fastfield::{self, FastFieldNotAvailableError};
use fastfield::{BytesFastFieldReader, FastValue, MultiValueIntFastFieldReader}; use fastfield::{BytesFastFieldReader, FastValue, MultiValueIntFastFieldReader};
use fieldnorm::FieldNormReader; use fieldnorm::FieldNormReader;
use schema::Cardinality; use schema::Cardinality;
use schema::Document;
use schema::Field; use schema::Field;
use schema::FieldType; use schema::FieldType;
use schema::Schema; use schema::Schema;
use space_usage::SegmentSpaceUsage;
use std::collections::HashMap; use std::collections::HashMap;
use std::fmt; use std::fmt;
use std::sync::Arc; use std::sync::Arc;
@@ -54,7 +53,7 @@ pub struct SegmentReader {
fast_fields_composite: CompositeFile, fast_fields_composite: CompositeFile,
fieldnorms_composite: CompositeFile, fieldnorms_composite: CompositeFile,
store_source: ReadOnlySource, store_reader: StoreReader,
delete_bitset_opt: Option<DeleteBitSet>, delete_bitset_opt: Option<DeleteBitSet>,
schema: Schema, schema: Schema,
} }
@@ -197,7 +196,8 @@ impl SegmentReader {
/// Accessor to the segment's `Field norms`'s reader. /// Accessor to the segment's `Field norms`'s reader.
/// ///
/// Field norms are the length (in tokens) of the fields. /// Field norms are the length (in tokens) of the fields.
/// It is used in the computation of the [TfIdf](https://fulmicoton.gitbooks.io/tantivy-doc/content/tfidf.html). /// It is used in the computation of the [TfIdf]
/// (https://fulmicoton.gitbooks.io/tantivy-doc/content/tfidf.html).
/// ///
/// They are simply stored as a fast field, serialized in /// They are simply stored as a fast field, serialized in
/// the `.fieldnorm` file of the segment. /// the `.fieldnorm` file of the segment.
@@ -215,8 +215,8 @@ impl SegmentReader {
} }
/// Accessor to the segment's `StoreReader`. /// Accessor to the segment's `StoreReader`.
pub fn get_store_reader(&self) -> StoreReader { pub fn get_store_reader(&self) -> &StoreReader {
StoreReader::from_source(self.store_source.clone()) &self.store_reader
} }
/// Open a new segment for reading. /// Open a new segment for reading.
@@ -225,6 +225,7 @@ impl SegmentReader {
let termdict_composite = CompositeFile::open(&termdict_source)?; let termdict_composite = CompositeFile::open(&termdict_source)?;
let store_source = segment.open_read(SegmentComponent::STORE)?; let store_source = segment.open_read(SegmentComponent::STORE)?;
let store_reader = StoreReader::from_source(store_source);
fail_point!("SegmentReader::open#middle"); fail_point!("SegmentReader::open#middle");
@@ -270,7 +271,7 @@ impl SegmentReader {
fast_fields_composite, fast_fields_composite,
fieldnorms_composite, fieldnorms_composite,
segment_id: segment.id(), segment_id: segment.id(),
store_source, store_reader,
delete_bitset_opt, delete_bitset_opt,
positions_composite, positions_composite,
positions_idx_composite, positions_idx_composite,
@@ -349,6 +350,14 @@ impl SegmentReader {
inv_idx_reader inv_idx_reader
} }
/// Returns the document (or to be accurate, its stored field)
/// bearing the given doc id.
/// This method is slow and should seldom be called from
/// within a collector.
pub fn doc(&self, doc_id: DocId) -> Result<Document> {
self.store_reader.get(doc_id)
}
/// Returns the segment id /// Returns the segment id
pub fn segment_id(&self) -> SegmentId { pub fn segment_id(&self) -> SegmentId {
self.segment_id self.segment_id
@@ -372,24 +381,6 @@ impl SegmentReader {
pub fn doc_ids_alive(&self) -> SegmentReaderAliveDocsIterator { pub fn doc_ids_alive(&self) -> SegmentReaderAliveDocsIterator {
SegmentReaderAliveDocsIterator::new(&self) SegmentReaderAliveDocsIterator::new(&self)
} }
/// Summarize total space usage of this segment.
pub fn space_usage(&self) -> SegmentSpaceUsage {
SegmentSpaceUsage::new(
self.num_docs(),
self.termdict_composite.space_usage(),
self.postings_composite.space_usage(),
self.positions_composite.space_usage(),
self.positions_idx_composite.space_usage(),
self.fast_fields_composite.space_usage(),
self.fieldnorms_composite.space_usage(),
self.get_store_reader().space_usage(),
self.delete_bitset_opt
.as_ref()
.map(|x| x.space_usage())
.unwrap_or(0),
)
}
} }
impl fmt::Debug for SegmentReader { impl fmt::Debug for SegmentReader {
@@ -447,12 +438,12 @@ impl<'a> Iterator for SegmentReaderAliveDocsIterator<'a> {
#[cfg(test)] #[cfg(test)]
mod test { mod test {
use core::Index; use core::Index;
use schema::{Schema, Term, STORED, TEXT}; use schema::{SchemaBuilder, Term, STORED, TEXT};
use DocId; use DocId;
#[test] #[test]
fn test_alive_docs_iterator() { fn test_alive_docs_iterator() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
schema_builder.add_text_field("name", TEXT | STORED); schema_builder.add_text_field("name", TEXT | STORED);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema.clone()); let index = Index::create_in_ram(schema.clone());
@@ -477,7 +468,9 @@ mod test {
// ok, now we should have a deleted doc // ok, now we should have a deleted doc
index_writer2.commit().unwrap(); index_writer2.commit().unwrap();
} }
let searcher = index.reader().unwrap().searcher();
index.load_searchers().unwrap();
let searcher = index.searcher();
let docs: Vec<DocId> = searcher.segment_reader(0).doc_ids_alive().collect(); let docs: Vec<DocId> = searcher.segment_reader(0).doc_ids_alive().collect();
assert_eq!(vec![0u32, 2u32], docs); assert_eq!(vec![0u32, 2u32], docs);
} }

142
src/directory/directory.rs
View File

@@ -1,104 +1,11 @@
use directory::directory_lock::Lock;
use directory::error::LockError;
use directory::error::{DeleteError, OpenReadError, OpenWriteError}; use directory::error::{DeleteError, OpenReadError, OpenWriteError};
use directory::WatchCallback;
use directory::WatchHandle;
use directory::{ReadOnlySource, WritePtr}; use directory::{ReadOnlySource, WritePtr};
use std::fmt; use std::fmt;
use std::io; use std::io;
use std::io::Write;
use std::marker::Send; use std::marker::Send;
use std::marker::Sync; use std::marker::Sync;
use std::path::Path; use std::path::Path;
use std::path::PathBuf;
use std::result; use std::result;
use std::thread;
use std::time::Duration;
/// Retry the logic of acquiring locks is pretty simple.
/// We just retry `n` times after a given `duratio`, both
/// depending on the type of lock.
struct RetryPolicy {
num_retries: usize,
wait_in_ms: u64,
}
impl RetryPolicy {
fn no_retry() -> RetryPolicy {
RetryPolicy {
num_retries: 0,
wait_in_ms: 0,
}
}
fn wait_and_retry(&mut self) -> bool {
if self.num_retries == 0 {
false
} else {
self.num_retries -= 1;
let wait_duration = Duration::from_millis(self.wait_in_ms);
thread::sleep(wait_duration);
true
}
}
}
/// The `DirectoryLock` is an object that represents a file lock.
/// See [`LockType`](struct.LockType.html)
///
/// It is transparently associated to a lock file, that gets deleted
/// on `Drop.` The lock is released automatically on `Drop`.
pub struct DirectoryLock(Box<Drop + Send + 'static>);
struct DirectoryLockGuard {
directory: Box<Directory>,
path: PathBuf,
}
impl<T: Drop + Send + 'static> From<Box<T>> for DirectoryLock {
fn from(underlying: Box<T>) -> Self {
DirectoryLock(underlying)
}
}
impl Drop for DirectoryLockGuard {
fn drop(&mut self) {
if let Err(e) = self.directory.delete(&*self.path) {
error!("Failed to remove the lock file. {:?}", e);
}
}
}
enum TryAcquireLockError {
FileExists,
IOError(io::Error),
}
fn try_acquire_lock(
filepath: &Path,
directory: &mut Directory,
) -> Result<DirectoryLock, TryAcquireLockError> {
let mut write = directory.open_write(filepath).map_err(|e| match e {
OpenWriteError::FileAlreadyExists(_) => TryAcquireLockError::FileExists,
OpenWriteError::IOError(io_error) => TryAcquireLockError::IOError(io_error.into()),
})?;
write.flush().map_err(TryAcquireLockError::IOError)?;
Ok(DirectoryLock::from(Box::new(DirectoryLockGuard {
directory: directory.box_clone(),
path: filepath.to_owned(),
})))
}
fn retry_policy(is_blocking: bool) -> RetryPolicy {
if is_blocking {
RetryPolicy {
num_retries: 100,
wait_in_ms: 100,
}
} else {
RetryPolicy::no_retry()
}
}
/// Write-once read many (WORM) abstraction for where /// Write-once read many (WORM) abstraction for where
/// tantivy's data should be stored. /// tantivy's data should be stored.
@@ -166,55 +73,6 @@ pub trait Directory: DirectoryClone + fmt::Debug + Send + Sync + 'static {
/// ///
/// The file may or may not previously exist. /// The file may or may not previously exist.
fn atomic_write(&mut self, path: &Path, data: &[u8]) -> io::Result<()>; fn atomic_write(&mut self, path: &Path, data: &[u8]) -> io::Result<()>;
/// Acquire a lock in the given directory.
///
/// The method is blocking or not depending on the `Lock` object.
fn acquire_lock(&self, lock: &Lock) -> Result<DirectoryLock, LockError> {
let mut box_directory = self.box_clone();
let mut retry_policy = retry_policy(lock.is_blocking);
loop {
match try_acquire_lock(&lock.filepath, &mut *box_directory) {
Ok(result) => {
return Ok(result);
}
Err(TryAcquireLockError::FileExists) => {
if !retry_policy.wait_and_retry() {
return Err(LockError::LockBusy);
}
}
Err(TryAcquireLockError::IOError(io_error)) => {
return Err(LockError::IOError(io_error));
}
}
}
}
/// Registers a callback that will be called whenever a change on the `meta.json`
/// using the `atomic_write` API is detected.
///
/// The behavior when using `.watch()` on a file using `.open_write(...)` is, on the other
/// hand, undefined.
///
/// The file will be watched for the lifetime of the returned `WatchHandle`. The caller is
/// required to keep it.
/// It does not override previous callbacks. When the file is modified, all callback that are
/// registered (and whose `WatchHandle` is still alive) are triggered.
///
/// Internally, tantivy only uses this API to detect new commits to implement the
/// `OnCommit` `ReloadPolicy`. Not implementing watch in a `Directory` only prevents the
/// `OnCommit` `ReloadPolicy` to work properly.
fn watch(&self, watch_callback: WatchCallback) -> WatchHandle;
/// Ensure that all volatile files reach are persisted (in directory where that makes sense.)
///
/// In order to make Near Real Time efficient, tantivy introduced the notion of soft_commit vs
/// commit. Commit will call `.flush()`, while softcommit won't.
///
/// `meta.json` should be the last file to be flushed.
fn flush(&self) -> io::Result<()> {
Ok(())
}
} }
/// DirectoryClone /// DirectoryClone

56
src/directory/directory_lock.rs
View File

@@ -1,56 +0,0 @@
use std::path::PathBuf;
/// A directory lock.
///
/// A lock is associated to a specific path and some
/// [`LockParams`](./enum.LockParams.html).
/// Tantivy itself uses only two locks but client application
/// can use the directory facility to define their own locks.
/// - [INDEX_WRITER_LOCK](./struct.INDEX_WRITER_LOCK.html)
/// - [META_LOCK](./struct.META_LOCK.html)
///
/// Check out these locks documentation for more information.
///
#[derive(Debug)]
pub struct Lock {
/// The lock needs to be associated with its own file `path`.
/// Depending on the platform, the lock might rely on the creation
/// and deletion of this filepath.
pub filepath: PathBuf,
/// `lock_params` describes whether acquiring the lock is meant
/// to be a blocking operation or a non-blocking.
///
/// Acquiring a blocking lock blocks until the lock is
/// available.
/// Acquiring a blocking lock returns rapidly, either successfully
/// or with an error signifying that someone is already holding
/// the lock.
pub is_blocking: bool,
}
lazy_static! {
/// Only one process should be able to write tantivy's index at a time.
/// This lock file, when present, is in charge of preventing other processes to open an IndexWriter.
///
/// If the process is killed and this file remains, it is safe to remove it manually.
///
/// Failing to acquire this lock usually means a misuse of tantivy's API,
/// (creating more than one instance of the `IndexWriter`), are a spurious
/// lock file remaining after a crash. In the latter case, removing the file after
/// checking no process running tantivy is running is safe.
pub static ref INDEX_WRITER_LOCK: Lock = Lock {
filepath: PathBuf::from(".tantivy-writer.lock"),
is_blocking: false
};
/// The meta lock file is here to protect the segment files being opened by
/// `IndexReader::reload()` from being garbage collected.
/// It makes it possible for another process to safely consume
/// our index in-writing. Ideally, we may have prefered `RWLock` semantics
/// here, but it is difficult to achieve on Windows.
///
/// Opening segment readers is a very fast process.
pub static ref META_LOCK: Lock = Lock {
filepath: PathBuf::from(".tantivy-meta.lock"),
is_blocking: true
};
}

35
src/directory/error.rs
View File

@@ -3,22 +3,6 @@ use std::fmt;
use std::io; use std::io;
use std::path::PathBuf; use std::path::PathBuf;
/// Error while trying to acquire a directory lock.
#[derive(Debug, Fail)]
pub enum LockError {
/// Failed to acquired a lock as it is already hold by another
/// client.
/// - In the context of a blocking lock, this means the lock was not released within some `timeout` period.
/// - In the context of a non-blocking lock, this means the lock was busy at the moment of the call.
#[fail(
display = "Could not acquire lock as it is already held, possibly by a different process."
)]
LockBusy,
/// Trying to acquire a lock failed with an `IOError`
#[fail(display = "Failed to acquire the lock due to an io:Error.")]
IOError(io::Error),
}
/// General IO error with an optional path to the offending file. /// General IO error with an optional path to the offending file.
#[derive(Debug)] #[derive(Debug)]
pub struct IOError { pub struct IOError {
@@ -26,12 +10,6 @@ pub struct IOError {
err: io::Error, err: io::Error,
} }
impl Into<io::Error> for IOError {
fn into(self) -> io::Error {
self.err
}
}
impl fmt::Display for IOError { impl fmt::Display for IOError {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match self.path { match self.path {
@@ -73,14 +51,6 @@ pub enum OpenDirectoryError {
DoesNotExist(PathBuf), DoesNotExist(PathBuf),
/// The path exists but is not a directory. /// The path exists but is not a directory.
NotADirectory(PathBuf), NotADirectory(PathBuf),
/// IoError
IoError(io::Error),
}
impl From<io::Error> for OpenDirectoryError {
fn from(io_err: io::Error) -> Self {
OpenDirectoryError::IoError(io_err)
}
} }
impl fmt::Display for OpenDirectoryError { impl fmt::Display for OpenDirectoryError {
@@ -92,11 +62,6 @@ impl fmt::Display for OpenDirectoryError {
OpenDirectoryError::NotADirectory(ref path) => { OpenDirectoryError::NotADirectory(ref path) => {
write!(f, "the path '{:?}' exists but is not a directory", path) write!(f, "the path '{:?}' exists but is not a directory", path)
} }
OpenDirectoryError::IoError(ref err) => write!(
f,
"IOError while trying to open/create the directory. {:?}",
err
),
} }
} }
} }

194
src/directory/managed_directory.rs
View File

@@ -1,11 +1,8 @@
use core::MANAGED_FILEPATH; use core::MANAGED_FILEPATH;
use directory::error::{DeleteError, IOError, LockError, OpenReadError, OpenWriteError}; use directory::error::{DeleteError, IOError, OpenReadError, OpenWriteError};
use directory::DirectoryLock;
use directory::Lock;
use directory::META_LOCK;
use directory::{ReadOnlySource, WritePtr}; use directory::{ReadOnlySource, WritePtr};
use directory::{WatchCallback, WatchHandle}; use error::TantivyError;
use error::DataCorruption; use indexer::LockType;
use serde_json; use serde_json;
use std::collections::HashSet; use std::collections::HashSet;
use std::io; use std::io;
@@ -62,17 +59,12 @@ fn save_managed_paths(
impl ManagedDirectory { impl ManagedDirectory {
/// Wraps a directory as managed directory. /// Wraps a directory as managed directory.
pub fn wrap<Dir: Directory>(directory: Dir) -> Result<ManagedDirectory> { pub fn new<Dir: Directory>(directory: Dir) -> Result<ManagedDirectory> {
match directory.atomic_read(&MANAGED_FILEPATH) { match directory.atomic_read(&MANAGED_FILEPATH) {
Ok(data) => { Ok(data) => {
let managed_files_json = String::from_utf8_lossy(&data); let managed_files_json = String::from_utf8_lossy(&data);
let managed_files: HashSet<PathBuf> = serde_json::from_str(&managed_files_json) let managed_files: HashSet<PathBuf> = serde_json::from_str(&managed_files_json)
.map_err(|e| { .map_err(|_| TantivyError::CorruptedFile(MANAGED_FILEPATH.clone()))?;
DataCorruption::new(
MANAGED_FILEPATH.clone(),
format!("Managed file cannot be deserialized: {:?}. ", e),
)
})?;
Ok(ManagedDirectory { Ok(ManagedDirectory {
directory: Box::new(directory), directory: Box::new(directory),
meta_informations: Arc::new(RwLock::new(MetaInformation { meta_informations: Arc::new(RwLock::new(MetaInformation {
@@ -95,9 +87,6 @@ impl ManagedDirectory {
/// ///
/// * `living_files` - List of files that are still used by the index. /// * `living_files` - List of files that are still used by the index.
/// ///
/// The use a callback ensures that the list of living_files is computed
/// while we hold the lock on meta.
///
/// This method does not panick nor returns errors. /// This method does not panick nor returns errors.
/// If a file cannot be deleted (for permission reasons for instance) /// If a file cannot be deleted (for permission reasons for instance)
/// an error is simply logged, and the file remains in the list of managed /// an error is simply logged, and the file remains in the list of managed
@@ -128,7 +117,7 @@ impl ManagedDirectory {
// 2) writer change meta.json (for instance after a merge or a commit) // 2) writer change meta.json (for instance after a merge or a commit)
// 3) gc kicks in. // 3) gc kicks in.
// 4) gc removes a file that was useful for process B, before process B opened it. // 4) gc removes a file that was useful for process B, before process B opened it.
if let Ok(_meta_lock) = self.acquire_lock(&META_LOCK) { if let Ok(_meta_lock) = LockType::MetaLock.acquire_lock(self) {
let living_files = get_living_files(); let living_files = get_living_files();
for managed_path in &meta_informations_rlock.managed_paths { for managed_path in &meta_informations_rlock.managed_paths {
if !living_files.contains(managed_path) { if !living_files.contains(managed_path) {
@@ -238,14 +227,6 @@ impl Directory for ManagedDirectory {
fn exists(&self, path: &Path) -> bool { fn exists(&self, path: &Path) -> bool {
self.directory.exists(path) self.directory.exists(path)
} }
fn acquire_lock(&self, lock: &Lock) -> result::Result<DirectoryLock, LockError> {
self.directory.acquire_lock(lock)
}
fn watch(&self, watch_callback: WatchCallback) -> WatchHandle {
self.directory.watch(watch_callback)
}
} }
impl Clone for ManagedDirectory { impl Clone for ManagedDirectory {
@@ -260,98 +241,95 @@ impl Clone for ManagedDirectory {
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::*;
#[cfg(feature = "mmap")] #[cfg(feature = "mmap")]
mod mmap_specific { use directory::MmapDirectory;
use std::io::Write;
use std::path::Path;
use tempdir::TempDir;
use super::super::*; lazy_static! {
use std::path::Path; static ref TEST_PATH1: &'static Path = Path::new("some_path_for_test");
use tempdir::TempDir; static ref TEST_PATH2: &'static Path = Path::new("some_path_for_test2");
}
lazy_static! {
static ref TEST_PATH1: &'static Path = Path::new("some_path_for_test");
static ref TEST_PATH2: &'static Path = Path::new("some_path_for_test2");
}
use directory::MmapDirectory;
use std::io::Write;
#[test]
fn test_managed_directory() {
let tempdir = TempDir::new("index").unwrap();
let tempdir_path = PathBuf::from(tempdir.path());
{
let mmap_directory = MmapDirectory::open(&tempdir_path).unwrap();
let mut managed_directory = ManagedDirectory::wrap(mmap_directory).unwrap();
{
let mut write_file = managed_directory.open_write(*TEST_PATH1).unwrap();
write_file.flush().unwrap();
}
{
managed_directory
.atomic_write(*TEST_PATH2, &vec![0u8, 1u8])
.unwrap();
}
{
assert!(managed_directory.exists(*TEST_PATH1));
assert!(managed_directory.exists(*TEST_PATH2));
}
{
let living_files: HashSet<PathBuf> =
[TEST_PATH1.to_owned()].into_iter().cloned().collect();
managed_directory.garbage_collect(|| living_files);
}
{
assert!(managed_directory.exists(*TEST_PATH1));
assert!(!managed_directory.exists(*TEST_PATH2));
}
}
{
let mmap_directory = MmapDirectory::open(&tempdir_path).unwrap();
let mut managed_directory = ManagedDirectory::wrap(mmap_directory).unwrap();
{
assert!(managed_directory.exists(*TEST_PATH1));
assert!(!managed_directory.exists(*TEST_PATH2));
}
{
let living_files: HashSet<PathBuf> = HashSet::new();
managed_directory.garbage_collect(|| living_files);
}
{
assert!(!managed_directory.exists(*TEST_PATH1));
assert!(!managed_directory.exists(*TEST_PATH2));
}
}
}
#[test]
fn test_managed_directory_gc_while_mmapped() {
let tempdir = TempDir::new("index").unwrap();
let tempdir_path = PathBuf::from(tempdir.path());
let living_files = HashSet::new();
#[test]
#[cfg(feature = "mmap")]
fn test_managed_directory() {
let tempdir = TempDir::new("index").unwrap();
let tempdir_path = PathBuf::from(tempdir.path());
{
let mmap_directory = MmapDirectory::open(&tempdir_path).unwrap(); let mmap_directory = MmapDirectory::open(&tempdir_path).unwrap();
let mut managed_directory = ManagedDirectory::wrap(mmap_directory).unwrap(); let mut managed_directory = ManagedDirectory::new(mmap_directory).unwrap();
managed_directory {
.atomic_write(*TEST_PATH1, &vec![0u8, 1u8]) let mut write_file = managed_directory.open_write(*TEST_PATH1).unwrap();
.unwrap(); write_file.flush().unwrap();
assert!(managed_directory.exists(*TEST_PATH1)); }
{
let _mmap_read = managed_directory.open_read(*TEST_PATH1).unwrap(); managed_directory
managed_directory.garbage_collect(|| living_files.clone()); .atomic_write(*TEST_PATH2, &vec![0u8, 1u8])
if cfg!(target_os = "windows") { .unwrap();
// On Windows, gc should try and fail the file as it is mmapped. }
{
assert!(managed_directory.exists(*TEST_PATH1)); assert!(managed_directory.exists(*TEST_PATH1));
// unmap should happen here. assert!(managed_directory.exists(*TEST_PATH2));
drop(_mmap_read); }
// The file should still be in the list of managed file and {
// eventually be deleted once mmap is released. let living_files: HashSet<PathBuf> =
[TEST_PATH1.to_owned()].into_iter().cloned().collect();
managed_directory.garbage_collect(|| living_files); managed_directory.garbage_collect(|| living_files);
assert!(!managed_directory.exists(*TEST_PATH1)); }
} else { {
assert!(!managed_directory.exists(*TEST_PATH1)); assert!(managed_directory.exists(*TEST_PATH1));
assert!(!managed_directory.exists(*TEST_PATH2));
} }
} }
{
let mmap_directory = MmapDirectory::open(&tempdir_path).unwrap();
let mut managed_directory = ManagedDirectory::new(mmap_directory).unwrap();
{
assert!(managed_directory.exists(*TEST_PATH1));
assert!(!managed_directory.exists(*TEST_PATH2));
}
{
let living_files: HashSet<PathBuf> = HashSet::new();
managed_directory.garbage_collect(|| living_files);
}
{
assert!(!managed_directory.exists(*TEST_PATH1));
assert!(!managed_directory.exists(*TEST_PATH2));
}
}
}
#[test]
#[cfg(feature = "mmap ")]
fn test_managed_directory_gc_while_mmapped() {
let tempdir = TempDir::new("index").unwrap();
let tempdir_path = PathBuf::from(tempdir.path());
let living_files = HashSet::new();
let mmap_directory = MmapDirectory::open(&tempdir_path).unwrap();
let mut managed_directory = ManagedDirectory::new(mmap_directory).unwrap();
managed_directory
.atomic_write(*TEST_PATH1, &vec![0u8, 1u8])
.unwrap();
assert!(managed_directory.exists(*TEST_PATH1));
let _mmap_read = managed_directory.open_read(*TEST_PATH1).unwrap();
managed_directory.garbage_collect(|| living_files.clone());
if cfg!(target_os = "windows") {
// On Windows, gc should try and fail the file as it is mmapped.
assert!(managed_directory.exists(*TEST_PATH1));
// unmap should happen here.
drop(_mmap_read);
// The file should still be in the list of managed file and
// eventually be deleted once mmap is released.
managed_directory.garbage_collect(|| living_files);
assert!(!managed_directory.exists(*TEST_PATH1));
} else {
assert!(!managed_directory.exists(*TEST_PATH1));
}
} }
} }

470
src/directory/mmap_directory.rs
View File

@@ -1,24 +1,12 @@
extern crate fs2;
extern crate notify;
use self::fs2::FileExt;
use self::notify::RawEvent;
use self::notify::RecursiveMode;
use self::notify::Watcher;
use atomicwrites; use atomicwrites;
use core::META_FILEPATH; use common::make_io_err;
use directory::error::LockError;
use directory::error::{DeleteError, IOError, OpenDirectoryError, OpenReadError, OpenWriteError}; use directory::error::{DeleteError, IOError, OpenDirectoryError, OpenReadError, OpenWriteError};
use directory::read_only_source::BoxedData; use directory::shared_vec_slice::SharedVecSlice;
use directory::Directory; use directory::Directory;
use directory::DirectoryLock;
use directory::Lock;
use directory::ReadOnlySource; use directory::ReadOnlySource;
use directory::WatchCallback;
use directory::WatchCallbackList;
use directory::WatchHandle;
use directory::WritePtr; use directory::WritePtr;
use memmap::Mmap; use fst::raw::MmapReadOnly;
use std::collections::hash_map::Entry as HashMapEntry;
use std::collections::HashMap; use std::collections::HashMap;
use std::convert::From; use std::convert::From;
use std::fmt; use std::fmt;
@@ -28,22 +16,14 @@ use std::io::{self, Seek, SeekFrom};
use std::io::{BufWriter, Read, Write}; use std::io::{BufWriter, Read, Write};
use std::path::{Path, PathBuf}; use std::path::{Path, PathBuf};
use std::result; use std::result;
use std::sync::mpsc::{channel, Receiver, Sender};
use std::sync::Arc; use std::sync::Arc;
use std::sync::Mutex;
use std::sync::RwLock; use std::sync::RwLock;
use std::sync::Weak;
use std::thread;
use tempdir::TempDir; use tempdir::TempDir;
/// Create a default io error given a string.
pub(crate) fn make_io_err(msg: String) -> io::Error {
io::Error::new(io::ErrorKind::Other, msg)
}
/// Returns None iff the file exists, can be read, but is empty (and hence /// Returns None iff the file exists, can be read, but is empty (and hence
/// cannot be mmapped) /// cannot be mmapped).
fn open_mmap(full_path: &Path) -> result::Result<Option<Mmap>, OpenReadError> { ///
fn open_mmap(full_path: &Path) -> result::Result<Option<MmapReadOnly>, OpenReadError> {
let file = File::open(full_path).map_err(|e| { let file = File::open(full_path).map_err(|e| {
if e.kind() == io::ErrorKind::NotFound { if e.kind() == io::ErrorKind::NotFound {
OpenReadError::FileDoesNotExist(full_path.to_owned()) OpenReadError::FileDoesNotExist(full_path.to_owned())
@@ -62,7 +42,7 @@ fn open_mmap(full_path: &Path) -> result::Result<Option<Mmap>, OpenReadError> {
return Ok(None); return Ok(None);
} }
unsafe { unsafe {
memmap::Mmap::map(&file) MmapReadOnly::open(&file)
.map(Some) .map(Some)
.map_err(|e| From::from(IOError::with_path(full_path.to_owned(), e))) .map_err(|e| From::from(IOError::with_path(full_path.to_owned(), e)))
} }
@@ -85,7 +65,7 @@ pub struct CacheInfo {
struct MmapCache { struct MmapCache {
counters: CacheCounters, counters: CacheCounters,
cache: HashMap<PathBuf, Weak<BoxedData>>, cache: HashMap<PathBuf, MmapReadOnly>,
} }
impl Default for MmapCache { impl Default for MmapCache {
@@ -98,7 +78,12 @@ impl Default for MmapCache {
} }
impl MmapCache { impl MmapCache {
fn get_info(&self) -> CacheInfo { /// Removes a `MmapReadOnly` entry from the mmap cache.
fn discard_from_cache(&mut self, full_path: &Path) -> bool {
self.cache.remove(full_path).is_some()
}
fn get_info(&mut self) -> CacheInfo {
let paths: Vec<PathBuf> = self.cache.keys().cloned().collect(); let paths: Vec<PathBuf> = self.cache.keys().cloned().collect();
CacheInfo { CacheInfo {
counters: self.counters.clone(), counters: self.counters.clone(),
@@ -106,105 +91,23 @@ impl MmapCache {
} }
} }
fn remove_weak_ref(&mut self) { fn get_mmap(&mut self, full_path: &Path) -> Result<Option<MmapReadOnly>, OpenReadError> {
let keys_to_remove: Vec<PathBuf> = self Ok(match self.cache.entry(full_path.to_owned()) {
.cache HashMapEntry::Occupied(occupied_entry) => {
.iter() let mmap = occupied_entry.get();
.filter(|(_, mmap_weakref)| mmap_weakref.upgrade().is_none())
.map(|(key, _)| key.clone())
.collect();
for key in keys_to_remove {
self.cache.remove(&key);
}
}
// Returns None if the file exists but as a len of 0 (and hence is not mmappable).
fn get_mmap(&mut self, full_path: &Path) -> Result<Option<Arc<BoxedData>>, OpenReadError> {
if let Some(mmap_weak) = self.cache.get(full_path) {
if let Some(mmap_arc) = mmap_weak.upgrade() {
self.counters.hit += 1; self.counters.hit += 1;
return Ok(Some(mmap_arc)); Some(mmap.clone())
} }
} HashMapEntry::Vacant(vacant_entry) => {
self.cache.remove(full_path); self.counters.miss += 1;
self.counters.miss += 1; if let Some(mmap) = open_mmap(full_path)? {
Ok(if let Some(mmap) = open_mmap(full_path)? { vacant_entry.insert(mmap.clone());
let mmap_arc: Arc<BoxedData> = Arc::new(Box::new(mmap)); Some(mmap)
let mmap_weak = Arc::downgrade(&mmap_arc); } else {
self.cache.insert(full_path.to_owned(), mmap_weak); None
Some(mmap_arc)
} else {
None
})
}
}
struct InnerWatcherWrapper {
_watcher: Mutex<notify::RecommendedWatcher>,
watcher_router: WatchCallbackList,
}
impl InnerWatcherWrapper {
pub fn new(path: &Path) -> Result<(Self, Receiver<notify::RawEvent>), notify::Error> {
let (tx, watcher_recv): (Sender<RawEvent>, Receiver<RawEvent>) = channel();
// We need to initialize the
let mut watcher = notify::raw_watcher(tx)?;
watcher.watch(path, RecursiveMode::Recursive)?;
let inner = InnerWatcherWrapper {
_watcher: Mutex::new(watcher),
watcher_router: Default::default(),
};
Ok((inner, watcher_recv))
}
}
#[derive(Clone)]
pub(crate) struct WatcherWrapper {
inner: Arc<InnerWatcherWrapper>,
}
impl WatcherWrapper {
pub fn new(path: &Path) -> Result<Self, OpenDirectoryError> {
let (inner, watcher_recv) = InnerWatcherWrapper::new(path).map_err(|err| match err {
notify::Error::PathNotFound => OpenDirectoryError::DoesNotExist(path.to_owned()),
_ => {
panic!("Unknown error while starting watching directory {:?}", path);
}
})?;
let watcher_wrapper = WatcherWrapper {
inner: Arc::new(inner),
};
let watcher_wrapper_clone = watcher_wrapper.clone();
thread::Builder::new()
.name("meta-file-watch-thread".to_string())
.spawn(move || {
loop {
match watcher_recv.recv().map(|evt| evt.path) {
Ok(Some(changed_path)) => {
// ... Actually subject to false positive.
// We might want to be more accurate than this at one point.
if let Some(filename) = changed_path.file_name() {
if filename == *META_FILEPATH {
watcher_wrapper_clone.inner.watcher_router.broadcast();
}
}
}
Ok(None) => {
// not an event we are interested in.
}
Err(_e) => {
// the watch send channel was dropped
break;
}
}
} }
}) }
.expect("Failed to spawn thread to watch meta.json"); })
Ok(watcher_wrapper)
}
pub fn watch(&mut self, watch_callback: WatchCallback) -> WatchHandle {
self.inner.watcher_router.subscribe(watch_callback)
} }
} }
@@ -212,72 +115,33 @@ impl WatcherWrapper {
/// ///
/// The Mmap object are cached to limit the /// The Mmap object are cached to limit the
/// system calls. /// system calls.
///
/// In the `MmapDirectory`, locks are implemented using the `fs2` crate definition of locks.
///
/// On MacOS & linux, it relies on `flock` (aka `BSD Lock`). These locks solve most of the
/// problems related to POSIX Locks, but may their contract may not be respected on `NFS`
/// depending on the implementation.
///
/// On Windows the semantics are again different.
#[derive(Clone)] #[derive(Clone)]
pub struct MmapDirectory { pub struct MmapDirectory {
inner: Arc<MmapDirectoryInner>,
}
struct MmapDirectoryInner {
root_path: PathBuf, root_path: PathBuf,
mmap_cache: RwLock<MmapCache>, mmap_cache: Arc<RwLock<MmapCache>>,
_temp_directory: Option<TempDir>, _temp_directory: Arc<Option<TempDir>>,
watcher: RwLock<WatcherWrapper>,
}
impl MmapDirectoryInner {
fn new(
root_path: PathBuf,
temp_directory: Option<TempDir>,
) -> Result<MmapDirectoryInner, OpenDirectoryError> {
let watch_wrapper = WatcherWrapper::new(&root_path)?;
let mmap_directory_inner = MmapDirectoryInner {
root_path,
mmap_cache: Default::default(),
_temp_directory: temp_directory,
watcher: RwLock::new(watch_wrapper),
};
Ok(mmap_directory_inner)
}
fn watch(&self, watch_callback: WatchCallback) -> WatchHandle {
let mut wlock = self.watcher.write().unwrap();
wlock.watch(watch_callback)
}
} }
impl fmt::Debug for MmapDirectory { impl fmt::Debug for MmapDirectory {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "MmapDirectory({:?})", self.inner.root_path) write!(f, "MmapDirectory({:?})", self.root_path)
} }
} }
impl MmapDirectory { impl MmapDirectory {
fn new(
root_path: PathBuf,
temp_directory: Option<TempDir>,
) -> Result<MmapDirectory, OpenDirectoryError> {
let inner = MmapDirectoryInner::new(root_path, temp_directory)?;
Ok(MmapDirectory {
inner: Arc::new(inner),
})
}
/// Creates a new MmapDirectory in a temporary directory. /// Creates a new MmapDirectory in a temporary directory.
/// ///
/// This is mostly useful to test the MmapDirectory itself. /// This is mostly useful to test the MmapDirectory itself.
/// For your unit tests, prefer the RAMDirectory. /// For your unit tests, prefer the RAMDirectory.
pub fn create_from_tempdir() -> Result<MmapDirectory, OpenDirectoryError> { pub fn create_from_tempdir() -> io::Result<MmapDirectory> {
let tempdir = TempDir::new("index").map_err(OpenDirectoryError::IoError)?; let tempdir = TempDir::new("index")?;
let tempdir_path = PathBuf::from(tempdir.path()); let tempdir_path = PathBuf::from(tempdir.path());
MmapDirectory::new(tempdir_path, Some(tempdir)) let directory = MmapDirectory {
root_path: tempdir_path,
mmap_cache: Arc::new(RwLock::new(MmapCache::default())),
_temp_directory: Arc::new(Some(tempdir)),
};
Ok(directory)
} }
/// Opens a MmapDirectory in a directory. /// Opens a MmapDirectory in a directory.
@@ -295,14 +159,18 @@ impl MmapDirectory {
directory_path, directory_path,
))) )))
} else { } else {
Ok(MmapDirectory::new(PathBuf::from(directory_path), None)?) Ok(MmapDirectory {
root_path: PathBuf::from(directory_path),
mmap_cache: Arc::new(RwLock::new(MmapCache::default())),
_temp_directory: Arc::new(None),
})
} }
} }
/// Joins a relative_path to the directory `root_path` /// Joins a relative_path to the directory `root_path`
/// to create a proper complete `filepath`. /// to create a proper complete `filepath`.
fn resolve_path(&self, relative_path: &Path) -> PathBuf { fn resolve_path(&self, relative_path: &Path) -> PathBuf {
self.inner.root_path.join(relative_path) self.root_path.join(relative_path)
} }
/// Sync the root directory. /// Sync the root directory.
@@ -327,7 +195,7 @@ impl MmapDirectory {
.custom_flags(winbase::FILE_FLAG_BACKUP_SEMANTICS); .custom_flags(winbase::FILE_FLAG_BACKUP_SEMANTICS);
} }
let fd = open_opts.open(&self.inner.root_path)?; let fd = open_opts.open(&self.root_path)?;
fd.sync_all()?; fd.sync_all()?;
Ok(()) Ok(())
} }
@@ -337,38 +205,17 @@ impl MmapDirectory {
/// ///
/// The `MmapDirectory` embeds a `MmapDirectory` /// The `MmapDirectory` embeds a `MmapDirectory`
/// to avoid multiplying the `mmap` system calls. /// to avoid multiplying the `mmap` system calls.
pub fn get_cache_info(&self) -> CacheInfo { pub fn get_cache_info(&mut self) -> CacheInfo {
self.inner self.mmap_cache
.mmap_cache
.write() .write()
.expect("mmap cache lock is poisoned")
.remove_weak_ref();
self.inner
.mmap_cache
.read()
.expect("Mmap cache lock is poisoned.") .expect("Mmap cache lock is poisoned.")
.get_info() .get_info()
} }
} }
/// We rely on fs2 for file locking. On Windows & MacOS this
/// uses BSD locks (`flock`). The lock is actually released when
/// the `File` object is dropped and its associated file descriptor
/// is closed.
struct ReleaseLockFile {
_file: File,
path: PathBuf,
}
impl Drop for ReleaseLockFile {
fn drop(&mut self) {
debug!("Releasing lock {:?}", self.path);
}
}
/// This Write wraps a File, but has the specificity of /// This Write wraps a File, but has the specificity of
/// call `sync_all` on flush. /// call `sync_all` on flush.
pub struct SafeFileWriter(File); struct SafeFileWriter(File);
impl SafeFileWriter { impl SafeFileWriter {
fn new(file: File) -> SafeFileWriter { fn new(file: File) -> SafeFileWriter {
@@ -398,7 +245,7 @@ impl Directory for MmapDirectory {
debug!("Open Read {:?}", path); debug!("Open Read {:?}", path);
let full_path = self.resolve_path(path); let full_path = self.resolve_path(path);
let mut mmap_cache = self.inner.mmap_cache.write().map_err(|_| { let mut mmap_cache = self.mmap_cache.write().map_err(|_| {
let msg = format!( let msg = format!(
"Failed to acquired write lock \ "Failed to acquired write lock \
on mmap cache while reading {:?}", on mmap cache while reading {:?}",
@@ -406,34 +253,11 @@ impl Directory for MmapDirectory {
); );
IOError::with_path(path.to_owned(), make_io_err(msg)) IOError::with_path(path.to_owned(), make_io_err(msg))
})?; })?;
Ok(mmap_cache Ok(mmap_cache
.get_mmap(&full_path)? .get_mmap(&full_path)?
.map(ReadOnlySource::from) .map(ReadOnlySource::Mmap)
.unwrap_or_else(ReadOnlySource::empty)) .unwrap_or_else(|| ReadOnlySource::Anonymous(SharedVecSlice::empty())))
}
/// Any entry associated to the path in the mmap will be
/// removed before the file is deleted.
fn delete(&self, path: &Path) -> result::Result<(), DeleteError> {
debug!("Deleting file {:?}", path);
let full_path = self.resolve_path(path);
match fs::remove_file(&full_path) {
Ok(_) => self
.sync_directory()
.map_err(|e| IOError::with_path(path.to_owned(), e).into()),
Err(e) => {
if e.kind() == io::ErrorKind::NotFound {
Err(DeleteError::FileDoesNotExist(path.to_owned()))
} else {
Err(IOError::with_path(path.to_owned(), e).into())
}
}
}
}
fn exists(&self, path: &Path) -> bool {
let full_path = self.resolve_path(path);
full_path.exists()
} }
fn open_write(&mut self, path: &Path) -> Result<WritePtr, OpenWriteError> { fn open_write(&mut self, path: &Path) -> Result<WritePtr, OpenWriteError> {
@@ -466,6 +290,44 @@ impl Directory for MmapDirectory {
Ok(BufWriter::new(Box::new(writer))) Ok(BufWriter::new(Box::new(writer)))
} }
/// Any entry associated to the path in the mmap will be
/// removed before the file is deleted.
fn delete(&self, path: &Path) -> result::Result<(), DeleteError> {
debug!("Deleting file {:?}", path);
let full_path = self.resolve_path(path);
let mut mmap_cache = self.mmap_cache.write().map_err(|_| {
let msg = format!(
"Failed to acquired write lock \
on mmap cache while deleting {:?}",
path
);
IOError::with_path(path.to_owned(), make_io_err(msg))
})?;
mmap_cache.discard_from_cache(path);
// Removing the entry in the MMap cache.
// The munmap will appear on Drop,
// when the last reference is gone.
mmap_cache.cache.remove(&full_path);
match fs::remove_file(&full_path) {
Ok(_) => self
.sync_directory()
.map_err(|e| IOError::with_path(path.to_owned(), e).into()),
Err(e) => {
if e.kind() == io::ErrorKind::NotFound {
Err(DeleteError::FileDoesNotExist(path.to_owned()))
} else {
Err(IOError::with_path(path.to_owned(), e).into())
}
}
}
}
fn exists(&self, path: &Path) -> bool {
let full_path = self.resolve_path(path);
full_path.exists()
}
fn atomic_read(&self, path: &Path) -> Result<Vec<u8>, OpenReadError> { fn atomic_read(&self, path: &Path) -> Result<Vec<u8>, OpenReadError> {
let full_path = self.resolve_path(path); let full_path = self.resolve_path(path);
let mut buffer = Vec::new(); let mut buffer = Vec::new();
@@ -492,30 +354,6 @@ impl Directory for MmapDirectory {
meta_file.write(|f| f.write_all(data))?; meta_file.write(|f| f.write_all(data))?;
Ok(()) Ok(())
} }
fn acquire_lock(&self, lock: &Lock) -> Result<DirectoryLock, LockError> {
let full_path = self.resolve_path(&lock.filepath);
// We make sure that the file exists.
let file: File = OpenOptions::new()
.write(true)
.create(true) //< if the file does not exist yet, create it.
.open(&full_path)
.map_err(LockError::IOError)?;
if lock.is_blocking {
file.lock_exclusive().map_err(LockError::IOError)?;
} else {
file.try_lock_exclusive().map_err(|_| LockError::LockBusy)?
}
// dropping the file handle will release the lock.
Ok(DirectoryLock::from(Box::new(ReleaseLockFile {
path: lock.filepath.clone(),
_file: file,
})))
}
fn watch(&self, watch_callback: WatchCallback) -> WatchHandle {
self.inner.watch(watch_callback)
}
} }
#[cfg(test)] #[cfg(test)]
@@ -525,18 +363,6 @@ mod tests {
// The following tests are specific to the MmapDirectory // The following tests are specific to the MmapDirectory
use super::*; use super::*;
use schema::{Schema, SchemaBuilder, TEXT};
use std::fs;
use std::sync::atomic::{AtomicUsize, Ordering};
use std::thread;
use std::time::Duration;
use Index;
use ReloadPolicy;
#[test]
fn test_open_non_existant_path() {
assert!(MmapDirectory::open(PathBuf::from("./nowhere")).is_err());
}
#[test] #[test]
fn test_open_empty() { fn test_open_empty() {
@@ -556,7 +382,7 @@ mod tests {
#[test] #[test]
fn test_cache() { fn test_cache() {
let content = b"abc"; let content = "abc".as_bytes();
// here we test if the cache releases // here we test if the cache releases
// mmaps correctly. // mmaps correctly.
@@ -572,104 +398,26 @@ mod tests {
w.flush().unwrap(); w.flush().unwrap();
} }
} }
{
let mut keep = vec![]; for (i, path) in paths.iter().enumerate() {
for (i, path) in paths.iter().enumerate() { let _r = mmap_directory.open_read(path).unwrap();
keep.push(mmap_directory.open_read(path).unwrap()); assert_eq!(mmap_directory.get_cache_info().mmapped.len(), i + 1);
assert_eq!(mmap_directory.get_cache_info().mmapped.len(), i + 1); }
} for path in paths.iter() {
assert_eq!(mmap_directory.get_cache_info().counters.hit, 0); let _r = mmap_directory.open_read(path).unwrap();
assert_eq!(mmap_directory.get_cache_info().counters.miss, 10); assert_eq!(mmap_directory.get_cache_info().mmapped.len(), num_paths);
assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 10); }
for path in paths.iter() { for (i, path) in paths.iter().enumerate() {
let _r = mmap_directory.open_read(path).unwrap(); mmap_directory.delete(path).unwrap();
assert_eq!(mmap_directory.get_cache_info().mmapped.len(), num_paths); assert_eq!(
mmap_directory.get_cache_info().mmapped.len(),
num_paths - i - 1
);
}
} }
assert_eq!(mmap_directory.get_cache_info().counters.hit, 10); assert_eq!(mmap_directory.get_cache_info().counters.hit, 10);
assert_eq!(mmap_directory.get_cache_info().counters.miss, 10); assert_eq!(mmap_directory.get_cache_info().counters.miss, 10);
assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 10);
for path in paths.iter() {
let _r = mmap_directory.open_read(path).unwrap();
assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 10);
}
assert_eq!(mmap_directory.get_cache_info().counters.hit, 20);
assert_eq!(mmap_directory.get_cache_info().counters.miss, 10);
assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 10);
drop(keep);
for path in paths.iter() {
let _r = mmap_directory.open_read(path).unwrap();
assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 1);
}
assert_eq!(mmap_directory.get_cache_info().counters.hit, 20);
assert_eq!(mmap_directory.get_cache_info().counters.miss, 20);
assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 0);
for path in &paths {
mmap_directory.delete(path).unwrap();
}
assert_eq!(mmap_directory.get_cache_info().counters.hit, 20);
assert_eq!(mmap_directory.get_cache_info().counters.miss, 20);
assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 0);
for path in paths.iter() {
assert!(mmap_directory.open_read(path).is_err());
}
assert_eq!(mmap_directory.get_cache_info().counters.hit, 20);
assert_eq!(mmap_directory.get_cache_info().counters.miss, 30);
assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 0); assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 0);
} }
#[test]
fn test_watch_wrapper() {
let counter: Arc<AtomicUsize> = Default::default();
let counter_clone = counter.clone();
let tmp_dir: TempDir = tempdir::TempDir::new("test_watch_wrapper").unwrap();
let tmp_dirpath = tmp_dir.path().to_owned();
let mut watch_wrapper = WatcherWrapper::new(&tmp_dirpath).unwrap();
let tmp_file = tmp_dirpath.join("coucou");
let _handle = watch_wrapper.watch(Box::new(move || {
counter_clone.fetch_add(1, Ordering::SeqCst);
}));
assert_eq!(counter.load(Ordering::SeqCst), 0);
fs::write(&tmp_file, b"whateverwilldo").unwrap();
thread::sleep(Duration::new(0, 1_000u32));
}
#[test]
fn test_mmap_released() {
let mmap_directory = MmapDirectory::create_from_tempdir().unwrap();
let mut schema_builder: SchemaBuilder = Schema::builder();
let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build();
{
let index = Index::create(mmap_directory.clone(), schema).unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
for _num_commits in 0..16 {
for _ in 0..10 {
index_writer.add_document(doc!(text_field=>"abc"));
}
index_writer.commit().unwrap();
}
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::Manual)
.try_into()
.unwrap();
for _ in 0..30 {
index_writer.add_document(doc!(text_field=>"abc"));
index_writer.commit().unwrap();
reader.reload().unwrap();
}
index_writer.wait_merging_threads().unwrap();
reader.reload().unwrap();
let num_segments = reader.searcher().segment_readers().len();
assert_eq!(num_segments, 4);
assert_eq!(
num_segments * 7,
mmap_directory.get_cache_info().mmapped.len()
);
}
assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 0);
}
} }

139
src/directory/mod.rs
View File

@@ -8,24 +8,21 @@ WORM directory abstraction.
mod mmap_directory; mod mmap_directory;
mod directory; mod directory;
mod directory_lock;
mod managed_directory; mod managed_directory;
mod ram_directory; mod ram_directory;
mod read_only_source; mod read_only_source;
mod watch_event_router; mod shared_vec_slice;
mod nrt_directory; mod static_dictionnary;
/// Errors specific to the directory module. /// Errors specific to the directory module.
pub mod error; pub mod error;
pub use self::directory::DirectoryLock; use std::io::{BufWriter, Seek, Write};
pub use self::directory::{Directory, DirectoryClone}; pub use self::directory::{Directory, DirectoryClone};
pub use self::directory_lock::{Lock, INDEX_WRITER_LOCK, META_LOCK};
pub use self::ram_directory::RAMDirectory; pub use self::ram_directory::RAMDirectory;
pub use self::read_only_source::ReadOnlySource; pub use self::read_only_source::ReadOnlySource;
pub(crate) use self::watch_event_router::WatchCallbackList; pub use self::static_dictionnary::StaticDirectory;
pub use self::watch_event_router::{WatchCallback, WatchHandle};
use std::io::{BufWriter, Seek, Write};
#[cfg(feature = "mmap")] #[cfg(feature = "mmap")]
pub use self::mmap_directory::MmapDirectory; pub use self::mmap_directory::MmapDirectory;
@@ -43,4 +40,128 @@ impl<T: Seek + Write> SeekableWrite for T {}
pub type WritePtr = BufWriter<Box<SeekableWrite>>; pub type WritePtr = BufWriter<Box<SeekableWrite>>;
#[cfg(test)] #[cfg(test)]
mod tests; mod tests {
use super::*;
use std::io::{Seek, SeekFrom, Write};
use std::path::Path;
lazy_static! {
static ref TEST_PATH: &'static Path = Path::new("some_path_for_test");
}
#[test]
fn test_ram_directory() {
let mut ram_directory = RAMDirectory::create();
test_directory(&mut ram_directory);
}
#[test]
#[cfg(feature = "mmap")]
fn test_mmap_directory() {
let mut mmap_directory = MmapDirectory::create_from_tempdir().unwrap();
test_directory(&mut mmap_directory);
}
#[test]
#[should_panic]
fn ram_directory_panics_if_flush_forgotten() {
let mut ram_directory = RAMDirectory::create();
let mut write_file = ram_directory.open_write(*TEST_PATH).unwrap();
assert!(write_file.write_all(&[4]).is_ok());
}
fn test_simple(directory: &mut Directory) {
{
{
let mut write_file = directory.open_write(*TEST_PATH).unwrap();
assert!(directory.exists(*TEST_PATH));
write_file.write_all(&[4]).unwrap();
write_file.write_all(&[3]).unwrap();
write_file.write_all(&[7, 3, 5]).unwrap();
write_file.flush().unwrap();
}
let read_file = directory.open_read(*TEST_PATH).unwrap();
let data: &[u8] = &*read_file;
assert_eq!(data, &[4u8, 3u8, 7u8, 3u8, 5u8]);
}
assert!(directory.delete(*TEST_PATH).is_ok());
assert!(!directory.exists(*TEST_PATH));
}
fn test_seek(directory: &mut Directory) {
{
{
let mut write_file = directory.open_write(*TEST_PATH).unwrap();
write_file.write_all(&[4, 3, 7, 3, 5]).unwrap();
write_file.seek(SeekFrom::Start(0)).unwrap();
write_file.write_all(&[3, 1]).unwrap();
write_file.flush().unwrap();
}
let read_file = directory.open_read(*TEST_PATH).unwrap();
let data: &[u8] = &*read_file;
assert_eq!(data, &[3u8, 1u8, 7u8, 3u8, 5u8]);
}
assert!(directory.delete(*TEST_PATH).is_ok());
}
fn test_rewrite_forbidden(directory: &mut Directory) {
{
directory.open_write(*TEST_PATH).unwrap();
assert!(directory.exists(*TEST_PATH));
}
{
assert!(directory.open_write(*TEST_PATH).is_err());
}
assert!(directory.delete(*TEST_PATH).is_ok());
}
fn test_write_create_the_file(directory: &mut Directory) {
{
assert!(directory.open_read(*TEST_PATH).is_err());
let _w = directory.open_write(*TEST_PATH).unwrap();
assert!(directory.exists(*TEST_PATH));
assert!(directory.open_read(*TEST_PATH).is_ok());
assert!(directory.delete(*TEST_PATH).is_ok());
}
}
fn test_directory_delete(directory: &mut Directory) {
assert!(directory.open_read(*TEST_PATH).is_err());
let mut write_file = directory.open_write(*TEST_PATH).unwrap();
write_file.write_all(&[1, 2, 3, 4]).unwrap();
write_file.flush().unwrap();
{
let read_handle = directory.open_read(*TEST_PATH).unwrap();
{
assert_eq!(&*read_handle, &[1u8, 2u8, 3u8, 4u8]);
// Mapped files can't be deleted on Windows
if !cfg!(windows) {
assert!(directory.delete(*TEST_PATH).is_ok());
assert_eq!(&*read_handle, &[1u8, 2u8, 3u8, 4u8]);
}
assert!(directory.delete(Path::new("SomeOtherPath")).is_err());
}
}
if cfg!(windows) {
assert!(directory.delete(*TEST_PATH).is_ok());
}
assert!(directory.open_read(*TEST_PATH).is_err());
assert!(directory.delete(*TEST_PATH).is_err());
}
fn test_directory(directory: &mut Directory) {
test_simple(directory);
test_seek(directory);
test_rewrite_forbidden(directory);
test_write_create_the_file(directory);
test_directory_delete(directory);
}
}

195
src/directory/nrt_directory.rs
View File

@@ -1,195 +0,0 @@
use directory::Directory;
use std::path::{PathBuf, Path};
use directory::ReadOnlySource;
use directory::error::OpenReadError;
use directory::error::DeleteError;
use std::io::{BufWriter, Cursor};
use directory::SeekableWrite;
use directory::error::OpenWriteError;
use directory::WatchHandle;
use directory::ram_directory::InnerRamDirectory;
use std::sync::RwLock;
use std::sync::Arc;
use directory::WatchCallback;
use std::fmt;
use std::io;
use std::io::{Seek, Write};
use directory::DirectoryClone;
const BUFFER_LEN: usize = 1_000_000;
pub enum NRTWriter {
InRam {
buffer: Cursor<Vec<u8>>,
path: PathBuf,
nrt_directory: NRTDirectory
},
UnderlyingFile(BufWriter<Box<SeekableWrite>>)
}
impl NRTWriter {
pub fn new(path: PathBuf, nrt_directory: NRTDirectory) -> NRTWriter {
NRTWriter::InRam {
buffer: Cursor::new(Vec::with_capacity(BUFFER_LEN)),
path,
nrt_directory,
}
}
}
impl io::Seek for NRTWriter {
fn seek(&mut self, pos: io::SeekFrom) -> io::Result<u64> {
match self {
NRTWriter::InRam { buffer, path, nrt_directory } => {
buffer.seek(pos)
}
NRTWriter::UnderlyingFile(file) => {
file.seek(pos)
}
}
}
}
impl io::Write for NRTWriter {
fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
self.write_all(buf)?;
Ok(buf.len())
}
fn flush(&mut self) -> io::Result<()> {
match self {
NRTWriter::InRam { buffer, path, nrt_directory } => {
let mut cache_wlock = nrt_directory.cache.write().unwrap();
cache_wlock.write(path.clone(), buffer.get_ref());
Ok(())
}
NRTWriter::UnderlyingFile(file) => {
file.flush()
}
}
}
fn write_all(&mut self, buf: &[u8]) -> io::Result<()> {
// Working around the borrow checker.
let mut underlying_write_opt: Option<BufWriter<Box<SeekableWrite>>> = None;
if let NRTWriter::InRam { buffer, path, nrt_directory } = self {
if buffer.get_ref().len() + buf.len() > BUFFER_LEN {
// We can't keep this in RAM. Let's move it to the underlying directory.
underlying_write_opt = Some(nrt_directory.open_write(path)
.map_err(|open_err| {
io::Error::new(io::ErrorKind::Other, open_err)
})?);
}
}
if let Some(underlying_write) = underlying_write_opt {
*self = NRTWriter::UnderlyingFile(underlying_write);
}
match self {
NRTWriter::InRam { buffer, path, nrt_directory } => {
assert!(buffer.get_ref().len() + buf.len() <= BUFFER_LEN);
buffer.write_all(buf)
}
NRTWriter::UnderlyingFile(file) => {
file.write_all(buf)
}
}
}
}
pub struct NRTDirectory {
underlying: Box<Directory>,
cache: Arc<RwLock<InnerRamDirectory>>,
}
impl Clone for NRTDirectory {
fn clone(&self) -> Self {
NRTDirectory {
underlying: self.underlying.box_clone(),
cache: self.cache.clone()
}
}
}
impl NRTDirectory {
fn wrap(underlying: Box<Directory>) -> NRTDirectory {
NRTDirectory {
underlying,
cache: Default::default()
}
}
}
impl fmt::Debug for NRTDirectory {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "NRTDirectory({:?})", self.underlying)
}
}
impl Directory for NRTDirectory {
fn open_read(&self, path: &Path) -> Result<ReadOnlySource, OpenReadError> {
unimplemented!()
}
fn delete(&self, path: &Path) -> Result<(), DeleteError> {
// We explicitly release the lock, to prevent a panic on the underlying directory
// to poison the lock.
//
// File can only go from cache to underlying so the result does not lead to
// any inconsistency.
{
let mut cache_wlock = self.cache.write().unwrap();
if cache_wlock.exists(path) {
return cache_wlock.delete(path);
}
}
self.underlying.delete(path)
}
fn exists(&self, path: &Path) -> bool {
// We explicitly release the lock, to prevent a panic on the underlying directory
// to poison the lock.
//
// File can only go from cache to underlying so the result does not lead to
// any inconsistency.
{
let rlock_cache = self.cache.read().unwrap();
if rlock_cache.exists(path) {
return true;
}
}
self.underlying.exists(path)
}
fn open_write(&mut self, path: &Path) -> Result<BufWriter<Box<SeekableWrite>>, OpenWriteError> {
let mut cache_wlock = self.cache.write().unwrap();
// TODO might poison our lock. I don't know have a sound solution yet.
let path_buf = path.to_owned();
if self.underlying.exists(path) {
return Err(OpenWriteError::FileAlreadyExists(path_buf));
}
let exists = cache_wlock.write(path_buf.clone(), &[]);
// force the creation of the file to mimic the MMap directory.
if exists {
Err(OpenWriteError::FileAlreadyExists(path_buf))
} else {
let vec_writer = NRTWriter::new(path_buf.clone(), self.clone());
Ok(BufWriter::new(Box::new(vec_writer)))
}
}
fn atomic_read(&self, path: &Path) -> Result<Vec<u8>, OpenReadError> {
self.underlying.atomic_read(path)
}
fn atomic_write(&mut self, path: &Path, data: &[u8]) -> io::Result<()> {
self.underlying.atomic_write(path, data)
}
fn watch(&self, watch_callback: WatchCallback) -> WatchHandle {
self.underlying.watch(watch_callback)
}
}

143
src/directory/ram_directory.rs
View File

@@ -1,8 +1,8 @@
use core::META_FILEPATH; use super::shared_vec_slice::SharedVecSlice;
use directory::error::{DeleteError, OpenReadError, OpenWriteError}; use common::make_io_err;
use directory::WatchCallbackList; use directory::error::{DeleteError, IOError, OpenReadError, OpenWriteError};
use directory::WritePtr; use directory::WritePtr;
use directory::{Directory, ReadOnlySource, WatchCallback, WatchHandle}; use directory::{Directory, ReadOnlySource};
use std::collections::HashMap; use std::collections::HashMap;
use std::fmt; use std::fmt;
use std::io::{self, BufWriter, Cursor, Seek, SeekFrom, Write}; use std::io::{self, BufWriter, Cursor, Seek, SeekFrom, Write};
@@ -22,13 +22,13 @@ use std::sync::{Arc, RwLock};
/// ///
struct VecWriter { struct VecWriter {
path: PathBuf, path: PathBuf,
shared_directory: RAMDirectory, shared_directory: InnerDirectory,
data: Cursor<Vec<u8>>, data: Cursor<Vec<u8>>,
is_flushed: bool, is_flushed: bool,
} }
impl VecWriter { impl VecWriter {
fn new(path_buf: PathBuf, shared_directory: RAMDirectory) -> VecWriter { fn new(path_buf: PathBuf, shared_directory: InnerDirectory) -> VecWriter {
VecWriter { VecWriter {
path: path_buf, path: path_buf,
data: Cursor::new(Vec::new()), data: Cursor::new(Vec::new()),
@@ -64,44 +64,73 @@ impl Write for VecWriter {
fn flush(&mut self) -> io::Result<()> { fn flush(&mut self) -> io::Result<()> {
self.is_flushed = true; self.is_flushed = true;
let mut fs = self.shared_directory.fs.write().unwrap(); self.shared_directory
fs.write(self.path.clone(), self.data.get_ref()); .write(self.path.clone(), self.data.get_ref())?;
Ok(()) Ok(())
} }
} }
#[derive(Default)] #[derive(Clone)]
pub(crate) struct InnerRamDirectory { struct InnerDirectory(Arc<RwLock<HashMap<PathBuf, Arc<Vec<u8>>>>>);
fs: HashMap<PathBuf, ReadOnlySource>,
watch_router: WatchCallbackList,
}
impl InnerRamDirectory { impl InnerDirectory {
pub fn write(&mut self, path: PathBuf, data: &[u8]) -> bool { fn new() -> InnerDirectory {
let data = ReadOnlySource::new(Vec::from(data)); InnerDirectory(Arc::new(RwLock::new(HashMap::new())))
self.fs.insert(path, data).is_some()
} }
pub fn open_read(&self, path: &Path) -> Result<ReadOnlySource, OpenReadError> { fn write(&self, path: PathBuf, data: &[u8]) -> io::Result<bool> {
self.fs let mut map = self.0.write().map_err(|_| {
.get(path) make_io_err(format!(
.ok_or_else(|| OpenReadError::FileDoesNotExist(PathBuf::from(path))) "Failed to lock the directory, when trying to write {:?}",
.map(|el| el.clone()) path
))
})?;
let prev_value = map.insert(path, Arc::new(Vec::from(data)));
Ok(prev_value.is_some())
} }
pub fn delete(&mut self, path: &Path) -> result::Result<(), DeleteError> { fn open_read(&self, path: &Path) -> Result<ReadOnlySource, OpenReadError> {
match self.fs.remove(path) { self.0
Some(_) => Ok(()), .read()
None => Err(DeleteError::FileDoesNotExist(PathBuf::from(path))), .map_err(|_| {
} let msg = format!(
"Failed to acquire read lock for the \
directory when trying to read {:?}",
path
);
let io_err = make_io_err(msg);
OpenReadError::IOError(IOError::with_path(path.to_owned(), io_err))
}).and_then(|readable_map| {
readable_map
.get(path)
.ok_or_else(|| OpenReadError::FileDoesNotExist(PathBuf::from(path)))
.map(Arc::clone)
.map(|data| ReadOnlySource::Anonymous(SharedVecSlice::new(data)))
})
} }
pub fn exists(&self, path: &Path) -> bool { fn delete(&self, path: &Path) -> result::Result<(), DeleteError> {
self.fs.contains_key(path) self.0
.write()
.map_err(|_| {
let msg = format!(
"Failed to acquire write lock for the \
directory when trying to delete {:?}",
path
);
let io_err = make_io_err(msg);
DeleteError::IOError(IOError::with_path(path.to_owned(), io_err))
}).and_then(|mut writable_map| match writable_map.remove(path) {
Some(_) => Ok(()),
None => Err(DeleteError::FileDoesNotExist(PathBuf::from(path))),
})
} }
pub fn watch(&mut self, watch_handle: WatchCallback) -> WatchHandle { fn exists(&self, path: &Path) -> bool {
self.watch_router.subscribe(watch_handle) self.0
.read()
.expect("Failed to get read lock directory.")
.contains_key(path)
} }
} }
@@ -116,36 +145,33 @@ impl fmt::Debug for RAMDirectory {
/// It is mainly meant for unit testing. /// It is mainly meant for unit testing.
/// Writes are only made visible upon flushing. /// Writes are only made visible upon flushing.
/// ///
#[derive(Clone, Default)] #[derive(Clone)]
pub struct RAMDirectory { pub struct RAMDirectory {
fs: Arc<RwLock<InnerRamDirectory>>, fs: InnerDirectory,
} }
impl RAMDirectory { impl RAMDirectory {
/// Constructor /// Constructor
pub fn create() -> RAMDirectory { pub fn create() -> RAMDirectory {
Self::default() RAMDirectory {
fs: InnerDirectory::new(),
}
} }
} }
impl Directory for RAMDirectory { impl Directory for RAMDirectory {
fn open_read(&self, path: &Path) -> result::Result<ReadOnlySource, OpenReadError> { fn open_read(&self, path: &Path) -> result::Result<ReadOnlySource, OpenReadError> {
self.fs.read().unwrap().open_read(path) self.fs.open_read(path)
}
fn delete(&self, path: &Path) -> result::Result<(), DeleteError> {
self.fs.write().unwrap().delete(path)
}
fn exists(&self, path: &Path) -> bool {
self.fs.read().unwrap().exists(path)
} }
fn open_write(&mut self, path: &Path) -> Result<WritePtr, OpenWriteError> { fn open_write(&mut self, path: &Path) -> Result<WritePtr, OpenWriteError> {
let mut fs = self.fs.write().unwrap();
let path_buf = PathBuf::from(path); let path_buf = PathBuf::from(path);
let vec_writer = VecWriter::new(path_buf.clone(), self.clone()); let vec_writer = VecWriter::new(path_buf.clone(), self.fs.clone());
let exists = fs.write(path_buf.clone(), &[]);
let exists = self
.fs
.write(path_buf.clone(), &Vec::new())
.map_err(|err| IOError::with_path(path.to_owned(), err))?;
// force the creation of the file to mimic the MMap directory. // force the creation of the file to mimic the MMap directory.
if exists { if exists {
Err(OpenWriteError::FileAlreadyExists(path_buf)) Err(OpenWriteError::FileAlreadyExists(path_buf))
@@ -154,8 +180,17 @@ impl Directory for RAMDirectory {
} }
} }
fn delete(&self, path: &Path) -> result::Result<(), DeleteError> {
self.fs.delete(path)
}
fn exists(&self, path: &Path) -> bool {
self.fs.exists(path)
}
fn atomic_read(&self, path: &Path) -> Result<Vec<u8>, OpenReadError> { fn atomic_read(&self, path: &Path) -> Result<Vec<u8>, OpenReadError> {
Ok(self.open_read(path)?.as_slice().to_owned()) let read = self.open_read(path)?;
Ok(read.as_slice().to_owned())
} }
fn atomic_write(&mut self, path: &Path, data: &[u8]) -> io::Result<()> { fn atomic_write(&mut self, path: &Path, data: &[u8]) -> io::Result<()> {
@@ -164,20 +199,10 @@ impl Directory for RAMDirectory {
msg.unwrap_or("Undefined".to_string()) msg.unwrap_or("Undefined".to_string())
))); )));
let path_buf = PathBuf::from(path); let path_buf = PathBuf::from(path);
let mut vec_writer = VecWriter::new(path_buf.clone(), self.fs.clone());
// Reserve the path to prevent calls to .write() to succeed. self.fs.write(path_buf, &Vec::new())?;
self.fs.write().unwrap().write(path_buf.clone(), &[]);
let mut vec_writer = VecWriter::new(path_buf.clone(), self.clone());
vec_writer.write_all(data)?; vec_writer.write_all(data)?;
vec_writer.flush()?; vec_writer.flush()?;
if path == Path::new(&*META_FILEPATH) {
self.fs.write().unwrap().watch_router.broadcast();
}
Ok(()) Ok(())
} }
fn watch(&self, watch_callback: WatchCallback) -> WatchHandle {
self.fs.write().unwrap().watch(watch_callback)
}
} }

92
src/directory/read_only_source.rs
View File

@@ -1,9 +1,12 @@
use super::shared_vec_slice::SharedVecSlice;
use common::HasLen; use common::HasLen;
#[cfg(feature = "mmap")]
use fst::raw::MmapReadOnly;
use stable_deref_trait::{CloneStableDeref, StableDeref}; use stable_deref_trait::{CloneStableDeref, StableDeref};
use std::ops::Deref; use std::ops::Deref;
use std::sync::Arc;
pub type BoxedData = Box<Deref<Target = [u8]> + Send + Sync + 'static>;
const EMPTY_SLICE: [u8; 0] = [];
/// Read object that represents files in tantivy. /// Read object that represents files in tantivy.
/// ///
@@ -11,10 +14,14 @@ pub type BoxedData = Box<Deref<Target = [u8]> + Send + Sync + 'static>;
/// the data in the form of a constant read-only `&[u8]`. /// the data in the form of a constant read-only `&[u8]`.
/// Whatever happens to the directory file, the data /// Whatever happens to the directory file, the data
/// hold by this object should never be altered or destroyed. /// hold by this object should never be altered or destroyed.
pub struct ReadOnlySource { pub enum ReadOnlySource {
data: Arc<BoxedData>, /// Mmap source of data
start: usize, #[cfg(feature = "mmap")]
stop: usize, Mmap(MmapReadOnly),
/// Wrapping a `Vec<u8>`
Anonymous(SharedVecSlice),
/// Wrapping a static slice
Static(&'static [u8])
} }
unsafe impl StableDeref for ReadOnlySource {} unsafe impl StableDeref for ReadOnlySource {}
@@ -28,38 +35,20 @@ impl Deref for ReadOnlySource {
} }
} }
impl From<Arc<BoxedData>> for ReadOnlySource {
fn from(data: Arc<BoxedData>) -> Self {
let len = data.len();
ReadOnlySource {
data,
start: 0,
stop: len,
}
}
}
impl ReadOnlySource { impl ReadOnlySource {
pub(crate) fn new<D>(data: D) -> ReadOnlySource
where
D: Deref<Target = [u8]> + Send + Sync + 'static,
{
let len = data.len();
ReadOnlySource {
data: Arc::new(Box::new(data)),
start: 0,
stop: len,
}
}
/// Creates an empty ReadOnlySource /// Creates an empty ReadOnlySource
pub fn empty() -> ReadOnlySource { pub fn empty() -> ReadOnlySource {
ReadOnlySource::new(&[][..]) ReadOnlySource::Static(&EMPTY_SLICE)
} }
/// Returns the data underlying the ReadOnlySource object. /// Returns the data underlying the ReadOnlySource object.
pub fn as_slice(&self) -> &[u8] { pub fn as_slice(&self) -> &[u8] {
&self.data[self.start..self.stop] match *self {
#[cfg(feature = "mmap")]
ReadOnlySource::Mmap(ref mmap_read_only) => mmap_read_only.as_slice(),
ReadOnlySource::Anonymous(ref shared_vec) => shared_vec.as_slice(),
ReadOnlySource::Static(data) => data,
}
} }
/// Splits into 2 `ReadOnlySource`, at the offset given /// Splits into 2 `ReadOnlySource`, at the offset given
@@ -80,18 +69,25 @@ impl ReadOnlySource {
/// worth of data in anonymous memory, and only a /// worth of data in anonymous memory, and only a
/// 1KB slice is remaining, the whole `500MBs` /// 1KB slice is remaining, the whole `500MBs`
/// are retained in memory. /// are retained in memory.
pub fn slice(&self, start: usize, stop: usize) -> ReadOnlySource { pub fn slice(&self, from_offset: usize, to_offset: usize) -> ReadOnlySource {
assert!( assert!(
start <= stop, from_offset <= to_offset,
"Requested negative slice [{}..{}]", "Requested negative slice [{}..{}]",
start, from_offset,
stop to_offset
); );
assert!(stop <= self.len()); match *self {
ReadOnlySource { #[cfg(feature = "mmap")]
data: self.data.clone(), ReadOnlySource::Mmap(ref mmap_read_only) => {
start: self.start + start, let sliced_mmap = mmap_read_only.range(from_offset, to_offset - from_offset);
stop: self.start + stop, ReadOnlySource::Mmap(sliced_mmap)
}
ReadOnlySource::Anonymous(ref shared_vec) => {
ReadOnlySource::Anonymous(shared_vec.slice(from_offset, to_offset))
}
ReadOnlySource::Static(data) => {
ReadOnlySource::Static(&data[from_offset..to_offset])
}
} }
} }
@@ -100,7 +96,8 @@ impl ReadOnlySource {
/// ///
/// Equivalent to `.slice(from_offset, self.len())` /// Equivalent to `.slice(from_offset, self.len())`
pub fn slice_from(&self, from_offset: usize) -> ReadOnlySource { pub fn slice_from(&self, from_offset: usize) -> ReadOnlySource {
self.slice(from_offset, self.len()) let len = self.len();
self.slice(from_offset, len)
} }
/// Like `.slice(...)` but enforcing only the `to` /// Like `.slice(...)` but enforcing only the `to`
@@ -114,18 +111,25 @@ impl ReadOnlySource {
impl HasLen for ReadOnlySource { impl HasLen for ReadOnlySource {
fn len(&self) -> usize { fn len(&self) -> usize {
self.stop - self.start self.as_slice().len()
} }
} }
impl Clone for ReadOnlySource { impl Clone for ReadOnlySource {
fn clone(&self) -> Self { fn clone(&self) -> Self {
self.slice_from(0) self.slice(0, self.len())
} }
} }
impl From<Vec<u8>> for ReadOnlySource { impl From<Vec<u8>> for ReadOnlySource {
fn from(data: Vec<u8>) -> ReadOnlySource { fn from(data: Vec<u8>) -> ReadOnlySource {
ReadOnlySource::new(data) let shared_data = SharedVecSlice::from(data);
ReadOnlySource::Anonymous(shared_data)
}
}
impl From<&'static [u8]> for ReadOnlySource {
fn from(data: &'static [u8]) -> ReadOnlySource {
ReadOnlySource::Static(data)
} }
} }

41
src/directory/shared_vec_slice.rs Normal file
View File

@@ -0,0 +1,41 @@
use std::sync::Arc;
#[derive(Clone)]
pub struct SharedVecSlice {
pub data: Arc<Vec<u8>>,
pub start: usize,
pub len: usize,
}
impl SharedVecSlice {
pub fn empty() -> SharedVecSlice {
SharedVecSlice::new(Arc::new(Vec::new()))
}
pub fn new(data: Arc<Vec<u8>>) -> SharedVecSlice {
let data_len = data.len();
SharedVecSlice {
data,
start: 0,
len: data_len,
}
}
pub fn as_slice(&self) -> &[u8] {
&self.data[self.start..self.start + self.len]
}
pub fn slice(&self, from_offset: usize, to_offset: usize) -> SharedVecSlice {
SharedVecSlice {
data: Arc::clone(&self.data),
start: self.start + from_offset,
len: to_offset - from_offset,
}
}
}
impl From<Vec<u8>> for SharedVecSlice {
fn from(data: Vec<u8>) -> SharedVecSlice {
SharedVecSlice::new(Arc::new(data))
}
}

222
src/directory/tests.rs
View File

@@ -1,222 +0,0 @@
use super::*;
use std::io::{Seek, SeekFrom, Write};
use std::mem;
use std::path::{Path, PathBuf};
use std::sync::atomic::AtomicUsize;
use std::sync::atomic::Ordering;
use std::sync::Arc;
use std::thread;
use std::time;
use std::time::Duration;
lazy_static! {
static ref TEST_PATH: &'static Path = Path::new("some_path_for_test");
}
#[test]
fn test_ram_directory() {
let mut ram_directory = RAMDirectory::create();
test_directory(&mut ram_directory);
}
#[test]
#[cfg(feature = "mmap")]
fn test_mmap_directory() {
let mut mmap_directory = MmapDirectory::create_from_tempdir().unwrap();
test_directory(&mut mmap_directory);
}
#[test]
#[should_panic]
fn ram_directory_panics_if_flush_forgotten() {
let mut ram_directory = RAMDirectory::create();
let mut write_file = ram_directory.open_write(*TEST_PATH).unwrap();
assert!(write_file.write_all(&[4]).is_ok());
}
fn test_simple(directory: &mut Directory) {
{
let mut write_file = directory.open_write(*TEST_PATH).unwrap();
assert!(directory.exists(*TEST_PATH));
write_file.write_all(&[4]).unwrap();
write_file.write_all(&[3]).unwrap();
write_file.write_all(&[7, 3, 5]).unwrap();
write_file.flush().unwrap();
}
{
let read_file = directory.open_read(*TEST_PATH).unwrap();
let data: &[u8] = &*read_file;
assert_eq!(data, &[4u8, 3u8, 7u8, 3u8, 5u8]);
}
assert!(directory.delete(*TEST_PATH).is_ok());
assert!(!directory.exists(*TEST_PATH));
}
fn test_seek(directory: &mut Directory) {
{
{
let mut write_file = directory.open_write(*TEST_PATH).unwrap();
write_file.write_all(&[4, 3, 7, 3, 5]).unwrap();
write_file.seek(SeekFrom::Start(0)).unwrap();
write_file.write_all(&[3, 1]).unwrap();
write_file.flush().unwrap();
}
let read_file = directory.open_read(*TEST_PATH).unwrap();
let data: &[u8] = &*read_file;
assert_eq!(data, &[3u8, 1u8, 7u8, 3u8, 5u8]);
}
assert!(directory.delete(*TEST_PATH).is_ok());
}
fn test_rewrite_forbidden(directory: &mut Directory) {
{
directory.open_write(*TEST_PATH).unwrap();
assert!(directory.exists(*TEST_PATH));
}
{
assert!(directory.open_write(*TEST_PATH).is_err());
}
assert!(directory.delete(*TEST_PATH).is_ok());
}
fn test_write_create_the_file(directory: &mut Directory) {
{
assert!(directory.open_read(*TEST_PATH).is_err());
let _w = directory.open_write(*TEST_PATH).unwrap();
assert!(directory.exists(*TEST_PATH));
assert!(directory.open_read(*TEST_PATH).is_ok());
assert!(directory.delete(*TEST_PATH).is_ok());
}
}
fn test_directory_delete(directory: &mut Directory) {
assert!(directory.open_read(*TEST_PATH).is_err());
let mut write_file = directory.open_write(*TEST_PATH).unwrap();
write_file.write_all(&[1, 2, 3, 4]).unwrap();
write_file.flush().unwrap();
{
let read_handle = directory.open_read(*TEST_PATH).unwrap();
{
assert_eq!(&*read_handle, &[1u8, 2u8, 3u8, 4u8]);
// Mapped files can't be deleted on Windows
if !cfg!(windows) {
assert!(directory.delete(*TEST_PATH).is_ok());
assert_eq!(&*read_handle, &[1u8, 2u8, 3u8, 4u8]);
}
assert!(directory.delete(Path::new("SomeOtherPath")).is_err());
}
}
if cfg!(windows) {
assert!(directory.delete(*TEST_PATH).is_ok());
}
assert!(directory.open_read(*TEST_PATH).is_err());
assert!(directory.delete(*TEST_PATH).is_err());
}
fn test_directory(directory: &mut Directory) {
test_simple(directory);
test_seek(directory);
test_rewrite_forbidden(directory);
test_write_create_the_file(directory);
test_directory_delete(directory);
test_lock_non_blocking(directory);
test_lock_blocking(directory);
test_watch(directory);
}
fn test_watch(directory: &mut Directory) {
let counter: Arc<AtomicUsize> = Default::default();
let counter_clone = counter.clone();
let watch_callback = Box::new(move || {
counter_clone.fetch_add(1, Ordering::SeqCst);
});
assert!(directory
.atomic_write(Path::new("meta.json"), b"random_test_data")
.is_ok());
thread::sleep(Duration::new(0, 10_000));
assert_eq!(0, counter.load(Ordering::SeqCst));
let watch_handle = directory.watch(watch_callback);
for i in 0..10 {
assert_eq!(i, counter.load(Ordering::SeqCst));
assert!(directory
.atomic_write(Path::new("meta.json"), b"random_test_data_2")
.is_ok());
for _ in 0..100 {
if counter.load(Ordering::SeqCst) > i {
break;
}
thread::sleep(Duration::from_millis(10));
}
assert_eq!(i + 1, counter.load(Ordering::SeqCst));
}
mem::drop(watch_handle);
assert!(directory
.atomic_write(Path::new("meta.json"), b"random_test_data")
.is_ok());
thread::sleep(Duration::from_millis(200));
assert_eq!(10, counter.load(Ordering::SeqCst));
}
fn test_lock_non_blocking(directory: &mut Directory) {
{
let lock_a_res = directory.acquire_lock(&Lock {
filepath: PathBuf::from("a.lock"),
is_blocking: false,
});
assert!(lock_a_res.is_ok());
let lock_b_res = directory.acquire_lock(&Lock {
filepath: PathBuf::from("b.lock"),
is_blocking: false,
});
assert!(lock_b_res.is_ok());
let lock_a_res2 = directory.acquire_lock(&Lock {
filepath: PathBuf::from("a.lock"),
is_blocking: false,
});
assert!(lock_a_res2.is_err());
}
let lock_a_res = directory.acquire_lock(&Lock {
filepath: PathBuf::from("a.lock"),
is_blocking: false,
});
assert!(lock_a_res.is_ok());
}
fn test_lock_blocking(directory: &mut Directory) {
let lock_a_res = directory.acquire_lock(&Lock {
filepath: PathBuf::from("a.lock"),
is_blocking: true,
});
assert!(lock_a_res.is_ok());
std::thread::spawn(move || {
//< lock_a_res is sent to the thread.
std::thread::sleep(time::Duration::from_millis(10));
// explicitely droping lock_a_res. It would have been sufficient to just force it
// to be part of the move, but the intent seems clearer that way.
drop(lock_a_res);
});
{
// A non-blocking call should fail, as the thread is running and holding the lock.
let lock_a_res = directory.acquire_lock(&Lock {
filepath: PathBuf::from("a.lock"),
is_blocking: false,
});
assert!(lock_a_res.is_err());
}
{
// the blocking call should wait for at least 10ms.
let start = time::Instant::now();
let lock_a_res = directory.acquire_lock(&Lock {
filepath: PathBuf::from("a.lock"),
is_blocking: true,
});
assert!(lock_a_res.is_ok());
assert!(start.elapsed().subsec_millis() >= 10);
}
}

156
src/directory/watch_event_router.rs
View File

@@ -1,156 +0,0 @@
use std::sync::Arc;
use std::sync::RwLock;
use std::sync::Weak;
/// Type alias for callbacks registered when watching files of a `Directory`.
pub type WatchCallback = Box<Fn() -> () + Sync + Send>;
/// Helper struct to implement the watch method in `Directory` implementations.
///
/// It registers callbacks (See `.subscribe(...)`) and
/// calls them upon calls to `.broadcast(...)`.
#[derive(Default)]
pub struct WatchCallbackList {
router: RwLock<Vec<Weak<WatchCallback>>>,
}
/// Controls how long a directory should watch for a file change.
///
/// After all the clones of `WatchHandle` are dropped, the associated will not be called when a
/// file change is detected.
#[must_use = "This `WatchHandle` controls the lifetime of the watch and should therefore be used."]
#[derive(Clone)]
pub struct WatchHandle(Arc<WatchCallback>);
impl WatchCallbackList {
/// Suscribes a new callback and returns a handle that controls the lifetime of the callback.
pub fn subscribe(&self, watch_callback: WatchCallback) -> WatchHandle {
let watch_callback_arc = Arc::new(watch_callback);
let watch_callback_weak = Arc::downgrade(&watch_callback_arc);
self.router.write().unwrap().push(watch_callback_weak);
WatchHandle(watch_callback_arc)
}
fn list_callback(&self) -> Vec<Arc<WatchCallback>> {
let mut callbacks = vec![];
let mut router_wlock = self.router.write().unwrap();
let mut i = 0;
while i < router_wlock.len() {
if let Some(watch) = router_wlock[i].upgrade() {
callbacks.push(watch);
i += 1;
} else {
router_wlock.swap_remove(i);
}
}
callbacks
}
/// Triggers all callbacks
pub fn broadcast(&self) {
let callbacks = self.list_callback();
let spawn_res = std::thread::Builder::new()
.name("watch-callbacks".to_string())
.spawn(move || {
for callback in callbacks {
callback();
}
});
if let Err(err) = spawn_res {
error!(
"Failed to spawn thread to call watch callbacks. Cause: {:?}",
err
);
}
}
}
#[cfg(test)]
mod tests {
use directory::WatchCallbackList;
use std::mem;
use std::sync::atomic::{AtomicUsize, Ordering};
use std::sync::Arc;
use std::thread;
use std::time::Duration;
const WAIT_TIME: u64 = 20;
#[test]
fn test_watch_event_router_simple() {
let watch_event_router = WatchCallbackList::default();
let counter: Arc<AtomicUsize> = Default::default();
let counter_clone = counter.clone();
let inc_callback = Box::new(move || {
counter_clone.fetch_add(1, Ordering::SeqCst);
});
watch_event_router.broadcast();
assert_eq!(0, counter.load(Ordering::SeqCst));
let handle_a = watch_event_router.subscribe(inc_callback);
thread::sleep(Duration::from_millis(WAIT_TIME));
assert_eq!(0, counter.load(Ordering::SeqCst));
watch_event_router.broadcast();
thread::sleep(Duration::from_millis(WAIT_TIME));
assert_eq!(1, counter.load(Ordering::SeqCst));
watch_event_router.broadcast();
watch_event_router.broadcast();
watch_event_router.broadcast();
thread::sleep(Duration::from_millis(WAIT_TIME));
assert_eq!(4, counter.load(Ordering::SeqCst));
mem::drop(handle_a);
watch_event_router.broadcast();
thread::sleep(Duration::from_millis(WAIT_TIME));
assert_eq!(4, counter.load(Ordering::SeqCst));
}
#[test]
fn test_watch_event_router_multiple_callback_same_key() {
let watch_event_router = WatchCallbackList::default();
let counter: Arc<AtomicUsize> = Default::default();
let inc_callback = |inc: usize| {
let counter_clone = counter.clone();
Box::new(move || {
counter_clone.fetch_add(inc, Ordering::SeqCst);
})
};
let handle_a = watch_event_router.subscribe(inc_callback(1));
let handle_a2 = watch_event_router.subscribe(inc_callback(10));
thread::sleep(Duration::from_millis(WAIT_TIME));
assert_eq!(0, counter.load(Ordering::SeqCst));
watch_event_router.broadcast();
watch_event_router.broadcast();
thread::sleep(Duration::from_millis(WAIT_TIME));
assert_eq!(22, counter.load(Ordering::SeqCst));
mem::drop(handle_a);
watch_event_router.broadcast();
thread::sleep(Duration::from_millis(WAIT_TIME));
assert_eq!(32, counter.load(Ordering::SeqCst));
mem::drop(handle_a2);
watch_event_router.broadcast();
watch_event_router.broadcast();
thread::sleep(Duration::from_millis(WAIT_TIME));
assert_eq!(32, counter.load(Ordering::SeqCst));
}
#[test]
fn test_watch_event_router_multiple_callback_different_key() {
let watch_event_router = WatchCallbackList::default();
let counter: Arc<AtomicUsize> = Default::default();
let counter_clone = counter.clone();
let inc_callback = Box::new(move || {
counter_clone.fetch_add(1, Ordering::SeqCst);
});
let handle_a = watch_event_router.subscribe(inc_callback);
assert_eq!(0, counter.load(Ordering::SeqCst));
watch_event_router.broadcast();
watch_event_router.broadcast();
thread::sleep(Duration::from_millis(WAIT_TIME));
assert_eq!(2, counter.load(Ordering::SeqCst));
thread::sleep(Duration::from_millis(WAIT_TIME));
mem::drop(handle_a);
watch_event_router.broadcast();
thread::sleep(Duration::from_millis(WAIT_TIME));
assert_eq!(2, counter.load(Ordering::SeqCst));
}
}

83
src/error.rs
View File

@@ -2,93 +2,53 @@
use std::io; use std::io;
use directory::error::LockError;
use directory::error::{IOError, OpenDirectoryError, OpenReadError, OpenWriteError}; use directory::error::{IOError, OpenDirectoryError, OpenReadError, OpenWriteError};
use fastfield::FastFieldNotAvailableError; use fastfield::FastFieldNotAvailableError;
use indexer::LockType;
use query; use query;
use schema; use schema;
use serde_json; use serde_json;
use std::fmt;
use std::path::PathBuf; use std::path::PathBuf;
use std::sync::PoisonError; use std::sync::PoisonError;
pub struct DataCorruption {
filepath: Option<PathBuf>,
comment: String,
}
impl DataCorruption {
pub fn new(filepath: PathBuf, comment: String) -> DataCorruption {
DataCorruption {
filepath: Some(filepath),
comment,
}
}
pub fn comment_only(comment: String) -> DataCorruption {
DataCorruption {
filepath: None,
comment,
}
}
}
impl fmt::Debug for DataCorruption {
fn fmt(&self, f: &mut fmt::Formatter) -> Result<(), fmt::Error> {
write!(f, "Data corruption: ")?;
if let Some(ref filepath) = &self.filepath {
write!(f, "(in file `{:?}`)", filepath)?;
}
write!(f, ": {}.", self.comment)?;
Ok(())
}
}
/// The library's failure based error enum /// The library's failure based error enum
#[derive(Debug, Fail)] #[derive(Debug, Fail)]
pub enum TantivyError { pub enum TantivyError {
/// Path does not exist. /// Path does not exist.
#[fail(display = "Path does not exist: '{:?}'", _0)] #[fail(display = "path does not exist: '{:?}'", _0)]
PathDoesNotExist(PathBuf), PathDoesNotExist(PathBuf),
/// File already exists, this is a problem when we try to write into a new file. /// File already exists, this is a problem when we try to write into a new file.
#[fail(display = "File already exists: '{:?}'", _0)] #[fail(display = "file already exists: '{:?}'", _0)]
FileAlreadyExists(PathBuf), FileAlreadyExists(PathBuf),
/// Index already exists in this directory
#[fail(display = "Index already exists")]
IndexAlreadyExists,
/// Failed to acquire file lock /// Failed to acquire file lock
#[fail(display = "Failed to acquire Lockfile: {:?}. {:?}", _0, _1)] #[fail(
LockFailure(LockError, Option<String>), display = "Failed to acquire Lockfile: {:?}. Possible causes: another IndexWriter instance or panic during previous lock drop.",
_0
)]
LockFailure(LockType),
/// IO Error. /// IO Error.
#[fail(display = "An IO error occurred: '{}'", _0)] #[fail(display = "an IO error occurred: '{}'", _0)]
IOError(#[cause] IOError), IOError(#[cause] IOError),
/// Data corruption. /// The data within is corrupted.
#[fail(display = "{:?}", _0)] ///
DataCorruption(DataCorruption), /// For instance, it contains invalid JSON.
#[fail(display = "file contains corrupted data: '{:?}'", _0)]
CorruptedFile(PathBuf),
/// A thread holding the locked panicked and poisoned the lock. /// A thread holding the locked panicked and poisoned the lock.
#[fail(display = "A thread holding the locked panicked and poisoned the lock")] #[fail(display = "a thread holding the locked panicked and poisoned the lock")]
Poisoned, Poisoned,
/// Invalid argument was passed by the user. /// Invalid argument was passed by the user.
#[fail(display = "An invalid argument was passed: '{}'", _0)] #[fail(display = "an invalid argument was passed: '{}'", _0)]
InvalidArgument(String), InvalidArgument(String),
/// An Error happened in one of the thread. /// An Error happened in one of the thread.
#[fail(display = "An error occurred in a thread: '{}'", _0)] #[fail(display = "an error occurred in a thread: '{}'", _0)]
ErrorInThread(String), ErrorInThread(String),
/// An Error appeared related to the schema. /// An Error appeared related to the schema.
#[fail(display = "Schema error: '{}'", _0)] #[fail(display = "Schema error: '{}'", _0)]
SchemaError(String), SchemaError(String),
/// Tried to access a fastfield reader for a field not configured accordingly. /// Tried to access a fastfield reader for a field not configured accordingly.
#[fail(display = "Fast field not available: '{:?}'", _0)] #[fail(display = "fast field not available: '{:?}'", _0)]
FastFieldError(#[cause] FastFieldNotAvailableError), FastFieldError(#[cause] FastFieldNotAvailableError),
/// System error. (e.g.: We failed spawning a new thread)
#[fail(display = "System error.'{}'", _0)]
SystemError(String),
}
impl From<DataCorruption> for TantivyError {
fn from(data_corruption: DataCorruption) -> TantivyError {
TantivyError::DataCorruption(data_corruption)
}
} }
impl From<FastFieldNotAvailableError> for TantivyError { impl From<FastFieldNotAvailableError> for TantivyError {
@@ -97,12 +57,6 @@ impl From<FastFieldNotAvailableError> for TantivyError {
} }
} }
impl From<LockError> for TantivyError {
fn from(lock_error: LockError) -> TantivyError {
TantivyError::LockFailure(lock_error, None)
}
}
impl From<IOError> for TantivyError { impl From<IOError> for TantivyError {
fn from(io_error: IOError) -> TantivyError { fn from(io_error: IOError) -> TantivyError {
TantivyError::IOError(io_error) TantivyError::IOError(io_error)
@@ -162,7 +116,6 @@ impl From<OpenDirectoryError> for TantivyError {
OpenDirectoryError::NotADirectory(directory_path) => { OpenDirectoryError::NotADirectory(directory_path) => {
TantivyError::InvalidArgument(format!("{:?} is not a directory", directory_path)) TantivyError::InvalidArgument(format!("{:?} is not a directory", directory_path))
} }
OpenDirectoryError::IoError(err) => TantivyError::IOError(IOError::from(err)),
} }
} }
} }

8
src/fastfield/bytes/mod.rs
View File

@@ -6,12 +6,12 @@ pub use self::writer::BytesFastFieldWriter;
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use schema::Schema; use schema::SchemaBuilder;
use Index; use Index;
#[test] #[test]
fn test_bytes() { fn test_bytes() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let field = schema_builder.add_bytes_field("bytesfield"); let field = schema_builder.add_bytes_field("bytesfield");
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
@@ -22,7 +22,9 @@ mod tests {
index_writer.add_document(doc!(field=>vec![1u8, 3, 5, 7, 9])); index_writer.add_document(doc!(field=>vec![1u8, 3, 5, 7, 9]));
index_writer.add_document(doc!(field=>vec![0u8; 1000])); index_writer.add_document(doc!(field=>vec![0u8; 1000]));
assert!(index_writer.commit().is_ok()); assert!(index_writer.commit().is_ok());
let searcher = index.reader().unwrap().searcher();
index.load_searchers().unwrap();
let searcher = index.searcher();
let reader = searcher.segment_reader(0); let reader = searcher.segment_reader(0);
let bytes_reader = reader.bytes_fast_field_reader(field).unwrap(); let bytes_reader = reader.bytes_fast_field_reader(field).unwrap();

6
src/fastfield/delete.rs
View File

@@ -2,7 +2,6 @@ use bit_set::BitSet;
use common::HasLen; use common::HasLen;
use directory::ReadOnlySource; use directory::ReadOnlySource;
use directory::WritePtr; use directory::WritePtr;
use space_usage::ByteCount;
use std::io; use std::io;
use std::io::Write; use std::io::Write;
use DocId; use DocId;
@@ -64,11 +63,6 @@ impl DeleteBitSet {
b & (1u8 << shift) != 0 b & (1u8 << shift) != 0
} }
} }
/// Summarize total space usage of this bitset.
pub fn space_usage(&self) -> ByteCount {
self.data.len()
}
} }
impl HasLen for DeleteBitSet { impl HasLen for DeleteBitSet {

14
src/fastfield/facet_reader.rs
View File

@@ -1,6 +1,5 @@
use super::MultiValueIntFastFieldReader; use super::MultiValueIntFastFieldReader;
use schema::Facet; use schema::Facet;
use std::str;
use termdict::TermDictionary; use termdict::TermDictionary;
use termdict::TermOrdinal; use termdict::TermOrdinal;
use DocId; use DocId;
@@ -21,7 +20,6 @@ use DocId;
pub struct FacetReader { pub struct FacetReader {
term_ords: MultiValueIntFastFieldReader<u64>, term_ords: MultiValueIntFastFieldReader<u64>,
term_dict: TermDictionary, term_dict: TermDictionary,
buffer: Vec<u8>,
} }
impl FacetReader { impl FacetReader {
@@ -39,7 +37,6 @@ impl FacetReader {
FacetReader { FacetReader {
term_ords, term_ords,
term_dict, term_dict,
buffer: vec![],
} }
} }
@@ -58,18 +55,11 @@ impl FacetReader {
} }
/// Given a term ordinal returns the term associated to it. /// Given a term ordinal returns the term associated to it.
pub fn facet_from_ord( pub fn facet_from_ord(&self, facet_ord: TermOrdinal, output: &mut Facet) {
&mut self,
facet_ord: TermOrdinal,
output: &mut Facet,
) -> Result<(), str::Utf8Error> {
let found_term = self let found_term = self
.term_dict .term_dict
.ord_to_term(facet_ord as u64, &mut self.buffer); .ord_to_term(facet_ord as u64, output.inner_buffer_mut());
assert!(found_term, "Term ordinal {} no found.", facet_ord); assert!(found_term, "Term ordinal {} no found.", facet_ord);
let facet_str = str::from_utf8(&self.buffer[..])?;
output.set_facet_str(facet_str);
Ok(())
} }
/// Return the list of facet ordinals associated to a document. /// Return the list of facet ordinals associated to a document.

17
src/fastfield/mod.rs
View File

@@ -127,19 +127,19 @@ mod tests {
use common::CompositeFile; use common::CompositeFile;
use directory::{Directory, RAMDirectory, WritePtr}; use directory::{Directory, RAMDirectory, WritePtr};
use fastfield::FastFieldReader; use fastfield::FastFieldReader;
use rand::prelude::SliceRandom; use rand::Rng;
use rand::rngs::StdRng;
use rand::SeedableRng; use rand::SeedableRng;
use rand::XorShiftRng;
use schema::Document; use schema::Document;
use schema::Field; use schema::Field;
use schema::Schema;
use schema::FAST; use schema::FAST;
use schema::{Schema, SchemaBuilder};
use std::collections::HashMap; use std::collections::HashMap;
use std::path::Path; use std::path::Path;
lazy_static! { lazy_static! {
pub static ref SCHEMA: Schema = { pub static ref SCHEMA: Schema = {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
schema_builder.add_u64_field("field", FAST); schema_builder.add_u64_field("field", FAST);
schema_builder.build() schema_builder.build()
}; };
@@ -298,7 +298,7 @@ mod tests {
fn test_signed_intfastfield() { fn test_signed_intfastfield() {
let path = Path::new("test"); let path = Path::new("test");
let mut directory: RAMDirectory = RAMDirectory::create(); let mut directory: RAMDirectory = RAMDirectory::create();
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
let i64_field = schema_builder.add_i64_field("field", FAST); let i64_field = schema_builder.add_i64_field("field", FAST);
let schema = schema_builder.build(); let schema = schema_builder.build();
@@ -342,7 +342,7 @@ mod tests {
fn test_signed_intfastfield_default_val() { fn test_signed_intfastfield_default_val() {
let path = Path::new("test"); let path = Path::new("test");
let mut directory: RAMDirectory = RAMDirectory::create(); let mut directory: RAMDirectory = RAMDirectory::create();
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
let i64_field = schema_builder.add_i64_field("field", FAST); let i64_field = schema_builder.add_i64_field("field", FAST);
let schema = schema_builder.build(); let schema = schema_builder.build();
@@ -367,10 +367,11 @@ mod tests {
} }
} }
// Warning: this generates the same permutation at each call
pub fn generate_permutation() -> Vec<u64> { pub fn generate_permutation() -> Vec<u64> {
let seed: [u8; 16] = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16];
let mut rng = XorShiftRng::from_seed(seed);
let mut permutation: Vec<u64> = (0u64..100_000u64).collect(); let mut permutation: Vec<u64> = (0u64..100_000u64).collect();
permutation.shuffle(&mut StdRng::from_seed([1u8; 32])); rng.shuffle(&mut permutation);
permutation permutation
} }

164
src/fastfield/multivalued/mod.rs
View File

@@ -7,20 +7,14 @@ pub use self::writer::MultiValueIntFastFieldWriter;
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
extern crate time;
use self::time::Duration;
use collector::TopDocs;
use query::QueryParser;
use schema::Cardinality; use schema::Cardinality;
use schema::Facet;
use schema::IntOptions; use schema::IntOptions;
use schema::Schema; use schema::SchemaBuilder;
use Index; use Index;
#[test] #[test]
fn test_multivalued_u64() { fn test_multivalued_u64() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let field = schema_builder.add_u64_field( let field = schema_builder.add_u64_field(
"multifield", "multifield",
IntOptions::default().set_fast(Cardinality::MultiValues), IntOptions::default().set_fast(Cardinality::MultiValues),
@@ -34,12 +28,11 @@ mod tests {
index_writer.add_document(doc!(field=>5u64, field=>20u64,field=>1u64)); index_writer.add_document(doc!(field=>5u64, field=>20u64,field=>1u64));
assert!(index_writer.commit().is_ok()); assert!(index_writer.commit().is_ok());
let searcher = index.reader().unwrap().searcher(); index.load_searchers().unwrap();
let segment_reader = searcher.segment_reader(0); let searcher = index.searcher();
let reader = searcher.segment_reader(0);
let mut vals = Vec::new(); let mut vals = Vec::new();
let multi_value_reader = segment_reader let multi_value_reader = reader.multi_fast_field_reader::<u64>(field).unwrap();
.multi_fast_field_reader::<u64>(field)
.unwrap();
{ {
multi_value_reader.get_vals(2, &mut vals); multi_value_reader.get_vals(2, &mut vals);
assert_eq!(&vals, &[4u64]); assert_eq!(&vals, &[4u64]);
@@ -54,136 +47,9 @@ mod tests {
} }
} }
#[test]
fn test_multivalued_date() {
let mut schema_builder = Schema::builder();
let date_field = schema_builder.add_date_field(
"multi_date_field",
IntOptions::default()
.set_fast(Cardinality::MultiValues)
.set_indexed()
.set_stored(),
);
let time_i =
schema_builder.add_i64_field("time_stamp_i", IntOptions::default().set_stored());
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
let first_time_stamp = chrono::Utc::now();
index_writer.add_document(
doc!(date_field=>first_time_stamp, date_field=>first_time_stamp, time_i=>1i64),
);
index_writer.add_document(doc!(time_i=>0i64));
// add one second
index_writer
.add_document(doc!(date_field=>first_time_stamp + Duration::seconds(1), time_i=>2i64));
// add another second
let two_secs_ahead = first_time_stamp + Duration::seconds(2);
index_writer.add_document(doc!(date_field=>two_secs_ahead, date_field=>two_secs_ahead,date_field=>two_secs_ahead, time_i=>3i64));
assert!(index_writer.commit().is_ok());
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let reader = searcher.segment_reader(0);
assert_eq!(reader.num_docs(), 4);
{
let parser = QueryParser::for_index(&index, vec![date_field]);
let query = parser
.parse_query(&format!("\"{}\"", first_time_stamp.to_rfc3339()).to_string())
.expect("could not parse query");
let results = searcher
.search(&query, &TopDocs::with_limit(5))
.expect("could not query index");
assert_eq!(results.len(), 1);
for (_score, doc_address) in results {
let retrieved_doc = searcher.doc(doc_address).expect("cannot fetch doc");
assert_eq!(
retrieved_doc
.get_first(date_field)
.expect("cannot find value")
.date_value()
.timestamp(),
first_time_stamp.timestamp()
);
assert_eq!(
retrieved_doc
.get_first(time_i)
.expect("cannot find value")
.i64_value(),
1i64
);
}
}
{
let parser = QueryParser::for_index(&index, vec![date_field]);
let query = parser
.parse_query(&format!("\"{}\"", two_secs_ahead.to_rfc3339()).to_string())
.expect("could not parse query");
let results = searcher
.search(&query, &TopDocs::with_limit(5))
.expect("could not query index");
assert_eq!(results.len(), 1);
for (_score, doc_address) in results {
let retrieved_doc = searcher.doc(doc_address).expect("cannot fetch doc");
assert_eq!(
retrieved_doc
.get_first(date_field)
.expect("cannot find value")
.date_value()
.timestamp(),
two_secs_ahead.timestamp()
);
assert_eq!(
retrieved_doc
.get_first(time_i)
.expect("cannot find value")
.i64_value(),
3i64
);
}
}
// TODO: support Date range queries
// {
// let parser = QueryParser::for_index(&index, vec![date_field]);
// let range_q = format!("\"{}\"..\"{}\"",
// (first_time_stamp + Duration::seconds(1)).to_rfc3339(),
// (first_time_stamp + Duration::seconds(3)).to_rfc3339()
// );
// let query = parser.parse_query(&range_q)
// .expect("could not parse query");
// let results = searcher.search(&query, &TopDocs::with_limit(5))
// .expect("could not query index");
//
//
// assert_eq!(results.len(), 2);
// for (i, doc_pair) in results.iter().enumerate() {
// let retrieved_doc = searcher.doc(doc_pair.1).expect("cannot fetch doc");
// let offset_sec = match i {
// 0 => 1,
// 1 => 3,
// _ => panic!("should not have more than 2 docs")
// };
// let time_i_val = match i {
// 0 => 2,
// 1 => 3,
// _ => panic!("should not have more than 2 docs")
// };
// assert_eq!(retrieved_doc.get_first(date_field).expect("cannot find value").date_value().timestamp(),
// (first_time_stamp + Duration::seconds(offset_sec)).timestamp());
// assert_eq!(retrieved_doc.get_first(time_i).expect("cannot find value").i64_value(), time_i_val);
// }
// }
}
#[test] #[test]
fn test_multivalued_i64() { fn test_multivalued_i64() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let field = schema_builder.add_i64_field( let field = schema_builder.add_i64_field(
"multifield", "multifield",
IntOptions::default().set_fast(Cardinality::MultiValues), IntOptions::default().set_fast(Cardinality::MultiValues),
@@ -197,7 +63,8 @@ mod tests {
index_writer.add_document(doc!(field=> -5i64, field => -20i64, field=>1i64)); index_writer.add_document(doc!(field=> -5i64, field => -20i64, field=>1i64));
assert!(index_writer.commit().is_ok()); assert!(index_writer.commit().is_ok());
let searcher = index.reader().unwrap().searcher(); index.load_searchers().unwrap();
let searcher = index.searcher();
let reader = searcher.segment_reader(0); let reader = searcher.segment_reader(0);
let mut vals = Vec::new(); let mut vals = Vec::new();
let multi_value_reader = reader.multi_fast_field_reader::<i64>(field).unwrap(); let multi_value_reader = reader.multi_fast_field_reader::<i64>(field).unwrap();
@@ -218,17 +85,4 @@ mod tests {
assert_eq!(&vals, &[-5i64, -20i64, 1i64]); assert_eq!(&vals, &[-5i64, -20i64, 1i64]);
} }
} }
#[test]
#[ignore]
fn test_many_facets() {
let mut schema_builder = Schema::builder();
let field = schema_builder.add_facet_field("facetfield");
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
for i in 0..100_000 {
index_writer.add_document(doc!(field=> Facet::from(format!("/lang/{}", i).as_str())));
}
assert!(index_writer.commit().is_ok());
}
} }

17
src/fastfield/multivalued/reader.rs
View File

@@ -39,7 +39,7 @@ impl<Item: FastValue> MultiValueIntFastFieldReader<Item> {
let (start, stop) = self.range(doc); let (start, stop) = self.range(doc);
let len = (stop - start) as usize; let len = (stop - start) as usize;
vals.resize(len, Item::default()); vals.resize(len, Item::default());
self.vals_reader.get_range_u64(start, &mut vals[..]); self.vals_reader.get_range(start as u32, &mut vals[..]);
} }
} }
@@ -47,11 +47,11 @@ impl<Item: FastValue> MultiValueIntFastFieldReader<Item> {
mod tests { mod tests {
use core::Index; use core::Index;
use schema::{Document, Facet, Schema}; use schema::{Document, Facet, SchemaBuilder};
#[test] #[test]
fn test_multifastfield_reader() { fn test_multifastfield_reader() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
let facet_field = schema_builder.add_facet_field("facets"); let facet_field = schema_builder.add_facet_field("facets");
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
@@ -75,26 +75,27 @@ mod tests {
index_writer.add_document(doc); index_writer.add_document(doc);
} }
index_writer.commit().expect("Commit failed"); index_writer.commit().expect("Commit failed");
let searcher = index.reader().unwrap().searcher(); index.load_searchers().expect("Reloading searchers");
let searcher = index.searcher();
let segment_reader = searcher.segment_reader(0); let segment_reader = searcher.segment_reader(0);
let mut facet_reader = segment_reader.facet_reader(facet_field).unwrap(); let mut facet_reader = segment_reader.facet_reader(facet_field).unwrap();
let mut facet = Facet::root(); let mut facet = Facet::root();
{ {
facet_reader.facet_from_ord(1, &mut facet).unwrap(); facet_reader.facet_from_ord(1, &mut facet);
assert_eq!(facet, Facet::from("/category")); assert_eq!(facet, Facet::from("/category"));
} }
{ {
facet_reader.facet_from_ord(2, &mut facet).unwrap(); facet_reader.facet_from_ord(2, &mut facet);
assert_eq!(facet, Facet::from("/category/cat1")); assert_eq!(facet, Facet::from("/category/cat1"));
} }
{ {
facet_reader.facet_from_ord(3, &mut facet).unwrap(); facet_reader.facet_from_ord(3, &mut facet);
assert_eq!(format!("{}", facet), "/category/cat2"); assert_eq!(format!("{}", facet), "/category/cat2");
assert_eq!(facet, Facet::from("/category/cat2")); assert_eq!(facet, Facet::from("/category/cat2"));
} }
{ {
facet_reader.facet_from_ord(4, &mut facet).unwrap(); facet_reader.facet_from_ord(4, &mut facet);
assert_eq!(facet, Facet::from("/category/cat3")); assert_eq!(facet, Facet::from("/category/cat3"));
} }

2
src/fastfield/multivalued/writer.rs
View File

@@ -32,7 +32,7 @@ use DocId;
/// term ids when the segment is getting serialized. /// term ids when the segment is getting serialized.
pub struct MultiValueIntFastFieldWriter { pub struct MultiValueIntFastFieldWriter {
field: Field, field: Field,
vals: Vec<UnorderedTermId>, vals: Vec<u64>,
doc_index: Vec<u64>, doc_index: Vec<u64>,
is_facet: bool, is_facet: bool,
} }

40
src/fastfield/reader.rs
View File

@@ -7,7 +7,7 @@ use directory::ReadOnlySource;
use directory::{Directory, RAMDirectory, WritePtr}; use directory::{Directory, RAMDirectory, WritePtr};
use fastfield::{FastFieldSerializer, FastFieldsWriter}; use fastfield::{FastFieldSerializer, FastFieldsWriter};
use owning_ref::OwningRef; use owning_ref::OwningRef;
use schema::Schema; use schema::SchemaBuilder;
use schema::FAST; use schema::FAST;
use std::collections::HashMap; use std::collections::HashMap;
use std::marker::PhantomData; use std::marker::PhantomData;
@@ -59,29 +59,7 @@ impl<Item: FastValue> FastFieldReader<Item> {
/// May panic if `doc` is greater than the segment /// May panic if `doc` is greater than the segment
// `maxdoc`. // `maxdoc`.
pub fn get(&self, doc: DocId) -> Item { pub fn get(&self, doc: DocId) -> Item {
self.get_u64(u64::from(doc)) Item::from_u64(self.min_value_u64 + self.bit_unpacker.get(doc as usize))
}
pub(crate) fn get_u64(&self, doc: u64) -> Item {
Item::from_u64(self.min_value_u64 + self.bit_unpacker.get(doc))
}
/// Internally `multivalued` also use SingleValue Fast fields.
/// It works as follows... A first column contains the list of start index
/// for each document, a second column contains the actual values.
///
/// The values associated to a given doc, are then
/// `second_column[first_column.get(doc)..first_column.get(doc+1)]`.
///
/// Which means single value fast field reader can be indexed internally with
/// something different from a `DocId`. For this use case, we want to use `u64`
/// values.
///
/// See `get_range` for an actual documentation about this method.
pub(crate) fn get_range_u64(&self, start: u64, output: &mut [Item]) {
for (i, out) in output.iter_mut().enumerate() {
*out = self.get_u64(start + (i as u64));
}
} }
/// Fills an output buffer with the fast field values /// Fills an output buffer with the fast field values
@@ -97,8 +75,16 @@ impl<Item: FastValue> FastFieldReader<Item> {
/// ///
/// May panic if `start + output.len()` is greater than /// May panic if `start + output.len()` is greater than
/// the segment's `maxdoc`. /// the segment's `maxdoc`.
pub fn get_range(&self, start: DocId, output: &mut [Item]) { ///
self.get_range_u64(u64::from(start), output); // TODO change start to `u64`.
// For multifastfield, start is an index in a second fastfield, not a `DocId`
pub fn get_range(&self, start: u32, output: &mut [Item]) {
// ok: Item is either `u64` or `i64`
let output_u64: &mut [u64] = unsafe { &mut *(output as *mut [Item] as *mut [u64]) };
self.bit_unpacker.get_range(start, output_u64);
for out in output_u64.iter_mut() {
*out = Item::from_u64(*out + self.min_value_u64).as_u64();
}
} }
/// Returns the minimum value for this fast field. /// Returns the minimum value for this fast field.
@@ -122,7 +108,7 @@ impl<Item: FastValue> FastFieldReader<Item> {
impl<Item: FastValue> From<Vec<Item>> for FastFieldReader<Item> { impl<Item: FastValue> From<Vec<Item>> for FastFieldReader<Item> {
fn from(vals: Vec<Item>) -> FastFieldReader<Item> { fn from(vals: Vec<Item>) -> FastFieldReader<Item> {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let field = schema_builder.add_u64_field("field", FAST); let field = schema_builder.add_u64_field("field", FAST);
let schema = schema_builder.build(); let schema = schema_builder.build();
let path = Path::new("__dummy__"); let path = Path::new("__dummy__");

2
src/fieldnorm/mod.rs
View File

@@ -15,7 +15,7 @@
//! precompute computationally expensive functions of the fieldnorm //! precompute computationally expensive functions of the fieldnorm
//! in a very short array. //! in a very short array.
//! //!
//! This trick is used by the BM25 similarity. //! This trick is used by the [BM25 similarity]().
mod code; mod code;
mod reader; mod reader;
mod serializer; mod serializer;

16
src/functional_test.rs
View File

@@ -1,6 +1,7 @@
use rand::thread_rng; use rand::thread_rng;
use std::collections::HashSet; use std::collections::HashSet;
use rand::distributions::Range;
use rand::Rng; use rand::Rng;
use schema::*; use schema::*;
use Index; use Index;
@@ -13,16 +14,17 @@ fn check_index_content(searcher: &Searcher, vals: &HashSet<u64>) {
#[test] #[test]
#[ignore] #[ignore]
#[cfg(feature = "mmap")]
fn test_indexing() { fn test_indexing() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let id_field = schema_builder.add_u64_field("id", INDEXED); let id_field = schema_builder.add_u64_field("id", INT_INDEXED);
let multiples_field = schema_builder.add_u64_field("multiples", INDEXED); let multiples_field = schema_builder.add_u64_field("multiples", INT_INDEXED);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_from_tempdir(schema).unwrap(); let index = Index::create_from_tempdir(schema).unwrap();
let reader = index.reader().unwrap();
let universe = Range::new(0u64, 20u64);
let mut rng = thread_rng(); let mut rng = thread_rng();
let mut index_writer = index.writer_with_num_threads(3, 120_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(3, 120_000_000).unwrap();
@@ -31,13 +33,13 @@ fn test_indexing() {
let mut uncommitted_docs: HashSet<u64> = HashSet::new(); let mut uncommitted_docs: HashSet<u64> = HashSet::new();
for _ in 0..200 { for _ in 0..200 {
let random_val = rng.gen_range(0, 20); let random_val = rng.sample(&universe);
if random_val == 0 { if random_val == 0 {
index_writer.commit().expect("Commit failed"); index_writer.commit().expect("Commit failed");
committed_docs.extend(&uncommitted_docs); committed_docs.extend(&uncommitted_docs);
uncommitted_docs.clear(); uncommitted_docs.clear();
reader.reload().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
// check that everything is correct. // check that everything is correct.
check_index_content(&searcher, &committed_docs); check_index_content(&searcher, &committed_docs);
} else { } else {

10
src/indexer/delete_queue.rs
View File

@@ -179,11 +179,6 @@ pub struct DeleteCursor {
} }
impl DeleteCursor { impl DeleteCursor {
pub fn empty() -> DeleteCursor {
DeleteQueue::new().cursor()
}
/// Skips operations and position it so that /// Skips operations and position it so that
/// - either all of the delete operation currently in the /// - either all of the delete operation currently in the
/// queue are consume and the next get will return None. /// queue are consume and the next get will return None.
@@ -196,7 +191,10 @@ impl DeleteCursor {
} }
} }
#[cfg_attr(feature = "cargo-clippy", allow(clippy::wrong_self_convention))] #[cfg_attr(
feature = "cargo-clippy",
allow(clippy::wrong_self_convention)
)]
fn is_behind_opstamp(&mut self, target_opstamp: u64) -> bool { fn is_behind_opstamp(&mut self, target_opstamp: u64) -> bool {
self.get() self.get()
.map(|operation| operation.opstamp < target_opstamp) .map(|operation| operation.opstamp < target_opstamp)

131
src/indexer/directory_lock.rs Normal file
View File

@@ -0,0 +1,131 @@
use directory::error::OpenWriteError;
use std::io::Write;
use std::path::{Path, PathBuf};
use std::thread;
use std::time::Duration;
use Directory;
use TantivyError;
#[derive(Debug, Clone, Copy)]
pub enum LockType {
/// Only one process should be able to write tantivy's index at a time.
/// This lock file, when present, is in charge of preventing other processes to open an IndexWriter.
///
/// If the process is killed and this file remains, it is safe to remove it manually.
///
/// Failing to acquire this lock usually means a misuse of tantivy's API,
/// (creating more than one instance of the `IndexWriter`), are a spurious
/// lock file remaining after a crash. In the latter case, removing the file after
/// checking no process running tantivy is running is safe.
IndexWriterLock,
/// The meta lock file is here to protect the segment files being opened by
/// `.load_searchers()` from being garbage collected.
/// It makes it possible for another process to safely consume
/// our index in-writing. Ideally, we may have prefered `RWLock` semantics
/// here, but it is difficult to achieve on Windows.
///
/// Opening segment readers is a very fast process.
/// Right now if the lock cannot be acquire on the first attempt, the logic
/// is very simplistic. We retry after `100ms` until we effectively
/// acquire the lock.
/// This lock should not have much contention in normal usage.
MetaLock,
}
/// Retry the logic of acquiring locks is pretty simple.
/// We just retry `n` times after a given `duratio`, both
/// depending on the type of lock.
struct RetryPolicy {
num_retries: usize,
wait_in_ms: u64,
}
impl RetryPolicy {
fn no_retry() -> RetryPolicy {
RetryPolicy {
num_retries: 0,
wait_in_ms: 0,
}
}
fn wait_and_retry(&mut self) -> bool {
if self.num_retries == 0 {
false
} else {
self.num_retries -= 1;
let wait_duration = Duration::from_millis(self.wait_in_ms);
thread::sleep(wait_duration);
true
}
}
}
impl LockType {
fn retry_policy(self) -> RetryPolicy {
match self {
LockType::IndexWriterLock => RetryPolicy::no_retry(),
LockType::MetaLock => RetryPolicy {
num_retries: 100,
wait_in_ms: 100,
},
}
}
fn try_acquire_lock(self, directory: &mut Directory) -> Result<DirectoryLock, TantivyError> {
let path = self.filename();
let mut write = directory.open_write(path).map_err(|e| match e {
OpenWriteError::FileAlreadyExists(_) => TantivyError::LockFailure(self),
OpenWriteError::IOError(io_error) => TantivyError::IOError(io_error),
})?;
write.flush()?;
Ok(DirectoryLock {
directory: directory.box_clone(),
path: path.to_owned(),
})
}
/// Acquire a lock in the given directory.
pub fn acquire_lock(self, directory: &Directory) -> Result<DirectoryLock, TantivyError> {
let mut box_directory = directory.box_clone();
let mut retry_policy = self.retry_policy();
loop {
let lock_result = self.try_acquire_lock(&mut *box_directory);
match lock_result {
Ok(result) => {
return Ok(result);
}
Err(TantivyError::LockFailure(ref filepath)) => {
if !retry_policy.wait_and_retry() {
return Err(TantivyError::LockFailure(filepath.to_owned()));
}
}
Err(_) => {}
}
}
}
fn filename(&self) -> &Path {
match *self {
LockType::MetaLock => Path::new(".tantivy-meta.lock"),
LockType::IndexWriterLock => Path::new(".tantivy-indexer.lock"),
}
}
}
/// The `DirectoryLock` is an object that represents a file lock.
/// See [`LockType`](struct.LockType.html)
///
/// It is transparently associated to a lock file, that gets deleted
/// on `Drop.` The lock is release automatically on `Drop`.
pub struct DirectoryLock {
directory: Box<Directory>,
path: PathBuf,
}
impl Drop for DirectoryLock {
fn drop(&mut self) {
if let Err(e) = self.directory.delete(&*self.path) {
error!("Failed to remove the lock file. {:?}", e);
}
}
}

458
src/indexer/index_writer.rs
View File

@@ -1,4 +1,4 @@
use super::operation::{AddOperation, UserOperation}; use super::operation::AddOperation;
use super::segment_updater::SegmentUpdater; use super::segment_updater::SegmentUpdater;
use super::PreparedCommit; use super::PreparedCommit;
use bit_set::BitSet; use bit_set::BitSet;
@@ -8,16 +8,16 @@ use core::SegmentComponent;
use core::SegmentId; use core::SegmentId;
use core::SegmentMeta; use core::SegmentMeta;
use core::SegmentReader; use core::SegmentReader;
use crossbeam::channel; use crossbeam_channel as channel;
use directory::DirectoryLock;
use docset::DocSet; use docset::DocSet;
use error::TantivyError; use error::TantivyError;
use fastfield::write_delete_bitset; use fastfield::write_delete_bitset;
use futures::{Canceled, Future}; use futures::sync::oneshot::Receiver;
use indexer::delete_queue::{DeleteCursor, DeleteQueue}; use indexer::delete_queue::{DeleteCursor, DeleteQueue};
use indexer::doc_opstamp_mapping::DocToOpstampMapping; use indexer::doc_opstamp_mapping::DocToOpstampMapping;
use indexer::operation::DeleteOperation; use indexer::operation::DeleteOperation;
use indexer::stamper::Stamper; use indexer::stamper::Stamper;
use indexer::DirectoryLock;
use indexer::MergePolicy; use indexer::MergePolicy;
use indexer::SegmentEntry; use indexer::SegmentEntry;
use indexer::SegmentWriter; use indexer::SegmentWriter;
@@ -26,8 +26,7 @@ use schema::Document;
use schema::IndexRecordOption; use schema::IndexRecordOption;
use schema::Term; use schema::Term;
use std::mem; use std::mem;
use std::ops::Range; use std::mem::swap;
use std::sync::Arc;
use std::thread; use std::thread;
use std::thread::JoinHandle; use std::thread::JoinHandle;
use Result; use Result;
@@ -44,8 +43,8 @@ pub const HEAP_SIZE_MAX: usize = u32::max_value() as usize - MARGIN_IN_BYTES;
// reaches `PIPELINE_MAX_SIZE_IN_DOCS` // reaches `PIPELINE_MAX_SIZE_IN_DOCS`
const PIPELINE_MAX_SIZE_IN_DOCS: usize = 10_000; const PIPELINE_MAX_SIZE_IN_DOCS: usize = 10_000;
type OperationSender = channel::Sender<Vec<AddOperation>>; type DocumentSender = channel::Sender<AddOperation>;
type OperationReceiver = channel::Receiver<Vec<AddOperation>>; type DocumentReceiver = channel::Receiver<AddOperation>;
/// Split the thread memory budget into /// Split the thread memory budget into
/// - the heap size /// - the heap size
@@ -53,19 +52,16 @@ type OperationReceiver = channel::Receiver<Vec<AddOperation>>;
/// ///
/// Returns (the heap size in bytes, the hash table size in number of bits) /// Returns (the heap size in bytes, the hash table size in number of bits)
fn initial_table_size(per_thread_memory_budget: usize) -> usize { fn initial_table_size(per_thread_memory_budget: usize) -> usize {
assert!(per_thread_memory_budget > 1_000);
let table_size_limit: usize = per_thread_memory_budget / 3; let table_size_limit: usize = per_thread_memory_budget / 3;
if let Some(limit) = (1..) (1..)
.take_while(|num_bits: &usize| compute_table_size(*num_bits) < table_size_limit) .take_while(|num_bits: &usize| compute_table_size(*num_bits) < table_size_limit)
.last() .last()
{ .unwrap_or_else(|| {
limit.min(19) // we cap it at 2^19 = 512K. panic!(
} else { "Per thread memory is too small: {}",
unreachable!( per_thread_memory_budget
"Per thread memory is too small: {}", )
per_thread_memory_budget }).min(19) // we cap it at 512K
);
}
} }
/// `IndexWriter` is the user entry-point to add document to an index. /// `IndexWriter` is the user entry-point to add document to an index.
@@ -85,8 +81,8 @@ pub struct IndexWriter {
workers_join_handle: Vec<JoinHandle<Result<()>>>, workers_join_handle: Vec<JoinHandle<Result<()>>>,
operation_receiver: OperationReceiver, document_receiver: DocumentReceiver,
operation_sender: OperationSender, document_sender: DocumentSender,
segment_updater: SegmentUpdater, segment_updater: SegmentUpdater,
@@ -133,7 +129,7 @@ pub fn open_index_writer(
let err_msg = format!("The heap size per thread cannot exceed {}", HEAP_SIZE_MAX); let err_msg = format!("The heap size per thread cannot exceed {}", HEAP_SIZE_MAX);
return Err(TantivyError::InvalidArgument(err_msg)); return Err(TantivyError::InvalidArgument(err_msg));
} }
let (document_sender, document_receiver): (OperationSender, OperationReceiver) = let (document_sender, document_receiver): (DocumentSender, DocumentReceiver) =
channel::bounded(PIPELINE_MAX_SIZE_IN_DOCS); channel::bounded(PIPELINE_MAX_SIZE_IN_DOCS);
let delete_queue = DeleteQueue::new(); let delete_queue = DeleteQueue::new();
@@ -143,7 +139,7 @@ pub fn open_index_writer(
let stamper = Stamper::new(current_opstamp); let stamper = Stamper::new(current_opstamp);
let segment_updater = let segment_updater =
SegmentUpdater::create(index.clone(), stamper.clone(), &delete_queue.cursor())?; SegmentUpdater::new(index.clone(), stamper.clone(), &delete_queue.cursor())?;
let mut index_writer = IndexWriter { let mut index_writer = IndexWriter {
_directory_lock: Some(directory_lock), _directory_lock: Some(directory_lock),
@@ -151,8 +147,8 @@ pub fn open_index_writer(
heap_size_in_bytes_per_thread, heap_size_in_bytes_per_thread,
index: index.clone(), index: index.clone(),
operation_receiver: document_receiver, document_receiver,
operation_sender: document_sender, document_sender,
segment_updater, segment_updater,
@@ -259,7 +255,7 @@ pub fn advance_deletes(
write_delete_bitset(&delete_bitset, &mut delete_file)?; write_delete_bitset(&delete_bitset, &mut delete_file)?;
} }
} }
segment_entry.set_meta(target_opstamp, segment.meta().clone()); segment_entry.set_meta((*segment.meta()).clone());
Ok(()) Ok(())
} }
@@ -267,7 +263,7 @@ fn index_documents(
memory_budget: usize, memory_budget: usize,
segment: &Segment, segment: &Segment,
generation: usize, generation: usize,
document_iterator: &mut Iterator<Item = Vec<AddOperation>>, document_iterator: &mut Iterator<Item = AddOperation>,
segment_updater: &mut SegmentUpdater, segment_updater: &mut SegmentUpdater,
mut delete_cursor: DeleteCursor, mut delete_cursor: DeleteCursor,
) -> Result<bool> { ) -> Result<bool> {
@@ -275,11 +271,11 @@ fn index_documents(
let segment_id = segment.id(); let segment_id = segment.id();
let table_size = initial_table_size(memory_budget); let table_size = initial_table_size(memory_budget);
let mut segment_writer = SegmentWriter::for_segment(table_size, segment.clone(), &schema)?; let mut segment_writer = SegmentWriter::for_segment(table_size, segment.clone(), &schema)?;
for documents in document_iterator { for doc in document_iterator {
for doc in documents { segment_writer.add_document(doc, &schema)?;
segment_writer.add_document(doc, &schema)?;
}
let mem_usage = segment_writer.mem_usage(); let mem_usage = segment_writer.mem_usage();
if mem_usage >= memory_budget - MARGIN_IN_BYTES { if mem_usage >= memory_budget - MARGIN_IN_BYTES {
info!( info!(
"Buffer limit reached, flushing segment with maxdoc={}.", "Buffer limit reached, flushing segment with maxdoc={}.",
@@ -305,7 +301,7 @@ fn index_documents(
let last_docstamp: u64 = *(doc_opstamps.last().unwrap()); let last_docstamp: u64 = *(doc_opstamps.last().unwrap());
let delete_bitset_opt = if delete_cursor.get().is_some() { let segment_entry: SegmentEntry = if delete_cursor.get().is_some() {
let doc_to_opstamps = DocToOpstampMapping::from(doc_opstamps); let doc_to_opstamps = DocToOpstampMapping::from(doc_opstamps);
let segment_reader = SegmentReader::open(segment)?; let segment_reader = SegmentReader::open(segment)?;
let mut deleted_bitset = BitSet::with_capacity(num_docs as usize); let mut deleted_bitset = BitSet::with_capacity(num_docs as usize);
@@ -316,22 +312,18 @@ fn index_documents(
&doc_to_opstamps, &doc_to_opstamps,
last_docstamp, last_docstamp,
)?; )?;
if may_have_deletes { SegmentEntry::new(segment_meta, delete_cursor, {
Some(deleted_bitset) if may_have_deletes {
} else { Some(deleted_bitset)
None } else {
} None
}
})
} else { } else {
// if there are no delete operation in the queue, no need // if there are no delete operation in the queue, no need
// to even open the segment. // to even open the segment.
None SegmentEntry::new(segment_meta, delete_cursor, None)
}; };
let segment_entry = SegmentEntry::new(
segment_meta,
delete_cursor,
delete_bitset_opt,
last_docstamp,
);
Ok(segment_updater.add_segment(generation, segment_entry)) Ok(segment_updater.add_segment(generation, segment_entry))
} }
@@ -340,7 +332,7 @@ impl IndexWriter {
pub fn wait_merging_threads(mut self) -> Result<()> { pub fn wait_merging_threads(mut self) -> Result<()> {
// this will stop the indexing thread, // this will stop the indexing thread,
// dropping the last reference to the segment_updater. // dropping the last reference to the segment_updater.
drop(self.operation_sender); drop(self.document_sender);
let former_workers_handles = mem::replace(&mut self.workers_join_handle, vec![]); let former_workers_handles = mem::replace(&mut self.workers_join_handle, vec![]);
for join_handle in former_workers_handles { for join_handle in former_workers_handles {
@@ -366,30 +358,27 @@ impl IndexWriter {
} }
#[doc(hidden)] #[doc(hidden)]
pub fn add_segment(&mut self, segment_meta: SegmentMeta, opstamp: u64) { pub fn add_segment(&mut self, segment_meta: SegmentMeta) {
let delete_cursor = self.delete_queue.cursor(); let delete_cursor = self.delete_queue.cursor();
let segment_entry = SegmentEntry::new(segment_meta, delete_cursor, None, opstamp); let segment_entry = SegmentEntry::new(segment_meta, delete_cursor, None);
self.segment_updater self.segment_updater
.add_segment(self.generation, segment_entry); .add_segment(self.generation, segment_entry);
} }
/// Creates a new segment. /// *Experimental & Advanced API* Creates a new segment.
/// and marks it as currently in write.
/// ///
/// This method is useful only for users trying to do complex /// This method is useful only for users trying to do complex
/// operations, like converting an index format to another. /// operations, like converting an index format to another.
///
/// It is safe to start writing file associated to the new `Segment`.
/// These will not be garbage collected as long as an instance object of
/// `SegmentMeta` object associated to the new `Segment` is "alive".
pub fn new_segment(&self) -> Segment { pub fn new_segment(&self) -> Segment {
self.index.new_segment() self.segment_updater.new_segment()
} }
/// Spawns a new worker thread for indexing. /// Spawns a new worker thread for indexing.
/// The thread consumes documents from the pipeline. /// The thread consumes documents from the pipeline.
/// ///
fn add_indexing_worker(&mut self) -> Result<()> { fn add_indexing_worker(&mut self) -> Result<()> {
let document_receiver_clone = self.operation_receiver.clone(); let document_receiver_clone = self.document_receiver.clone();
let mut segment_updater = self.segment_updater.clone(); let mut segment_updater = self.segment_updater.clone();
let generation = self.generation; let generation = self.generation;
@@ -397,16 +386,13 @@ impl IndexWriter {
let mut delete_cursor = self.delete_queue.cursor(); let mut delete_cursor = self.delete_queue.cursor();
let mem_budget = self.heap_size_in_bytes_per_thread; let mem_budget = self.heap_size_in_bytes_per_thread;
let index = self.index.clone();
let join_handle: JoinHandle<Result<()>> = thread::Builder::new() let join_handle: JoinHandle<Result<()>> = thread::Builder::new()
.name(format!( .name(format!(
"thrd-tantivy-index{}-gen{}", "indexing thread {} for gen {}",
self.worker_id, generation self.worker_id, generation
)) )).spawn(move || {
.spawn(move || {
loop { loop {
let mut document_iterator = let mut document_iterator = document_receiver_clone.clone().peekable();
document_receiver_clone.clone().into_iter().peekable();
// the peeking here is to avoid // the peeking here is to avoid
// creating a new segment's files // creating a new segment's files
@@ -415,19 +401,15 @@ impl IndexWriter {
// this is a valid guarantee as the // this is a valid guarantee as the
// peeked document now belongs to // peeked document now belongs to
// our local iterator. // our local iterator.
if let Some(operations) = document_iterator.peek() { if let Some(operation) = document_iterator.peek() {
if let Some(first) = operations.first() { delete_cursor.skip_to(operation.opstamp);
delete_cursor.skip_to(first.opstamp);
} else {
return Ok(());
}
} else { } else {
// No more documents. // No more documents.
// Happens when there is a commit, or if the `IndexWriter` // Happens when there is a commit, or if the `IndexWriter`
// was dropped. // was dropped.
return Ok(()); return Ok(());
} }
let segment = index.new_segment(); let segment = segment_updater.new_segment();
index_documents( index_documents(
mem_budget, mem_budget,
&segment, &segment,
@@ -444,7 +426,7 @@ impl IndexWriter {
} }
/// Accessor to the merge policy. /// Accessor to the merge policy.
pub fn get_merge_policy(&self) -> Arc<Box<MergePolicy>> { pub fn get_merge_policy(&self) -> Box<MergePolicy> {
self.segment_updater.get_merge_policy() self.segment_updater.get_merge_policy()
} }
@@ -469,10 +451,7 @@ impl IndexWriter {
/// Merges a given list of segments /// Merges a given list of segments
/// ///
/// `segment_ids` is required to be non-empty. /// `segment_ids` is required to be non-empty.
pub fn merge( pub fn merge(&mut self, segment_ids: &[SegmentId]) -> Result<Receiver<SegmentMeta>> {
&mut self,
segment_ids: &[SegmentId],
) -> Result<impl Future<Item = SegmentMeta, Error = Canceled>> {
self.segment_updater.start_merge(segment_ids) self.segment_updater.start_merge(segment_ids)
} }
@@ -484,11 +463,14 @@ impl IndexWriter {
/// when no documents are remaining. /// when no documents are remaining.
/// ///
/// Returns the former segment_ready channel. /// Returns the former segment_ready channel.
fn recreate_document_channel(&mut self) -> OperationReceiver { fn recreate_document_channel(&mut self) -> DocumentReceiver {
let (document_sender, document_receiver): (OperationSender, OperationReceiver) = let (mut document_sender, mut document_receiver): (
channel::bounded(PIPELINE_MAX_SIZE_IN_DOCS); DocumentSender,
mem::replace(&mut self.operation_sender, document_sender); DocumentReceiver,
mem::replace(&mut self.operation_receiver, document_receiver) ) = channel::bounded(PIPELINE_MAX_SIZE_IN_DOCS);
swap(&mut self.document_sender, &mut document_sender);
swap(&mut self.document_receiver, &mut document_receiver);
document_receiver
} }
/// Rollback to the last commit /// Rollback to the last commit
@@ -506,7 +488,7 @@ impl IndexWriter {
// segment updates will be ignored. // segment updates will be ignored.
self.segment_updater.kill(); self.segment_updater.kill();
let document_receiver = self.operation_receiver.clone(); let document_receiver = self.document_receiver.clone();
// take the directory lock to create a new index_writer. // take the directory lock to create a new index_writer.
let directory_lock = self let directory_lock = self
@@ -532,7 +514,7 @@ impl IndexWriter {
// //
// This will reach an end as the only document_sender // This will reach an end as the only document_sender
// was dropped with the index_writer. // was dropped with the index_writer.
for _ in document_receiver.iter() {} for _ in document_receiver.clone() {}
Ok(()) Ok(())
} }
@@ -559,16 +541,6 @@ impl IndexWriter {
/// using this API. /// using this API.
/// See [`PreparedCommit::set_payload()`](PreparedCommit.html) /// See [`PreparedCommit::set_payload()`](PreparedCommit.html)
pub fn prepare_commit(&mut self) -> Result<PreparedCommit> { pub fn prepare_commit(&mut self) -> Result<PreparedCommit> {
info!("Preparing commit");
self.prepare_commit_internal(false)
}
pub fn prepare_commit_soft(&mut self) -> Result<PreparedCommit> {
info!("Preparing soft commit");
self.prepare_commit_internal(true)
}
pub(crate) fn prepare_commit_internal(&mut self, soft: bool) -> Result<PreparedCommit> {
// Here, because we join all of the worker threads, // Here, because we join all of the worker threads,
// all of the segment update for this commit have been // all of the segment update for this commit have been
// sent. // sent.
@@ -585,19 +557,24 @@ impl IndexWriter {
// and recreate a new one channels. // and recreate a new one channels.
self.recreate_document_channel(); self.recreate_document_channel();
let former_workers_join_handle = mem::replace(&mut self.workers_join_handle, Vec::new()); let mut former_workers_join_handle = Vec::new();
swap(
&mut former_workers_join_handle,
&mut self.workers_join_handle,
);
for worker_handle in former_workers_join_handle { for worker_handle in former_workers_join_handle {
let indexing_worker_result = worker_handle let indexing_worker_result = worker_handle
.join() .join()
.map_err(|e| TantivyError::ErrorInThread(format!("{:?}", e)))?; .map_err(|e| TantivyError::ErrorInThread(format!("{:?}", e)))?;
// add a new worker for the next generation, whether the worker failed or not.
self.add_indexing_worker()?;
indexing_worker_result?; indexing_worker_result?;
// add a new worker for the next generation.
self.add_indexing_worker()?;
} }
let commit_opstamp = self.stamper.stamp(); let commit_opstamp = self.stamper.stamp();
let prepared_commit = PreparedCommit::new(self, commit_opstamp, soft); let prepared_commit = PreparedCommit::new(self, commit_opstamp);
info!("Prepared commit {}", commit_opstamp); info!("Prepared commit {}", commit_opstamp);
Ok(prepared_commit) Ok(prepared_commit)
} }
@@ -620,11 +597,6 @@ impl IndexWriter {
self.prepare_commit()?.commit() self.prepare_commit()?.commit()
} }
pub fn soft_commit(&mut self) -> Result<u64> {
self.prepare_commit_soft()?.commit()
}
pub(crate) fn segment_updater(&self) -> &SegmentUpdater { pub(crate) fn segment_updater(&self) -> &SegmentUpdater {
&self.segment_updater &self.segment_updater
} }
@@ -668,189 +640,42 @@ impl IndexWriter {
pub fn add_document(&mut self, document: Document) -> u64 { pub fn add_document(&mut self, document: Document) -> u64 {
let opstamp = self.stamper.stamp(); let opstamp = self.stamper.stamp();
let add_operation = AddOperation { opstamp, document }; let add_operation = AddOperation { opstamp, document };
let send_result = self.operation_sender.send(vec![add_operation]); self.document_sender.send(add_operation);
if let Err(e) = send_result {
panic!("Failed to index document. Sending to indexing channel failed. This probably means all of the indexing threads have panicked. {:?}", e);
}
opstamp opstamp
} }
/// Gets a range of stamps from the stamper and "pops" the last stamp
/// from the range returning a tuple of the last optstamp and the popped
/// range.
///
/// The total number of stamps generated by this method is `count + 1`;
/// each operation gets a stamp from the `stamps` iterator and `last_opstamp`
/// is for the batch itself.
fn get_batch_opstamps(&mut self, count: u64) -> (u64, Range<u64>) {
let Range { start, end } = self.stamper.stamps(count + 1u64);
let last_opstamp = end - 1;
let stamps = Range {
start,
end: last_opstamp,
};
(last_opstamp, stamps)
}
/// Runs a group of document operations ensuring that the operations are
/// assigned contigous u64 opstamps and that add operations of the same
/// group are flushed into the same segment.
///
/// If the indexing pipeline is full, this call may block.
///
/// Each operation of the given `user_operations` will receive an in-order,
/// contiguous u64 opstamp. The entire batch itself is also given an
/// opstamp that is 1 greater than the last given operation. This
/// `batch_opstamp` is the return value of `run`. An empty group of
/// `user_operations`, an empty `Vec<UserOperation>`, still receives
/// a valid opstamp even though no changes were _actually_ made to the index.
///
/// Like adds and deletes (see `IndexWriter.add_document` and
/// `IndexWriter.delete_term`), the changes made by calling `run` will be
/// visible to readers only after calling `commit()`.
pub fn run(&mut self, user_operations: Vec<UserOperation>) -> u64 {
let count = user_operations.len() as u64;
if count == 0 {
return self.stamper.stamp();
}
let (batch_opstamp, stamps) = self.get_batch_opstamps(count);
let mut adds: Vec<AddOperation> = Vec::new();
for (user_op, opstamp) in user_operations.into_iter().zip(stamps) {
match user_op {
UserOperation::Delete(term) => {
let delete_operation = DeleteOperation { opstamp, term };
self.delete_queue.push(delete_operation);
}
UserOperation::Add(document) => {
let add_operation = AddOperation { opstamp, document };
adds.push(add_operation);
}
}
}
let send_result = self.operation_sender.send(adds);
if let Err(e) = send_result {
panic!("Failed to index document. Sending to indexing channel failed. This probably means all of the indexing threads have panicked. {:?}", e);
};
batch_opstamp
}
} }
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::super::operation::UserOperation;
use super::initial_table_size; use super::initial_table_size;
use collector::TopDocs;
use directory::error::LockError;
use error::*; use error::*;
use indexer::NoMergePolicy; use indexer::NoMergePolicy;
use query::TermQuery; use schema::{self, Document};
use schema::{self, IndexRecordOption};
use Index; use Index;
use ReloadPolicy;
use Term; use Term;
use IndexReader;
#[test]
fn test_operations_group() {
// an operations group with 2 items should cause 3 opstamps 0, 1, and 2.
let mut schema_builder = schema::Schema::builder();
let text_field = schema_builder.add_text_field("text", schema::TEXT);
let index = Index::create_in_ram(schema_builder.build());
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
let operations = vec![
UserOperation::Add(doc!(text_field=>"a")),
UserOperation::Add(doc!(text_field=>"b")),
];
let batch_opstamp1 = index_writer.run(operations);
assert_eq!(batch_opstamp1, 2u64);
}
#[test]
fn test_ordered_batched_operations() {
// * one delete for `doc!(field=>"a")`
// * one add for `doc!(field=>"a")`
// * one add for `doc!(field=>"b")`
// * one delete for `doc!(field=>"b")`
// after commit there is one doc with "a" and 0 doc with "b"
let mut schema_builder = schema::Schema::builder();
let text_field = schema_builder.add_text_field("text", schema::TEXT);
let index = Index::create_in_ram(schema_builder.build());
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::Manual)
.try_into()
.unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
let a_term = Term::from_field_text(text_field, "a");
let b_term = Term::from_field_text(text_field, "b");
let operations = vec![
UserOperation::Delete(a_term),
UserOperation::Add(doc!(text_field=>"a")),
UserOperation::Add(doc!(text_field=>"b")),
UserOperation::Delete(b_term),
];
index_writer.run(operations);
index_writer.commit().expect("failed to commit");
reader.reload().expect("failed to load searchers");
let a_term = Term::from_field_text(text_field, "a");
let b_term = Term::from_field_text(text_field, "b");
let a_query = TermQuery::new(a_term, IndexRecordOption::Basic);
let b_query = TermQuery::new(b_term, IndexRecordOption::Basic);
let searcher = reader.searcher();
let a_docs = searcher
.search(&a_query, &TopDocs::with_limit(1))
.expect("search for a failed");
let b_docs = searcher
.search(&b_query, &TopDocs::with_limit(1))
.expect("search for b failed");
assert_eq!(a_docs.len(), 1);
assert_eq!(b_docs.len(), 0);
}
#[test]
fn test_empty_operations_group() {
let schema_builder = schema::Schema::builder();
let index = Index::create_in_ram(schema_builder.build());
let mut index_writer = index.writer(3_000_000).unwrap();
let operations1 = vec![];
let batch_opstamp1 = index_writer.run(operations1);
assert_eq!(batch_opstamp1, 0u64);
let operations2 = vec![];
let batch_opstamp2 = index_writer.run(operations2);
assert_eq!(batch_opstamp2, 1u64);
}
#[test] #[test]
fn test_lockfile_stops_duplicates() { fn test_lockfile_stops_duplicates() {
let schema_builder = schema::Schema::builder(); let schema_builder = schema::SchemaBuilder::default();
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
let _index_writer = index.writer(3_000_000).unwrap(); let _index_writer = index.writer(40_000_000).unwrap();
match index.writer(3_000_000) { match index.writer(40_000_000) {
Err(TantivyError::LockFailure(LockError::LockBusy, _)) => {} Err(TantivyError::LockFailure(_)) => {}
_ => panic!("Expected a `LockFailure` error"), _ => panic!("Expected FileAlreadyExists error"),
} }
} }
#[test] #[test]
fn test_lockfile_already_exists_error_msg() { fn test_lockfile_already_exists_error_msg() {
let schema_builder = schema::Schema::builder(); let schema_builder = schema::SchemaBuilder::default();
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
let _index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let _index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
match index.writer_with_num_threads(1, 3_000_000) { match index.writer_with_num_threads(1, 3_000_000) {
Err(err) => { Err(err) => {
let err_msg = err.to_string(); let err_msg = err.to_string();
assert!(err_msg.contains("already an `IndexWriter`")); assert!(err_msg.contains("Lockfile"));
assert!(err_msg.contains("Possible causes:"))
} }
_ => panic!("Expected LockfileAlreadyExists error"), _ => panic!("Expected LockfileAlreadyExists error"),
} }
@@ -858,9 +683,9 @@ mod tests {
#[test] #[test]
fn test_set_merge_policy() { fn test_set_merge_policy() {
let schema_builder = schema::Schema::builder(); let schema_builder = schema::SchemaBuilder::default();
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
let index_writer = index.writer(3_000_000).unwrap(); let index_writer = index.writer(40_000_000).unwrap();
assert_eq!( assert_eq!(
format!("{:?}", index_writer.get_merge_policy()), format!("{:?}", index_writer.get_merge_policy()),
"LogMergePolicy { min_merge_size: 8, min_layer_size: 10000, \ "LogMergePolicy { min_merge_size: 8, min_layer_size: 10000, \
@@ -876,133 +701,90 @@ mod tests {
#[test] #[test]
fn test_lockfile_released_on_drop() { fn test_lockfile_released_on_drop() {
let schema_builder = schema::Schema::builder(); let schema_builder = schema::SchemaBuilder::default();
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
{ {
let _index_writer = index.writer(3_000_000).unwrap(); let _index_writer = index.writer(40_000_000).unwrap();
// the lock should be released when the // the lock should be released when the
// index_writer leaves the scope. // index_writer leaves the scope.
} }
let _index_writer_two = index.writer(3_000_000).unwrap(); let _index_writer_two = index.writer(40_000_000).unwrap();
}
fn num_docs_containing_text(reader: &IndexReader, term: &str) -> u64 {
let searcher = reader.searcher();
let text_field = reader.schema().get_field("text").unwrap();
let term = Term::from_field_text(text_field, term);
searcher.doc_freq(&term)
} }
#[test] #[test]
fn test_commit_and_rollback() { fn test_commit_and_rollback() {
let mut schema_builder = schema::Schema::builder(); let mut schema_builder = schema::SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", schema::TEXT); let text_field = schema_builder.add_text_field("text", schema::TEXT);
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::Manual)
.try_into()
.unwrap();
let num_docs_containing = |s: &str| { let num_docs_containing = |s: &str| {
let searcher = reader.searcher(); let searcher = index.searcher();
let term = Term::from_field_text(text_field, s); let term = Term::from_field_text(text_field, s);
searcher.doc_freq(&term) searcher.doc_freq(&term)
}; };
let mut index_writer = index.writer(3_000_000).unwrap();
assert_eq!(index_writer.commit_opstamp(), 0u64);
assert_eq!(num_docs_containing_text(&reader, "a"), 0);
{ {
// writing the segment // writing the segment
let mut index_writer = index.writer(3_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"a")); index_writer.add_document(doc!(text_field=>"a"));
index_writer.rollback().unwrap(); index_writer.rollback().unwrap();
assert_eq!(index_writer.commit_opstamp(), 0u64); assert_eq!(index_writer.commit_opstamp(), 0u64);
assert_eq!(num_docs_containing("a"), 0); assert_eq!(num_docs_containing("a"), 0);
{ {
index_writer.add_document(doc!(text_field=>"b")); index_writer.add_document(doc!(text_field=>"b"));
index_writer.add_document(doc!(text_field=>"c")); index_writer.add_document(doc!(text_field=>"c"));
} }
assert!(index_writer.commit().is_ok()); assert_eq!(index_writer.commit().unwrap(), 2u64);
reader.reload().unwrap(); index.load_searchers().unwrap();
assert_eq!(num_docs_containing("a"), 0); assert_eq!(num_docs_containing("a"), 0);
assert_eq!(num_docs_containing("b"), 1); assert_eq!(num_docs_containing("b"), 1);
assert_eq!(num_docs_containing("c"), 1); assert_eq!(num_docs_containing("c"), 1);
} }
reader.reload().unwrap(); index.load_searchers().unwrap();
reader.searcher(); index.searcher();
}
#[test]
fn test_softcommit_and_rollback() {
let mut schema_builder = schema::Schema::builder();
let text_field = schema_builder.add_text_field("text", schema::TEXT);
let index = Index::create_in_ram(schema_builder.build());
let reader = index.reader().unwrap();
// writing the segment
let mut index_writer = index.writer(3_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"a"));
index_writer.rollback().unwrap();
assert_eq!(index_writer.commit_opstamp(), 0u64);
assert_eq!(num_docs_containing_text(&reader, "a"), 0u64);
{
index_writer.add_document(doc!(text_field=>"b"));
index_writer.add_document(doc!(text_field=>"c"));
}
assert!(index_writer.soft_commit().is_ok());
reader.reload().unwrap(); // we need to load soft committed stuff.
assert_eq!(num_docs_containing_text(&reader, "a"), 0u64);
assert_eq!(num_docs_containing_text(&reader, "b"), 1u64);
assert_eq!(num_docs_containing_text(&reader, "c"), 1u64);
index_writer.rollback().unwrap();
reader.reload().unwrap();
assert_eq!(num_docs_containing_text(&reader, "a"), 0u64);
assert_eq!(num_docs_containing_text(&reader, "b"), 0u64);
assert_eq!(num_docs_containing_text(&reader, "c"), 0u64);
} }
#[test] #[test]
fn test_with_merges() { fn test_with_merges() {
let mut schema_builder = schema::Schema::builder(); let mut schema_builder = schema::SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", schema::TEXT); let text_field = schema_builder.add_text_field("text", schema::TEXT);
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::Manual)
.try_into()
.unwrap();
let num_docs_containing = |s: &str| { let num_docs_containing = |s: &str| {
let searcher = index.searcher();
let term_a = Term::from_field_text(text_field, s); let term_a = Term::from_field_text(text_field, s);
reader.searcher().doc_freq(&term_a) searcher.doc_freq(&term_a)
}; };
{ {
// writing the segment // writing the segment
let mut index_writer = index.writer(12_000_000).unwrap(); let mut index_writer = index.writer(12_000_000).unwrap();
// create 8 segments with 100 tiny docs // create 8 segments with 100 tiny docs
for _doc in 0..100 { for _doc in 0..100 {
index_writer.add_document(doc!(text_field=>"a")); let mut doc = Document::default();
doc.add_text(text_field, "a");
index_writer.add_document(doc);
} }
index_writer.commit().expect("commit failed"); index_writer.commit().expect("commit failed");
for _doc in 0..100 { for _doc in 0..100 {
index_writer.add_document(doc!(text_field=>"a")); let mut doc = Document::default();
doc.add_text(text_field, "a");
index_writer.add_document(doc);
} }
// this should create 8 segments and trigger a merge. // this should create 8 segments and trigger a merge.
index_writer.commit().expect("commit failed"); index_writer.commit().expect("commit failed");
index_writer index_writer
.wait_merging_threads() .wait_merging_threads()
.expect("waiting merging thread failed"); .expect("waiting merging thread failed");
index.load_searchers().unwrap();
reader.reload().unwrap(); assert_eq!(num_docs_containing("a"), 200);
assert_eq!(num_docs_containing_text(&reader, "a"), 200);
assert!(index.searchable_segments().unwrap().len() < 8); assert!(index.searchable_segments().unwrap().len() < 8);
} }
} }
#[test] #[test]
fn test_prepare_with_commit_message() { fn test_prepare_with_commit_message() {
let mut schema_builder = schema::Schema::builder(); let mut schema_builder = schema::SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", schema::TEXT); let text_field = schema_builder.add_text_field("text", schema::TEXT);
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
@@ -1016,6 +798,7 @@ mod tests {
{ {
let mut prepared_commit = index_writer.prepare_commit().expect("commit failed"); let mut prepared_commit = index_writer.prepare_commit().expect("commit failed");
prepared_commit.set_payload("first commit"); prepared_commit.set_payload("first commit");
assert_eq!(prepared_commit.opstamp(), 100);
prepared_commit.commit().expect("commit failed"); prepared_commit.commit().expect("commit failed");
} }
{ {
@@ -1035,10 +818,10 @@ mod tests {
#[test] #[test]
fn test_prepare_but_rollback() { fn test_prepare_but_rollback() {
let mut schema_builder = schema::Schema::builder(); let mut schema_builder = schema::SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", schema::TEXT); let text_field = schema_builder.add_text_field("text", schema::TEXT);
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
let reader = index.reader();
{ {
// writing the segment // writing the segment
let mut index_writer = index.writer_with_num_threads(4, 12_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(4, 12_000_000).unwrap();
@@ -1049,6 +832,7 @@ mod tests {
{ {
let mut prepared_commit = index_writer.prepare_commit().expect("commit failed"); let mut prepared_commit = index_writer.prepare_commit().expect("commit failed");
prepared_commit.set_payload("first commit"); prepared_commit.set_payload("first commit");
assert_eq!(prepared_commit.opstamp(), 100);
prepared_commit.abort().expect("commit failed"); prepared_commit.abort().expect("commit failed");
} }
{ {
@@ -1060,15 +844,11 @@ mod tests {
} }
index_writer.commit().unwrap(); index_writer.commit().unwrap();
} }
index.load_searchers().unwrap();
let num_docs_containing = |s: &str| { let num_docs_containing = |s: &str| {
let searcher = index.searcher();
let term_a = Term::from_field_text(text_field, s); let term_a = Term::from_field_text(text_field, s);
index searcher.doc_freq(&term_a)
.reader_builder()
.reload_policy(ReloadPolicy::Manual)
.try_into()
.unwrap()
.searcher()
.doc_freq(&term_a)
}; };
assert_eq!(num_docs_containing("a"), 0); assert_eq!(num_docs_containing("a"), 0);
assert_eq!(num_docs_containing("b"), 100); assert_eq!(num_docs_containing("b"), 100);
@@ -1076,9 +856,9 @@ mod tests {
#[test] #[test]
fn test_hashmap_size() { fn test_hashmap_size() {
assert_eq!(initial_table_size(100_000), 11); assert_eq!(initial_table_size(100_000), 12);
assert_eq!(initial_table_size(1_000_000), 14); assert_eq!(initial_table_size(1_000_000), 15);
assert_eq!(initial_table_size(10_000_000), 17); assert_eq!(initial_table_size(10_000_000), 18);
assert_eq!(initial_table_size(1_000_000_000), 19); assert_eq!(initial_table_size(1_000_000_000), 19);
} }
@@ -1086,7 +866,7 @@ mod tests {
#[test] #[test]
fn test_write_commit_fails() { fn test_write_commit_fails() {
use fail; use fail;
let mut schema_builder = schema::Schema::builder(); let mut schema_builder = schema::SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", schema::TEXT); let text_field = schema_builder.add_text_field("text", schema::TEXT);
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
@@ -1100,9 +880,11 @@ mod tests {
index_writer.add_document(doc!(text_field => "b")); index_writer.add_document(doc!(text_field => "b"));
} }
assert!(index_writer.commit().is_err()); assert!(index_writer.commit().is_err());
index.load_searchers().unwrap();
let num_docs_containing = |s: &str| { let num_docs_containing = |s: &str| {
let searcher = index.searcher();
let term_a = Term::from_field_text(text_field, s); let term_a = Term::from_field_text(text_field, s);
index.reader().unwrap().searcher().doc_freq(&term_a) searcher.doc_freq(&term_a)
}; };
assert_eq!(num_docs_containing("a"), 100); assert_eq!(num_docs_containing("a"), 100);
assert_eq!(num_docs_containing("b"), 0); assert_eq!(num_docs_containing("b"), 0);

64
src/indexer/merge_operation.rs
View File

@@ -1,64 +0,0 @@
use census::{Inventory, TrackedObject};
use std::collections::HashSet;
use SegmentId;
#[derive(Default)]
pub struct MergeOperationInventory(Inventory<InnerMergeOperation>);
impl MergeOperationInventory {
pub fn segment_in_merge(&self) -> HashSet<SegmentId> {
let mut segment_in_merge = HashSet::default();
for merge_op in self.0.list() {
for &segment_id in &merge_op.segment_ids {
segment_in_merge.insert(segment_id);
}
}
segment_in_merge
}
}
/// A `MergeOperation` has two role.
/// It carries all of the information required to describe a merge :
/// - `target_opstamp` is the opstamp up to which we want to consume the
/// delete queue and reflect their deletes.
/// - `segment_ids` is the list of segment to be merged.
///
/// The second role is to ensure keep track of the fact that these
/// segments are in merge and avoid starting a merge operation that
/// may conflict with this one.
///
/// This works by tracking merge operations. When considering computing
/// merge candidates, we simply list tracked merge operations and remove
/// their segments from possible merge candidates.
pub struct MergeOperation {
inner: TrackedObject<InnerMergeOperation>,
}
struct InnerMergeOperation {
target_opstamp: u64,
segment_ids: Vec<SegmentId>,
}
impl MergeOperation {
pub fn new(
inventory: &MergeOperationInventory,
target_opstamp: u64,
segment_ids: Vec<SegmentId>,
) -> MergeOperation {
let inner_merge_operation = InnerMergeOperation {
target_opstamp,
segment_ids,
};
MergeOperation {
inner: inventory.0.track(inner_merge_operation),
}
}
pub fn target_opstamp(&self) -> u64 {
self.inner.target_opstamp
}
pub fn segment_ids(&self) -> &[SegmentId] {
&self.inner.segment_ids[..]
}
}

17
src/indexer/merge_policy.rs
View File

@@ -11,7 +11,7 @@ pub struct MergeCandidate(pub Vec<SegmentId>);
/// ///
/// Every time a the list of segments changes, the segment updater /// Every time a the list of segments changes, the segment updater
/// asks the merge policy if some segments should be merged. /// asks the merge policy if some segments should be merged.
pub trait MergePolicy: marker::Send + marker::Sync + Debug { pub trait MergePolicy: MergePolicyClone + marker::Send + marker::Sync + Debug {
/// Given the list of segment metas, returns the list of merge candidates. /// Given the list of segment metas, returns the list of merge candidates.
/// ///
/// This call happens on the segment updater thread, and will block /// This call happens on the segment updater thread, and will block
@@ -19,6 +19,21 @@ pub trait MergePolicy: marker::Send + marker::Sync + Debug {
fn compute_merge_candidates(&self, segments: &[SegmentMeta]) -> Vec<MergeCandidate>; fn compute_merge_candidates(&self, segments: &[SegmentMeta]) -> Vec<MergeCandidate>;
} }
/// MergePolicyClone
pub trait MergePolicyClone {
/// Returns a boxed clone of the MergePolicy.
fn box_clone(&self) -> Box<MergePolicy>;
}
impl<T> MergePolicyClone for T
where
T: 'static + MergePolicy + Clone,
{
fn box_clone(&self) -> Box<MergePolicy> {
Box::new(self.clone())
}
}
/// Never merge segments. /// Never merge segments.
#[derive(Debug, Clone)] #[derive(Debug, Clone)]
pub struct NoMergePolicy; pub struct NoMergePolicy;

397
src/indexer/merger.rs
View File

@@ -1,4 +1,3 @@
use common::MAX_DOC_LIMIT;
use core::Segment; use core::Segment;
use core::SegmentReader; use core::SegmentReader;
use core::SerializableSegment; use core::SerializableSegment;
@@ -24,7 +23,6 @@ use termdict::TermMerger;
use termdict::TermOrdinal; use termdict::TermOrdinal;
use DocId; use DocId;
use Result; use Result;
use TantivyError;
fn compute_total_num_tokens(readers: &[SegmentReader], field: Field) -> u64 { fn compute_total_num_tokens(readers: &[SegmentReader], field: Field) -> u64 {
let mut total_tokens = 0u64; let mut total_tokens = 0u64;
@@ -42,15 +40,13 @@ fn compute_total_num_tokens(readers: &[SegmentReader], field: Field) -> u64 {
total_tokens += reader.inverted_index(field).total_num_tokens(); total_tokens += reader.inverted_index(field).total_num_tokens();
} }
} }
total_tokens total_tokens + count
+ count .iter()
.iter() .cloned()
.cloned() .enumerate()
.enumerate() .map(|(fieldnorm_ord, count)| {
.map(|(fieldnorm_ord, count)| { count as u64 * u64::from(FieldNormReader::id_to_fieldnorm(fieldnorm_ord as u8))
count as u64 * u64::from(FieldNormReader::id_to_fieldnorm(fieldnorm_ord as u8)) }).sum::<u64>()
})
.sum::<u64>()
} }
pub struct IndexMerger { pub struct IndexMerger {
@@ -152,14 +148,6 @@ impl IndexMerger {
readers.push(reader); readers.push(reader);
} }
} }
if max_doc >= MAX_DOC_LIMIT {
let err_msg = format!(
"The segment resulting from this merge would have {} docs,\
which exceeds the limit {}.",
max_doc, MAX_DOC_LIMIT
);
return Err(TantivyError::InvalidArgument(err_msg));
}
Ok(IndexMerger { Ok(IndexMerger {
schema, schema,
readers, readers,
@@ -204,17 +192,17 @@ impl IndexMerger {
fast_field_serializer, fast_field_serializer,
)?; )?;
} }
FieldType::U64(ref options) FieldType::U64(ref options) | FieldType::I64(ref options) => {
| FieldType::I64(ref options) match options.get_fastfield_cardinality() {
| FieldType::Date(ref options) => match options.get_fastfield_cardinality() { Some(Cardinality::SingleValue) => {
Some(Cardinality::SingleValue) => { self.write_single_fast_field(field, fast_field_serializer)?;
self.write_single_fast_field(field, fast_field_serializer)?; }
Some(Cardinality::MultiValues) => {
self.write_multi_fast_field(field, fast_field_serializer)?;
}
None => {}
} }
Some(Cardinality::MultiValues) => { }
self.write_multi_fast_field(field, fast_field_serializer)?;
}
None => {}
},
FieldType::Str(_) => { FieldType::Str(_) => {
// We don't handle str fast field for the moment // We don't handle str fast field for the moment
// They can be implemented using what is done // They can be implemented using what is done
@@ -535,8 +523,7 @@ impl IndexMerger {
} }
} }
None None
}) }).collect();
.collect();
// At this point, `segment_postings` contains the posting list // At this point, `segment_postings` contains the posting list
// of all of the segments containing the given term. // of all of the segments containing the given term.
@@ -627,7 +614,7 @@ impl IndexMerger {
store_writer.store(&doc)?; store_writer.store(&doc)?;
} }
} else { } else {
store_writer.stack(&store_reader)?; store_writer.stack(store_reader)?;
} }
} }
Ok(()) Ok(())
@@ -648,9 +635,10 @@ impl SerializableSegment for IndexMerger {
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use byteorder::{BigEndian, ReadBytesExt, WriteBytesExt}; use byteorder::{BigEndian, ReadBytesExt, WriteBytesExt};
use collector::chain;
use collector::tests::TestCollector; use collector::tests::TestCollector;
use collector::tests::{BytesFastFieldTestCollector, FastFieldTestCollector}; use collector::tests::{BytesFastFieldTestCollector, FastFieldTestCollector};
use collector::{Count, FacetCollector}; use collector::FacetCollector;
use core::Index; use core::Index;
use futures::Future; use futures::Future;
use query::AllQuery; use query::AllQuery;
@@ -659,12 +647,10 @@ mod tests {
use schema; use schema;
use schema::Cardinality; use schema::Cardinality;
use schema::Document; use schema::Document;
use schema::Facet;
use schema::IndexRecordOption; use schema::IndexRecordOption;
use schema::IntOptions; use schema::IntOptions;
use schema::Term; use schema::Term;
use schema::TextFieldIndexing; use schema::TextFieldIndexing;
use schema::INDEXED;
use std::io::Cursor; use std::io::Cursor;
use DocAddress; use DocAddress;
use IndexWriter; use IndexWriter;
@@ -672,22 +658,19 @@ mod tests {
#[test] #[test]
fn test_index_merger_no_deletes() { fn test_index_merger_no_deletes() {
let mut schema_builder = schema::Schema::builder(); let mut schema_builder = schema::SchemaBuilder::default();
let text_fieldtype = schema::TextOptions::default() let text_fieldtype = schema::TextOptions::default()
.set_indexing_options( .set_indexing_options(
TextFieldIndexing::default() TextFieldIndexing::default()
.set_tokenizer("default") .set_tokenizer("default")
.set_index_option(IndexRecordOption::WithFreqs), .set_index_option(IndexRecordOption::WithFreqs),
) ).set_stored();
.set_stored();
let text_field = schema_builder.add_text_field("text", text_fieldtype); let text_field = schema_builder.add_text_field("text", text_fieldtype);
let date_field = schema_builder.add_date_field("date", INDEXED);
let score_fieldtype = schema::IntOptions::default().set_fast(Cardinality::SingleValue); let score_fieldtype = schema::IntOptions::default().set_fast(Cardinality::SingleValue);
let score_field = schema_builder.add_u64_field("score", score_fieldtype); let score_field = schema_builder.add_u64_field("score", score_fieldtype);
let bytes_score_field = schema_builder.add_bytes_field("score_bytes"); let bytes_score_field = schema_builder.add_bytes_field("score_bytes");
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
let reader = index.reader().unwrap();
let curr_time = chrono::Utc::now();
let add_score_bytes = |doc: &mut Document, score: u32| { let add_score_bytes = |doc: &mut Document, score: u32| {
let mut bytes = Vec::new(); let mut bytes = Vec::new();
bytes bytes
@@ -704,7 +687,6 @@ mod tests {
let mut doc = Document::default(); let mut doc = Document::default();
doc.add_text(text_field, "af b"); doc.add_text(text_field, "af b");
doc.add_u64(score_field, 3); doc.add_u64(score_field, 3);
doc.add_date(date_field, &curr_time);
add_score_bytes(&mut doc, 3); add_score_bytes(&mut doc, 3);
index_writer.add_document(doc); index_writer.add_document(doc);
} }
@@ -730,7 +712,6 @@ mod tests {
{ {
let mut doc = Document::default(); let mut doc = Document::default();
doc.add_text(text_field, "af b"); doc.add_text(text_field, "af b");
doc.add_date(date_field, &curr_time);
doc.add_u64(score_field, 11); doc.add_u64(score_field, 11);
add_score_bytes(&mut doc, 11); add_score_bytes(&mut doc, 11);
index_writer.add_document(doc); index_writer.add_document(doc);
@@ -758,39 +739,30 @@ mod tests {
index_writer.wait_merging_threads().unwrap(); index_writer.wait_merging_threads().unwrap();
} }
{ {
reader.reload().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
let get_doc_ids = |terms: Vec<Term>| { let get_doc_ids = |terms: Vec<Term>| {
let mut collector = TestCollector::default();
let query = BooleanQuery::new_multiterms_query(terms); let query = BooleanQuery::new_multiterms_query(terms);
let top_docs = searcher.search(&query, &TestCollector).unwrap(); assert!(searcher.search(&query, &mut collector).is_ok());
top_docs.docs().to_vec() collector.docs()
}; };
{ {
assert_eq!( assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "a")]), get_doc_ids(vec![Term::from_field_text(text_field, "a")]),
vec![DocAddress(0, 1), DocAddress(0, 2), DocAddress(0, 4)] vec![1, 2, 4]
); );
assert_eq!( assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "af")]), get_doc_ids(vec![Term::from_field_text(text_field, "af")]),
vec![DocAddress(0, 0), DocAddress(0, 3)] vec![0, 3]
); );
assert_eq!( assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "g")]), get_doc_ids(vec![Term::from_field_text(text_field, "g")]),
vec![DocAddress(0, 4)] vec![4]
); );
assert_eq!( assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "b")]), get_doc_ids(vec![Term::from_field_text(text_field, "b")]),
vec![ vec![0, 1, 2, 3, 4]
DocAddress(0, 0),
DocAddress(0, 1),
DocAddress(0, 2),
DocAddress(0, 3),
DocAddress(0, 4)
]
);
assert_eq!(
get_doc_ids(vec![Term::from_field_date(date_field, &curr_time)]),
vec![DocAddress(0, 0), DocAddress(0, 3)]
); );
} }
{ {
@@ -816,18 +788,17 @@ mod tests {
{ {
let get_fast_vals = |terms: Vec<Term>| { let get_fast_vals = |terms: Vec<Term>| {
let query = BooleanQuery::new_multiterms_query(terms); let query = BooleanQuery::new_multiterms_query(terms);
searcher let mut collector = FastFieldTestCollector::for_field(score_field);
.search(&query, &FastFieldTestCollector::for_field(score_field)) assert!(searcher.search(&query, &mut collector).is_ok());
.unwrap() collector.vals()
}; };
let get_fast_vals_bytes = |terms: Vec<Term>| { let get_fast_vals_bytes = |terms: Vec<Term>| {
let query = BooleanQuery::new_multiterms_query(terms); let query = BooleanQuery::new_multiterms_query(terms);
let mut collector = BytesFastFieldTestCollector::for_field(bytes_score_field);
searcher searcher
.search( .search(&query, &mut collector)
&query, .expect("failed to search");
&BytesFastFieldTestCollector::for_field(bytes_score_field), collector.vals()
)
.expect("failed to search")
}; };
assert_eq!( assert_eq!(
get_fast_vals(vec![Term::from_field_text(text_field, "a")]), get_fast_vals(vec![Term::from_field_text(text_field, "a")]),
@@ -843,27 +814,34 @@ mod tests {
#[test] #[test]
fn test_index_merger_with_deletes() { fn test_index_merger_with_deletes() {
let mut schema_builder = schema::Schema::builder(); let mut schema_builder = schema::SchemaBuilder::default();
let text_fieldtype = schema::TextOptions::default() let text_fieldtype = schema::TextOptions::default()
.set_indexing_options( .set_indexing_options(
TextFieldIndexing::default().set_index_option(IndexRecordOption::WithFreqs), TextFieldIndexing::default().set_index_option(IndexRecordOption::WithFreqs),
) ).set_stored();
.set_stored();
let text_field = schema_builder.add_text_field("text", text_fieldtype); let text_field = schema_builder.add_text_field("text", text_fieldtype);
let score_fieldtype = schema::IntOptions::default().set_fast(Cardinality::SingleValue); let score_fieldtype = schema::IntOptions::default().set_fast(Cardinality::SingleValue);
let score_field = schema_builder.add_u64_field("score", score_fieldtype); let score_field = schema_builder.add_u64_field("score", score_fieldtype);
let bytes_score_field = schema_builder.add_bytes_field("score_bytes"); let bytes_score_field = schema_builder.add_bytes_field("score_bytes");
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
let reader = index.reader().unwrap();
let search_term = |searcher: &Searcher, term: Term| { let search_term = |searcher: &Searcher, term: Term| {
let collector = FastFieldTestCollector::for_field(score_field); let mut collector = FastFieldTestCollector::for_field(score_field);
let bytes_collector = BytesFastFieldTestCollector::for_field(bytes_score_field); let mut bytes_collector = BytesFastFieldTestCollector::for_field(bytes_score_field);
let term_query = TermQuery::new(term, IndexRecordOption::Basic); let term_query = TermQuery::new(term, IndexRecordOption::Basic);
let (scores, bytes) = searcher
.search(&term_query, &(collector, bytes_collector)) {
.unwrap(); let mut combined_collector =
let mut score_bytes = Cursor::new(bytes); chain().push(&mut collector).push(&mut bytes_collector);
searcher
.search(&term_query, &mut combined_collector)
.unwrap();
}
let scores = collector.vals();
let mut score_bytes = Cursor::new(bytes_collector.vals());
for &score in &scores { for &score in &scores {
assert_eq!(score as u32, score_bytes.read_u32::<BigEndian>().unwrap()); assert_eq!(score as u32, score_bytes.read_u32::<BigEndian>().unwrap());
} }
@@ -876,24 +854,24 @@ mod tests {
{ {
// a first commit // a first commit
index_writer.add_document(doc!( index_writer.add_document(doc!(
text_field => "a b d", text_field => "a b d",
score_field => 1u64, score_field => 1u64,
bytes_score_field => vec![0u8, 0, 0, 1], bytes_score_field => vec![0u8, 0, 0, 1],
)); ));
index_writer.add_document(doc!( index_writer.add_document(doc!(
text_field => "b c", text_field => "b c",
score_field => 2u64, score_field => 2u64,
bytes_score_field => vec![0u8, 0, 0, 2], bytes_score_field => vec![0u8, 0, 0, 2],
)); ));
index_writer.delete_term(Term::from_field_text(text_field, "c")); index_writer.delete_term(Term::from_field_text(text_field, "c"));
index_writer.add_document(doc!( index_writer.add_document(doc!(
text_field => "c d", text_field => "c d",
score_field => 3u64, score_field => 3u64,
bytes_score_field => vec![0u8, 0, 0, 3], bytes_score_field => vec![0u8, 0, 0, 3],
)); ));
index_writer.commit().expect("committed"); index_writer.commit().expect("committed");
reader.reload().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let ref searcher = *index.searcher();
assert_eq!(searcher.num_docs(), 2); assert_eq!(searcher.num_docs(), 2);
assert_eq!(searcher.segment_readers()[0].num_docs(), 2); assert_eq!(searcher.segment_readers()[0].num_docs(), 2);
assert_eq!(searcher.segment_readers()[0].max_doc(), 3); assert_eq!(searcher.segment_readers()[0].max_doc(), 3);
@@ -917,37 +895,37 @@ mod tests {
{ {
// a second commit // a second commit
index_writer.add_document(doc!( index_writer.add_document(doc!(
text_field => "a d e", text_field => "a d e",
score_field => 4_000u64, score_field => 4_000u64,
bytes_score_field => vec![0u8, 0, 0, 4], bytes_score_field => vec![0u8, 0, 0, 4],
)); ));
index_writer.add_document(doc!( index_writer.add_document(doc!(
text_field => "e f", text_field => "e f",
score_field => 5_000u64, score_field => 5_000u64,
bytes_score_field => vec![0u8, 0, 0, 5], bytes_score_field => vec![0u8, 0, 0, 5],
)); ));
index_writer.delete_term(Term::from_field_text(text_field, "a")); index_writer.delete_term(Term::from_field_text(text_field, "a"));
index_writer.delete_term(Term::from_field_text(text_field, "f")); index_writer.delete_term(Term::from_field_text(text_field, "f"));
index_writer.add_document(doc!( index_writer.add_document(doc!(
text_field => "f g", text_field => "f g",
score_field => 6_000u64, score_field => 6_000u64,
bytes_score_field => vec![0u8, 0, 23, 112], bytes_score_field => vec![0u8, 0, 23, 112],
)); ));
index_writer.add_document(doc!( index_writer.add_document(doc!(
text_field => "g h", text_field => "g h",
score_field => 7_000u64, score_field => 7_000u64,
bytes_score_field => vec![0u8, 0, 27, 88], bytes_score_field => vec![0u8, 0, 27, 88],
)); ));
index_writer.commit().expect("committed"); index_writer.commit().expect("committed");
reader.reload().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
assert_eq!(searcher.segment_readers().len(), 2); assert_eq!(searcher.segment_readers().len(), 2);
assert_eq!(searcher.num_docs(), 3); assert_eq!(searcher.num_docs(), 3);
assert_eq!(searcher.segment_readers()[0].num_docs(), 2); assert_eq!(searcher.segment_readers()[0].num_docs(), 1);
assert_eq!(searcher.segment_readers()[0].max_doc(), 4); assert_eq!(searcher.segment_readers()[0].max_doc(), 3);
assert_eq!(searcher.segment_readers()[1].num_docs(), 1); assert_eq!(searcher.segment_readers()[1].num_docs(), 2);
assert_eq!(searcher.segment_readers()[1].max_doc(), 3); assert_eq!(searcher.segment_readers()[1].max_doc(), 4);
assert_eq!( assert_eq!(
search_term(&searcher, Term::from_field_text(text_field, "a")), search_term(&searcher, Term::from_field_text(text_field, "a")),
empty_vec empty_vec
@@ -981,15 +959,15 @@ mod tests {
.segment_reader(0) .segment_reader(0)
.fast_field_reader::<u64>(score_field) .fast_field_reader::<u64>(score_field)
.unwrap(); .unwrap();
assert_eq!(score_field_reader.min_value(), 4000); assert_eq!(score_field_reader.min_value(), 1);
assert_eq!(score_field_reader.max_value(), 7000); assert_eq!(score_field_reader.max_value(), 3);
let score_field_reader = searcher let score_field_reader = searcher
.segment_reader(1) .segment_reader(1)
.fast_field_reader::<u64>(score_field) .fast_field_reader::<u64>(score_field)
.unwrap(); .unwrap();
assert_eq!(score_field_reader.min_value(), 1); assert_eq!(score_field_reader.min_value(), 4000);
assert_eq!(score_field_reader.max_value(), 3); assert_eq!(score_field_reader.max_value(), 7000);
} }
{ {
// merging the segments // merging the segments
@@ -1001,8 +979,8 @@ mod tests {
.expect("Failed to initiate merge") .expect("Failed to initiate merge")
.wait() .wait()
.expect("Merging failed"); .expect("Merging failed");
reader.reload().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let ref searcher = *index.searcher();
assert_eq!(searcher.segment_readers().len(), 1); assert_eq!(searcher.segment_readers().len(), 1);
assert_eq!(searcher.num_docs(), 3); assert_eq!(searcher.num_docs(), 3);
assert_eq!(searcher.segment_readers()[0].num_docs(), 3); assert_eq!(searcher.segment_readers()[0].num_docs(), 3);
@@ -1047,8 +1025,8 @@ mod tests {
index_writer.delete_term(Term::from_field_text(text_field, "c")); index_writer.delete_term(Term::from_field_text(text_field, "c"));
index_writer.commit().unwrap(); index_writer.commit().unwrap();
reader.reload().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let ref searcher = *index.searcher();
assert_eq!(searcher.segment_readers().len(), 1); assert_eq!(searcher.segment_readers().len(), 1);
assert_eq!(searcher.num_docs(), 2); assert_eq!(searcher.num_docs(), 2);
assert_eq!(searcher.segment_readers()[0].num_docs(), 2); assert_eq!(searcher.segment_readers()[0].num_docs(), 2);
@@ -1098,9 +1076,9 @@ mod tests {
.expect("Failed to initiate merge") .expect("Failed to initiate merge")
.wait() .wait()
.expect("Merging failed"); .expect("Merging failed");
reader.reload().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let ref searcher = *index.searcher();
assert_eq!(searcher.segment_readers().len(), 1); assert_eq!(searcher.segment_readers().len(), 1);
assert_eq!(searcher.num_docs(), 2); assert_eq!(searcher.num_docs(), 2);
assert_eq!(searcher.segment_readers()[0].num_docs(), 2); assert_eq!(searcher.segment_readers()[0].num_docs(), 2);
@@ -1144,27 +1122,30 @@ mod tests {
{ {
// Test removing all docs // Test removing all docs
index_writer.delete_term(Term::from_field_text(text_field, "g")); index_writer.delete_term(Term::from_field_text(text_field, "g"));
index_writer.commit().unwrap();
let segment_ids = index let segment_ids = index
.searchable_segment_ids() .searchable_segment_ids()
.expect("Searchable segments failed."); .expect("Searchable segments failed.");
reader.reload().unwrap(); index_writer
.merge(&segment_ids)
.expect("Failed to initiate merge")
.wait()
.expect("Merging failed");
index.load_searchers().unwrap();
let searcher = reader.searcher(); let ref searcher = *index.searcher();
assert!(segment_ids.is_empty()); assert_eq!(searcher.segment_readers().len(), 1);
assert!(searcher.segment_readers().is_empty());
assert_eq!(searcher.num_docs(), 0); assert_eq!(searcher.num_docs(), 0);
} }
} }
#[test] #[test]
fn test_merge_facets() { fn test_merge_facets() {
let mut schema_builder = schema::Schema::builder(); let mut schema_builder = schema::SchemaBuilder::default();
let facet_field = schema_builder.add_facet_field("facet"); let facet_field = schema_builder.add_facet_field("facet");
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
let reader = index.reader().unwrap(); use schema::Facet;
{ {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
let index_doc = |index_writer: &mut IndexWriter, doc_facets: &[&str]| { let index_doc = |index_writer: &mut IndexWriter, doc_facets: &[&str]| {
let mut doc = Document::default(); let mut doc = Document::default();
for facet in doc_facets { for facet in doc_facets {
@@ -1191,16 +1172,20 @@ mod tests {
index_doc(&mut index_writer, &["/top/e", "/top/f"]); index_doc(&mut index_writer, &["/top/e", "/top/f"]);
index_writer.commit().expect("committed"); index_writer.commit().expect("committed");
} }
index.load_searchers().unwrap();
reader.reload().unwrap();
let test_searcher = |expected_num_docs: usize, expected: &[(&str, u64)]| { let test_searcher = |expected_num_docs: usize, expected: &[(&str, u64)]| {
let searcher = reader.searcher(); let searcher = index.searcher();
let mut facet_collector = FacetCollector::for_field(facet_field); let mut facet_collector = FacetCollector::for_field(facet_field);
facet_collector.add_facet(Facet::from("/top")); facet_collector.add_facet(Facet::from("/top"));
let (count, facet_counts) = searcher use collector::{CountCollector, MultiCollector};
.search(&AllQuery, &(Count, facet_collector)) let mut count_collector = CountCollector::default();
.unwrap(); {
assert_eq!(count, expected_num_docs); let mut multi_collectors =
MultiCollector::from(vec![&mut count_collector, &mut facet_collector]);
searcher.search(&AllQuery, &mut multi_collectors).unwrap();
}
assert_eq!(count_collector.count(), expected_num_docs);
let facet_counts = facet_collector.harvest();
let facets: Vec<(String, u64)> = facet_counts let facets: Vec<(String, u64)> = facet_counts
.get("/top") .get("/top")
.map(|(facet, count)| (facet.to_string(), count)) .map(|(facet, count)| (facet.to_string(), count))
@@ -1224,19 +1209,21 @@ mod tests {
("/top/f", 1), ("/top/f", 1),
], ],
); );
// Merging the segments // Merging the segments
{ {
let segment_ids = index let segment_ids = index
.searchable_segment_ids() .searchable_segment_ids()
.expect("Searchable segments failed."); .expect("Searchable segments failed.");
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
index_writer index_writer
.merge(&segment_ids) .merge(&segment_ids)
.expect("Failed to initiate merge") .expect("Failed to initiate merge")
.wait() .wait()
.expect("Merging failed"); .expect("Merging failed");
index_writer.wait_merging_threads().unwrap(); index_writer.wait_merging_threads().unwrap();
reader.reload().unwrap();
index.load_searchers().unwrap();
test_searcher( test_searcher(
11, 11,
&[ &[
@@ -1252,12 +1239,12 @@ mod tests {
// Deleting one term // Deleting one term
{ {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
let facet = Facet::from_path(vec!["top", "a", "firstdoc"]); let facet = Facet::from_path(vec!["top", "a", "firstdoc"]);
let facet_term = Term::from_facet(facet_field, &facet); let facet_term = Term::from_facet(facet_field, &facet);
index_writer.delete_term(facet_term); index_writer.delete_term(facet_term);
index_writer.commit().unwrap(); index_writer.commit().unwrap();
reader.reload().unwrap(); index.load_searchers().unwrap();
test_searcher( test_searcher(
9, 9,
&[ &[
@@ -1272,45 +1259,17 @@ mod tests {
} }
} }
#[test]
fn test_bug_merge() {
let mut schema_builder = schema::Schema::builder();
let int_field = schema_builder.add_u64_field("intvals", INDEXED);
let index = Index::create_in_ram(schema_builder.build());
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(int_field => 1u64));
index_writer.commit().expect("commit failed");
index_writer.add_document(doc!(int_field => 1u64));
index_writer.commit().expect("commit failed");
let reader = index.reader().unwrap();
let searcher = reader.searcher();
assert_eq!(searcher.num_docs(), 2);
index_writer.delete_term(Term::from_field_u64(int_field, 1));
let segment_ids = index
.searchable_segment_ids()
.expect("Searchable segments failed.");
index_writer
.merge(&segment_ids)
.expect("Failed to initiate merge")
.wait()
.expect("Merging failed");
reader.reload().unwrap();
// commit has not been called yet. The document should still be
// there.
assert_eq!(reader.searcher().num_docs(), 2);
}
#[test] #[test]
fn test_merge_multivalued_int_fields_all_deleted() { fn test_merge_multivalued_int_fields_all_deleted() {
let mut schema_builder = schema::Schema::builder(); let mut schema_builder = schema::SchemaBuilder::default();
let int_options = IntOptions::default() let int_options = IntOptions::default()
.set_fast(Cardinality::MultiValues) .set_fast(Cardinality::MultiValues)
.set_indexed(); .set_indexed();
let int_field = schema_builder.add_u64_field("intvals", int_options); let int_field = schema_builder.add_u64_field("intvals", int_options);
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
let reader = index.reader().unwrap();
{ {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
let mut doc = Document::default(); let mut doc = Document::default();
doc.add_u64(int_field, 1); doc.add_u64(int_field, 1);
index_writer.add_document(doc.clone()); index_writer.add_document(doc.clone());
@@ -1318,34 +1277,32 @@ mod tests {
index_writer.add_document(doc); index_writer.add_document(doc);
index_writer.commit().expect("commit failed"); index_writer.commit().expect("commit failed");
index_writer.delete_term(Term::from_field_u64(int_field, 1)); index_writer.delete_term(Term::from_field_u64(int_field, 1));
index_writer.commit().expect("commit failed");
}
index.load_searchers().unwrap();
let searcher = index.searcher();
assert_eq!(searcher.num_docs(), 0);
// Merging the segments
{
let segment_ids = index let segment_ids = index
.searchable_segment_ids() .searchable_segment_ids()
.expect("Searchable segments failed."); .expect("Searchable segments failed.");
let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
index_writer index_writer
.merge(&segment_ids) .merge(&segment_ids)
.expect("Failed to initiate merge") .expect("Failed to initiate merge")
.wait() .wait()
.expect("Merging failed"); .expect("Merging failed");
// assert delete has not been committed
reader.reload().expect("failed to load searcher 1");
let searcher = reader.searcher();
assert_eq!(searcher.num_docs(), 2);
index_writer.commit().unwrap();
index_writer.wait_merging_threads().unwrap(); index_writer.wait_merging_threads().unwrap();
} }
index.load_searchers().unwrap();
reader.reload().unwrap(); let searcher = index.searcher();
let searcher = reader.searcher();
assert_eq!(searcher.num_docs(), 0); assert_eq!(searcher.num_docs(), 0);
} }
#[test] #[test]
fn test_merge_multivalued_int_fields_simple() { fn test_merge_multivalued_int_fields() {
let mut schema_builder = schema::Schema::builder(); let mut schema_builder = schema::SchemaBuilder::default();
let int_options = IntOptions::default() let int_options = IntOptions::default()
.set_fast(Cardinality::MultiValues) .set_fast(Cardinality::MultiValues)
.set_indexed(); .set_indexed();
@@ -1353,7 +1310,7 @@ mod tests {
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
{ {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
let index_doc = |index_writer: &mut IndexWriter, int_vals: &[u64]| { let index_doc = |index_writer: &mut IndexWriter, int_vals: &[u64]| {
let mut doc = Document::default(); let mut doc = Document::default();
for &val in int_vals { for &val in int_vals {
@@ -1361,6 +1318,7 @@ mod tests {
} }
index_writer.add_document(doc); index_writer.add_document(doc);
}; };
index_doc(&mut index_writer, &[1, 2]); index_doc(&mut index_writer, &[1, 2]);
index_doc(&mut index_writer, &[1, 2, 3]); index_doc(&mut index_writer, &[1, 2, 3]);
index_doc(&mut index_writer, &[4, 5]); index_doc(&mut index_writer, &[4, 5]);
@@ -1369,14 +1327,19 @@ mod tests {
index_doc(&mut index_writer, &[3]); index_doc(&mut index_writer, &[3]);
index_doc(&mut index_writer, &[17]); index_doc(&mut index_writer, &[17]);
index_writer.commit().expect("committed"); index_writer.commit().expect("committed");
index_doc(&mut index_writer, &[20]); index_doc(&mut index_writer, &[20]);
index_writer.commit().expect("committed"); index_writer.commit().expect("committed");
index_doc(&mut index_writer, &[28, 27]); index_doc(&mut index_writer, &[28, 27]);
index_doc(&mut index_writer, &[1_000]); index_doc(&mut index_writer, &[1_000]);
index_writer.commit().expect("committed"); index_writer.commit().expect("committed");
} }
let reader = index.reader().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher();
let searcher = index.searcher();
let mut vals: Vec<u64> = Vec::new(); let mut vals: Vec<u64> = Vec::new();
{ {
@@ -1405,59 +1368,41 @@ mod tests {
assert_eq!(&vals, &[17]); assert_eq!(&vals, &[17]);
} }
println!(
"{:?}",
searcher
.segment_readers()
.iter()
.map(|reader| reader.max_doc())
.collect::<Vec<_>>()
);
{ {
let segment = searcher.segment_reader(1u32); let segment = searcher.segment_reader(1u32);
let ff_reader = segment.multi_fast_field_reader(int_field).unwrap(); let ff_reader = segment.multi_fast_field_reader(int_field).unwrap();
ff_reader.get_vals(0, &mut vals); ff_reader.get_vals(0, &mut vals);
assert_eq!(&vals, &[20]);
}
{
let segment = searcher.segment_reader(2u32);
let ff_reader = segment.multi_fast_field_reader(int_field).unwrap();
ff_reader.get_vals(0, &mut vals);
assert_eq!(&vals, &[28, 27]); assert_eq!(&vals, &[28, 27]);
ff_reader.get_vals(1, &mut vals); ff_reader.get_vals(1, &mut vals);
assert_eq!(&vals, &[1_000]); assert_eq!(&vals, &[1_000]);
} }
{
let segment = searcher.segment_reader(2u32);
let ff_reader = segment.multi_fast_field_reader(int_field).unwrap();
ff_reader.get_vals(0, &mut vals);
assert_eq!(&vals, &[20]);
}
// Merging the segments // Merging the segments
{ {
let segment_ids = index let segment_ids = index
.searchable_segment_ids() .searchable_segment_ids()
.expect("Searchable segments failed."); .expect("Searchable segments failed.");
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
index_writer index_writer
.merge(&segment_ids) .merge(&segment_ids)
.expect("Failed to initiate merge") .expect("Failed to initiate merge")
.wait() .wait()
.expect("Merging failed"); .expect("Merging failed");
index_writer index_writer.wait_merging_threads().unwrap();
.wait_merging_threads()
.expect("Wait for merging threads");
} }
reader.reload().expect("Load searcher");
index.load_searchers().unwrap();
{ {
let searcher = reader.searcher(); let searcher = index.searcher();
println!(
"{:?}",
searcher
.segment_readers()
.iter()
.map(|reader| reader.max_doc())
.collect::<Vec<_>>()
);
let segment = searcher.segment_reader(0u32); let segment = searcher.segment_reader(0u32);
let ff_reader = segment.multi_fast_field_reader(int_field).unwrap(); let ff_reader = segment.multi_fast_field_reader(int_field).unwrap();
@@ -1483,13 +1428,13 @@ mod tests {
assert_eq!(&vals, &[17]); assert_eq!(&vals, &[17]);
ff_reader.get_vals(7, &mut vals); ff_reader.get_vals(7, &mut vals);
assert_eq!(&vals, &[28, 27]); assert_eq!(&vals, &[20]);
ff_reader.get_vals(8, &mut vals); ff_reader.get_vals(8, &mut vals);
assert_eq!(&vals, &[1_000]); assert_eq!(&vals, &[28, 27]);
ff_reader.get_vals(9, &mut vals); ff_reader.get_vals(9, &mut vals);
assert_eq!(&vals, &[20]); assert_eq!(&vals, &[1_000]);
} }
} }
} }

9
src/indexer/mod.rs
View File

@@ -1,9 +1,8 @@
pub mod delete_queue; pub mod delete_queue;
mod directory_lock;
mod doc_opstamp_mapping; mod doc_opstamp_mapping;
pub mod index_writer; pub mod index_writer;
mod log_merge_policy; mod log_merge_policy;
mod merge_operation;
pub mod merge_policy; pub mod merge_policy;
pub mod merger; pub mod merger;
pub mod operation; pub mod operation;
@@ -16,12 +15,14 @@ pub mod segment_updater;
mod segment_writer; mod segment_writer;
mod stamper; mod stamper;
pub(crate) use self::directory_lock::DirectoryLock;
pub use self::directory_lock::LockType;
pub use self::index_writer::IndexWriter; pub use self::index_writer::IndexWriter;
pub use self::log_merge_policy::LogMergePolicy; pub use self::log_merge_policy::LogMergePolicy;
pub use self::merge_operation::{MergeOperation, MergeOperationInventory};
pub use self::merge_policy::{MergeCandidate, MergePolicy, NoMergePolicy}; pub use self::merge_policy::{MergeCandidate, MergePolicy, NoMergePolicy};
pub use self::prepared_commit::PreparedCommit; pub use self::prepared_commit::PreparedCommit;
pub use self::segment_entry::SegmentEntry; pub use self::segment_entry::{SegmentEntry, SegmentState};
pub use self::segment_manager::SegmentManager; pub use self::segment_manager::SegmentManager;
pub use self::segment_serializer::SegmentSerializer; pub use self::segment_serializer::SegmentSerializer;
pub use self::segment_writer::SegmentWriter; pub use self::segment_writer::SegmentWriter;

7
src/indexer/operation.rs
View File

@@ -14,10 +14,3 @@ pub struct AddOperation {
pub opstamp: u64, pub opstamp: u64,
pub document: Document, pub document: Document,
} }
/// UserOperation is an enum type that encapsulates other operation types.
#[derive(Eq, PartialEq, Debug)]
pub enum UserOperation {
Add(Document),
Delete(Term),
}

10
src/indexer/prepared_commit.rs
View File

@@ -6,20 +6,14 @@ pub struct PreparedCommit<'a> {
index_writer: &'a mut IndexWriter, index_writer: &'a mut IndexWriter,
payload: Option<String>, payload: Option<String>,
opstamp: u64, opstamp: u64,
soft: bool,
} }
impl<'a> PreparedCommit<'a> { impl<'a> PreparedCommit<'a> {
pub(crate) fn new( pub(crate) fn new(index_writer: &'a mut IndexWriter, opstamp: u64) -> PreparedCommit {
index_writer: &'a mut IndexWriter,
opstamp: u64,
soft: bool,
) -> PreparedCommit {
PreparedCommit { PreparedCommit {
index_writer, index_writer,
payload: None, payload: None,
opstamp, opstamp,
soft,
} }
} }
@@ -39,7 +33,7 @@ impl<'a> PreparedCommit<'a> {
info!("committing {}", self.opstamp); info!("committing {}", self.opstamp);
self.index_writer self.index_writer
.segment_updater() .segment_updater()
.commit(self.opstamp, self.payload, self.soft)?; .commit(self.opstamp, self.payload)?;
Ok(self.opstamp) Ok(self.opstamp)
} }
} }

60
src/indexer/segment_entry.rs
View File

@@ -4,6 +4,21 @@ use core::SegmentMeta;
use indexer::delete_queue::DeleteCursor; use indexer::delete_queue::DeleteCursor;
use std::fmt; use std::fmt;
#[derive(Clone, Copy, PartialEq, Eq, Debug)]
pub enum SegmentState {
Ready,
InMerge,
}
impl SegmentState {
pub fn letter_code(self) -> char {
match self {
SegmentState::InMerge => 'M',
SegmentState::Ready => 'R',
}
}
}
/// A segment entry describes the state of /// A segment entry describes the state of
/// a given segment, at a given instant. /// a given segment, at a given instant.
/// ///
@@ -20,9 +35,9 @@ use std::fmt;
#[derive(Clone)] #[derive(Clone)]
pub struct SegmentEntry { pub struct SegmentEntry {
meta: SegmentMeta, meta: SegmentMeta,
state: SegmentState,
delete_bitset: Option<BitSet>, delete_bitset: Option<BitSet>,
delete_cursor: DeleteCursor, delete_cursor: DeleteCursor,
opstamp: u64,
} }
impl SegmentEntry { impl SegmentEntry {
@@ -31,20 +46,15 @@ impl SegmentEntry {
segment_meta: SegmentMeta, segment_meta: SegmentMeta,
delete_cursor: DeleteCursor, delete_cursor: DeleteCursor,
delete_bitset: Option<BitSet>, delete_bitset: Option<BitSet>,
opstamp: u64,
) -> SegmentEntry { ) -> SegmentEntry {
SegmentEntry { SegmentEntry {
meta: segment_meta, meta: segment_meta,
state: SegmentState::Ready,
delete_bitset, delete_bitset,
delete_cursor, delete_cursor,
opstamp,
} }
} }
pub fn opstamp(&self) -> u64 {
self.opstamp
}
/// Return a reference to the segment entry deleted bitset. /// Return a reference to the segment entry deleted bitset.
/// ///
/// `DocId` in this bitset are flagged as deleted. /// `DocId` in this bitset are flagged as deleted.
@@ -53,8 +63,7 @@ impl SegmentEntry {
} }
/// Set the `SegmentMeta` for this segment. /// Set the `SegmentMeta` for this segment.
pub fn set_meta(&mut self, opstamp: u64, segment_meta: SegmentMeta) { pub fn set_meta(&mut self, segment_meta: SegmentMeta) {
self.opstamp = opstamp;
self.meta = segment_meta; self.meta = segment_meta;
} }
@@ -63,6 +72,14 @@ impl SegmentEntry {
&mut self.delete_cursor &mut self.delete_cursor
} }
/// Return the `SegmentEntry`.
///
/// The state describes whether the segment is available for
/// a merge or not.
pub fn state(&self) -> SegmentState {
self.state
}
/// Returns the segment id. /// Returns the segment id.
pub fn segment_id(&self) -> SegmentId { pub fn segment_id(&self) -> SegmentId {
self.meta.id() self.meta.id()
@@ -72,10 +89,33 @@ impl SegmentEntry {
pub fn meta(&self) -> &SegmentMeta { pub fn meta(&self) -> &SegmentMeta {
&self.meta &self.meta
} }
/// Mark the `SegmentEntry` as in merge.
///
/// Only segments that are not already
/// in a merge are elligible for future merge.
pub fn start_merge(&mut self) {
self.state = SegmentState::InMerge;
}
/// Cancel a merge
///
/// If a merge fails, it is important to switch
/// the segment back to a idle state, so that it
/// may be elligible for future merges.
pub fn cancel_merge(&mut self) {
self.state = SegmentState::Ready;
}
/// Returns true iff a segment should
/// be considered for a merge.
pub fn is_ready(&self) -> bool {
self.state == SegmentState::Ready
}
} }
impl fmt::Debug for SegmentEntry { impl fmt::Debug for SegmentEntry {
fn fmt(&self, formatter: &mut fmt::Formatter) -> fmt::Result { fn fmt(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
write!(formatter, "SegmentEntry({:?})", self.meta) write!(formatter, "SegmentEntry({:?}, {:?})", self.meta, self.state)
} }
} }

251
src/indexer/segment_manager.rs
View File

@@ -11,47 +11,12 @@ use std::path::PathBuf;
use std::sync::RwLock; use std::sync::RwLock;
use std::sync::{RwLockReadGuard, RwLockWriteGuard}; use std::sync::{RwLockReadGuard, RwLockWriteGuard};
use Result as TantivyResult; use Result as TantivyResult;
use std::sync::Arc;
use std::collections::HashMap;
/// Provides a read-only view of the available segments.
#[derive(Clone)]
pub struct AvailableSegments {
registers: Arc<RwLock<SegmentRegisters>>,
}
impl AvailableSegments {
pub fn committed(&self) -> Vec<SegmentMeta> {
self.registers
.read()
.unwrap()
.committed
.segment_metas()
}
pub fn soft_committed(&self) -> Vec<SegmentMeta> {
self.registers
.read()
.unwrap()
.soft_committed
.segment_metas()
}
}
#[derive(Default)]
struct SegmentRegisters { struct SegmentRegisters {
uncommitted: HashMap<SegmentId, SegmentEntry>, uncommitted: SegmentRegister,
committed: SegmentRegister, committed: SegmentRegister,
/// soft commits can advance committed segment to a future delete writing: HashSet<SegmentId>,
/// opstamp.
///
/// In that case the same `SegmentId` can appear in both `committed`
/// and in `committed_in_the_future`.
///
/// We do not consider these segments for merges.
soft_committed: SegmentRegister,
/// `DeleteCursor`, positionned on the soft commit.
delete_cursor: DeleteCursor,
} }
/// The segment manager stores the list of segments /// The segment manager stores the list of segments
@@ -59,8 +24,9 @@ struct SegmentRegisters {
/// ///
/// It guarantees the atomicity of the /// It guarantees the atomicity of the
/// changes (merges especially) /// changes (merges especially)
#[derive(Default)]
pub struct SegmentManager { pub struct SegmentManager {
registers: Arc<RwLock<SegmentRegisters>> registers: RwLock<SegmentRegisters>,
} }
impl Debug for SegmentManager { impl Debug for SegmentManager {
@@ -75,23 +41,12 @@ impl Debug for SegmentManager {
} }
pub fn get_mergeable_segments( pub fn get_mergeable_segments(
in_merge_segment_ids: &HashSet<SegmentId>,
segment_manager: &SegmentManager, segment_manager: &SegmentManager,
) -> (Vec<SegmentMeta>, Vec<SegmentMeta>) { ) -> (Vec<SegmentMeta>, Vec<SegmentMeta>) {
let registers_lock = segment_manager.read(); let registers_lock = segment_manager.read();
( (
registers_lock registers_lock.committed.get_mergeable_segments(),
.soft_committed registers_lock.uncommitted.get_mergeable_segments(),
.get_mergeable_segments(in_merge_segment_ids),
registers_lock
.uncommitted
.values()
.map(|segment_entry| segment_entry.meta())
.filter(|segment_meta| {
!in_merge_segment_ids.contains(&segment_meta.id())
})
.cloned()
.collect::<Vec<_>>()
) )
} }
@@ -99,22 +54,28 @@ impl SegmentManager {
pub fn from_segments( pub fn from_segments(
segment_metas: Vec<SegmentMeta>, segment_metas: Vec<SegmentMeta>,
delete_cursor: &DeleteCursor, delete_cursor: &DeleteCursor,
opstamp: u64,
) -> SegmentManager { ) -> SegmentManager {
SegmentManager { SegmentManager {
registers: Arc::new(RwLock::new(SegmentRegisters { registers: RwLock::new(SegmentRegisters {
uncommitted: HashMap::default(), uncommitted: SegmentRegister::default(),
committed: SegmentRegister::new(segment_metas.clone(), opstamp), committed: SegmentRegister::new(segment_metas, delete_cursor),
soft_committed: SegmentRegister::new(segment_metas, opstamp), writing: HashSet::new(),
delete_cursor: delete_cursor.clone(), }),
}))
} }
} }
pub fn available_segments_view(&self) -> AvailableSegments { /// Returns all of the segment entries (committed or uncommitted)
AvailableSegments { pub fn segment_entries(&self) -> Vec<SegmentEntry> {
registers: self.registers.clone() let registers_lock = self.read();
} let mut segment_entries = registers_lock.uncommitted.segment_entries();
segment_entries.extend(registers_lock.committed.segment_entries());
segment_entries
}
/// Returns the overall number of segments in the `SegmentManager`
pub fn num_segments(&self) -> usize {
let registers_lock = self.read();
registers_lock.committed.len() + registers_lock.uncommitted.len()
} }
/// List the files that are useful to the index. /// List the files that are useful to the index.
@@ -145,84 +106,40 @@ impl SegmentManager {
.expect("Failed to acquire write lock on SegmentManager.") .expect("Failed to acquire write lock on SegmentManager.")
} }
/// Deletes all empty segments pub fn commit(&self, segment_entries: Vec<SegmentEntry>) {
fn remove_empty_segments(&self) {
let mut registers_lock = self.write();
registers_lock
.committed
.segment_metas()
.iter()
.filter(|segment_meta| segment_meta.num_docs() == 0)
.for_each(|segment_meta| {
registers_lock
.committed
.remove_segment(&segment_meta.id())
});
registers_lock
.soft_committed
.segment_metas()
.iter()
.filter(|segment_meta| segment_meta.num_docs() == 0)
.for_each(|segment_meta| {
registers_lock
.committed
.remove_segment(&segment_meta.id())
});
}
/// Returns all of the segment entries (soft committed or uncommitted)
pub fn segment_entries(&self) -> Vec<SegmentEntry> {
let registers_lock = self.read();
let mut segment_entries: Vec<SegmentEntry > = registers_lock.uncommitted.values().cloned().collect();
segment_entries.extend(registers_lock.soft_committed.segment_entries(&registers_lock.delete_cursor).into_iter());
segment_entries
}
pub fn commit(&self, opstamp: u64, segment_entries: Vec<SegmentEntry>) {
let mut registers_lock = self.write(); let mut registers_lock = self.write();
registers_lock.committed.clear();
registers_lock.uncommitted.clear(); registers_lock.uncommitted.clear();
registers_lock for segment_entry in segment_entries {
.committed registers_lock.committed.add_segment_entry(segment_entry);
.set_commit(opstamp, segment_entries.clone()); }
registers_lock
.soft_committed
.set_commit(opstamp, segment_entries);
registers_lock.delete_cursor.skip_to(opstamp);
} }
pub fn soft_commit(&self, opstamp: u64, segment_entries: Vec<SegmentEntry>) { /// Marks a list of segments as in merge.
let mut registers_lock = self.write();
registers_lock.uncommitted.clear();
registers_lock
.soft_committed
.set_commit(opstamp, segment_entries);
registers_lock.delete_cursor.skip_to(opstamp);
}
/// Gets the list of segment_entries associated to a list of `segment_ids`.
/// This method is used when starting a merge operations.
/// ///
/// Returns an error if some segments are missing, or if /// Returns an error if some segments are missing, or if
/// the `segment_ids` are not either all soft_committed or all /// the `segment_ids` are not either all committed or all
/// uncommitted. /// uncommitted.
pub fn start_merge(&self, segment_ids: &[SegmentId]) -> TantivyResult<Vec<SegmentEntry>> { pub fn start_merge(&self, segment_ids: &[SegmentId]) -> TantivyResult<Vec<SegmentEntry>> {
let registers_lock = self.read(); let mut registers_lock = self.write();
let mut segment_entries = vec![]; let mut segment_entries = vec![];
if segment_ids.iter().all(|segment_id| registers_lock.uncommitted.contains_key(segment_id)) { if registers_lock.uncommitted.contains_all(segment_ids) {
for segment_id in segment_ids { for segment_id in segment_ids {
let segment_entry = registers_lock.uncommitted let segment_entry = registers_lock.uncommitted
.get(segment_id) .start_merge(segment_id)
.expect("Segment id not found {}. Should never happen because of the contains all if-block.");
segment_entries.push(segment_entry.clone());
}
} else if registers_lock.soft_committed.contains_all(segment_ids) {
for segment_id in segment_ids {
let segment_entry = registers_lock.soft_committed
.get(segment_id, &registers_lock.delete_cursor)
.expect("Segment id not found {}. Should never happen because of the contains all if-block."); .expect("Segment id not found {}. Should never happen because of the contains all if-block.");
segment_entries.push(segment_entry); segment_entries.push(segment_entry);
} }
} else if registers_lock.committed.contains_all(segment_ids) {
for segment_id in segment_ids {
let segment_entry = registers_lock.committed
.start_merge(segment_id)
.expect("Segment id not found {}. Should never happen because of the contains all if-block.");
segment_entries.push(segment_entry);
}
for segment_id in segment_ids {
registers_lock.committed.start_merge(segment_id);
}
} else { } else {
let error_msg = "Merge operation sent for segments that are not \ let error_msg = "Merge operation sent for segments that are not \
all uncommited or commited." all uncommited or commited."
@@ -232,38 +149,86 @@ impl SegmentManager {
Ok(segment_entries) Ok(segment_entries)
} }
pub fn cancel_merge(
&self,
before_merge_segment_ids: &[SegmentId],
after_merge_segment_id: SegmentId,
) {
let mut registers_lock = self.write();
// we mark all segments are ready for merge.
{
let target_segment_register: &mut SegmentRegister;
target_segment_register = {
if registers_lock
.uncommitted
.contains_all(before_merge_segment_ids)
{
&mut registers_lock.uncommitted
} else if registers_lock
.committed
.contains_all(before_merge_segment_ids)
{
&mut registers_lock.committed
} else {
warn!("couldn't find segment in SegmentManager");
return;
}
};
for segment_id in before_merge_segment_ids {
target_segment_register.cancel_merge(segment_id);
}
}
// ... and we make sure the target segment entry
// can be garbage collected.
registers_lock.writing.remove(&after_merge_segment_id);
}
pub fn write_segment(&self, segment_id: SegmentId) {
let mut registers_lock = self.write();
registers_lock.writing.insert(segment_id);
}
pub fn add_segment(&self, segment_entry: SegmentEntry) { pub fn add_segment(&self, segment_entry: SegmentEntry) {
let mut registers_lock = self.write(); let mut registers_lock = self.write();
registers_lock registers_lock.writing.remove(&segment_entry.segment_id());
.uncommitted registers_lock.uncommitted.add_segment_entry(segment_entry);
.insert(segment_entry.segment_id(), segment_entry);
} }
pub fn end_merge( pub fn end_merge(
&self, &self,
before_merge_segment_ids: &[SegmentId], before_merge_segment_ids: &[SegmentId],
after_merge_segment_entry: SegmentEntry after_merge_segment_entry: SegmentEntry,
) { ) {
let mut registers_lock = self.write(); let mut registers_lock = self.write();
registers_lock
.writing
.remove(&after_merge_segment_entry.segment_id());
if before_merge_segment_ids.iter().all(|seg_id| let target_register: &mut SegmentRegister = {
registers_lock if registers_lock
.uncommitted .uncommitted
.contains_key(seg_id)) .contains_all(before_merge_segment_ids)
{ {
for segment_id in before_merge_segment_ids { &mut registers_lock.uncommitted
registers_lock.uncommitted.remove(&segment_id); } else if registers_lock
.committed
.contains_all(before_merge_segment_ids)
{
&mut registers_lock.committed
} else {
warn!("couldn't find segment in SegmentManager");
return;
} }
registers_lock.uncommitted.insert(after_merge_segment_entry.segment_id(), };
after_merge_segment_entry); for segment_id in before_merge_segment_ids {
} else { target_register.remove_segment(segment_id);
registers_lock.committed.receive_merge(&before_merge_segment_ids, &after_merge_segment_entry);
registers_lock.soft_committed.receive_merge(&before_merge_segment_ids, &after_merge_segment_entry)
} }
target_register.add_segment_entry(after_merge_segment_entry);
} }
pub fn committed_segment_metas(&self) -> Vec<SegmentMeta> { pub fn committed_segment_metas(&self) -> Vec<SegmentMeta> {
self.remove_empty_segments();
let registers_lock = self.read(); let registers_lock = self.read();
registers_lock.committed.segment_metas() registers_lock.committed.segment_metas()
} }

182
src/indexer/segment_register.rs
View File

@@ -3,7 +3,6 @@ use core::SegmentMeta;
use indexer::delete_queue::DeleteCursor; use indexer::delete_queue::DeleteCursor;
use indexer::segment_entry::SegmentEntry; use indexer::segment_entry::SegmentEntry;
use std::collections::HashMap; use std::collections::HashMap;
use std::collections::HashSet;
use std::fmt::{self, Debug, Formatter}; use std::fmt::{self, Debug, Formatter};
/// The segment register keeps track /// The segment register keeps track
@@ -16,15 +15,14 @@ use std::fmt::{self, Debug, Formatter};
/// merge candidates. /// merge candidates.
#[derive(Default)] #[derive(Default)]
pub struct SegmentRegister { pub struct SegmentRegister {
segment_states: HashMap<SegmentId, SegmentMeta>, segment_states: HashMap<SegmentId, SegmentEntry>,
opstamp_constraint: u64,
} }
impl Debug for SegmentRegister { impl Debug for SegmentRegister {
fn fmt(&self, f: &mut Formatter) -> Result<(), fmt::Error> { fn fmt(&self, f: &mut Formatter) -> Result<(), fmt::Error> {
write!(f, "SegmentRegister(")?; write!(f, "SegmentRegister(")?;
for k in self.segment_states.keys() { for (k, v) in &self.segment_states {
write!(f, "{}, ", k.short_uuid_string())?; write!(f, "{}:{}, ", k.short_uuid_string(), v.state().letter_code())?;
} }
write!(f, ")")?; write!(f, ")")?;
Ok(()) Ok(())
@@ -36,113 +34,76 @@ impl SegmentRegister {
self.segment_states.clear(); self.segment_states.clear();
} }
pub fn get_mergeable_segments( pub fn len(&self) -> usize {
&self, self.segment_states.len()
in_merge_segment_ids: &HashSet<SegmentId>, }
) -> Vec<SegmentMeta> {
pub fn get_mergeable_segments(&self) -> Vec<SegmentMeta> {
self.segment_states self.segment_states
.values() .values()
.filter(|segment_meta| !in_merge_segment_ids.contains(&segment_meta.id())) .filter(|segment_entry| segment_entry.is_ready())
.cloned() .map(|segment_entry| segment_entry.meta().clone())
.collect() .collect()
} }
pub fn segment_entries(&self) -> Vec<SegmentEntry> {
self.segment_states.values().cloned().collect()
}
pub fn segment_metas(&self) -> Vec<SegmentMeta> { pub fn segment_metas(&self) -> Vec<SegmentMeta> {
let mut segment_metas: Vec<SegmentMeta> = self let mut segment_ids: Vec<SegmentMeta> = self
.segment_states .segment_states
.values() .values()
.cloned() .map(|segment_entry| segment_entry.meta().clone())
.collect(); .collect();
segment_metas.sort_by_key(|meta| meta.id()); segment_ids.sort_by_key(|meta| meta.id());
segment_metas segment_ids
} }
pub fn segment_entries(&self, delete_cursor: &DeleteCursor) -> Vec<SegmentEntry> { pub fn contains_all(&mut self, segment_ids: &[SegmentId]) -> bool {
self.segment_states
.values()
.map(|segment_meta| {
SegmentEntry::new(segment_meta.clone(), delete_cursor.clone(), None, self.opstamp_constraint)
})
.collect()
}
pub fn contains_all(&self, segment_ids: &[SegmentId]) -> bool {
segment_ids segment_ids
.iter() .iter()
.all(|segment_id| self.segment_states.contains_key(segment_id)) .all(|segment_id| self.segment_states.contains_key(segment_id))
} }
pub fn receive_merge(&mut self, pub fn add_segment_entry(&mut self, segment_entry: SegmentEntry) {
before_merge_segment_ids: &[SegmentId],
after_merge_segment_entry: &SegmentEntry) {
if after_merge_segment_entry.opstamp() != self.opstamp_constraint {
return;
}
if !self.contains_all(before_merge_segment_ids) {
return;
}
for segment_id in before_merge_segment_ids {
self.segment_states.remove(segment_id);
}
self.register_segment_entry(after_merge_segment_entry.clone());
}
/// Registers a `SegmentEntry`.
///
/// If a segment entry associated to this `SegmentId` is already there,
/// override it with the new `SegmentEntry`.
pub fn register_segment_entry(&mut self, segment_entry: SegmentEntry) {
if self.opstamp_constraint != segment_entry.opstamp() {
panic!(format!(
"Invalid segment. Expect opstamp {}, got {}.",
self.opstamp_constraint,
segment_entry.opstamp()
));
}
if segment_entry.meta().num_docs() == 0 {
return;
}
let segment_id = segment_entry.segment_id(); let segment_id = segment_entry.segment_id();
// Check that we are ok with deletes. self.segment_states.insert(segment_id, segment_entry);
self.segment_states.insert(segment_id, segment_entry.meta().clone());
}
pub fn set_commit(&mut self, opstamp: u64, segment_entries: Vec<SegmentEntry>) {
self.segment_states.clear();
self.opstamp_constraint = opstamp;
for segment_entry in segment_entries {
self.register_segment_entry(segment_entry);
}
} }
pub fn remove_segment(&mut self, segment_id: &SegmentId) { pub fn remove_segment(&mut self, segment_id: &SegmentId) {
self.segment_states.remove(&segment_id); self.segment_states.remove(segment_id);
} }
pub fn get(&self, segment_id: &SegmentId, delete_cursor: &DeleteCursor) -> Option<SegmentEntry> { pub fn cancel_merge(&mut self, segment_id: &SegmentId) {
self.segment_states self.segment_states
.get(&segment_id) .get_mut(segment_id)
.map(|segment_meta| .expect("Received a merge notification for a segment that is not registered")
SegmentEntry::new( .cancel_merge();
segment_meta.clone(),
delete_cursor.clone(),
None,
self.opstamp_constraint
))
} }
pub fn new( pub fn start_merge(&mut self, segment_id: &SegmentId) -> Option<SegmentEntry> {
segment_metas: Vec<SegmentMeta>, if let Some(segment_entry) = self.segment_states.get_mut(segment_id) {
opstamp: u64, segment_entry.start_merge();
) -> SegmentRegister { Some(segment_entry.clone())
} else {
None
}
}
pub fn new(segment_metas: Vec<SegmentMeta>, delete_cursor: &DeleteCursor) -> SegmentRegister {
let mut segment_states = HashMap::new(); let mut segment_states = HashMap::new();
for segment_meta in segment_metas { for segment_meta in segment_metas {
segment_states.insert(segment_meta.id(), segment_meta); let segment_id = segment_meta.id();
} let segment_entry = SegmentEntry::new(segment_meta, delete_cursor.clone(), None);
SegmentRegister { segment_states.insert(segment_id, segment_entry);
segment_states,
opstamp_constraint: opstamp,
} }
SegmentRegister { segment_states }
}
#[cfg(test)]
pub fn segment_entry(&self, segment_id: &SegmentId) -> Option<SegmentEntry> {
self.segment_states.get(segment_id).cloned()
} }
} }
@@ -152,6 +113,7 @@ mod tests {
use core::SegmentId; use core::SegmentId;
use core::SegmentMeta; use core::SegmentMeta;
use indexer::delete_queue::*; use indexer::delete_queue::*;
use indexer::SegmentState;
fn segment_ids(segment_register: &SegmentRegister) -> Vec<SegmentId> { fn segment_ids(segment_register: &SegmentRegister) -> Vec<SegmentId> {
segment_register segment_register
@@ -171,22 +133,52 @@ mod tests {
let segment_id_merged = SegmentId::generate_random(); let segment_id_merged = SegmentId::generate_random();
{ {
let segment_meta = SegmentMeta::new(segment_id_a, 1u32); let segment_meta = SegmentMeta::new(segment_id_a, 0u32);
let segment_entry = SegmentEntry::new(segment_meta, delete_queue.cursor(), None, 0u64); let segment_entry = SegmentEntry::new(segment_meta, delete_queue.cursor(), None);
segment_register.register_segment_entry(segment_entry); segment_register.add_segment_entry(segment_entry);
} }
assert_eq!(
segment_register
.segment_entry(&segment_id_a)
.unwrap()
.state(),
SegmentState::Ready
);
assert_eq!(segment_ids(&segment_register), vec![segment_id_a]); assert_eq!(segment_ids(&segment_register), vec![segment_id_a]);
{ {
let segment_meta = SegmentMeta::new(segment_id_b, 2u32); let segment_meta = SegmentMeta::new(segment_id_b, 0u32);
let segment_entry = SegmentEntry::new(segment_meta, delete_queue.cursor(), None, 0u64); let segment_entry = SegmentEntry::new(segment_meta, delete_queue.cursor(), None);
segment_register.register_segment_entry(segment_entry); segment_register.add_segment_entry(segment_entry);
} }
assert_eq!(
segment_register
.segment_entry(&segment_id_b)
.unwrap()
.state(),
SegmentState::Ready
);
segment_register.start_merge(&segment_id_a);
segment_register.start_merge(&segment_id_b);
assert_eq!(
segment_register
.segment_entry(&segment_id_a)
.unwrap()
.state(),
SegmentState::InMerge
);
assert_eq!(
segment_register
.segment_entry(&segment_id_b)
.unwrap()
.state(),
SegmentState::InMerge
);
segment_register.remove_segment(&segment_id_a);
segment_register.remove_segment(&segment_id_b);
{ {
let segment_meta_merged = SegmentMeta::new(segment_id_merged, 3u32); let segment_meta_merged = SegmentMeta::new(segment_id_merged, 0u32);
let segment_entry = let segment_entry = SegmentEntry::new(segment_meta_merged, delete_queue.cursor(), None);
SegmentEntry::new(segment_meta_merged, delete_queue.cursor(), None, 0u64); segment_register.add_segment_entry(segment_entry);
segment_register.receive_merge(&[segment_id_a, segment_id_b], &segment_entry);
segment_register.register_segment_entry(segment_entry);
} }
assert_eq!(segment_ids(&segment_register), vec![segment_id_merged]); assert_eq!(segment_ids(&segment_register), vec![segment_id_merged]);
} }

384
src/indexer/segment_updater.rs
View File

@@ -16,10 +16,9 @@ use futures_cpupool::CpuFuture;
use futures_cpupool::CpuPool; use futures_cpupool::CpuPool;
use indexer::delete_queue::DeleteCursor; use indexer::delete_queue::DeleteCursor;
use indexer::index_writer::advance_deletes; use indexer::index_writer::advance_deletes;
use indexer::merge_operation::MergeOperationInventory;
use indexer::merger::IndexMerger; use indexer::merger::IndexMerger;
use indexer::stamper::Stamper; use indexer::stamper::Stamper;
use indexer::MergeOperation; use indexer::MergeCandidate;
use indexer::SegmentEntry; use indexer::SegmentEntry;
use indexer::SegmentSerializer; use indexer::SegmentSerializer;
use indexer::{DefaultMergePolicy, MergePolicy}; use indexer::{DefaultMergePolicy, MergePolicy};
@@ -27,7 +26,6 @@ use schema::Schema;
use serde_json; use serde_json;
use std::borrow::BorrowMut; use std::borrow::BorrowMut;
use std::collections::HashMap; use std::collections::HashMap;
use std::collections::HashSet;
use std::io::Write; use std::io::Write;
use std::mem; use std::mem;
use std::ops::DerefMut; use std::ops::DerefMut;
@@ -47,30 +45,33 @@ use Result;
/// and flushed. /// and flushed.
/// ///
/// This method is not part of tantivy's public API /// This method is not part of tantivy's public API
pub fn save_new_metas(schema: Schema, directory: &mut Directory) -> Result<()> { pub fn save_new_metas(schema: Schema, opstamp: u64, directory: &mut Directory) -> Result<()> {
save_metas( save_metas(vec![], schema, opstamp, None, directory)
&IndexMeta {
segments: Vec::new(),
schema,
opstamp: 0u64,
payload: None,
},
directory,
)
} }
/// Save the index meta file. /// Save the index meta file.
/// This operation is atomic: /// This operation is atomic:
/// Either /// Either
/// - it fails, in which case an error is returned, // - it fails, in which case an error is returned,
/// and the `meta.json` remains untouched, /// and the `meta.json` remains untouched,
/// - it success, and `meta.json` is written /// - it success, and `meta.json` is written
/// and flushed. /// and flushed.
/// ///
/// This method is not part of tantivy's public API /// This method is not part of tantivy's public API
fn save_metas(metas: &IndexMeta, directory: &mut Directory) -> Result<()> { pub fn save_metas(
let mut buffer = serde_json::to_vec_pretty(metas)?; segment_metas: Vec<SegmentMeta>,
// Just adding a new line at the end of the buffer. schema: Schema,
opstamp: u64,
payload: Option<String>,
directory: &mut Directory,
) -> Result<()> {
let metas = IndexMeta {
segments: segment_metas,
schema,
opstamp,
payload,
};
let mut buffer = serde_json::to_vec_pretty(&metas)?;
writeln!(&mut buffer)?; writeln!(&mut buffer)?;
directory.atomic_write(&META_FILEPATH, &buffer[..])?; directory.atomic_write(&META_FILEPATH, &buffer[..])?;
debug!("Saved metas {:?}", serde_json::to_string_pretty(&metas)); debug!("Saved metas {:?}", serde_json::to_string_pretty(&metas));
@@ -82,21 +83,16 @@ fn save_metas(metas: &IndexMeta, directory: &mut Directory) -> Result<()> {
// //
// All this processing happens on a single thread // All this processing happens on a single thread
// consuming a common queue. // consuming a common queue.
//
// We voluntarily pass a merge_operation ref to guarantee that
// the merge_operation is alive during the process
#[derive(Clone)] #[derive(Clone)]
pub struct SegmentUpdater(Arc<InnerSegmentUpdater>); pub struct SegmentUpdater(Arc<InnerSegmentUpdater>);
fn perform_merge( fn perform_merge(
merge_operation: &MergeOperation,
index: &Index, index: &Index,
mut segment_entries: Vec<SegmentEntry>, mut segment_entries: Vec<SegmentEntry>,
mut merged_segment: Segment,
target_opstamp: u64,
) -> Result<SegmentEntry> { ) -> Result<SegmentEntry> {
let target_opstamp = merge_operation.target_opstamp();
// first we need to apply deletes to our segment. // first we need to apply deletes to our segment.
let mut merged_segment = index.new_segment();
// TODO add logging // TODO add logging
let schema = index.schema(); let schema = index.schema();
@@ -125,68 +121,60 @@ fn perform_merge(
let segment_meta = SegmentMeta::new(merged_segment.id(), num_docs); let segment_meta = SegmentMeta::new(merged_segment.id(), num_docs);
let after_merge_segment_entry = SegmentEntry::new(segment_meta.clone(), delete_cursor, None, target_opstamp); let after_merge_segment_entry = SegmentEntry::new(segment_meta.clone(), delete_cursor, None);
Ok(after_merge_segment_entry) Ok(after_merge_segment_entry)
} }
struct InnerSegmentUpdater { struct InnerSegmentUpdater {
// we keep a copy of the current active IndexMeta to
// avoid loading the file everytime we need it in the
// `SegmentUpdater`.
//
// This should be up to date as all update happen through
// the unique active `SegmentUpdater`.
active_metas: RwLock<Arc<IndexMeta>>,
pool: CpuPool, pool: CpuPool,
index: Index, index: Index,
segment_manager: SegmentManager, segment_manager: SegmentManager,
merge_policy: RwLock<Arc<Box<MergePolicy>>>, merge_policy: RwLock<Box<MergePolicy>>,
merging_thread_id: AtomicUsize, merging_thread_id: AtomicUsize,
merging_threads: RwLock<HashMap<usize, JoinHandle<Result<()>>>>, merging_threads: RwLock<HashMap<usize, JoinHandle<Result<()>>>>,
generation: AtomicUsize, generation: AtomicUsize,
killed: AtomicBool, killed: AtomicBool,
stamper: Stamper, stamper: Stamper,
merge_operations: MergeOperationInventory,
} }
impl SegmentUpdater { impl SegmentUpdater {
pub fn create( pub fn new(
index: Index, index: Index,
stamper: Stamper, stamper: Stamper,
delete_cursor: &DeleteCursor, delete_cursor: &DeleteCursor,
) -> Result<SegmentUpdater> { ) -> Result<SegmentUpdater> {
let index_meta = index.load_metas()?;
let segments = index.searchable_segment_metas()?; let segments = index.searchable_segment_metas()?;
let opstamp = index_meta.opstamp; let segment_manager = SegmentManager::from_segments(segments, delete_cursor);
let segment_manager = SegmentManager::from_segments(segments, delete_cursor, opstamp);
let pool = CpuPoolBuilder::new() let pool = CpuPoolBuilder::new()
.name_prefix("segment_updater") .name_prefix("segment_updater")
.pool_size(1) .pool_size(1)
.create(); .create();
let index_meta = index.load_metas()?;
Ok(SegmentUpdater(Arc::new(InnerSegmentUpdater { Ok(SegmentUpdater(Arc::new(InnerSegmentUpdater {
active_metas: RwLock::new(Arc::new(index_meta)),
pool, pool,
index, index,
segment_manager, segment_manager,
merge_policy: RwLock::new(Arc::new(Box::new(DefaultMergePolicy::default()))), merge_policy: RwLock::new(Box::new(DefaultMergePolicy::default())),
merging_thread_id: AtomicUsize::default(), merging_thread_id: AtomicUsize::default(),
merging_threads: RwLock::new(HashMap::new()), merging_threads: RwLock::new(HashMap::new()),
generation: AtomicUsize::default(), generation: AtomicUsize::default(),
killed: AtomicBool::new(false), killed: AtomicBool::new(false),
stamper, stamper,
merge_operations: Default::default(),
}))) })))
} }
pub fn get_merge_policy(&self) -> Arc<Box<MergePolicy>> { pub fn new_segment(&self) -> Segment {
self.0.merge_policy.read().unwrap().clone() let new_segment = self.0.index.new_segment();
let segment_id = new_segment.id();
self.0.segment_manager.write_segment(segment_id);
new_segment
}
pub fn get_merge_policy(&self) -> Box<MergePolicy> {
self.0.merge_policy.read().unwrap().box_clone()
} }
pub fn set_merge_policy(&self, merge_policy: Box<MergePolicy>) { pub fn set_merge_policy(&self, merge_policy: Box<MergePolicy>) {
let arc_merge_policy = Arc::new(merge_policy); *self.0.merge_policy.write().unwrap() = merge_policy;
*self.0.merge_policy.write().unwrap() = arc_merge_policy;
} }
fn get_merging_thread_id(&self) -> usize { fn get_merging_thread_id(&self) -> usize {
@@ -207,8 +195,7 @@ impl SegmentUpdater {
segment_updater.0.segment_manager.add_segment(segment_entry); segment_updater.0.segment_manager.add_segment(segment_entry);
segment_updater.consider_merge_options(); segment_updater.consider_merge_options();
true true
}) }).forget();
.forget();
true true
} else { } else {
false false
@@ -240,39 +227,20 @@ impl SegmentUpdater {
if self.is_alive() { if self.is_alive() {
let index = &self.0.index; let index = &self.0.index;
let directory = index.directory(); let directory = index.directory();
let mut commited_segment_metas = self.0.segment_manager.committed_segment_metas(); save_metas(
self.0.segment_manager.committed_segment_metas(),
// We sort segment_readers by number of documents. index.schema(),
// This is an heuristic to make multithreading more efficient.
//
// This is not done at the searcher level because I had a strange
// use case in which I was dealing with a large static index,
// dispatched over 5 SSD drives.
//
// A `UnionDirectory` makes it possible to read from these
// 5 different drives and creates a meta.json on the fly.
// In order to optimize the throughput, it creates a lasagna of segments
// from the different drives.
//
// Segment 1 from disk 1, Segment 1 from disk 2, etc.
commited_segment_metas.sort_by_key(|segment_meta| -(segment_meta.max_doc() as i32));
let index_meta = IndexMeta {
segments: commited_segment_metas,
schema: index.schema(),
opstamp, opstamp,
payload: commit_message, commit_message,
}; directory.box_clone().borrow_mut(),
save_metas(&index_meta, directory.box_clone().borrow_mut()) ).expect("Could not save metas.");
.expect("Could not save metas.");
self.store_meta(&index_meta);
} }
} }
pub fn garbage_collect_files(&self) -> Result<()> { pub fn garbage_collect_files(&self) -> Result<()> {
self.run_async(move |segment_updater| { self.run_async(move |segment_updater| {
segment_updater.garbage_collect_files_exec(); segment_updater.garbage_collect_files_exec();
}) }).wait()
.wait()
} }
fn garbage_collect_files_exec(&self) { fn garbage_collect_files_exec(&self) {
@@ -283,94 +251,65 @@ impl SegmentUpdater {
.garbage_collect(|| self.0.segment_manager.list_files()); .garbage_collect(|| self.0.segment_manager.list_files());
} }
pub fn commit(&self, opstamp: u64, payload: Option<String>, soft: bool) -> Result<()> { pub fn commit(&self, opstamp: u64, payload: Option<String>) -> Result<()> {
self.run_async(move |segment_updater| { self.run_async(move |segment_updater| {
if segment_updater.is_alive() { if segment_updater.is_alive() {
let segment_entries = segment_updater let segment_entries = segment_updater
.purge_deletes(opstamp) .purge_deletes(opstamp)
.expect("Failed purge deletes"); .expect("Failed purge deletes");
if soft { segment_updater.0.segment_manager.commit(segment_entries);
// Soft commit. segment_updater.save_metas(opstamp, payload);
//
// The list `segment_entries` above is what we might want to use as searchable
// segment. However, we do not want to mark them as committed, and we want
// to keep the current set of committed segment.
segment_updater.0.segment_manager.soft_commit(opstamp, segment_entries);
// ... We do not save the meta file.
} else {
// Hard_commit. We register the new segment entries as committed.
segment_updater
.0
.segment_manager
.commit(opstamp, segment_entries);
// TODO error handling.
segment_updater.save_metas(opstamp, payload);
segment_updater.0.index.directory().flush().unwrap();
}
segment_updater.garbage_collect_files_exec(); segment_updater.garbage_collect_files_exec();
segment_updater.consider_merge_options(); segment_updater.consider_merge_options();
} }
}) }).wait()
.wait()
} }
pub fn start_merge(&self, segment_ids: &[SegmentId]) -> Result<Receiver<SegmentMeta>> { pub fn start_merge(&self, segment_ids: &[SegmentId]) -> Result<Receiver<SegmentMeta>> {
let commit_opstamp = self.load_metas().opstamp; //let future_merged_segment = */
let merge_operation = MergeOperation::new( let segment_ids_vec = segment_ids.to_vec();
&self.0.merge_operations, self.run_async(move |segment_updater| {
commit_opstamp, segment_updater.start_merge_impl(&segment_ids_vec[..])
segment_ids.to_vec(), }).wait()?
);
self.run_async(move |segment_updater| segment_updater.start_merge_impl(merge_operation))
.wait()?
}
fn store_meta(&self, index_meta: &IndexMeta) {
*self.0.active_metas.write().unwrap() = Arc::new(index_meta.clone());
}
fn load_metas(&self) -> Arc<IndexMeta> {
self.0.active_metas.read().unwrap().clone()
} }
// `segment_ids` is required to be non-empty. // `segment_ids` is required to be non-empty.
fn start_merge_impl(&self, merge_operation: MergeOperation) -> Result<Receiver<SegmentMeta>> { fn start_merge_impl(&self, segment_ids: &[SegmentId]) -> Result<Receiver<SegmentMeta>> {
assert!( assert!(!segment_ids.is_empty(), "Segment_ids cannot be empty.");
!merge_operation.segment_ids().is_empty(),
"Segment_ids cannot be empty."
);
let segment_updater_clone = self.clone(); let segment_updater_clone = self.clone();
let segment_entries: Vec<SegmentEntry> = self let segment_entries: Vec<SegmentEntry> = self.0.segment_manager.start_merge(segment_ids)?;
.0
.segment_manager
.start_merge(merge_operation.segment_ids())?;
// let segment_ids_vec = merge_operation.segment_ids.to_vec(); let segment_ids_vec = segment_ids.to_vec();
let merging_thread_id = self.get_merging_thread_id(); let merging_thread_id = self.get_merging_thread_id();
info!( info!(
"Starting merge thread #{} - {:?}", "Starting merge thread #{} - {:?}",
merging_thread_id, merging_thread_id, segment_ids
merge_operation.segment_ids()
); );
let (merging_future_send, merging_future_recv) = oneshot(); let (merging_future_send, merging_future_recv) = oneshot();
let target_opstamp = self.0.stamper.stamp();
// first we need to apply deletes to our segment. // first we need to apply deletes to our segment.
let merging_join_handle = thread::Builder::new() let merging_join_handle = thread::Builder::new()
.name(format!("mergingthread-{}", merging_thread_id)) .name(format!("mergingthread-{}", merging_thread_id))
.spawn(move || { .spawn(move || {
// first we need to apply deletes to our segment. // first we need to apply deletes to our segment.
let merged_segment = segment_updater_clone.new_segment();
let merged_segment_id = merged_segment.id();
let merge_result = perform_merge( let merge_result = perform_merge(
&merge_operation,
&segment_updater_clone.0.index, &segment_updater_clone.0.index,
segment_entries, segment_entries,
merged_segment,
target_opstamp,
); );
match merge_result { match merge_result {
Ok(after_merge_segment_entry) => { Ok(after_merge_segment_entry) => {
let merged_segment_meta = after_merge_segment_entry.meta().clone(); let merged_segment_meta = after_merge_segment_entry.meta().clone();
segment_updater_clone segment_updater_clone
.end_merge(merge_operation, after_merge_segment_entry) .end_merge(segment_ids_vec, after_merge_segment_entry)
.expect("Segment updater thread is corrupted."); .expect("Segment updater thread is corrupted.");
// the future may fail if the listener of the oneshot future // the future may fail if the listener of the oneshot future
@@ -381,18 +320,13 @@ impl SegmentUpdater {
let _merging_future_res = merging_future_send.send(merged_segment_meta); let _merging_future_res = merging_future_send.send(merged_segment_meta);
} }
Err(e) => { Err(e) => {
warn!( warn!("Merge of {:?} was cancelled: {:?}", segment_ids_vec, e);
"Merge of {:?} was cancelled: {:?}",
merge_operation.segment_ids(),
e
);
// ... cancel merge // ... cancel merge
if cfg!(test) { if cfg!(test) {
panic!("Merge failed."); panic!("Merge failed.");
} }
// As `merge_operation` will be dropped, the segment in merge state will segment_updater_clone.cancel_merge(&segment_ids_vec, merged_segment_id);
// be available for merge again. // merging_future_send will be dropped, sending an error to the future.
// `merging_future_send` will be dropped, sending an error to the future.
} }
} }
segment_updater_clone segment_updater_clone
@@ -402,8 +336,7 @@ impl SegmentUpdater {
.unwrap() .unwrap()
.remove(&merging_thread_id); .remove(&merging_thread_id);
Ok(()) Ok(())
}) }).expect("Failed to spawn a thread.");
.expect("Failed to spawn a thread.");
self.0 self.0
.merging_threads .merging_threads
.write() .write()
@@ -413,35 +346,16 @@ impl SegmentUpdater {
} }
fn consider_merge_options(&self) { fn consider_merge_options(&self) {
let merge_segment_ids: HashSet<SegmentId> = self.0.merge_operations.segment_in_merge();
let (committed_segments, uncommitted_segments) = let (committed_segments, uncommitted_segments) =
get_mergeable_segments(&merge_segment_ids, &self.0.segment_manager); get_mergeable_segments(&self.0.segment_manager);
// Committed segments cannot be merged with uncommitted_segments. // Committed segments cannot be merged with uncommitted_segments.
// We therefore consider merges using these two sets of segments independently. // We therefore consider merges using these two sets of segments independently.
let merge_policy = self.get_merge_policy(); let merge_policy = self.get_merge_policy();
let mut merge_candidates = merge_policy.compute_merge_candidates(&uncommitted_segments);
let current_opstamp = self.0.stamper.stamp(); let committed_merge_candidates = merge_policy.compute_merge_candidates(&committed_segments);
let mut merge_candidates: Vec<MergeOperation> = merge_policy merge_candidates.extend_from_slice(&committed_merge_candidates[..]);
.compute_merge_candidates(&uncommitted_segments) for MergeCandidate(segment_metas) in merge_candidates {
.into_iter() match self.start_merge_impl(&segment_metas) {
.map(|merge_candidate| {
MergeOperation::new(&self.0.merge_operations, current_opstamp, merge_candidate.0)
})
.collect();
let commit_opstamp = self.load_metas().opstamp;
let committed_merge_candidates = merge_policy
.compute_merge_candidates(&committed_segments)
.into_iter()
.map(|merge_candidate| {
MergeOperation::new(&self.0.merge_operations, commit_opstamp, merge_candidate.0)
})
.collect::<Vec<_>>();
merge_candidates.extend(committed_merge_candidates.into_iter());
for merge_operation in merge_candidates {
match self.start_merge_impl(merge_operation) {
Ok(merge_future) => { Ok(merge_future) => {
if let Err(e) = merge_future.fuse().poll() { if let Err(e) = merge_future.fuse().poll() {
error!("The merge task failed quickly after starting: {:?}", e); error!("The merge task failed quickly after starting: {:?}", e);
@@ -457,16 +371,31 @@ impl SegmentUpdater {
} }
} }
fn cancel_merge(
&self,
before_merge_segment_ids: &[SegmentId],
after_merge_segment_entry: SegmentId,
) {
self.0
.segment_manager
.cancel_merge(before_merge_segment_ids, after_merge_segment_entry);
}
fn end_merge( fn end_merge(
&self, &self,
merge_operation: MergeOperation, before_merge_segment_ids: Vec<SegmentId>,
mut after_merge_segment_entry: SegmentEntry, mut after_merge_segment_entry: SegmentEntry,
) -> Result<()> { ) -> Result<()> {
self.run_async(move |segment_updater| { self.run_async(move |segment_updater| {
info!("End merge {:?}", after_merge_segment_entry.meta()); info!("End merge {:?}", after_merge_segment_entry.meta());
let mut delete_cursor = after_merge_segment_entry.delete_cursor().clone(); let mut delete_cursor = after_merge_segment_entry.delete_cursor().clone();
if let Some(delete_operation) = delete_cursor.get() { if let Some(delete_operation) = delete_cursor.get() {
let committed_opstamp = segment_updater.load_metas().opstamp; let committed_opstamp = segment_updater
.0
.index
.load_metas()
.expect("Failed to read opstamp")
.opstamp;
if delete_operation.opstamp < committed_opstamp { if delete_operation.opstamp < committed_opstamp {
let index = &segment_updater.0.index; let index = &segment_updater.0.index;
let segment = index.segment(after_merge_segment_entry.meta().clone()); let segment = index.segment(after_merge_segment_entry.meta().clone());
@@ -475,15 +404,16 @@ impl SegmentUpdater {
{ {
error!( error!(
"Merge of {:?} was cancelled (advancing deletes failed): {:?}", "Merge of {:?} was cancelled (advancing deletes failed): {:?}",
merge_operation.segment_ids(), before_merge_segment_ids, e
e
); );
// ... cancel merge
if cfg!(test) { if cfg!(test) {
panic!("Merge failed."); panic!("Merge failed.");
} }
// ... cancel merge segment_updater.cancel_merge(
// `merge_operations` are tracked. As it is dropped, the &before_merge_segment_ids,
// the segment_ids will be available again for merge. after_merge_segment_entry.segment_id(),
);
return; return;
} }
} }
@@ -491,14 +421,13 @@ impl SegmentUpdater {
segment_updater segment_updater
.0 .0
.segment_manager .segment_manager
.end_merge(merge_operation.segment_ids(), after_merge_segment_entry); .end_merge(&before_merge_segment_ids, after_merge_segment_entry);
segment_updater.consider_merge_options(); segment_updater.consider_merge_options();
info!("save metas"); info!("save metas");
let previous_metas = segment_updater.load_metas(); let previous_metas = segment_updater.0.index.load_metas().unwrap();
segment_updater.save_metas(previous_metas.opstamp, previous_metas.payload.clone()); segment_updater.save_metas(previous_metas.opstamp, previous_metas.payload);
segment_updater.garbage_collect_files_exec(); segment_updater.garbage_collect_files_exec();
}) }).wait()
.wait()
} }
/// Wait for current merging threads. /// Wait for current merging threads.
@@ -517,25 +446,32 @@ impl SegmentUpdater {
/// Obsolete files will eventually be cleaned up /// Obsolete files will eventually be cleaned up
/// by the directory garbage collector. /// by the directory garbage collector.
pub fn wait_merging_thread(&self) -> Result<()> { pub fn wait_merging_thread(&self) -> Result<()> {
let mut num_segments: usize;
loop { loop {
let merging_threads: HashMap<usize, JoinHandle<Result<()>>> = { num_segments = self.0.segment_manager.num_segments();
let mut new_merging_threads = HashMap::new();
{
let mut merging_threads = self.0.merging_threads.write().unwrap(); let mut merging_threads = self.0.merging_threads.write().unwrap();
mem::replace(merging_threads.deref_mut(), HashMap::new()) mem::swap(&mut new_merging_threads, merging_threads.deref_mut());
};
if merging_threads.is_empty() {
return Ok(());
} }
debug!("wait merging thread {}", merging_threads.len()); debug!("wait merging thread {}", new_merging_threads.len());
for (_, merging_thread_handle) in merging_threads { for (_, merging_thread_handle) in new_merging_threads {
merging_thread_handle merging_thread_handle
.join() .join()
.map(|_| ()) .map(|_| ())
.map_err(|_| TantivyError::ErrorInThread("Merging thread failed.".into()))?; .map_err(|_| TantivyError::ErrorInThread("Merging thread failed.".into()))?;
} }
// Our merging thread may have queued their completed merged segment. // Our merging thread may have queued their completed
// Let's wait for that too.
self.run_async(move |_| {}).wait()?; self.run_async(move |_| {}).wait()?;
let new_num_segments = self.0.segment_manager.num_segments();
if new_num_segments >= num_segments {
break;
}
} }
Ok(())
} }
} }
@@ -548,14 +484,14 @@ mod tests {
#[test] #[test]
fn test_delete_during_merge() { fn test_delete_during_merge() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
// writing the segment // writing the segment
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
index_writer.set_merge_policy(Box::new(MergeWheneverPossible)); index_writer.set_merge_policy(Box::new(MergeWheneverPossible));
{ {
@@ -585,8 +521,9 @@ mod tests {
index_writer.delete_term(term); index_writer.delete_term(term);
assert!(index_writer.commit().is_ok()); assert!(index_writer.commit().is_ok());
} }
let reader = index.reader().unwrap();
assert_eq!(reader.searcher().num_docs(), 302); index.load_searchers().unwrap();
assert_eq!(index.searcher().num_docs(), 302);
{ {
index_writer index_writer
@@ -594,79 +531,8 @@ mod tests {
.expect("waiting for merging threads"); .expect("waiting for merging threads");
} }
reader.reload().unwrap(); index.load_searchers().unwrap();
assert_eq!(reader.searcher().segment_readers().len(), 1); assert_eq!(index.searcher().segment_readers().len(), 1);
assert_eq!(reader.searcher().num_docs(), 302); assert_eq!(index.searcher().num_docs(), 302);
}
#[test]
fn delete_all_docs() {
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
// writing the segment
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
{
for _ in 0..100 {
index_writer.add_document(doc!(text_field=>"a"));
index_writer.add_document(doc!(text_field=>"b"));
}
assert!(index_writer.commit().is_ok());
}
{
for _ in 0..100 {
index_writer.add_document(doc!(text_field=>"c"));
index_writer.add_document(doc!(text_field=>"d"));
}
assert!(index_writer.commit().is_ok());
}
{
index_writer.add_document(doc!(text_field=>"e"));
index_writer.add_document(doc!(text_field=>"f"));
assert!(index_writer.commit().is_ok());
}
{
let seg_ids = index
.searchable_segment_ids()
.expect("Searchable segments failed.");
// docs exist, should have at least 1 segment
assert!(seg_ids.len() > 0);
}
{
let term_vals = vec!["a", "b", "c", "d", "e", "f"];
for term_val in term_vals {
let term = Term::from_field_text(text_field, term_val);
index_writer.delete_term(term);
assert!(index_writer.commit().is_ok());
}
}
{
index_writer
.wait_merging_threads()
.expect("waiting for merging threads");
}
let reader = index.reader().unwrap();
assert_eq!(reader.searcher().num_docs(), 0);
let seg_ids = index
.searchable_segment_ids()
.expect("Searchable segments failed.");
assert!(seg_ids.is_empty());
reader.reload().unwrap();
assert_eq!(reader.searcher().num_docs(), 0);
// empty segments should be erased
assert!(index.searchable_segment_metas().unwrap().is_empty());
assert!(reader.searcher().segment_readers().is_empty());
} }
} }

27
src/indexer/segment_writer.rs
View File

@@ -62,8 +62,7 @@ impl SegmentWriter {
segment.index().tokenizers().get(tokenizer_name) segment.index().tokenizers().get(tokenizer_name)
}), }),
_ => None, _ => None,
}) }).collect();
.collect();
Ok(SegmentWriter { Ok(SegmentWriter {
max_doc: 0, max_doc: 0,
multifield_postings, multifield_postings,
@@ -111,18 +110,18 @@ impl SegmentWriter {
} }
match *field_options.field_type() { match *field_options.field_type() {
FieldType::HierarchicalFacet => { FieldType::HierarchicalFacet => {
let facets: Vec<&str> = field_values let facets: Vec<&[u8]> = field_values
.iter() .iter()
.flat_map(|field_value| match *field_value.value() { .flat_map(|field_value| match *field_value.value() {
Value::Facet(ref facet) => Some(facet.encoded_str()), Value::Facet(ref facet) => Some(facet.encoded_bytes()),
_ => { _ => {
panic!("Expected hierarchical facet"); panic!("Expected hierarchical facet");
} }
}) }).collect();
.collect();
let mut term = Term::for_field(field); // we set the Term let mut term = Term::for_field(field); // we set the Term
for fake_str in facets { for facet_bytes in facets {
let mut unordered_term_id_opt = None; let mut unordered_term_id_opt = None;
let fake_str = unsafe { str::from_utf8_unchecked(facet_bytes) };
FacetTokenizer.token_stream(fake_str).process(&mut |token| { FacetTokenizer.token_stream(fake_str).process(&mut |token| {
term.set_text(&token.text); term.set_text(&token.text);
let unordered_term_id = let unordered_term_id =
@@ -146,8 +145,7 @@ impl SegmentWriter {
.flat_map(|field_value| match *field_value.value() { .flat_map(|field_value| match *field_value.value() {
Value::Str(ref text) => Some(text.as_str()), Value::Str(ref text) => Some(text.as_str()),
_ => None, _ => None,
}) }).collect();
.collect();
if texts.is_empty() { if texts.is_empty() {
0 0
} else { } else {
@@ -171,17 +169,6 @@ impl SegmentWriter {
} }
} }
} }
FieldType::Date(ref int_option) => {
if int_option.is_indexed() {
for field_value in field_values {
let term = Term::from_field_i64(
field_value.field(),
field_value.value().date_value().timestamp(),
);
self.multifield_postings.subscribe(doc_id, &term);
}
}
}
FieldType::I64(ref int_option) => { FieldType::I64(ref int_option) => {
if int_option.is_indexed() { if int_option.is_indexed() {
for field_value in field_values { for field_value in field_values {

73
src/indexer/stamper.rs
View File

@@ -1,77 +1,50 @@
use std::ops::Range;
use std::sync::atomic::Ordering;
use std::sync::Arc;
// AtomicU64 have not landed in stable. // AtomicU64 have not landed in stable.
// For the moment let's just use AtomicUsize on // For the moment let's just use AtomicUsize on
// x86/64 bit platform, and a mutex on other platform. // x86/64 bit platform, and a mutex on other platform.
#[cfg(target_arch = "x86_64")]
#[cfg(target = "x86_64")]
mod archicture_impl { mod archicture_impl {
use std::sync::atomic::{AtomicUsize, Ordering}; use std::sync::atomic::{AtomicUsize, Ordering};
use std::sync::Arc;
#[derive(Default)] #[derive(Clone, Default)]
pub struct AtomicU64Ersatz(AtomicUsize); pub struct Stamper(Arc<AtomicU64>);
impl AtomicU64Ersatz { impl Stamper {
pub fn new(first_opstamp: u64) -> AtomicU64Ersatz { pub fn new(first_opstamp: u64) -> Stamper {
AtomicU64Ersatz(AtomicUsize::new(first_opstamp as usize)) Stamper(Arc::new(AtomicU64::new(first_opstamp)))
} }
pub fn fetch_add(&self, val: u64, order: Ordering) -> u64 { pub fn stamp(&self) -> u64 {
self.0.fetch_add(val as usize, order) as u64 self.0.fetch_add(1u64, Ordering::SeqCst) as u64
} }
} }
} }
#[cfg(not(target_arch = "x86_64"))] #[cfg(not(target = "x86_64"))]
mod archicture_impl { mod archicture_impl {
use std::sync::atomic::Ordering; use std::sync::{Arc, Mutex};
/// Under other architecture, we rely on a mutex.
use std::sync::RwLock;
#[derive(Default)] #[derive(Clone, Default)]
pub struct AtomicU64Ersatz(RwLock<u64>); pub struct Stamper(Arc<Mutex<u64>>);
impl AtomicU64Ersatz { impl Stamper {
pub fn new(first_opstamp: u64) -> AtomicU64Ersatz { pub fn new(first_opstamp: u64) -> Stamper {
AtomicU64Ersatz(RwLock::new(first_opstamp)) Stamper(Arc::new(Mutex::new(first_opstamp)))
} }
pub fn fetch_add(&self, incr: u64, _order: Ordering) -> u64 { pub fn stamp(&self) -> u64 {
let mut lock = self.0.write().unwrap(); let mut guard = self.0.lock().expect("Failed to lock the stamper");
let previous_val = *lock; let previous_val = *guard;
*lock = previous_val + incr; *guard = previous_val + 1;
previous_val previous_val
} }
} }
} }
use self::archicture_impl::AtomicU64Ersatz; pub use self::archicture_impl::Stamper;
#[derive(Clone, Default)]
pub struct Stamper(Arc<AtomicU64Ersatz>);
impl Stamper {
pub fn new(first_opstamp: u64) -> Stamper {
Stamper(Arc::new(AtomicU64Ersatz::new(first_opstamp)))
}
pub fn stamp(&self) -> u64 {
self.0.fetch_add(1u64, Ordering::SeqCst) as u64
}
/// Given a desired count `n`, `stamps` returns an iterator that
/// will supply `n` number of u64 stamps.
pub fn stamps(&self, n: u64) -> Range<u64> {
let start = self.0.fetch_add(n, Ordering::SeqCst);
Range {
start,
end: start + n,
}
}
}
#[cfg(test)] #[cfg(test)]
mod test { mod test {
@@ -89,7 +62,5 @@ mod test {
assert_eq!(stamper.stamp(), 10u64); assert_eq!(stamper.stamp(), 10u64);
assert_eq!(stamper_clone.stamp(), 11u64); assert_eq!(stamper_clone.stamp(), 11u64);
assert_eq!(stamper.stamps(3u64), (12..15));
assert_eq!(stamper.stamp(), 15u64);
} }
} }

460
src/lib.rs
View File

@@ -1,5 +1,6 @@
#![doc(html_logo_url = "http://fulmicoton.com/tantivy-logo/tantivy-logo.png")] #![doc(html_logo_url = "http://fulmicoton.com/tantivy-logo/tantivy-logo.png")]
#![cfg_attr(all(feature = "unstable", test), feature(test))] #![cfg_attr(all(feature = "unstable", test), feature(test))]
#![cfg_attr(feature = "cargo-clippy", feature(tool_lints))]
#![cfg_attr(feature = "cargo-clippy", allow(clippy::module_inception))] #![cfg_attr(feature = "cargo-clippy", allow(clippy::module_inception))]
#![doc(test(attr(allow(unused_variables), deny(warnings))))] #![doc(test(attr(allow(unused_variables), deny(warnings))))]
#![warn(missing_docs)] #![warn(missing_docs)]
@@ -23,8 +24,7 @@
//! # use tempdir::TempDir; //! # use tempdir::TempDir;
//! # use tantivy::Index; //! # use tantivy::Index;
//! # use tantivy::schema::*; //! # use tantivy::schema::*;
//! # use tantivy::{Score, DocAddress}; //! # use tantivy::collector::TopCollector;
//! # use tantivy::collector::TopDocs;
//! # use tantivy::query::QueryParser; //! # use tantivy::query::QueryParser;
//! # //! #
//! # fn main() { //! # fn main() {
@@ -46,7 +46,7 @@
//! // in a compressed, row-oriented key-value store. //! // in a compressed, row-oriented key-value store.
//! // This store is useful to reconstruct the //! // This store is useful to reconstruct the
//! // documents that were selected during the search phase. //! // documents that were selected during the search phase.
//! let mut schema_builder = Schema::builder(); //! let mut schema_builder = SchemaBuilder::default();
//! let title = schema_builder.add_text_field("title", TEXT | STORED); //! let title = schema_builder.add_text_field("title", TEXT | STORED);
//! let body = schema_builder.add_text_field("body", TEXT); //! let body = schema_builder.add_text_field("body", TEXT);
//! let schema = schema_builder.build(); //! let schema = schema_builder.build();
@@ -75,9 +75,9 @@
//! //!
//! // # Searching //! // # Searching
//! //!
//! let reader = index.reader()?; //! index.load_searchers()?;
//! //!
//! let searcher = reader.searcher(); //! let searcher = index.searcher();
//! //!
//! let query_parser = QueryParser::for_index(&index, vec![title, body]); //! let query_parser = QueryParser::for_index(&index, vec![title, body]);
//! //!
@@ -86,13 +86,13 @@
//! // A ticket has been opened regarding this problem. //! // A ticket has been opened regarding this problem.
//! let query = query_parser.parse_query("sea whale")?; //! let query = query_parser.parse_query("sea whale")?;
//! //!
//! // Perform search. //! let mut top_collector = TopCollector::with_limit(10);
//! // `topdocs` contains the 10 most relevant doc ids, sorted by decreasing scores... //! searcher.search(&*query, &mut top_collector)?;
//! let top_docs: Vec<(Score, DocAddress)> =
//! searcher.search(&query, &TopDocs::with_limit(10))?;
//! //!
//! for (_score, doc_address) in top_docs { //! // Our top collector now contains the 10
//! // Retrieve the actual content of documents given its `doc_address`. //! // most relevant doc ids...
//! let doc_addresses = top_collector.docs();
//! for doc_address in doc_addresses {
//! let retrieved_doc = searcher.doc(doc_address)?; //! let retrieved_doc = searcher.doc(doc_address)?;
//! println!("{}", schema.to_json(&retrieved_doc)); //! println!("{}", schema.to_json(&retrieved_doc));
//! } //! }
@@ -129,28 +129,31 @@ extern crate base64;
extern crate bit_set; extern crate bit_set;
extern crate bitpacking; extern crate bitpacking;
extern crate byteorder; extern crate byteorder;
extern crate combine; extern crate combine;
extern crate crossbeam; extern crate crossbeam;
extern crate crossbeam_channel;
extern crate fnv; extern crate fnv;
extern crate fst;
extern crate futures; extern crate futures;
extern crate futures_cpupool; extern crate futures_cpupool;
extern crate htmlescape; extern crate htmlescape;
extern crate itertools; extern crate itertools;
extern crate levenshtein_automata; extern crate levenshtein_automata;
#[cfg(feature = "mmap")]
extern crate memmap;
extern crate num_cpus; extern crate num_cpus;
extern crate owning_ref; extern crate owning_ref;
extern crate regex; extern crate regex;
extern crate rust_stemmers; extern crate rust_stemmers;
extern crate scoped_pool;
extern crate serde; extern crate serde;
extern crate stable_deref_trait; extern crate stable_deref_trait;
extern crate tantivy_fst;
extern crate tempdir; extern crate tempdir;
extern crate tempfile; extern crate tempfile;
extern crate uuid; extern crate uuid;
#[cfg(test)] #[cfg(test)]
#[macro_use] #[macro_use]
extern crate matches; extern crate matches;
@@ -169,12 +172,11 @@ extern crate maplit;
extern crate test; extern crate test;
#[macro_use] #[macro_use]
extern crate downcast_rs; extern crate downcast;
#[macro_use] #[macro_use]
extern crate fail; extern crate fail;
#[cfg(feature = "mmap")]
#[cfg(test)] #[cfg(test)]
mod functional_test; mod functional_test;
@@ -183,19 +185,18 @@ mod macros;
pub use error::TantivyError; pub use error::TantivyError;
#[deprecated(since = "0.7.0", note = "please use `tantivy::TantivyError` instead")] #[deprecated(
since = "0.7.0",
note = "please use `tantivy::TantivyError` instead"
)]
pub use error::TantivyError as Error; pub use error::TantivyError as Error;
extern crate census; extern crate census;
pub extern crate chrono;
extern crate owned_read; extern crate owned_read;
/// Tantivy result. /// Tantivy result.
pub type Result<T> = std::result::Result<T, error::TantivyError>; pub type Result<T> = std::result::Result<T, error::TantivyError>;
/// Tantivy DateTime
pub type DateTime = chrono::DateTime<chrono::Utc>;
mod common; mod common;
mod core; mod core;
mod indexer; mod indexer;
@@ -212,15 +213,11 @@ pub(crate) mod positions;
pub mod postings; pub mod postings;
pub mod query; pub mod query;
pub mod schema; pub mod schema;
pub mod space_usage;
pub mod store; pub mod store;
pub mod termdict; pub mod termdict;
mod reader;
pub use self::reader::{IndexReader, IndexReaderBuilder, ReloadPolicy};
mod snippet; mod snippet;
pub use self::snippet::{Snippet, SnippetGenerator}; pub use self::snippet::SnippetGenerator;
mod docset; mod docset;
pub use self::docset::{DocSet, SkipResult}; pub use self::docset::{DocSet, SkipResult};
@@ -238,7 +235,11 @@ pub use common::{i64_to_u64, u64_to_i64};
/// Expose the current version of tantivy, as well /// Expose the current version of tantivy, as well
/// whether it was compiled with the simd compression. /// whether it was compiled with the simd compression.
pub fn version() -> &'static str { pub fn version() -> &'static str {
env!("CARGO_PKG_VERSION") if cfg!(feature = "simdcompression") {
concat!(env!("CARGO_PKG_VERSION"), "-simd")
} else {
concat!(env!("CARGO_PKG_VERSION"), "-nosimd")
}
} }
/// Defines tantivy's merging strategy /// Defines tantivy's merging strategy
@@ -299,15 +300,12 @@ mod tests {
use docset::DocSet; use docset::DocSet;
use query::BooleanQuery; use query::BooleanQuery;
use rand::distributions::Bernoulli; use rand::distributions::Bernoulli;
use rand::distributions::Uniform; use rand::distributions::Range;
use rand::rngs::StdRng; use rand::{Rng, SeedableRng, XorShiftRng};
use rand::{Rng, SeedableRng};
use schema::*; use schema::*;
use DocAddress;
use Index; use Index;
use IndexWriter; use IndexWriter;
use Postings; use Postings;
use ReloadPolicy;
pub fn assert_nearly_equals(expected: f32, val: f32) { pub fn assert_nearly_equals(expected: f32, val: f32) {
assert!( assert!(
@@ -323,15 +321,16 @@ mod tests {
} }
pub fn generate_nonunique_unsorted(max_value: u32, n_elems: usize) -> Vec<u32> { pub fn generate_nonunique_unsorted(max_value: u32, n_elems: usize) -> Vec<u32> {
let seed: [u8; 32] = [1; 32]; let seed: [u8; 16] = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15];
StdRng::from_seed(seed) XorShiftRng::from_seed(seed)
.sample_iter(&Uniform::new(0u32, max_value)) .sample_iter(&Range::new(0u32, max_value))
.take(n_elems) .take(n_elems)
.collect::<Vec<u32>>() .collect::<Vec<u32>>()
} }
pub fn sample_with_seed(n: u32, ratio: f64, seed_val: u8) -> Vec<u32> { pub fn sample_with_seed(n: u32, ratio: f64, seed_val: u8) -> Vec<u32> {
StdRng::from_seed([seed_val; 32]) let seed: [u8; 16] = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, seed_val];
XorShiftRng::from_seed(seed)
.sample_iter(&Bernoulli::new(ratio)) .sample_iter(&Bernoulli::new(ratio))
.take(n as usize) .take(n as usize)
.enumerate() .enumerate()
@@ -346,13 +345,13 @@ mod tests {
#[test] #[test]
#[cfg(feature = "mmap")] #[cfg(feature = "mmap")]
fn test_indexing() { fn test_indexing() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_from_tempdir(schema).unwrap(); let index = Index::create_from_tempdir(schema).unwrap();
{ {
// writing the segment // writing the segment
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
{ {
let doc = doc!(text_field=>"af b"); let doc = doc!(text_field=>"af b");
index_writer.add_document(doc); index_writer.add_document(doc);
@@ -371,10 +370,10 @@ mod tests {
#[test] #[test]
fn test_docfreq1() { fn test_docfreq1() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
{ {
index_writer.add_document(doc!(text_field=>"a b c")); index_writer.add_document(doc!(text_field=>"a b c"));
index_writer.commit().unwrap(); index_writer.commit().unwrap();
@@ -396,8 +395,8 @@ mod tests {
index_writer.commit().unwrap(); index_writer.commit().unwrap();
} }
{ {
let reader = index.reader().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
let term_a = Term::from_field_text(text_field, "a"); let term_a = Term::from_field_text(text_field, "a");
assert_eq!(searcher.doc_freq(&term_a), 3); assert_eq!(searcher.doc_freq(&term_a), 3);
let term_b = Term::from_field_text(text_field, "b"); let term_b = Term::from_field_text(text_field, "b");
@@ -411,12 +410,12 @@ mod tests {
#[test] #[test]
fn test_fieldnorm_no_docs_with_field() { fn test_fieldnorm_no_docs_with_field() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let title_field = schema_builder.add_text_field("title", TEXT); let title_field = schema_builder.add_text_field("title", TEXT);
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
{ {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
{ {
let doc = doc!(text_field=>"a b c"); let doc = doc!(text_field=>"a b c");
index_writer.add_document(doc); index_writer.add_document(doc);
@@ -424,8 +423,8 @@ mod tests {
index_writer.commit().unwrap(); index_writer.commit().unwrap();
} }
{ {
let index_reader = index.reader().unwrap(); index.load_searchers().unwrap();
let searcher = index_reader.searcher(); let searcher = index.searcher();
let reader = searcher.segment_reader(0); let reader = searcher.segment_reader(0);
{ {
let fieldnorm_reader = reader.get_fieldnorms_reader(text_field); let fieldnorm_reader = reader.get_fieldnorms_reader(text_field);
@@ -440,11 +439,11 @@ mod tests {
#[test] #[test]
fn test_fieldnorm() { fn test_fieldnorm() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let index = Index::create_in_ram(schema_builder.build()); let index = Index::create_in_ram(schema_builder.build());
{ {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
{ {
let doc = doc!(text_field=>"a b c"); let doc = doc!(text_field=>"a b c");
index_writer.add_document(doc); index_writer.add_document(doc);
@@ -460,8 +459,8 @@ mod tests {
index_writer.commit().unwrap(); index_writer.commit().unwrap();
} }
{ {
let reader = index.reader().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
let segment_reader: &SegmentReader = searcher.segment_reader(0); let segment_reader: &SegmentReader = searcher.segment_reader(0);
let fieldnorms_reader = segment_reader.get_fieldnorms_reader(text_field); let fieldnorms_reader = segment_reader.get_fieldnorms_reader(text_field);
assert_eq!(fieldnorms_reader.fieldnorm(0), 3); assert_eq!(fieldnorms_reader.fieldnorm(0), 3);
@@ -481,7 +480,7 @@ mod tests {
#[test] #[test]
fn test_delete_postings1() { fn test_delete_postings1() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let term_abcd = Term::from_field_text(text_field, "abcd"); let term_abcd = Term::from_field_text(text_field, "abcd");
let term_a = Term::from_field_text(text_field, "a"); let term_a = Term::from_field_text(text_field, "a");
@@ -489,151 +488,183 @@ mod tests {
let term_c = Term::from_field_text(text_field, "c"); let term_c = Term::from_field_text(text_field, "c");
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::Manual)
.try_into()
.unwrap();
{ {
// writing the segment // writing the segment
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
// 0 {
index_writer.add_document(doc!(text_field=>"a b")); // 0
// 1 let doc = doc!(text_field=>"a b");
index_writer.add_document(doc!(text_field=>" a c")); index_writer.add_document(doc);
// 2 }
index_writer.add_document(doc!(text_field=>" b c")); {
// 3 // 1
index_writer.add_document(doc!(text_field=>" b d")); let doc = doc!(text_field=>" a c");
index_writer.add_document(doc);
index_writer.delete_term(Term::from_field_text(text_field, "c")); }
index_writer.delete_term(Term::from_field_text(text_field, "a")); {
// 4 // 2
index_writer.add_document(doc!(text_field=>" b c")); let doc = doc!(text_field=>" b c");
// 5 index_writer.add_document(doc);
index_writer.add_document(doc!(text_field=>" a")); }
{
// 3
let doc = doc!(text_field=>" b d");
index_writer.add_document(doc);
}
{
index_writer.delete_term(Term::from_field_text(text_field, "c"));
}
{
index_writer.delete_term(Term::from_field_text(text_field, "a"));
}
{
// 4
let doc = doc!(text_field=>" b c");
index_writer.add_document(doc);
}
{
// 5
let doc = doc!(text_field=>" a");
index_writer.add_document(doc);
}
index_writer.commit().unwrap(); index_writer.commit().unwrap();
} }
{ {
reader.reload().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
let segment_reader = searcher.segment_reader(0); let reader = searcher.segment_reader(0);
let inverted_index = segment_reader.inverted_index(text_field); let inverted_index = reader.inverted_index(text_field);
assert!(inverted_index assert!(
.read_postings(&term_abcd, IndexRecordOption::WithFreqsAndPositions) inverted_index
.is_none()); .read_postings(&term_abcd, IndexRecordOption::WithFreqsAndPositions)
.is_none()
);
{ {
let mut postings = inverted_index let mut postings = inverted_index
.read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions) .read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions)
.unwrap(); .unwrap();
assert!(advance_undeleted(&mut postings, segment_reader)); assert!(advance_undeleted(&mut postings, reader));
assert_eq!(postings.doc(), 5); assert_eq!(postings.doc(), 5);
assert!(!advance_undeleted(&mut postings, segment_reader)); assert!(!advance_undeleted(&mut postings, reader));
} }
{ {
let mut postings = inverted_index let mut postings = inverted_index
.read_postings(&term_b, IndexRecordOption::WithFreqsAndPositions) .read_postings(&term_b, IndexRecordOption::WithFreqsAndPositions)
.unwrap(); .unwrap();
assert!(advance_undeleted(&mut postings, segment_reader)); assert!(advance_undeleted(&mut postings, reader));
assert_eq!(postings.doc(), 3); assert_eq!(postings.doc(), 3);
assert!(advance_undeleted(&mut postings, segment_reader)); assert!(advance_undeleted(&mut postings, reader));
assert_eq!(postings.doc(), 4); assert_eq!(postings.doc(), 4);
assert!(!advance_undeleted(&mut postings, segment_reader)); assert!(!advance_undeleted(&mut postings, reader));
} }
} }
{ {
// writing the segment // writing the segment
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
// 0 {
index_writer.add_document(doc!(text_field=>"a b")); // 0
// 1 let doc = doc!(text_field=>"a b");
index_writer.delete_term(Term::from_field_text(text_field, "c")); index_writer.add_document(doc);
}
{
// 1
index_writer.delete_term(Term::from_field_text(text_field, "c"));
}
index_writer.rollback().unwrap(); index_writer.rollback().unwrap();
} }
{ {
reader.reload().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
let seg_reader = searcher.segment_reader(0); let reader = searcher.segment_reader(0);
let inverted_index = seg_reader.inverted_index(term_abcd.field()); let inverted_index = reader.inverted_index(term_abcd.field());
assert!(inverted_index assert!(
.read_postings(&term_abcd, IndexRecordOption::WithFreqsAndPositions) inverted_index
.is_none()); .read_postings(&term_abcd, IndexRecordOption::WithFreqsAndPositions)
.is_none()
);
{ {
let mut postings = inverted_index let mut postings = inverted_index
.read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions) .read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions)
.unwrap(); .unwrap();
assert!(advance_undeleted(&mut postings, seg_reader)); assert!(advance_undeleted(&mut postings, reader));
assert_eq!(postings.doc(), 5); assert_eq!(postings.doc(), 5);
assert!(!advance_undeleted(&mut postings, seg_reader)); assert!(!advance_undeleted(&mut postings, reader));
} }
{ {
let mut postings = inverted_index let mut postings = inverted_index
.read_postings(&term_b, IndexRecordOption::WithFreqsAndPositions) .read_postings(&term_b, IndexRecordOption::WithFreqsAndPositions)
.unwrap(); .unwrap();
assert!(advance_undeleted(&mut postings, seg_reader)); assert!(advance_undeleted(&mut postings, reader));
assert_eq!(postings.doc(), 3); assert_eq!(postings.doc(), 3);
assert!(advance_undeleted(&mut postings, seg_reader)); assert!(advance_undeleted(&mut postings, reader));
assert_eq!(postings.doc(), 4); assert_eq!(postings.doc(), 4);
assert!(!advance_undeleted(&mut postings, seg_reader)); assert!(!advance_undeleted(&mut postings, reader));
} }
} }
{ {
// writing the segment // writing the segment
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"a b")); {
index_writer.delete_term(Term::from_field_text(text_field, "c")); let doc = doc!(text_field=>"a b");
index_writer.add_document(doc);
}
{
index_writer.delete_term(Term::from_field_text(text_field, "c"));
}
index_writer.rollback().unwrap(); index_writer.rollback().unwrap();
index_writer.delete_term(Term::from_field_text(text_field, "a")); index_writer.delete_term(Term::from_field_text(text_field, "a"));
index_writer.commit().unwrap(); index_writer.commit().unwrap();
} }
{ {
reader.reload().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
let segment_reader = searcher.segment_reader(0); let reader = searcher.segment_reader(0);
let inverted_index = segment_reader.inverted_index(term_abcd.field()); let inverted_index = reader.inverted_index(term_abcd.field());
assert!(inverted_index assert!(
.read_postings(&term_abcd, IndexRecordOption::WithFreqsAndPositions) inverted_index
.is_none()); .read_postings(&term_abcd, IndexRecordOption::WithFreqsAndPositions)
.is_none()
);
{ {
let mut postings = inverted_index let mut postings = inverted_index
.read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions) .read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions)
.unwrap(); .unwrap();
assert!(!advance_undeleted(&mut postings, segment_reader)); assert!(!advance_undeleted(&mut postings, reader));
} }
{ {
let mut postings = inverted_index let mut postings = inverted_index
.read_postings(&term_b, IndexRecordOption::WithFreqsAndPositions) .read_postings(&term_b, IndexRecordOption::WithFreqsAndPositions)
.unwrap(); .unwrap();
assert!(advance_undeleted(&mut postings, segment_reader)); assert!(advance_undeleted(&mut postings, reader));
assert_eq!(postings.doc(), 3); assert_eq!(postings.doc(), 3);
assert!(advance_undeleted(&mut postings, segment_reader)); assert!(advance_undeleted(&mut postings, reader));
assert_eq!(postings.doc(), 4); assert_eq!(postings.doc(), 4);
assert!(!advance_undeleted(&mut postings, segment_reader)); assert!(!advance_undeleted(&mut postings, reader));
} }
{ {
let mut postings = inverted_index let mut postings = inverted_index
.read_postings(&term_c, IndexRecordOption::WithFreqsAndPositions) .read_postings(&term_c, IndexRecordOption::WithFreqsAndPositions)
.unwrap(); .unwrap();
assert!(advance_undeleted(&mut postings, segment_reader)); assert!(advance_undeleted(&mut postings, reader));
assert_eq!(postings.doc(), 4); assert_eq!(postings.doc(), 4);
assert!(!advance_undeleted(&mut postings, segment_reader)); assert!(!advance_undeleted(&mut postings, reader));
} }
} }
} }
#[test] #[test]
fn test_indexed_u64() { fn test_indexed_u64() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let field = schema_builder.add_u64_field("value", INDEXED); let field = schema_builder.add_u64_field("value", INT_INDEXED);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
index_writer.add_document(doc!(field=>1u64)); index_writer.add_document(doc!(field=>1u64));
index_writer.commit().unwrap(); index_writer.commit().unwrap();
let reader = index.reader().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
let term = Term::from_field_u64(field, 1u64); let term = Term::from_field_u64(field, 1u64);
let mut postings = searcher let mut postings = searcher
.segment_reader(0) .segment_reader(0)
@@ -647,17 +678,17 @@ mod tests {
#[test] #[test]
fn test_indexed_i64() { fn test_indexed_i64() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let value_field = schema_builder.add_i64_field("value", INDEXED); let value_field = schema_builder.add_i64_field("value", INT_INDEXED);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
let negative_val = -1i64; let negative_val = -1i64;
index_writer.add_document(doc!(value_field => negative_val)); index_writer.add_document(doc!(value_field => negative_val));
index_writer.commit().unwrap(); index_writer.commit().unwrap();
let reader = index.reader().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
let term = Term::from_field_i64(value_field, negative_val); let term = Term::from_field_i64(value_field, negative_val);
let mut postings = searcher let mut postings = searcher
.segment_reader(0) .segment_reader(0)
@@ -671,34 +702,29 @@ mod tests {
#[test] #[test]
fn test_indexedfield_not_in_documents() { fn test_indexedfield_not_in_documents() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let absent_field = schema_builder.add_text_field("text", TEXT); let absent_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_with_num_threads(2, 6_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(2, 40_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"a")); index_writer.add_document(doc!(text_field=>"a"));
assert!(index_writer.commit().is_ok()); assert!(index_writer.commit().is_ok());
let reader = index.reader().unwrap(); assert!(index.load_searchers().is_ok());
let searcher = reader.searcher(); let searcher = index.searcher();
let segment_reader = searcher.segment_reader(0); let segment_reader = searcher.segment_reader(0);
segment_reader.inverted_index(absent_field); //< should not panic segment_reader.inverted_index(absent_field); //< should not panic
} }
#[test] #[test]
fn test_delete_postings2() { fn test_delete_postings2() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::Manual)
.try_into()
.unwrap();
// writing the segment // writing the segment
let mut index_writer = index.writer_with_num_threads(2, 6_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(2, 40_000_000).unwrap();
let add_document = |index_writer: &mut IndexWriter, val: &'static str| { let add_document = |index_writer: &mut IndexWriter, val: &'static str| {
let doc = doc!(text_field=>val); let doc = doc!(text_field=>val);
@@ -721,20 +747,20 @@ mod tests {
remove_document(&mut index_writer, "38"); remove_document(&mut index_writer, "38");
remove_document(&mut index_writer, "34"); remove_document(&mut index_writer, "34");
index_writer.commit().unwrap(); index_writer.commit().unwrap();
reader.reload().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
assert_eq!(searcher.num_docs(), 6); assert_eq!(searcher.num_docs(), 6);
} }
#[test] #[test]
fn test_termfreq() { fn test_termfreq() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
{ {
// writing the segment // writing the segment
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
{ {
let doc = doc!(text_field=>"af af af bc bc"); let doc = doc!(text_field=>"af af af bc bc");
index_writer.add_document(doc); index_writer.add_document(doc);
@@ -742,14 +768,16 @@ mod tests {
index_writer.commit().unwrap(); index_writer.commit().unwrap();
} }
{ {
let index_reader = index.reader().unwrap(); index.load_searchers().unwrap();
let searcher = index_reader.searcher(); let searcher = index.searcher();
let reader = searcher.segment_reader(0); let reader = searcher.segment_reader(0);
let inverted_index = reader.inverted_index(text_field); let inverted_index = reader.inverted_index(text_field);
let term_abcd = Term::from_field_text(text_field, "abcd"); let term_abcd = Term::from_field_text(text_field, "abcd");
assert!(inverted_index assert!(
.read_postings(&term_abcd, IndexRecordOption::WithFreqsAndPositions) inverted_index
.is_none()); .read_postings(&term_abcd, IndexRecordOption::WithFreqsAndPositions)
.is_none()
);
let term_af = Term::from_field_text(text_field, "af"); let term_af = Term::from_field_text(text_field, "af");
let mut postings = inverted_index let mut postings = inverted_index
.read_postings(&term_af, IndexRecordOption::WithFreqsAndPositions) .read_postings(&term_af, IndexRecordOption::WithFreqsAndPositions)
@@ -763,84 +791,109 @@ mod tests {
#[test] #[test]
fn test_searcher_1() { fn test_searcher_1() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
let reader = index.reader().unwrap();
{ {
// writing the segment // writing the segment
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"af af af b")); {
index_writer.add_document(doc!(text_field=>"a b c")); let doc = doc!(text_field=>"af af af b");
index_writer.add_document(doc!(text_field=>"a b c d")); index_writer.add_document(doc);
}
{
let doc = doc!(text_field=>"a b c");
index_writer.add_document(doc);
}
{
let doc = doc!(text_field=>"a b c d");
index_writer.add_document(doc);
}
index_writer.commit().unwrap(); index_writer.commit().unwrap();
} }
{ {
reader.reload().unwrap(); index.load_searchers().unwrap();
let searcher = reader.searcher(); let searcher = index.searcher();
let get_doc_ids = |terms: Vec<Term>| { let get_doc_ids = |terms: Vec<Term>| {
let query = BooleanQuery::new_multiterms_query(terms); let query = BooleanQuery::new_multiterms_query(terms);
let topdocs = searcher.search(&query, &TestCollector).unwrap(); let mut collector = TestCollector::default();
topdocs.docs().to_vec() assert!(searcher.search(&query, &mut collector).is_ok());
collector.docs()
}; };
assert_eq!( {
get_doc_ids(vec![Term::from_field_text(text_field, "a")]), assert_eq!(
vec![DocAddress(0, 1), DocAddress(0, 2)] get_doc_ids(vec![Term::from_field_text(text_field, "a")]),
); vec![1, 2]
assert_eq!( );
get_doc_ids(vec![Term::from_field_text(text_field, "af")]), }
vec![DocAddress(0, 0)] {
); assert_eq!(
assert_eq!( get_doc_ids(vec![Term::from_field_text(text_field, "af")]),
get_doc_ids(vec![Term::from_field_text(text_field, "b")]), vec![0]
vec![DocAddress(0, 0), DocAddress(0, 1), DocAddress(0, 2)] );
); }
assert_eq!( {
get_doc_ids(vec![Term::from_field_text(text_field, "c")]), assert_eq!(
vec![DocAddress(0, 1), DocAddress(0, 2)] get_doc_ids(vec![Term::from_field_text(text_field, "b")]),
); vec![0, 1, 2]
assert_eq!( );
get_doc_ids(vec![Term::from_field_text(text_field, "d")]), }
vec![DocAddress(0, 2)] {
); assert_eq!(
assert_eq!( get_doc_ids(vec![Term::from_field_text(text_field, "c")]),
get_doc_ids(vec![ vec![1, 2]
Term::from_field_text(text_field, "b"), );
Term::from_field_text(text_field, "a"), }
]), {
vec![DocAddress(0, 0), DocAddress(0, 1), DocAddress(0, 2)] assert_eq!(
); get_doc_ids(vec![Term::from_field_text(text_field, "d")]),
vec![2]
);
}
{
assert_eq!(
get_doc_ids(vec![
Term::from_field_text(text_field, "b"),
Term::from_field_text(text_field, "a"),
]),
vec![0, 1, 2]
);
}
} }
} }
#[test] #[test]
fn test_searcher_2() { fn test_searcher_2() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::Manual)
.try_into()
.unwrap();
assert_eq!(reader.searcher().num_docs(), 0u64);
{ {
// writing the segment // writing the segment
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"af b")); {
index_writer.add_document(doc!(text_field=>"a b c")); let doc = doc!(text_field=>"af b");
index_writer.add_document(doc!(text_field=>"a b c d")); index_writer.add_document(doc);
}
{
let doc = doc!(text_field=>"a b c");
index_writer.add_document(doc);
}
{
let doc = doc!(text_field=>"a b c d");
index_writer.add_document(doc);
}
index_writer.commit().unwrap(); index_writer.commit().unwrap();
} }
reader.reload().unwrap(); index.searcher();
assert_eq!(reader.searcher().num_docs(), 3u64);
} }
#[test] #[test]
fn test_doc_macro() { fn test_doc_macro() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let other_text_field = schema_builder.add_text_field("text2", TEXT); let other_text_field = schema_builder.add_text_field("text2", TEXT);
let document = doc!(text_field => "tantivy", let document = doc!(text_field => "tantivy",
@@ -858,11 +911,11 @@ mod tests {
#[test] #[test]
fn test_wrong_fast_field_type() { fn test_wrong_fast_field_type() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let fast_field_unsigned = schema_builder.add_u64_field("unsigned", FAST); let fast_field_unsigned = schema_builder.add_u64_field("unsigned", FAST);
let fast_field_signed = schema_builder.add_i64_field("signed", FAST); let fast_field_signed = schema_builder.add_i64_field("signed", FAST);
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let stored_int_field = schema_builder.add_u64_field("text", STORED); let stored_int_field = schema_builder.add_u64_field("text", INT_STORED);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
@@ -872,8 +925,9 @@ mod tests {
index_writer.add_document(document); index_writer.add_document(document);
index_writer.commit().unwrap(); index_writer.commit().unwrap();
} }
let reader = index.reader().unwrap();
let searcher = reader.searcher(); index.load_searchers().unwrap();
let searcher = index.searcher();
let segment_reader: &SegmentReader = searcher.segment_reader(0); let segment_reader: &SegmentReader = searcher.segment_reader(0);
{ {
let fast_field_reader_res = segment_reader.fast_field_reader::<u64>(text_field); let fast_field_reader_res = segment_reader.fast_field_reader::<u64>(text_field);

28
src/macros.rs
View File

@@ -26,12 +26,12 @@
/// #[macro_use] /// #[macro_use]
/// extern crate tantivy; /// extern crate tantivy;
/// ///
/// use tantivy::schema::{Schema, TEXT, FAST}; /// use tantivy::schema::{SchemaBuilder, TEXT, FAST};
/// ///
/// //... /// //...
/// ///
/// # fn main() { /// # fn main() {
/// let mut schema_builder = Schema::builder(); /// let mut schema_builder = SchemaBuilder::new();
/// let title = schema_builder.add_text_field("title", TEXT); /// let title = schema_builder.add_text_field("title", TEXT);
/// let author = schema_builder.add_text_field("text", TEXT); /// let author = schema_builder.add_text_field("text", TEXT);
/// let likes = schema_builder.add_u64_field("num_u64", FAST); /// let likes = schema_builder.add_u64_field("num_u64", FAST);
@@ -61,39 +61,39 @@ macro_rules! doc(
}; };
// if there is a trailing comma retry with the trailing comma stripped. // if there is a trailing comma retry with the trailing comma stripped.
($($field:expr => $value:expr),+ ,) => { ($($field:expr => $value:expr),+ ,) => {
doc!( $( $field => $value ), *) doc!( $( $field => $value ), *);
}; };
); );
#[cfg(test)] #[cfg(test)]
mod test { mod test {
use schema::{Schema, FAST, TEXT}; use schema::{SchemaBuilder, FAST, TEXT};
#[test] #[test]
fn test_doc_basic() { fn test_doc_basic() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
let title = schema_builder.add_text_field("title", TEXT); let title = schema_builder.add_text_field("title", TEXT);
let author = schema_builder.add_text_field("text", TEXT); let author = schema_builder.add_text_field("text", TEXT);
let likes = schema_builder.add_u64_field("num_u64", FAST); let likes = schema_builder.add_u64_field("num_u64", FAST);
let _schema = schema_builder.build(); let _schema = schema_builder.build();
let _doc = doc!( let _doc = doc!(
title => "Life Aquatic", title => "Life Aquatic",
author => "Wes Anderson", author => "Wes Anderson",
likes => 4u64 likes => 4u64
); );
} }
#[test] #[test]
fn test_doc_trailing_comma() { fn test_doc_trailing_comma() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
let title = schema_builder.add_text_field("title", TEXT); let title = schema_builder.add_text_field("title", TEXT);
let author = schema_builder.add_text_field("text", TEXT); let author = schema_builder.add_text_field("text", TEXT);
let likes = schema_builder.add_u64_field("num_u64", FAST); let likes = schema_builder.add_u64_field("num_u64", FAST);
let _schema = schema_builder.build(); let _schema = schema_builder.build();
let _doc = doc!( let _doc = doc!(
title => "Life Aquatic", title => "Life Aquatic",
author => "Wes Anderson", author => "Wes Anderson",
likes => 4u64, likes => 4u64,
); );
} }
} }

4
src/positions/mod.rs
View File

@@ -34,6 +34,10 @@ const COMPRESSION_BLOCK_SIZE: usize = BitPacker4x::BLOCK_LEN;
const LONG_SKIP_IN_BLOCKS: usize = 1_024; const LONG_SKIP_IN_BLOCKS: usize = 1_024;
const LONG_SKIP_INTERVAL: u64 = (LONG_SKIP_IN_BLOCKS * COMPRESSION_BLOCK_SIZE) as u64; const LONG_SKIP_INTERVAL: u64 = (LONG_SKIP_IN_BLOCKS * COMPRESSION_BLOCK_SIZE) as u64;
lazy_static! {
static ref BIT_PACKER: BitPacker4x = BitPacker4x::new();
}
#[cfg(test)] #[cfg(test)]
pub mod tests { pub mod tests {

148
src/positions/reader.rs
View File

@@ -1,23 +1,4 @@
/// Positions works as a long sequence of compressed block. use super::BIT_PACKER;
/// All terms are chained one after the other.
///
/// When accessing the position of a term, we get a positions_idx from the `Terminfo`.
/// This means we need to skip to the `nth` positions efficiently.
///
/// This is done thanks to two levels of skiping that we refer to in the code
/// as `long_skip` and `short_skip`.
///
/// The `long_skip` makes it possible to skip every 1_024 compression blocks (= 131_072 positions).
/// Skipping offset are simply stored one after as an offset stored over 8 bytes.
///
/// We find the number of long skips, as `n / long_skip`.
///
/// Blocks are compressed using bitpacking, so `skip_read` contains the number of bytes
/// (values can go from 0bit to 32 bits) required to decompressed every block.
///
/// A given block obviously takes `(128 x num_bit_for_the_block / num_bits_in_a_byte)`,
/// so skipping a block without decompressing it is just a matter of advancing that many
/// bytes.
use bitpacking::{BitPacker, BitPacker4x}; use bitpacking::{BitPacker, BitPacker4x};
use common::{BinarySerializable, FixedSize}; use common::{BinarySerializable, FixedSize};
use directory::ReadOnlySource; use directory::ReadOnlySource;
@@ -27,65 +8,9 @@ use positions::LONG_SKIP_INTERVAL;
use positions::LONG_SKIP_IN_BLOCKS; use positions::LONG_SKIP_IN_BLOCKS;
use postings::compression::compressed_block_size; use postings::compression::compressed_block_size;
struct Positions {
bit_packer: BitPacker4x,
skip_source: ReadOnlySource,
position_source: ReadOnlySource,
long_skip_source: ReadOnlySource,
}
impl Positions {
pub fn new(position_source: ReadOnlySource, skip_source: ReadOnlySource) -> Positions {
let skip_len = skip_source.len();
let (body, footer) = skip_source.split(skip_len - u32::SIZE_IN_BYTES);
let num_long_skips = u32::deserialize(&mut footer.as_slice()).expect("Index corrupted");
let body_split = body.len() - u64::SIZE_IN_BYTES * (num_long_skips as usize);
let (skip_source, long_skip_source) = body.split(body_split);
Positions {
bit_packer: BitPacker4x::new(),
skip_source,
long_skip_source,
position_source,
}
}
/// Returns the offset of the block associated to the given `long_skip_id`.
///
/// One `long_skip_id` means `LONG_SKIP_IN_BLOCKS` blocks.
fn long_skip(&self, long_skip_id: usize) -> u64 {
if long_skip_id == 0 {
return 0;
}
let long_skip_slice = self.long_skip_source.as_slice();
let mut long_skip_blocks: &[u8] = &long_skip_slice[(long_skip_id - 1) * 8..][..8];
u64::deserialize(&mut long_skip_blocks).expect("Index corrupted")
}
fn reader(&self, offset: u64) -> PositionReader {
let long_skip_id = (offset / LONG_SKIP_INTERVAL) as usize;
let small_skip = (offset % LONG_SKIP_INTERVAL) as usize;
let offset_num_bytes: u64 = self.long_skip(long_skip_id);
let mut position_read = OwnedRead::new(self.position_source.clone());
position_read.advance(offset_num_bytes as usize);
let mut skip_read = OwnedRead::new(self.skip_source.clone());
skip_read.advance(long_skip_id * LONG_SKIP_IN_BLOCKS);
let mut position_reader = PositionReader {
bit_packer: self.bit_packer,
skip_read,
position_read,
inner_offset: 0,
buffer: Box::new([0u32; 128]),
ahead: None,
};
position_reader.skip(small_skip);
position_reader
}
}
pub struct PositionReader { pub struct PositionReader {
skip_read: OwnedRead, skip_read: OwnedRead,
position_read: OwnedRead, position_read: OwnedRead,
bit_packer: BitPacker4x,
inner_offset: usize, inner_offset: usize,
buffer: Box<[u32; 128]>, buffer: Box<[u32; 128]>,
ahead: Option<usize>, // if None, no block is loaded. ahead: Option<usize>, // if None, no block is loaded.
@@ -102,7 +27,6 @@ pub struct PositionReader {
// If the requested number of els ends exactly at a given block, the next // If the requested number of els ends exactly at a given block, the next
// block is not decompressed. // block is not decompressed.
fn read_impl( fn read_impl(
bit_packer: BitPacker4x,
mut position: &[u8], mut position: &[u8],
buffer: &mut [u32; 128], buffer: &mut [u32; 128],
mut inner_offset: usize, mut inner_offset: usize,
@@ -113,23 +37,21 @@ fn read_impl(
let mut output_len = output.len(); let mut output_len = output.len();
let mut ahead = 0; let mut ahead = 0;
loop { loop {
let available_len = COMPRESSION_BLOCK_SIZE - inner_offset; let available_len = 128 - inner_offset;
// We have enough elements in the current block.
// Let's copy the requested elements in the output buffer,
// and return.
if output_len <= available_len { if output_len <= available_len {
output[output_start..].copy_from_slice(&buffer[inner_offset..][..output_len]); output[output_start..].copy_from_slice(&buffer[inner_offset..][..output_len]);
return ahead; return ahead;
} else {
output[output_start..][..available_len].copy_from_slice(&buffer[inner_offset..]);
output_len -= available_len;
output_start += available_len;
inner_offset = 0;
let num_bits = num_bits[ahead];
BitPacker4x::new().decompress(position, &mut buffer[..], num_bits);
let block_len = compressed_block_size(num_bits);
position = &position[block_len..];
ahead += 1;
} }
output[output_start..][..available_len].copy_from_slice(&buffer[inner_offset..]);
output_len -= available_len;
output_start += available_len;
inner_offset = 0;
let num_bits = num_bits[ahead];
bit_packer.decompress(position, &mut buffer[..], num_bits);
let block_len = compressed_block_size(num_bits);
position = &position[block_len..];
ahead += 1;
} }
} }
@@ -139,7 +61,35 @@ impl PositionReader {
skip_source: ReadOnlySource, skip_source: ReadOnlySource,
offset: u64, offset: u64,
) -> PositionReader { ) -> PositionReader {
Positions::new(position_source, skip_source).reader(offset) let skip_len = skip_source.len();
let (body, footer) = skip_source.split(skip_len - u32::SIZE_IN_BYTES);
let num_long_skips = u32::deserialize(&mut footer.as_slice()).expect("Index corrupted");
let body_split = body.len() - u64::SIZE_IN_BYTES * (num_long_skips as usize);
let (skip_body, long_skips) = body.split(body_split);
let long_skip_id = (offset / LONG_SKIP_INTERVAL) as usize;
let small_skip = (offset - (long_skip_id as u64) * (LONG_SKIP_INTERVAL as u64)) as usize;
let offset_num_bytes: u64 = {
if long_skip_id > 0 {
let mut long_skip_blocks: &[u8] =
&long_skips.as_slice()[(long_skip_id - 1) * 8..][..8];
u64::deserialize(&mut long_skip_blocks).expect("Index corrupted") * 16
} else {
0
}
};
let mut position_read = OwnedRead::new(position_source);
position_read.advance(offset_num_bytes as usize);
let mut skip_read = OwnedRead::new(skip_body);
skip_read.advance(long_skip_id * LONG_SKIP_IN_BLOCKS);
let mut position_reader = PositionReader {
skip_read,
position_read,
inner_offset: 0,
buffer: Box::new([0u32; 128]),
ahead: None,
};
position_reader.skip(small_skip);
position_reader
} }
/// Fills a buffer with the next `output.len()` integers. /// Fills a buffer with the next `output.len()` integers.
@@ -151,13 +101,10 @@ impl PositionReader {
if self.ahead != Some(0) { if self.ahead != Some(0) {
// the block currently available is not the block // the block currently available is not the block
// for the current position // for the current position
self.bit_packer BIT_PACKER.decompress(position_data, self.buffer.as_mut(), num_bits);
.decompress(position_data, self.buffer.as_mut(), num_bits);
self.ahead = Some(0);
} }
let block_len = compressed_block_size(num_bits); let block_len = compressed_block_size(num_bits);
self.ahead = Some(read_impl( self.ahead = Some(read_impl(
self.bit_packer,
&position_data[block_len..], &position_data[block_len..],
self.buffer.as_mut(), self.buffer.as_mut(),
self.inner_offset, self.inner_offset,
@@ -186,13 +133,14 @@ impl PositionReader {
} }
}); });
let skip_len_in_bits = self.skip_read.as_ref()[..num_blocks_to_advance] let skip_len = self.skip_read.as_ref()[..num_blocks_to_advance]
.iter() .iter()
.map(|num_bits| *num_bits as usize) .cloned()
.map(|num_bit| num_bit as usize)
.sum::<usize>() .sum::<usize>()
* COMPRESSION_BLOCK_SIZE; * (COMPRESSION_BLOCK_SIZE / 8);
let skip_len_in_bytes = skip_len_in_bits / 8;
self.skip_read.advance(num_blocks_to_advance); self.skip_read.advance(num_blocks_to_advance);
self.position_read.advance(skip_len_in_bytes); self.position_read.advance(skip_len);
} }
} }

22
src/positions/serializer.rs
View File

@@ -1,30 +1,29 @@
use super::BIT_PACKER;
use bitpacking::BitPacker; use bitpacking::BitPacker;
use bitpacking::BitPacker4x;
use common::BinarySerializable; use common::BinarySerializable;
use common::CountingWriter;
use positions::{COMPRESSION_BLOCK_SIZE, LONG_SKIP_INTERVAL}; use positions::{COMPRESSION_BLOCK_SIZE, LONG_SKIP_INTERVAL};
use std::io::{self, Write}; use std::io;
pub struct PositionSerializer<W: io::Write> { pub struct PositionSerializer<W: io::Write> {
bit_packer: BitPacker4x, write_stream: W,
write_stream: CountingWriter<W>,
write_skiplist: W, write_skiplist: W,
block: Vec<u32>, block: Vec<u32>,
buffer: Vec<u8>, buffer: Vec<u8>,
num_ints: u64, num_ints: u64,
long_skips: Vec<u64>, long_skips: Vec<u64>,
cumulated_num_bits: u64,
} }
impl<W: io::Write> PositionSerializer<W> { impl<W: io::Write> PositionSerializer<W> {
pub fn new(write_stream: W, write_skiplist: W) -> PositionSerializer<W> { pub fn new(write_stream: W, write_skiplist: W) -> PositionSerializer<W> {
PositionSerializer { PositionSerializer {
bit_packer: BitPacker4x::new(), write_stream,
write_stream: CountingWriter::wrap(write_stream),
write_skiplist, write_skiplist,
block: Vec::with_capacity(128), block: Vec::with_capacity(128),
buffer: vec![0u8; 128 * 4], buffer: vec![0u8; 128 * 4],
num_ints: 0u64, num_ints: 0u64,
long_skips: Vec::new(), long_skips: Vec::new(),
cumulated_num_bits: 0u64,
} }
} }
@@ -51,15 +50,14 @@ impl<W: io::Write> PositionSerializer<W> {
} }
fn flush_block(&mut self) -> io::Result<()> { fn flush_block(&mut self) -> io::Result<()> {
let num_bits = self.bit_packer.num_bits(&self.block[..]); let num_bits = BIT_PACKER.num_bits(&self.block[..]);
self.cumulated_num_bits += u64::from(num_bits);
self.write_skiplist.write_all(&[num_bits])?; self.write_skiplist.write_all(&[num_bits])?;
let written_len = self let written_len = BIT_PACKER.compress(&self.block[..], &mut self.buffer, num_bits);
.bit_packer
.compress(&self.block[..], &mut self.buffer, num_bits);
self.write_stream.write_all(&self.buffer[..written_len])?; self.write_stream.write_all(&self.buffer[..written_len])?;
self.block.clear(); self.block.clear();
if (self.num_ints % LONG_SKIP_INTERVAL) == 0u64 { if (self.num_ints % LONG_SKIP_INTERVAL) == 0u64 {
self.long_skips.push(self.write_stream.written_bytes()); self.long_skips.push(self.cumulated_num_bits);
} }
Ok(()) Ok(())
} }

229
src/postings/block_search.rs
View File

@@ -1,229 +0,0 @@
/// This modules define the logic used to search for a doc in a given
/// block. (at most 128 docs)
///
/// Searching within a block is a hotspot when running intersection.
/// so it was worth defining it in its own module.
#[cfg(target_arch = "x86_64")]
mod sse2 {
use postings::compression::COMPRESSION_BLOCK_SIZE;
use std::arch::x86_64::__m128i as DataType;
use std::arch::x86_64::_mm_add_epi32 as op_add;
use std::arch::x86_64::_mm_cmplt_epi32 as op_lt;
use std::arch::x86_64::_mm_load_si128 as op_load; // requires 128-bits alignment
use std::arch::x86_64::_mm_set1_epi32 as set1;
use std::arch::x86_64::_mm_setzero_si128 as set0;
use std::arch::x86_64::_mm_sub_epi32 as op_sub;
use std::arch::x86_64::{_mm_cvtsi128_si32, _mm_shuffle_epi32};
const MASK1: i32 = 78;
const MASK2: i32 = 177;
/// Performs an exhaustive linear search over the
///
/// There is no early exit here. We simply count the
/// number of elements that are `< target`.
pub fn linear_search_sse2_128(arr: &[u32], target: u32) -> usize {
unsafe {
let ptr = arr.as_ptr() as *const DataType;
let vkey = set1(target as i32);
let mut cnt = set0();
// We work over 4 `__m128i` at a time.
// A single `__m128i` actual contains 4 `u32`.
for i in 0..(COMPRESSION_BLOCK_SIZE as isize) / (4 * 4) {
let cmp1 = op_lt(op_load(ptr.offset(i * 4)), vkey);
let cmp2 = op_lt(op_load(ptr.offset(i * 4 + 1)), vkey);
let cmp3 = op_lt(op_load(ptr.offset(i * 4 + 2)), vkey);
let cmp4 = op_lt(op_load(ptr.offset(i * 4 + 3)), vkey);
let sum = op_add(op_add(cmp1, cmp2), op_add(cmp3, cmp4));
cnt = op_sub(cnt, sum);
}
cnt = op_add(cnt, _mm_shuffle_epi32(cnt, MASK1));
cnt = op_add(cnt, _mm_shuffle_epi32(cnt, MASK2));
_mm_cvtsi128_si32(cnt) as usize
}
}
#[cfg(test)]
mod test {
use super::linear_search_sse2_128;
#[test]
fn test_linear_search_sse2_128_u32() {
for i in 0..23 {
dbg!(i);
let arr: Vec<u32> = (0..128).map(|el| el * 2 + 1 << 18).collect();
assert_eq!(linear_search_sse2_128(&arr, arr[64] + 1), 65);
}
}
}
}
/// This `linear search` browser exhaustively through the array.
/// but the early exit is very difficult to predict.
///
/// Coupled with `exponential search` this function is likely
/// to be called with the same `len`
fn linear_search(arr: &[u32], target: u32) -> usize {
arr.iter().map(|&el| if el < target { 1 } else { 0 }).sum()
}
fn exponential_search(arr: &[u32], target: u32) -> (usize, usize) {
let end = arr.len();
let mut begin = 0;
for &pivot in &[1, 3, 7, 15, 31, 63] {
if pivot >= end {
break;
}
if arr[pivot] > target {
return (begin, pivot);
}
begin = pivot;
}
(begin, end)
}
fn galloping(block_docs: &[u32], target: u32) -> usize {
let (start, end) = exponential_search(&block_docs, target);
start + linear_search(&block_docs[start..end], target)
}
/// Tantivy may rely on SIMD instructions to search for a specific document within
/// a given block.
#[derive(Clone, Copy, PartialEq)]
pub enum BlockSearcher {
#[cfg(target_arch = "x86_64")]
SSE2,
Scalar,
}
impl BlockSearcher {
/// Search the first index containing an element greater or equal to
/// the target.
///
/// The results should be equivalent to
/// ```ignore
/// block[..]
// .iter()
// .take_while(|&&val| val < target)
// .count()
/// ```
///
/// The `start` argument is just used to hint that the response is
/// greater than beyond `start`. The implementation may or may not use
/// it for optimization.
///
/// # Assumption
///
/// The array len is > start.
/// The block is sorted
/// The target is assumed greater or equal to the `arr[start]`.
/// The target is assumed smaller or equal to the last element of the block.
///
/// Currently the scalar implementation starts by an exponential search, and
/// then operates a linear search in the result subarray.
///
/// If SSE2 instructions are available in the `(platform, running CPU)`,
/// then we use a different implementation that does an exhaustive linear search over
/// the full block whenever the block is full (`len == 128`). It is surprisingly faster, most likely because of the lack
/// of branch.
pub fn search_in_block(&self, block_docs: &[u32], start: usize, target: u32) -> usize {
#[cfg(target_arch = "x86_64")]
{
use postings::compression::COMPRESSION_BLOCK_SIZE;
if *self == BlockSearcher::SSE2 {
if block_docs.len() == COMPRESSION_BLOCK_SIZE {
return sse2::linear_search_sse2_128(block_docs, target);
}
}
}
start + galloping(&block_docs[start..], target)
}
}
impl Default for BlockSearcher {
fn default() -> BlockSearcher {
#[cfg(target_arch = "x86_64")]
{
if is_x86_feature_detected!("sse2") {
return BlockSearcher::SSE2;
}
}
BlockSearcher::Scalar
}
}
#[cfg(test)]
mod tests {
use super::exponential_search;
use super::linear_search;
use super::BlockSearcher;
#[test]
fn test_linear_search() {
let len: usize = 50;
let arr: Vec<u32> = (0..len).map(|el| 1u32 + (el as u32) * 2).collect();
for target in 1..*arr.last().unwrap() {
let res = linear_search(&arr[..], target);
if res > 0 {
assert!(arr[res - 1] < target);
}
if res < len {
assert!(arr[res] >= target);
}
}
}
#[test]
fn test_exponentiel_search() {
assert_eq!(exponential_search(&[1, 2], 0), (0, 1));
assert_eq!(exponential_search(&[1, 2], 1), (0, 1));
assert_eq!(
exponential_search(&[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11], 7),
(3, 7)
);
}
fn util_test_search_in_block(block_searcher: BlockSearcher, block: &[u32], target: u32) {
let cursor = search_in_block_trivial_but_slow(block, target);
for i in 0..cursor {
assert_eq!(block_searcher.search_in_block(block, i, target), cursor);
}
}
fn util_test_search_in_block_all(block_searcher: BlockSearcher, block: &[u32]) {
use std::collections::HashSet;
let mut targets = HashSet::new();
for (i, val) in block.iter().cloned().enumerate() {
if i > 0 {
targets.insert(val - 1);
}
targets.insert(val);
}
for target in targets {
util_test_search_in_block(block_searcher, block, target);
}
}
fn search_in_block_trivial_but_slow(block: &[u32], target: u32) -> usize {
block.iter().take_while(|&&val| val < target).count()
}
fn test_search_in_block_util(block_searcher: BlockSearcher) {
for len in 1u32..128u32 {
let v: Vec<u32> = (0..len).map(|i| i * 2).collect();
util_test_search_in_block_all(block_searcher, &v[..]);
}
}
#[test]
fn test_search_in_block_scalar() {
test_search_in_block_util(BlockSearcher::Scalar);
}
#[cfg(target_arch = "x86_64")]
#[test]
fn test_search_in_block_sse2() {
test_search_in_block_util(BlockSearcher::SSE2);
}
}

50
src/postings/compression/mod.rs
View File

@@ -43,14 +43,9 @@ impl BlockEncoder {
} }
} }
/// We ensure that the OutputBuffer is align on 128 bits
/// in order to run SSE2 linear search on it.
#[repr(align(128))]
struct OutputBuffer([u32; COMPRESSION_BLOCK_SIZE + 1]);
pub struct BlockDecoder { pub struct BlockDecoder {
bitpacker: BitPacker4x, bitpacker: BitPacker4x,
output: OutputBuffer, pub output: [u32; COMPRESSION_BLOCK_SIZE + 1],
pub output_len: usize, pub output_len: usize,
} }
@@ -64,7 +59,7 @@ impl BlockDecoder {
output[COMPRESSION_BLOCK_SIZE] = 0u32; output[COMPRESSION_BLOCK_SIZE] = 0u32;
BlockDecoder { BlockDecoder {
bitpacker: BitPacker4x::new(), bitpacker: BitPacker4x::new(),
output: OutputBuffer(output), output,
output_len: 0, output_len: 0,
} }
} }
@@ -77,23 +72,23 @@ impl BlockDecoder {
) -> usize { ) -> usize {
self.output_len = COMPRESSION_BLOCK_SIZE; self.output_len = COMPRESSION_BLOCK_SIZE;
self.bitpacker self.bitpacker
.decompress_sorted(offset, &compressed_data, &mut self.output.0, num_bits) .decompress_sorted(offset, &compressed_data, &mut self.output, num_bits)
} }
pub fn uncompress_block_unsorted(&mut self, compressed_data: &[u8], num_bits: u8) -> usize { pub fn uncompress_block_unsorted(&mut self, compressed_data: &[u8], num_bits: u8) -> usize {
self.output_len = COMPRESSION_BLOCK_SIZE; self.output_len = COMPRESSION_BLOCK_SIZE;
self.bitpacker self.bitpacker
.decompress(&compressed_data, &mut self.output.0, num_bits) .decompress(&compressed_data, &mut self.output, num_bits)
} }
#[inline] #[inline]
pub fn output_array(&self) -> &[u32] { pub fn output_array(&self) -> &[u32] {
&self.output.0[..self.output_len] &self.output[..self.output_len]
} }
#[inline] #[inline]
pub fn output(&self, idx: usize) -> u32 { pub fn output(&self, idx: usize) -> u32 {
self.output.0[idx] self.output[idx]
} }
} }
@@ -164,12 +159,12 @@ impl VIntDecoder for BlockDecoder {
num_els: usize, num_els: usize,
) -> usize { ) -> usize {
self.output_len = num_els; self.output_len = num_els;
vint::uncompress_sorted(compressed_data, &mut self.output.0[..num_els], offset) vint::uncompress_sorted(compressed_data, &mut self.output[..num_els], offset)
} }
fn uncompress_vint_unsorted<'a>(&mut self, compressed_data: &'a [u8], num_els: usize) -> usize { fn uncompress_vint_unsorted<'a>(&mut self, compressed_data: &'a [u8], num_els: usize) -> usize {
self.output_len = num_els; self.output_len = num_els;
vint::uncompress_unsorted(compressed_data, &mut self.output.0[..num_els]) vint::uncompress_unsorted(compressed_data, &mut self.output[..num_els])
} }
} }
@@ -271,17 +266,21 @@ pub mod tests {
mod bench { mod bench {
use super::*; use super::*;
use rand::Rng;
use rand::SeedableRng; use rand::SeedableRng;
use rand::{Rng, XorShiftRng}; use rand::XorShiftRng;
use test::Bencher; use test::Bencher;
fn generate_array_with_seed(n: usize, ratio: f64, seed_val: u8) -> Vec<u32> { fn generate_array_with_seed(n: usize, ratio: f32, seed_val: u32) -> Vec<u32> {
let seed: &[u8; 16] = &[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, seed_val]; let seed: &[u32; 4] = &[1, 2, 3, seed_val];
let mut rng: XorShiftRng = XorShiftRng::from_seed(*seed); let mut rng: XorShiftRng = XorShiftRng::from_seed(*seed);
(0u32..).filter(|_| rng.gen_bool(ratio)).take(n).collect() (0..u32::max_value())
.filter(|_| rng.next_f32() < ratio)
.take(n)
.collect()
} }
pub fn generate_array(n: usize, ratio: f64) -> Vec<u32> { pub fn generate_array(n: usize, ratio: f32) -> Vec<u32> {
generate_array_with_seed(n, ratio, 4) generate_array_with_seed(n, ratio, 4)
} }
@@ -298,23 +297,24 @@ mod bench {
fn bench_uncompress(b: &mut Bencher) { fn bench_uncompress(b: &mut Bencher) {
let mut encoder = BlockEncoder::new(); let mut encoder = BlockEncoder::new();
let data = generate_array(COMPRESSION_BLOCK_SIZE, 0.1); let data = generate_array(COMPRESSION_BLOCK_SIZE, 0.1);
let (num_bits, compressed) = encoder.compress_block_sorted(&data, 0u32); let (_, compressed) = encoder.compress_block_sorted(&data, 0u32);
let mut decoder = BlockDecoder::new(); let mut decoder = BlockDecoder::new();
b.iter(|| { b.iter(|| {
decoder.uncompress_block_sorted(compressed, 0u32, num_bits); decoder.uncompress_block_sorted(compressed, 0u32);
}); });
} }
#[test] #[test]
fn test_all_docs_compression_numbits() { fn test_all_docs_compression_numbits() {
for expected_num_bits in 0u8.. { for num_bits in 0..33 {
let mut data = [0u32; 128]; let mut data = [0u32; 128];
if expected_num_bits > 0 { if num_bits > 0 {
data[0] = (1u64 << (expected_num_bits as usize) - 1) as u32; data[0] = 1 << (num_bits - 1);
} }
let mut encoder = BlockEncoder::new(); let mut encoder = BlockEncoder::new();
let (num_bits, compressed) = encoder.compress_block_unsorted(&data); let compressed = encoder.compress_block_unsorted(&data);
assert_eq!(compressed.len(), compressed_block_size(num_bits)); assert_eq!(compressed[0] as usize, num_bits);
assert_eq!(compressed.len(), compressed_block_size(compressed[0]));
} }
} }

89
src/postings/mod.rs
View File

@@ -2,7 +2,6 @@
Postings module (also called inverted index) Postings module (also called inverted index)
*/ */
mod block_search;
pub(crate) mod compression; pub(crate) mod compression;
/// Postings module /// Postings module
/// ///
@@ -17,8 +16,6 @@ mod skip;
mod stacker; mod stacker;
mod term_info; mod term_info;
pub(crate) use self::block_search::BlockSearcher;
pub(crate) use self::postings_writer::MultiFieldPostingsWriter; pub(crate) use self::postings_writer::MultiFieldPostingsWriter;
pub use self::serializer::{FieldSerializer, InvertedIndexSerializer}; pub use self::serializer::{FieldSerializer, InvertedIndexSerializer};
@@ -34,6 +31,7 @@ pub(crate) use self::stacker::compute_table_size;
pub use common::HasLen; pub use common::HasLen;
pub(crate) const USE_SKIP_INFO_LIMIT: u32 = COMPRESSION_BLOCK_SIZE as u32; pub(crate) const USE_SKIP_INFO_LIMIT: u32 = COMPRESSION_BLOCK_SIZE as u32;
pub(crate) type UnorderedTermId = u64; pub(crate) type UnorderedTermId = u64;
#[cfg_attr(feature = "cargo-clippy", allow(clippy::enum_variant_names))] #[cfg_attr(feature = "cargo-clippy", allow(clippy::enum_variant_names))]
@@ -56,18 +54,17 @@ pub mod tests {
use indexer::operation::AddOperation; use indexer::operation::AddOperation;
use indexer::SegmentWriter; use indexer::SegmentWriter;
use query::Scorer; use query::Scorer;
use rand::rngs::StdRng; use rand::{Rng, SeedableRng, XorShiftRng};
use rand::{Rng, SeedableRng};
use schema::Field; use schema::Field;
use schema::IndexRecordOption; use schema::IndexRecordOption;
use schema::{Document, Schema, Term, INDEXED, STRING, TEXT}; use schema::{Document, SchemaBuilder, Term, INT_INDEXED, STRING, TEXT};
use std::iter; use std::iter;
use DocId; use DocId;
use Score; use Score;
#[test] #[test]
pub fn test_position_write() { pub fn test_position_write() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
@@ -91,7 +88,7 @@ pub mod tests {
#[test] #[test]
pub fn test_skip_positions() { pub fn test_skip_positions() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::new();
let title = schema_builder.add_text_field("title", TEXT); let title = schema_builder.add_text_field("title", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
@@ -103,11 +100,14 @@ pub mod tests {
} }
index_writer.add_document(doc!(title => r#"abc be be be be abc"#)); index_writer.add_document(doc!(title => r#"abc be be be be abc"#));
index_writer.commit().unwrap(); index_writer.commit().unwrap();
index.load_searchers().unwrap();
let searcher = index.reader().unwrap().searcher(); let searcher = index.searcher();
let inverted_index = searcher.segment_reader(0u32).inverted_index(title); let inverted_index = searcher.segment_reader(0u32).inverted_index(title);
let term = Term::from_field_text(title, "abc"); let term = Term::from_field_text(title, "abc");
let mut positions = Vec::new(); let mut positions = Vec::new();
{ {
let mut postings = inverted_index let mut postings = inverted_index
.read_postings(&term, IndexRecordOption::WithFreqsAndPositions) .read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
@@ -163,7 +163,7 @@ pub mod tests {
#[test] #[test]
pub fn test_position_and_fieldnorm1() { pub fn test_position_and_fieldnorm1() {
let mut positions = Vec::new(); let mut positions = Vec::new();
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema.clone()); let index = Index::create_in_ram(schema.clone());
@@ -220,10 +220,12 @@ pub mod tests {
} }
{ {
let term_a = Term::from_field_text(text_field, "abcdef"); let term_a = Term::from_field_text(text_field, "abcdef");
assert!(segment_reader assert!(
.inverted_index(term_a.field()) segment_reader
.read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions) .inverted_index(term_a.field())
.is_none()); .read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions)
.is_none()
);
} }
{ {
let term_a = Term::from_field_text(text_field, "a"); let term_a = Term::from_field_text(text_field, "a");
@@ -274,12 +276,12 @@ pub mod tests {
#[test] #[test]
pub fn test_position_and_fieldnorm2() { pub fn test_position_and_fieldnorm2() {
let mut positions: Vec<u32> = Vec::new(); let mut positions: Vec<u32> = Vec::new();
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
{ {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
{ {
let mut doc = Document::default(); let mut doc = Document::default();
doc.add_text(text_field, "g b b d c g c"); doc.add_text(text_field, "g b b d c g c");
@@ -292,8 +294,9 @@ pub mod tests {
} }
assert!(index_writer.commit().is_ok()); assert!(index_writer.commit().is_ok());
} }
index.load_searchers().unwrap();
let term_a = Term::from_field_text(text_field, "a"); let term_a = Term::from_field_text(text_field, "a");
let searcher = index.reader().unwrap().searcher(); let searcher = index.searcher();
let segment_reader = searcher.segment_reader(0); let segment_reader = searcher.segment_reader(0);
let mut postings = segment_reader let mut postings = segment_reader
.inverted_index(text_field) .inverted_index(text_field)
@@ -314,13 +317,13 @@ pub mod tests {
let num_docs = 300u32; let num_docs = 300u32;
let index = { let index = {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let value_field = schema_builder.add_u64_field("value", INDEXED); let value_field = schema_builder.add_u64_field("value", INT_INDEXED);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
{ {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
for i in 0..num_docs { for i in 0..num_docs {
let mut doc = Document::default(); let mut doc = Document::default();
doc.add_u64(value_field, 2); doc.add_u64(value_field, 2);
@@ -330,9 +333,10 @@ pub mod tests {
} }
assert!(index_writer.commit().is_ok()); assert!(index_writer.commit().is_ok());
} }
index.load_searchers().unwrap();
index index
}; };
let searcher = index.reader().unwrap().searcher(); let searcher = index.searcher();
let segment_reader = searcher.segment_reader(0); let segment_reader = searcher.segment_reader(0);
// check that the basic usage works // check that the basic usage works
@@ -396,11 +400,12 @@ pub mod tests {
// delete some of the documents // delete some of the documents
{ {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
index_writer.delete_term(term_0); index_writer.delete_term(term_0);
assert!(index_writer.commit().is_ok()); assert!(index_writer.commit().is_ok());
} }
let searcher = index.reader().unwrap().searcher(); index.load_searchers().unwrap();
let searcher = index.searcher();
let segment_reader = searcher.segment_reader(0); let segment_reader = searcher.segment_reader(0);
// make sure seeking still works // make sure seeking still works
@@ -445,19 +450,33 @@ pub mod tests {
// delete everything else // delete everything else
{ {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
index_writer.delete_term(term_1); index_writer.delete_term(term_1);
assert!(index_writer.commit().is_ok()); assert!(index_writer.commit().is_ok());
} }
let searcher = index.reader().unwrap().searcher(); index.load_searchers().unwrap();
let searcher = index.searcher();
let segment_reader = searcher.segment_reader(0);
// finally, check that it's empty // finally, check that it's empty
{ {
let searchable_segment_ids = index let mut segment_postings = segment_reader
.searchable_segment_ids() .inverted_index(term_2.field())
.expect("could not get index segment ids"); .read_postings(&term_2, IndexRecordOption::Basic)
assert!(searchable_segment_ids.is_empty()); .unwrap();
assert_eq!(searcher.num_docs(), 0);
assert_eq!(segment_postings.skip_next(0), SkipResult::Reached);
assert_eq!(segment_postings.doc(), 0);
assert!(segment_reader.is_deleted(0));
let mut segment_postings = segment_reader
.inverted_index(term_2.field())
.read_postings(&term_2, IndexRecordOption::Basic)
.unwrap();
assert_eq!(segment_postings.skip_next(num_docs), SkipResult::End);
} }
} }
@@ -479,16 +498,17 @@ pub mod tests {
Term::from_field_text(field, "d") Term::from_field_text(field, "d")
}; };
pub static ref INDEX: Index = { pub static ref INDEX: Index = {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", STRING); let text_field = schema_builder.add_text_field("text", STRING);
let schema = schema_builder.build(); let schema = schema_builder.build();
let mut rng: StdRng = StdRng::from_seed([1u8; 32]); let seed: [u8; 16] = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16];
let mut rng: XorShiftRng = XorShiftRng::from_seed(seed);
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
let posting_list_size = 1_000_000; let posting_list_size = 1_000_000;
{ {
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
for _ in 0..posting_list_size { for _ in 0..posting_list_size {
let mut doc = Document::default(); let mut doc = Document::default();
if rng.gen_bool(1f64 / 15f64) { if rng.gen_bool(1f64 / 15f64) {
@@ -505,6 +525,7 @@ pub mod tests {
} }
assert!(index_writer.commit().is_ok()); assert!(index_writer.commit().is_ok());
} }
index.load_searchers().unwrap();
index index
}; };
} }
@@ -633,7 +654,7 @@ mod bench {
}); });
} }
fn bench_skip_next(p: f64, b: &mut Bencher) { fn bench_skip_next(p: f32, b: &mut Bencher) {
let searcher = INDEX.searcher(); let searcher = INDEX.searcher();
let segment_reader = searcher.segment_reader(0); let segment_reader = searcher.segment_reader(0);
let docs = tests::sample(segment_reader.num_docs(), p); let docs = tests::sample(segment_reader.num_docs(), p);

90
src/postings/postings_writer.rs
View File

@@ -1,8 +1,6 @@
use super::stacker::{Addr, MemoryArena, TermHashMap}; use super::stacker::{Addr, MemoryArena, TermHashMap};
use postings::recorder::{ use postings::recorder::{NothingRecorder, Recorder, TFAndPositionRecorder, TermFrequencyRecorder};
BufferLender, NothingRecorder, Recorder, TFAndPositionRecorder, TermFrequencyRecorder,
};
use postings::UnorderedTermId; use postings::UnorderedTermId;
use postings::{FieldSerializer, InvertedIndexSerializer}; use postings::{FieldSerializer, InvertedIndexSerializer};
use schema::IndexRecordOption; use schema::IndexRecordOption;
@@ -31,12 +29,10 @@ fn posting_from_field_entry(field_entry: &FieldEntry) -> Box<PostingsWriter> {
IndexRecordOption::WithFreqsAndPositions => { IndexRecordOption::WithFreqsAndPositions => {
SpecializedPostingsWriter::<TFAndPositionRecorder>::new_boxed() SpecializedPostingsWriter::<TFAndPositionRecorder>::new_boxed()
} }
}) }).unwrap_or_else(|| SpecializedPostingsWriter::<NothingRecorder>::new_boxed()),
.unwrap_or_else(|| SpecializedPostingsWriter::<NothingRecorder>::new_boxed()), FieldType::U64(_) | FieldType::I64(_) | FieldType::HierarchicalFacet => {
FieldType::U64(_) SpecializedPostingsWriter::<NothingRecorder>::new_boxed()
| FieldType::I64(_) }
| FieldType::Date(_)
| FieldType::HierarchicalFacet => SpecializedPostingsWriter::<NothingRecorder>::new_boxed(),
FieldType::Bytes => { FieldType::Bytes => {
// FieldType::Bytes cannot actually be indexed. // FieldType::Bytes cannot actually be indexed.
// TODO fix during the indexer refactoring described in #276 // TODO fix during the indexer refactoring described in #276
@@ -52,31 +48,6 @@ pub struct MultiFieldPostingsWriter {
per_field_postings_writers: Vec<Box<PostingsWriter>>, per_field_postings_writers: Vec<Box<PostingsWriter>>,
} }
fn make_field_partition(
term_offsets: &[(&[u8], Addr, UnorderedTermId)],
) -> Vec<(Field, usize, usize)> {
let term_offsets_it = term_offsets
.iter()
.map(|(key, _, _)| Term::wrap(key).field())
.enumerate();
let mut prev_field = Field(u32::max_value());
let mut fields = vec![];
let mut offsets = vec![];
for (offset, field) in term_offsets_it {
if field != prev_field {
prev_field = field;
fields.push(field);
offsets.push(offset);
}
}
offsets.push(term_offsets.len());
let mut field_offsets = vec![];
for i in 0..fields.len() {
field_offsets.push((fields[i], offsets[i], offsets[i + 1]));
}
field_offsets
}
impl MultiFieldPostingsWriter { impl MultiFieldPostingsWriter {
/// Create a new `MultiFieldPostingsWriter` given /// Create a new `MultiFieldPostingsWriter` given
/// a schema and a heap. /// a schema and a heap.
@@ -122,16 +93,38 @@ impl MultiFieldPostingsWriter {
&self, &self,
serializer: &mut InvertedIndexSerializer, serializer: &mut InvertedIndexSerializer,
) -> Result<HashMap<Field, HashMap<UnorderedTermId, TermOrdinal>>> { ) -> Result<HashMap<Field, HashMap<UnorderedTermId, TermOrdinal>>> {
let mut term_offsets: Vec<(&[u8], Addr, UnorderedTermId)> = let mut term_offsets: Vec<(&[u8], Addr, UnorderedTermId)> = self
self.term_index.iter().collect(); .term_index
.iter()
.map(|(term_bytes, addr, bucket_id)| (term_bytes, addr, bucket_id as UnorderedTermId))
.collect();
term_offsets.sort_unstable_by_key(|&(k, _, _)| k); term_offsets.sort_unstable_by_key(|&(k, _, _)| k);
let mut unordered_term_mappings: HashMap<Field, HashMap<UnorderedTermId, TermOrdinal>> = let mut offsets: Vec<(Field, usize)> = vec![];
HashMap::new(); let term_offsets_it = term_offsets
.iter()
.cloned()
.map(|(key, _, _)| Term::wrap(key).field())
.enumerate();
let field_offsets = make_field_partition(&term_offsets); let mut unordered_term_mappings: HashMap<
Field,
HashMap<UnorderedTermId, TermOrdinal>,
> = HashMap::new();
let mut prev_field = Field(u32::max_value());
for (offset, field) in term_offsets_it {
if field != prev_field {
offsets.push((field, offset));
prev_field = field;
}
}
offsets.push((Field(0), term_offsets.len()));
for i in 0..(offsets.len() - 1) {
let (field, start) = offsets[i];
let (_, stop) = offsets[i + 1];
for (field, start, stop) in field_offsets {
let field_entry = self.schema.get_field_entry(field); let field_entry = self.schema.get_field_entry(field);
match *field_entry.field_type() { match *field_entry.field_type() {
@@ -145,11 +138,10 @@ impl MultiFieldPostingsWriter {
.enumerate() .enumerate()
.map(|(term_ord, unord_term_id)| { .map(|(term_ord, unord_term_id)| {
(unord_term_id as UnorderedTermId, term_ord as TermOrdinal) (unord_term_id as UnorderedTermId, term_ord as TermOrdinal)
}) }).collect();
.collect();
unordered_term_mappings.insert(field, mapping); unordered_term_mappings.insert(field, mapping);
} }
FieldType::U64(_) | FieldType::I64(_) | FieldType::Date(_) => {} FieldType::U64(_) | FieldType::I64(_) => {}
FieldType::Bytes => {} FieldType::Bytes => {}
} }
@@ -221,7 +213,7 @@ pub trait PostingsWriter {
/// The `SpecializedPostingsWriter` is just here to remove dynamic /// The `SpecializedPostingsWriter` is just here to remove dynamic
/// dispatch to the recorder information. /// dispatch to the recorder information.
pub(crate) struct SpecializedPostingsWriter<Rec: Recorder + 'static> { pub struct SpecializedPostingsWriter<Rec: Recorder + 'static> {
total_num_tokens: u64, total_num_tokens: u64,
_recorder_type: PhantomData<Rec>, _recorder_type: PhantomData<Rec>,
} }
@@ -253,7 +245,8 @@ impl<Rec: Recorder + 'static> PostingsWriter for SpecializedPostingsWriter<Rec>
debug_assert!(term.as_slice().len() >= 4); debug_assert!(term.as_slice().len() >= 4);
self.total_num_tokens += 1; self.total_num_tokens += 1;
term_index.mutate_or_create(term, |opt_recorder: Option<Rec>| { term_index.mutate_or_create(term, |opt_recorder: Option<Rec>| {
if let Some(mut recorder) = opt_recorder { if opt_recorder.is_some() {
let mut recorder = opt_recorder.unwrap();
let current_doc = recorder.current_doc(); let current_doc = recorder.current_doc();
if current_doc != doc { if current_doc != doc {
recorder.close_doc(heap); recorder.close_doc(heap);
@@ -262,7 +255,7 @@ impl<Rec: Recorder + 'static> PostingsWriter for SpecializedPostingsWriter<Rec>
recorder.record_position(position, heap); recorder.record_position(position, heap);
recorder recorder
} else { } else {
let mut recorder = Rec::new(); let mut recorder = Rec::new(heap);
recorder.new_doc(doc, heap); recorder.new_doc(doc, heap);
recorder.record_position(position, heap); recorder.record_position(position, heap);
recorder recorder
@@ -277,11 +270,10 @@ impl<Rec: Recorder + 'static> PostingsWriter for SpecializedPostingsWriter<Rec>
termdict_heap: &MemoryArena, termdict_heap: &MemoryArena,
heap: &MemoryArena, heap: &MemoryArena,
) -> io::Result<()> { ) -> io::Result<()> {
let mut buffer_lender = BufferLender::default();
for &(term_bytes, addr, _) in term_addrs { for &(term_bytes, addr, _) in term_addrs {
let recorder: Rec = termdict_heap.read(addr); let recorder: Rec = unsafe { termdict_heap.read(addr) };
serializer.new_term(&term_bytes[4..])?; serializer.new_term(&term_bytes[4..])?;
recorder.serialize(&mut buffer_lender, serializer, heap)?; recorder.serialize(serializer, heap)?;
serializer.close_term()?; serializer.close_term()?;
} }
Ok(()) Ok(())

202
src/postings/recorder.rs
View File

@@ -1,50 +1,10 @@
use super::stacker::{ExpUnrolledLinkedList, MemoryArena}; use super::stacker::{ExpUnrolledLinkedList, MemoryArena};
use common::{read_u32_vint, write_u32_vint};
use postings::FieldSerializer; use postings::FieldSerializer;
use std::io; use std::{self, io};
use DocId; use DocId;
const POSITION_END: u32 = 0; const EMPTY_ARRAY: [u32; 0] = [0u32; 0];
const POSITION_END: u32 = std::u32::MAX;
#[derive(Default)]
pub(crate) struct BufferLender {
buffer_u8: Vec<u8>,
buffer_u32: Vec<u32>,
}
impl BufferLender {
pub fn lend_u8(&mut self) -> &mut Vec<u8> {
self.buffer_u8.clear();
&mut self.buffer_u8
}
pub fn lend_all(&mut self) -> (&mut Vec<u8>, &mut Vec<u32>) {
self.buffer_u8.clear();
self.buffer_u32.clear();
(&mut self.buffer_u8, &mut self.buffer_u32)
}
}
pub struct VInt32Reader<'a> {
data: &'a [u8],
}
impl<'a> VInt32Reader<'a> {
fn new(data: &'a [u8]) -> VInt32Reader<'a> {
VInt32Reader { data }
}
}
impl<'a> Iterator for VInt32Reader<'a> {
type Item = u32;
fn next(&mut self) -> Option<u32> {
if self.data.is_empty() {
None
} else {
Some(read_u32_vint(&mut self.data))
}
}
}
/// Recorder is in charge of recording relevant information about /// Recorder is in charge of recording relevant information about
/// the presence of a term in a document. /// the presence of a term in a document.
@@ -55,9 +15,9 @@ impl<'a> Iterator for VInt32Reader<'a> {
/// * the document id /// * the document id
/// * the term frequency /// * the term frequency
/// * the term positions /// * the term positions
pub(crate) trait Recorder: Copy + 'static { pub trait Recorder: Copy {
/// ///
fn new() -> Self; fn new(heap: &mut MemoryArena) -> Self;
/// Returns the current document /// Returns the current document
fn current_doc(&self) -> u32; fn current_doc(&self) -> u32;
/// Starts recording information about a new document /// Starts recording information about a new document
@@ -69,12 +29,7 @@ pub(crate) trait Recorder: Copy + 'static {
/// Close the document. It will help record the term frequency. /// Close the document. It will help record the term frequency.
fn close_doc(&mut self, heap: &mut MemoryArena); fn close_doc(&mut self, heap: &mut MemoryArena);
/// Pushes the postings information to the serializer. /// Pushes the postings information to the serializer.
fn serialize( fn serialize(&self, serializer: &mut FieldSerializer, heap: &MemoryArena) -> io::Result<()>;
&self,
buffer_lender: &mut BufferLender,
serializer: &mut FieldSerializer,
heap: &MemoryArena,
) -> io::Result<()>;
} }
/// Only records the doc ids /// Only records the doc ids
@@ -85,9 +40,9 @@ pub struct NothingRecorder {
} }
impl Recorder for NothingRecorder { impl Recorder for NothingRecorder {
fn new() -> Self { fn new(heap: &mut MemoryArena) -> Self {
NothingRecorder { NothingRecorder {
stack: ExpUnrolledLinkedList::new(), stack: ExpUnrolledLinkedList::new(heap),
current_doc: u32::max_value(), current_doc: u32::max_value(),
} }
} }
@@ -98,23 +53,16 @@ impl Recorder for NothingRecorder {
fn new_doc(&mut self, doc: DocId, heap: &mut MemoryArena) { fn new_doc(&mut self, doc: DocId, heap: &mut MemoryArena) {
self.current_doc = doc; self.current_doc = doc;
let _ = write_u32_vint(doc, &mut self.stack.writer(heap)); self.stack.push(doc, heap);
} }
fn record_position(&mut self, _position: u32, _heap: &mut MemoryArena) {} fn record_position(&mut self, _position: u32, _heap: &mut MemoryArena) {}
fn close_doc(&mut self, _heap: &mut MemoryArena) {} fn close_doc(&mut self, _heap: &mut MemoryArena) {}
fn serialize( fn serialize(&self, serializer: &mut FieldSerializer, heap: &MemoryArena) -> io::Result<()> {
&self, for doc in self.stack.iter(heap) {
buffer_lender: &mut BufferLender, serializer.write_doc(doc, 0u32, &EMPTY_ARRAY)?;
serializer: &mut FieldSerializer,
heap: &MemoryArena,
) -> io::Result<()> {
let buffer = buffer_lender.lend_u8();
self.stack.read_to_end(heap, buffer);
for doc in VInt32Reader::new(&buffer[..]) {
serializer.write_doc(doc as u32, 0u32, &[][..])?;
} }
Ok(()) Ok(())
} }
@@ -129,9 +77,9 @@ pub struct TermFrequencyRecorder {
} }
impl Recorder for TermFrequencyRecorder { impl Recorder for TermFrequencyRecorder {
fn new() -> Self { fn new(heap: &mut MemoryArena) -> Self {
TermFrequencyRecorder { TermFrequencyRecorder {
stack: ExpUnrolledLinkedList::new(), stack: ExpUnrolledLinkedList::new(heap),
current_doc: u32::max_value(), current_doc: u32::max_value(),
current_tf: 0u32, current_tf: 0u32,
} }
@@ -143,7 +91,7 @@ impl Recorder for TermFrequencyRecorder {
fn new_doc(&mut self, doc: DocId, heap: &mut MemoryArena) { fn new_doc(&mut self, doc: DocId, heap: &mut MemoryArena) {
self.current_doc = doc; self.current_doc = doc;
let _ = write_u32_vint(doc, &mut self.stack.writer(heap)); self.stack.push(doc, heap);
} }
fn record_position(&mut self, _position: u32, _heap: &mut MemoryArena) { fn record_position(&mut self, _position: u32, _heap: &mut MemoryArena) {
@@ -152,24 +100,24 @@ impl Recorder for TermFrequencyRecorder {
fn close_doc(&mut self, heap: &mut MemoryArena) { fn close_doc(&mut self, heap: &mut MemoryArena) {
debug_assert!(self.current_tf > 0); debug_assert!(self.current_tf > 0);
let _ = write_u32_vint(self.current_tf, &mut self.stack.writer(heap)); self.stack.push(self.current_tf, heap);
self.current_tf = 0; self.current_tf = 0;
} }
fn serialize( fn serialize(&self, serializer: &mut FieldSerializer, heap: &MemoryArena) -> io::Result<()> {
&self, // the last document has not been closed...
buffer_lender: &mut BufferLender, // its term freq is self.current_tf.
serializer: &mut FieldSerializer, let mut doc_iter = self
heap: &MemoryArena, .stack
) -> io::Result<()> { .iter(heap)
let buffer = buffer_lender.lend_u8(); .chain(Some(self.current_tf).into_iter());
self.stack.read_to_end(heap, buffer);
let mut u32_it = VInt32Reader::new(&buffer[..]);
while let Some(doc) = u32_it.next() {
let term_freq = u32_it.next().unwrap_or(self.current_tf);
serializer.write_doc(doc as u32, term_freq, &[][..])?;
}
while let Some(doc) = doc_iter.next() {
let term_freq = doc_iter
.next()
.expect("The IndexWriter recorded a doc without a term freq.");
serializer.write_doc(doc, term_freq, &EMPTY_ARRAY)?;
}
Ok(()) Ok(())
} }
} }
@@ -180,10 +128,11 @@ pub struct TFAndPositionRecorder {
stack: ExpUnrolledLinkedList, stack: ExpUnrolledLinkedList,
current_doc: DocId, current_doc: DocId,
} }
impl Recorder for TFAndPositionRecorder { impl Recorder for TFAndPositionRecorder {
fn new() -> Self { fn new(heap: &mut MemoryArena) -> Self {
TFAndPositionRecorder { TFAndPositionRecorder {
stack: ExpUnrolledLinkedList::new(), stack: ExpUnrolledLinkedList::new(heap),
current_doc: u32::max_value(), current_doc: u32::max_value(),
} }
} }
@@ -194,88 +143,33 @@ impl Recorder for TFAndPositionRecorder {
fn new_doc(&mut self, doc: DocId, heap: &mut MemoryArena) { fn new_doc(&mut self, doc: DocId, heap: &mut MemoryArena) {
self.current_doc = doc; self.current_doc = doc;
let _ = write_u32_vint(doc, &mut self.stack.writer(heap)); self.stack.push(doc, heap);
} }
fn record_position(&mut self, position: u32, heap: &mut MemoryArena) { fn record_position(&mut self, position: u32, heap: &mut MemoryArena) {
let _ = write_u32_vint(position + 1u32, &mut self.stack.writer(heap)); self.stack.push(position, heap);
} }
fn close_doc(&mut self, heap: &mut MemoryArena) { fn close_doc(&mut self, heap: &mut MemoryArena) {
let _ = write_u32_vint(POSITION_END, &mut self.stack.writer(heap)); self.stack.push(POSITION_END, heap);
} }
fn serialize( fn serialize(&self, serializer: &mut FieldSerializer, heap: &MemoryArena) -> io::Result<()> {
&self, let mut doc_positions = Vec::with_capacity(100);
buffer_lender: &mut BufferLender, let mut positions_iter = self.stack.iter(heap);
serializer: &mut FieldSerializer, while let Some(doc) = positions_iter.next() {
heap: &MemoryArena, let mut prev_position = 0;
) -> io::Result<()> { doc_positions.clear();
let (buffer_u8, buffer_positions) = buffer_lender.lend_all(); for position in &mut positions_iter {
self.stack.read_to_end(heap, buffer_u8); if position == POSITION_END {
let mut u32_it = VInt32Reader::new(&buffer_u8[..]); break;
while let Some(doc) = u32_it.next() { } else {
let mut prev_position_plus_one = 1u32; doc_positions.push(position - prev_position);
buffer_positions.clear(); prev_position = position;
loop {
match u32_it.next() {
Some(POSITION_END) | None => {
break;
}
Some(position_plus_one) => {
let delta_position = position_plus_one - prev_position_plus_one;
buffer_positions.push(delta_position);
prev_position_plus_one = position_plus_one;
}
} }
} }
serializer.write_doc(doc, buffer_positions.len() as u32, &buffer_positions)?; serializer.write_doc(doc, doc_positions.len() as u32, &doc_positions)?;
} }
Ok(()) Ok(())
} }
} }
#[cfg(test)]
mod tests {
use super::write_u32_vint;
use super::BufferLender;
use super::VInt32Reader;
#[test]
fn test_buffer_lender() {
let mut buffer_lender = BufferLender::default();
{
let buf = buffer_lender.lend_u8();
assert!(buf.is_empty());
buf.push(1u8);
}
{
let buf = buffer_lender.lend_u8();
assert!(buf.is_empty());
buf.push(1u8);
}
{
let (_, buf) = buffer_lender.lend_all();
assert!(buf.is_empty());
buf.push(1u32);
}
{
let (_, buf) = buffer_lender.lend_all();
assert!(buf.is_empty());
buf.push(1u32);
}
}
#[test]
fn test_vint_u32() {
let mut buffer = vec![];
let vals = [0, 1, 324_234_234, u32::max_value()];
for &i in &vals {
assert!(write_u32_vint(i, &mut buffer).is_ok());
}
assert_eq!(buffer.len(), 1 + 1 + 5 + 5);
let res: Vec<u32> = VInt32Reader::new(&buffer[..]).collect();
assert_eq!(&res[..], &vals[..]);
}
}

205
src/postings/segment_postings.rs
View File

@@ -2,21 +2,22 @@ use common::BitSet;
use common::HasLen; use common::HasLen;
use common::{BinarySerializable, VInt}; use common::{BinarySerializable, VInt};
use docset::{DocSet, SkipResult}; use docset::{DocSet, SkipResult};
use fst::Streamer;
use owned_read::OwnedRead; use owned_read::OwnedRead;
use positions::PositionReader; use positions::PositionReader;
use postings::compression::compressed_block_size; use postings::compression::compressed_block_size;
use postings::compression::{BlockDecoder, VIntDecoder, COMPRESSION_BLOCK_SIZE}; use postings::compression::{BlockDecoder, VIntDecoder, COMPRESSION_BLOCK_SIZE};
use postings::serializer::PostingsSerializer; use postings::serializer::PostingsSerializer;
use postings::BlockSearcher;
use postings::FreqReadingOption; use postings::FreqReadingOption;
use postings::Postings; use postings::Postings;
use postings::SkipReader; use postings::SkipReader;
use postings::USE_SKIP_INFO_LIMIT; use postings::USE_SKIP_INFO_LIMIT;
use schema::IndexRecordOption; use schema::IndexRecordOption;
use std::cmp::Ordering; use std::cmp::Ordering;
use tantivy_fst::Streamer;
use DocId; use DocId;
const EMPTY_ARR: [u8; 0] = [];
struct PositionComputer { struct PositionComputer {
// store the amount of position int // store the amount of position int
// before reading positions. // before reading positions.
@@ -61,7 +62,6 @@ pub struct SegmentPostings {
block_cursor: BlockSegmentPostings, block_cursor: BlockSegmentPostings,
cur: usize, cur: usize,
position_computer: Option<PositionComputer>, position_computer: Option<PositionComputer>,
block_searcher: BlockSearcher,
} }
impl SegmentPostings { impl SegmentPostings {
@@ -72,7 +72,6 @@ impl SegmentPostings {
block_cursor: empty_block_cursor, block_cursor: empty_block_cursor,
cur: COMPRESSION_BLOCK_SIZE, cur: COMPRESSION_BLOCK_SIZE,
position_computer: None, position_computer: None,
block_searcher: BlockSearcher::default(),
} }
} }
@@ -120,31 +119,46 @@ impl SegmentPostings {
block_cursor: segment_block_postings, block_cursor: segment_block_postings,
cur: COMPRESSION_BLOCK_SIZE, // cursor within the block cur: COMPRESSION_BLOCK_SIZE, // cursor within the block
position_computer: positions_stream_opt.map(PositionComputer::new), position_computer: positions_stream_opt.map(PositionComputer::new),
block_searcher: BlockSearcher::default(),
} }
} }
} }
impl DocSet for SegmentPostings { fn exponential_search(target: u32, arr: &[u32]) -> (usize, usize) {
// goes to the next element. let mut start = 0;
// next needs to be called a first time to point to the correct element. let end = arr.len();
#[inline] debug_assert!(target >= arr[start]);
fn advance(&mut self) -> bool { debug_assert!(target <= arr[end - 1]);
if self.position_computer.is_some() { let mut jump = 1;
let term_freq = self.term_freq() as usize; loop {
self.position_computer.as_mut().unwrap().add_skip(term_freq); let new = start + jump;
if new >= end {
return (start, end);
} }
self.cur += 1; if arr[new] > target {
if self.cur >= self.block_cursor.block_len() { return (start, new);
self.cur = 0;
if !self.block_cursor.advance() {
self.cur = COMPRESSION_BLOCK_SIZE;
return false;
}
} }
true start = new;
jump *= 2;
} }
}
/// Search the first index containing an element greater or equal to the target.
///
/// # Assumption
///
/// The array is assumed non empty.
/// The target is assumed greater or equal to the first element.
/// The target is assumed smaller or equal to the last element.
fn search_within_block(block_docs: &[u32], target: u32) -> usize {
let (start, end) = exponential_search(target, block_docs);
start.wrapping_add(
block_docs[start..end]
.binary_search(&target)
.unwrap_or_else(|e| e),
)
}
impl DocSet for SegmentPostings {
fn skip_next(&mut self, target: DocId) -> SkipResult { fn skip_next(&mut self, target: DocId) -> SkipResult {
if !self.advance() { if !self.advance() {
return SkipResult::End; return SkipResult::End;
@@ -202,9 +216,11 @@ impl DocSet for SegmentPostings {
// we're in the right block now, start with an exponential search // we're in the right block now, start with an exponential search
let block_docs = self.block_cursor.docs(); let block_docs = self.block_cursor.docs();
debug_assert!(target >= self.doc());
let new_cur = self let new_cur = self
.block_searcher .cur
.search_in_block(&block_docs, self.cur, target); .wrapping_add(search_within_block(&block_docs[self.cur..], target));
if need_positions { if need_positions {
sum_freqs_skipped += self.block_cursor.freqs()[self.cur..new_cur] sum_freqs_skipped += self.block_cursor.freqs()[self.cur..new_cur]
.iter() .iter()
@@ -226,6 +242,29 @@ impl DocSet for SegmentPostings {
} }
} }
// goes to the next element.
// next needs to be called a first time to point to the correct element.
#[inline]
fn advance(&mut self) -> bool {
if self.position_computer.is_some() {
let term_freq = self.term_freq() as usize;
self.position_computer.as_mut().unwrap().add_skip(term_freq);
}
self.cur += 1;
if self.cur >= self.block_cursor.block_len() {
self.cur = 0;
if !self.block_cursor.advance() {
self.cur = COMPRESSION_BLOCK_SIZE;
return false;
}
}
true
}
fn size_hint(&self) -> u32 {
self.len() as u32
}
/// Return the current document's `DocId`. /// Return the current document's `DocId`.
#[inline] #[inline]
fn doc(&self) -> DocId { fn doc(&self) -> DocId {
@@ -237,10 +276,6 @@ impl DocSet for SegmentPostings {
docs[self.cur] docs[self.cur]
} }
fn size_hint(&self) -> u32 {
self.len() as u32
}
fn append_to_bitset(&mut self, bitset: &mut BitSet) { fn append_to_bitset(&mut self, bitset: &mut BitSet) {
// finish the current block // finish the current block
if self.advance() { if self.advance() {
@@ -339,7 +374,7 @@ impl BlockSegmentPostings {
let (skip_data_opt, postings_data) = split_into_skips_and_postings(doc_freq, data); let (skip_data_opt, postings_data) = split_into_skips_and_postings(doc_freq, data);
let skip_reader = match skip_data_opt { let skip_reader = match skip_data_opt {
Some(skip_data) => SkipReader::new(skip_data, record_option), Some(skip_data) => SkipReader::new(skip_data, record_option),
None => SkipReader::new(OwnedRead::new(&[][..]), record_option), None => SkipReader::new(OwnedRead::new(&EMPTY_ARR[..]), record_option),
}; };
let doc_freq = doc_freq as usize; let doc_freq = doc_freq as usize;
let num_vint_docs = doc_freq % COMPRESSION_BLOCK_SIZE; let num_vint_docs = doc_freq % COMPRESSION_BLOCK_SIZE;
@@ -373,7 +408,7 @@ impl BlockSegmentPostings {
if let Some(skip_data) = skip_data_opt { if let Some(skip_data) = skip_data_opt {
self.skip_reader.reset(skip_data); self.skip_reader.reset(skip_data);
} else { } else {
self.skip_reader.reset(OwnedRead::new(&[][..])) self.skip_reader.reset(OwnedRead::new(&EMPTY_ARR[..]))
} }
self.doc_offset = 0; self.doc_offset = 0;
self.doc_freq = doc_freq as usize; self.doc_freq = doc_freq as usize;
@@ -498,8 +533,7 @@ impl BlockSegmentPostings {
} else { } else {
BlockSegmentPostingsSkipResult::Terminated BlockSegmentPostingsSkipResult::Terminated
} }
}) }).unwrap_or(BlockSegmentPostingsSkipResult::Terminated);
.unwrap_or(BlockSegmentPostingsSkipResult::Terminated);
} }
BlockSegmentPostingsSkipResult::Terminated BlockSegmentPostingsSkipResult::Terminated
} }
@@ -586,19 +620,20 @@ impl<'b> Streamer<'b> for BlockSegmentPostings {
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
use super::search_within_block;
use super::BlockSegmentPostings; use super::BlockSegmentPostings;
use super::BlockSegmentPostingsSkipResult; use super::BlockSegmentPostingsSkipResult;
use super::SegmentPostings; use super::SegmentPostings;
use common::HasLen; use common::HasLen;
use core::Index; use core::Index;
use docset::DocSet; use docset::DocSet;
use fst::Streamer;
use schema::IndexRecordOption; use schema::IndexRecordOption;
use schema::Schema; use schema::SchemaBuilder;
use schema::Term; use schema::Term;
use schema::INDEXED; use schema::INT_INDEXED;
use tantivy_fst::Streamer;
use DocId; use DocId;
use SkipResult;
#[test] #[test]
fn test_empty_segment_postings() { fn test_empty_segment_postings() {
@@ -615,9 +650,49 @@ mod tests {
assert_eq!(postings.doc_freq(), 0); assert_eq!(postings.doc_freq(), 0);
} }
fn search_within_block_trivial_but_slow(block: &[u32], target: u32) -> usize {
block
.iter()
.cloned()
.enumerate()
.filter(|&(_, ref val)| *val >= target)
.next()
.unwrap()
.0
}
fn util_test_search_within_block(block: &[u32], target: u32) {
assert_eq!(
search_within_block(block, target),
search_within_block_trivial_but_slow(block, target)
);
}
fn util_test_search_within_block_all(block: &[u32]) {
use std::collections::HashSet;
let mut targets = HashSet::new();
for (i, val) in block.iter().cloned().enumerate() {
if i > 0 {
targets.insert(val - 1);
}
targets.insert(val);
}
for target in targets {
util_test_search_within_block(block, target);
}
}
#[test]
fn test_search_within_block() {
for len in 1u32..128u32 {
let v: Vec<u32> = (0..len).map(|i| i * 2).collect();
util_test_search_within_block_all(&v[..]);
}
}
#[test] #[test]
fn test_block_segment_postings() { fn test_block_segment_postings() {
let mut block_segments = build_block_postings(&(0..100_000).collect::<Vec<u32>>()); let mut block_segments = build_block_postings((0..100_000).collect::<Vec<u32>>());
let mut offset: u32 = 0u32; let mut offset: u32 = 0u32;
// checking that the block before calling advance is empty // checking that the block before calling advance is empty
assert!(block_segments.docs().is_empty()); assert!(block_segments.docs().is_empty());
@@ -631,44 +706,14 @@ mod tests {
} }
} }
#[test] fn build_block_postings(docs: Vec<DocId>) -> BlockSegmentPostings {
fn test_skip_right_at_new_block() { let mut schema_builder = SchemaBuilder::default();
let mut doc_ids = (0..128).collect::<Vec<u32>>(); let int_field = schema_builder.add_u64_field("id", INT_INDEXED);
doc_ids.push(129);
doc_ids.push(130);
{
let block_segments = build_block_postings(&doc_ids);
let mut docset = SegmentPostings::from_block_postings(block_segments, None);
assert_eq!(docset.skip_next(128), SkipResult::OverStep);
assert_eq!(docset.doc(), 129);
assert!(docset.advance());
assert_eq!(docset.doc(), 130);
assert!(!docset.advance());
}
{
let block_segments = build_block_postings(&doc_ids);
let mut docset = SegmentPostings::from_block_postings(block_segments, None);
assert_eq!(docset.skip_next(129), SkipResult::Reached);
assert_eq!(docset.doc(), 129);
assert!(docset.advance());
assert_eq!(docset.doc(), 130);
assert!(!docset.advance());
}
{
let block_segments = build_block_postings(&doc_ids);
let mut docset = SegmentPostings::from_block_postings(block_segments, None);
assert_eq!(docset.skip_next(131), SkipResult::End);
}
}
fn build_block_postings(docs: &[DocId]) -> BlockSegmentPostings {
let mut schema_builder = Schema::builder();
let int_field = schema_builder.add_u64_field("id", INDEXED);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
let mut last_doc = 0u32; let mut last_doc = 0u32;
for &doc in docs { for doc in docs {
for _ in last_doc..doc { for _ in last_doc..doc {
index_writer.add_document(doc!(int_field=>1u64)); index_writer.add_document(doc!(int_field=>1u64));
} }
@@ -676,7 +721,8 @@ mod tests {
last_doc = doc + 1; last_doc = doc + 1;
} }
index_writer.commit().unwrap(); index_writer.commit().unwrap();
let searcher = index.reader().unwrap().searcher(); index.load_searchers().unwrap();
let searcher = index.searcher();
let segment_reader = searcher.segment_reader(0); let segment_reader = searcher.segment_reader(0);
let inverted_index = segment_reader.inverted_index(int_field); let inverted_index = segment_reader.inverted_index(int_field);
let term = Term::from_field_u64(int_field, 0u64); let term = Term::from_field_u64(int_field, 0u64);
@@ -687,7 +733,7 @@ mod tests {
#[test] #[test]
fn test_block_segment_postings_skip() { fn test_block_segment_postings_skip() {
for i in 0..4 { for i in 0..4 {
let mut block_postings = build_block_postings(&[3]); let mut block_postings = build_block_postings(vec![3]);
assert_eq!( assert_eq!(
block_postings.skip_to(i), block_postings.skip_to(i),
BlockSegmentPostingsSkipResult::Success(0u32) BlockSegmentPostingsSkipResult::Success(0u32)
@@ -697,7 +743,7 @@ mod tests {
BlockSegmentPostingsSkipResult::Terminated BlockSegmentPostingsSkipResult::Terminated
); );
} }
let mut block_postings = build_block_postings(&[3]); let mut block_postings = build_block_postings(vec![3]);
assert_eq!( assert_eq!(
block_postings.skip_to(4u32), block_postings.skip_to(4u32),
BlockSegmentPostingsSkipResult::Terminated BlockSegmentPostingsSkipResult::Terminated
@@ -710,7 +756,7 @@ mod tests {
for i in 0..1300 { for i in 0..1300 {
docs.push((i * i / 100) + i); docs.push((i * i / 100) + i);
} }
let mut block_postings = build_block_postings(&docs[..]); let mut block_postings = build_block_postings(docs.clone());
for i in vec![0, 424, 10000] { for i in vec![0, 424, 10000] {
assert_eq!( assert_eq!(
block_postings.skip_to(i), block_postings.skip_to(i),
@@ -732,11 +778,11 @@ mod tests {
#[test] #[test]
fn test_reset_block_segment_postings() { fn test_reset_block_segment_postings() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let int_field = schema_builder.add_u64_field("id", INDEXED); let int_field = schema_builder.add_u64_field("id", INT_INDEXED);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
// create two postings list, one containg even number, // create two postings list, one containg even number,
// the other containing odd numbers. // the other containing odd numbers.
for i in 0..6 { for i in 0..6 {
@@ -744,7 +790,8 @@ mod tests {
index_writer.add_document(doc); index_writer.add_document(doc);
} }
index_writer.commit().unwrap(); index_writer.commit().unwrap();
let searcher = index.reader().unwrap().searcher(); index.load_searchers().unwrap();
let searcher = index.searcher();
let segment_reader = searcher.segment_reader(0); let segment_reader = searcher.segment_reader(0);
let mut block_segments; let mut block_segments;

10
src/postings/serializer.rs
View File

@@ -55,7 +55,7 @@ pub struct InvertedIndexSerializer {
impl InvertedIndexSerializer { impl InvertedIndexSerializer {
/// Open a new `PostingsSerializer` for the given segment /// Open a new `PostingsSerializer` for the given segment
fn create( fn new(
terms_write: CompositeWrite<WritePtr>, terms_write: CompositeWrite<WritePtr>,
postings_write: CompositeWrite<WritePtr>, postings_write: CompositeWrite<WritePtr>,
positions_write: CompositeWrite<WritePtr>, positions_write: CompositeWrite<WritePtr>,
@@ -74,7 +74,7 @@ impl InvertedIndexSerializer {
/// Open a new `PostingsSerializer` for the given segment /// Open a new `PostingsSerializer` for the given segment
pub fn open(segment: &mut Segment) -> Result<InvertedIndexSerializer> { pub fn open(segment: &mut Segment) -> Result<InvertedIndexSerializer> {
use SegmentComponent::{POSITIONS, POSITIONSSKIP, POSTINGS, TERMS}; use SegmentComponent::{POSITIONS, POSITIONSSKIP, POSTINGS, TERMS};
InvertedIndexSerializer::create( InvertedIndexSerializer::new(
CompositeWrite::wrap(segment.open_write(TERMS)?), CompositeWrite::wrap(segment.open_write(TERMS)?),
CompositeWrite::wrap(segment.open_write(POSTINGS)?), CompositeWrite::wrap(segment.open_write(POSTINGS)?),
CompositeWrite::wrap(segment.open_write(POSITIONS)?), CompositeWrite::wrap(segment.open_write(POSITIONS)?),
@@ -99,7 +99,7 @@ impl InvertedIndexSerializer {
let positions_write = self.positions_write.for_field(field); let positions_write = self.positions_write.for_field(field);
let positionsidx_write = self.positionsidx_write.for_field(field); let positionsidx_write = self.positionsidx_write.for_field(field);
let field_type: FieldType = (*field_entry.field_type()).clone(); let field_type: FieldType = (*field_entry.field_type()).clone();
FieldSerializer::create( FieldSerializer::new(
&field_type, &field_type,
term_dictionary_write, term_dictionary_write,
postings_write, postings_write,
@@ -130,7 +130,7 @@ pub struct FieldSerializer<'a> {
} }
impl<'a> FieldSerializer<'a> { impl<'a> FieldSerializer<'a> {
fn create( fn new(
field_type: &FieldType, field_type: &FieldType,
term_dictionary_write: &'a mut CountingWriter<WritePtr>, term_dictionary_write: &'a mut CountingWriter<WritePtr>,
postings_write: &'a mut CountingWriter<WritePtr>, postings_write: &'a mut CountingWriter<WritePtr>,
@@ -152,7 +152,7 @@ impl<'a> FieldSerializer<'a> {
_ => (false, false), _ => (false, false),
}; };
let term_dictionary_builder = let term_dictionary_builder =
TermDictionaryBuilder::create(term_dictionary_write, &field_type)?; TermDictionaryBuilder::new(term_dictionary_write, &field_type)?;
let postings_serializer = let postings_serializer =
PostingsSerializer::new(postings_write, term_freq_enabled, position_enabled); PostingsSerializer::new(postings_write, term_freq_enabled, position_enabled);
let positions_serializer_opt = if position_enabled { let positions_serializer_opt = if position_enabled {

337
src/postings/stacker/expull.rs
View File

@@ -1,37 +1,28 @@
use super::{Addr, MemoryArena}; use super::{Addr, MemoryArena};
use postings::stacker::memory_arena::load; use common::is_power_of_2;
use postings::stacker::memory_arena::store;
use std::io;
use std::mem; use std::mem;
const MAX_BLOCK_LEN: u32 = 1u32 << 15; const MAX_BLOCK_LEN: u32 = 1u32 << 15;
const FIRST_BLOCK: usize = 16;
const INLINED_BLOCK_LEN: usize = FIRST_BLOCK + mem::size_of::<Addr>();
enum CapacityResult { const FIRST_BLOCK: u32 = 4u32;
Available(u32),
NeedAlloc(u32),
}
fn len_to_capacity(len: u32) -> CapacityResult { #[inline]
pub fn jump_needed(len: u32) -> Option<usize> {
match len { match len {
0...15 => CapacityResult::Available(FIRST_BLOCK as u32 - len), 0...3 => None,
16...MAX_BLOCK_LEN => { 4...MAX_BLOCK_LEN => {
let cap = 1 << (32u32 - (len - 1u32).leading_zeros()); if is_power_of_2(len as usize) {
let available = cap - len; Some(len as usize)
if available == 0 {
CapacityResult::NeedAlloc(len)
} else { } else {
CapacityResult::Available(available) None
} }
} }
n => { n => {
let available = n % MAX_BLOCK_LEN; if n % MAX_BLOCK_LEN == 0 {
if available == 0 { Some(MAX_BLOCK_LEN as usize)
CapacityResult::NeedAlloc(MAX_BLOCK_LEN)
} else { } else {
CapacityResult::Available(MAX_BLOCK_LEN - available) None
} }
} }
} }
@@ -61,119 +52,82 @@ fn len_to_capacity(len: u32) -> CapacityResult {
#[derive(Debug, Clone, Copy)] #[derive(Debug, Clone, Copy)]
pub struct ExpUnrolledLinkedList { pub struct ExpUnrolledLinkedList {
len: u32, len: u32,
head: Addr,
tail: Addr, tail: Addr,
inlined_data: [u8; INLINED_BLOCK_LEN as usize],
}
pub struct ExpUnrolledLinkedListWriter<'a> {
eull: &'a mut ExpUnrolledLinkedList,
heap: &'a mut MemoryArena,
}
fn ensure_capacity<'a>(
eull: &'a mut ExpUnrolledLinkedList,
heap: &'a mut MemoryArena,
) -> &'a mut [u8] {
if eull.len <= FIRST_BLOCK as u32 {
// We are still hitting the inline block.
if eull.len < FIRST_BLOCK as u32 {
return &mut eull.inlined_data[eull.len as usize..FIRST_BLOCK];
}
// We need to allocate a new block!
let new_block_addr: Addr = heap.allocate_space(FIRST_BLOCK + mem::size_of::<Addr>());
store(&mut eull.inlined_data[FIRST_BLOCK..], new_block_addr);
eull.tail = new_block_addr;
return heap.slice_mut(eull.tail, FIRST_BLOCK);
}
let len = match len_to_capacity(eull.len) {
CapacityResult::NeedAlloc(new_block_len) => {
let new_block_addr: Addr =
heap.allocate_space(new_block_len as usize + mem::size_of::<Addr>());
heap.write_at(eull.tail, new_block_addr);
eull.tail = new_block_addr;
new_block_len
}
CapacityResult::Available(available) => available,
};
heap.slice_mut(eull.tail, len as usize)
}
impl<'a> ExpUnrolledLinkedListWriter<'a> {
pub fn extend_from_slice(&mut self, mut buf: &[u8]) {
if buf.is_empty() {
// we need to cut early, because `ensure_capacity`
// allocates if there is no capacity at all right now.
return;
}
while !buf.is_empty() {
let add_len: usize;
{
let output_buf = ensure_capacity(self.eull, self.heap);
add_len = buf.len().min(output_buf.len());
output_buf[..add_len].copy_from_slice(&buf[..add_len]);
}
self.eull.len += add_len as u32;
self.eull.tail = self.eull.tail.offset(add_len as u32);
buf = &buf[add_len..];
}
}
}
impl<'a> io::Write for ExpUnrolledLinkedListWriter<'a> {
fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
// There is no use case to only write the capacity.
// This is not IO after all, so we write the whole
// buffer even if the contract of `.write` is looser.
self.extend_from_slice(buf);
Ok(buf.len())
}
fn write_all(&mut self, buf: &[u8]) -> io::Result<()> {
self.extend_from_slice(buf);
Ok(())
}
fn flush(&mut self) -> io::Result<()> {
Ok(())
}
} }
impl ExpUnrolledLinkedList { impl ExpUnrolledLinkedList {
pub fn new() -> ExpUnrolledLinkedList { pub fn new(heap: &mut MemoryArena) -> ExpUnrolledLinkedList {
let addr = heap.allocate_space((FIRST_BLOCK as usize) * mem::size_of::<u32>());
ExpUnrolledLinkedList { ExpUnrolledLinkedList {
len: 0u32, len: 0u32,
tail: Addr::null_pointer(), head: addr,
inlined_data: [0u8; INLINED_BLOCK_LEN as usize], tail: addr,
} }
} }
#[inline(always)] pub fn iter<'a>(&self, heap: &'a MemoryArena) -> ExpUnrolledLinkedListIterator<'a> {
pub fn writer<'a>(&'a mut self, heap: &'a mut MemoryArena) -> ExpUnrolledLinkedListWriter<'a> { ExpUnrolledLinkedListIterator {
ExpUnrolledLinkedListWriter { eull: self, heap } heap,
addr: self.head,
len: self.len,
consumed: 0,
}
} }
pub fn read_to_end(&self, heap: &MemoryArena, output: &mut Vec<u8>) { /// Appends a new element to the current stack.
let len = self.len as usize; ///
if len <= FIRST_BLOCK { /// If the current block end is reached, a new block is allocated.
output.extend_from_slice(&self.inlined_data[..len]); pub fn push(&mut self, val: u32, heap: &mut MemoryArena) {
return; self.len += 1;
if let Some(new_block_len) = jump_needed(self.len) {
// We need to allocate another block.
// We also allocate an extra `u32` to store the pointer
// to the future next block.
let new_block_size: usize = (new_block_len + 1) * mem::size_of::<u32>();
let new_block_addr: Addr = heap.allocate_space(new_block_size);
unsafe {
// logic
heap.write(self.tail, new_block_addr)
};
self.tail = new_block_addr;
} }
output.extend_from_slice(&self.inlined_data[..FIRST_BLOCK]); unsafe {
let mut cur = FIRST_BLOCK; // logic
let mut addr = load(&self.inlined_data[FIRST_BLOCK..]); heap.write(self.tail, val);
loop { self.tail = self.tail.offset(mem::size_of::<u32>() as u32);
let cap = match len_to_capacity(cur as u32) { }
CapacityResult::Available(capacity) => capacity, }
CapacityResult::NeedAlloc(capacity) => capacity, }
} as usize;
let data = heap.slice(addr, cap); pub struct ExpUnrolledLinkedListIterator<'a> {
if cur + cap >= len { heap: &'a MemoryArena,
output.extend_from_slice(&data[..(len - cur)]); addr: Addr,
return; len: u32,
} consumed: u32,
output.extend_from_slice(data); }
cur += cap;
addr = heap.read(addr.offset(cap as u32)); impl<'a> Iterator for ExpUnrolledLinkedListIterator<'a> {
type Item = u32;
fn next(&mut self) -> Option<u32> {
if self.consumed == self.len {
None
} else {
self.consumed += 1;
let addr: Addr = if jump_needed(self.consumed).is_some() {
unsafe {
// logic
self.heap.read(self.addr)
}
} else {
self.addr
};
self.addr = addr.offset(mem::size_of::<u32>() as u32);
Some(unsafe {
// logic
self.heap.read(addr)
})
} }
} }
} }
@@ -182,134 +136,46 @@ impl ExpUnrolledLinkedList {
mod tests { mod tests {
use super::super::MemoryArena; use super::super::MemoryArena;
use super::len_to_capacity; use super::jump_needed;
use super::*; use super::*;
use byteorder::{ByteOrder, LittleEndian, WriteBytesExt};
#[test]
#[test] #[test]
fn test_stack() { fn test_stack() {
let mut heap = MemoryArena::new(); let mut heap = MemoryArena::new();
let mut stack = ExpUnrolledLinkedList::new(); let mut stack = ExpUnrolledLinkedList::new(&mut heap);
stack.writer(&mut heap).extend_from_slice(&[1u8]); stack.push(1u32, &mut heap);
stack.writer(&mut heap).extend_from_slice(&[2u8]); stack.push(2u32, &mut heap);
stack.writer(&mut heap).extend_from_slice(&[3u8, 4u8]); stack.push(4u32, &mut heap);
stack.writer(&mut heap).extend_from_slice(&[5u8]); stack.push(8u32, &mut heap);
{ {
let mut buffer = Vec::new(); let mut it = stack.iter(&heap);
stack.read_to_end(&heap, &mut buffer); assert_eq!(it.next().unwrap(), 1u32);
assert_eq!(&buffer[..], &[1u8, 2u8, 3u8, 4u8, 5u8]); assert_eq!(it.next().unwrap(), 2u32);
assert_eq!(it.next().unwrap(), 4u32);
assert_eq!(it.next().unwrap(), 8u32);
assert!(it.next().is_none());
} }
} }
#[test]
fn test_stack_long() {
let mut heap = MemoryArena::new();
let mut stack = ExpUnrolledLinkedList::new();
let source: Vec<u32> = (0..100).collect();
for &el in &source {
assert!(stack
.writer(&mut heap)
.write_u32::<LittleEndian>(el)
.is_ok());
}
let mut buffer = Vec::new();
stack.read_to_end(&heap, &mut buffer);
let mut result = vec![];
let mut remaining = &buffer[..];
while !remaining.is_empty() {
result.push(LittleEndian::read_u32(&remaining[..4]));
remaining = &remaining[4..];
}
assert_eq!(&result[..], &source[..]);
}
#[test]
fn test_stack_interlaced() {
let mut heap = MemoryArena::new();
let mut stack = ExpUnrolledLinkedList::new();
let mut stack2 = ExpUnrolledLinkedList::new();
let mut vec1: Vec<u8> = vec![];
let mut vec2: Vec<u8> = vec![];
for i in 0..9 {
assert!(stack.writer(&mut heap).write_u32::<LittleEndian>(i).is_ok());
assert!(vec1.write_u32::<LittleEndian>(i).is_ok());
if i % 2 == 0 {
assert!(stack2
.writer(&mut heap)
.write_u32::<LittleEndian>(i)
.is_ok());
assert!(vec2.write_u32::<LittleEndian>(i).is_ok());
}
}
let mut res1 = vec![];
let mut res2 = vec![];
stack.read_to_end(&heap, &mut res1);
stack2.read_to_end(&heap, &mut res2);
assert_eq!(&vec1[..], &res1[..]);
assert_eq!(&vec2[..], &res2[..]);
}
#[test] #[test]
fn test_jump_if_needed() { fn test_jump_if_needed() {
let mut available = 16u32; let mut block_len = 4u32;
for i in 0..10_000_000 { let mut i = 0;
match len_to_capacity(i) { while i < 10_000_000 {
CapacityResult::NeedAlloc(cap) => { assert!(jump_needed(i + block_len - 1).is_none());
assert_eq!(available, 0, "Failed len={}: Expected 0 got {}", i, cap); assert!(jump_needed(i + block_len + 1).is_none());
available = cap; assert!(jump_needed(i + block_len).is_some());
} let new_block_len = jump_needed(i + block_len).unwrap();
CapacityResult::Available(cap) => { i += block_len;
assert_eq!( block_len = new_block_len as u32;
available, cap,
"Failed len={}: Expected {} Got {}",
i, available, cap
);
}
}
available -= 1;
} }
} }
#[test]
fn test_jump_if_needed_progression() {
let mut v = vec![];
for i in 0.. {
if v.len() >= 10 {
break;
}
match len_to_capacity(i) {
CapacityResult::NeedAlloc(cap) => {
v.push((i, cap));
}
_ => {}
}
}
assert_eq!(
&v[..],
&[
(16, 16),
(32, 32),
(64, 64),
(128, 128),
(256, 256),
(512, 512),
(1024, 1024),
(2048, 2048),
(4096, 4096),
(8192, 8192)
]
);
}
} }
#[cfg(all(test, feature = "unstable"))] #[cfg(all(test, feature = "unstable"))]
mod bench { mod bench {
use super::super::MemoryArena;
use super::ExpUnrolledLinkedList; use super::ExpUnrolledLinkedList;
use byteorder::{NativeEndian, WriteBytesExt}; use tantivy_memory_arena::MemoryArena;
use test::Bencher; use test::Bencher;
const NUM_STACK: usize = 10_000; const NUM_STACK: usize = 10_000;
@@ -333,19 +199,20 @@ mod bench {
#[bench] #[bench]
fn bench_push_stack(bench: &mut Bencher) { fn bench_push_stack(bench: &mut Bencher) {
let heap = MemoryArena::new();
bench.iter(|| { bench.iter(|| {
let mut heap = MemoryArena::new();
let mut stacks = Vec::with_capacity(100); let mut stacks = Vec::with_capacity(100);
for _ in 0..NUM_STACK { for _ in 0..NUM_STACK {
let mut stack = ExpUnrolledLinkedList::new(); let (_, stack) = heap.allocate_object::<ExpUnrolledLinkedList>();
stacks.push(stack); stacks.push(stack);
} }
for s in 0..NUM_STACK { for s in 0..NUM_STACK {
for i in 0u32..STACK_SIZE { for i in 0u32..STACK_SIZE {
let t = s * 392017 % NUM_STACK; let t = s * 392017 % NUM_STACK;
let _ = stacks[t].writer(&mut heap).write_u32::<NativeEndian>(i); stacks[t].push(i, &heap);
} }
} }
heap.clear();
}); });
} }
} }

148
src/postings/stacker/memory_arena.rs
View File

@@ -37,7 +37,7 @@ const PAGE_SIZE: usize = 1 << NUM_BITS_PAGE_ADDR; // pages are 1 MB large
/// page of memory. /// page of memory.
/// ///
/// The last 20 bits are an address within this page of memory. /// The last 20 bits are an address within this page of memory.
#[derive(Copy, Clone, Debug)] #[derive(Clone, Copy, Debug)]
pub struct Addr(u32); pub struct Addr(u32);
impl Addr { impl Addr {
@@ -69,16 +69,32 @@ impl Addr {
} }
} }
pub fn store<Item: Copy + 'static>(dest: &mut [u8], val: Item) { /// Trait required for an object to be `storable`.
assert_eq!(dest.len(), std::mem::size_of::<Item>()); ///
unsafe { /// # Warning
ptr::write_unaligned(dest.as_mut_ptr() as *mut Item, val); ///
} /// Most of the time you should not implement this trait,
/// and only use the `MemoryArena` with object implementing `Copy`.
///
/// `ArenaStorable` is used in `tantivy` to force
/// a `Copy` object and a `slice` of data to be stored contiguously.
pub trait ArenaStorable {
fn num_bytes(&self) -> usize;
unsafe fn write_into(self, arena: &mut MemoryArena, addr: Addr);
} }
pub fn load<Item: Copy + 'static>(data: &[u8]) -> Item { impl<V> ArenaStorable for V
assert_eq!(data.len(), std::mem::size_of::<Item>()); where
unsafe { ptr::read_unaligned(data.as_ptr() as *const Item) } V: Copy,
{
fn num_bytes(&self) -> usize {
mem::size_of::<V>()
}
unsafe fn write_into(self, arena: &mut MemoryArena, addr: Addr) {
let dst_ptr = arena.get_mut_ptr(addr) as *mut V;
ptr::write_unaligned(dst_ptr, self);
}
} }
/// The `MemoryArena` /// The `MemoryArena`
@@ -110,9 +126,47 @@ impl MemoryArena {
self.pages.len() * PAGE_SIZE self.pages.len() * PAGE_SIZE
} }
pub fn write_at<Item: Copy + 'static>(&mut self, addr: Addr, val: Item) { /// Writes a slice at the given address, assuming the
let dest = self.slice_mut(addr, std::mem::size_of::<Item>()); /// memory was allocated beforehands.
store(dest, val); ///
/// # Panics
///
/// May panic or corrupt the heap if he space was not
/// properly allocated beforehands.
pub fn write_bytes<B: AsRef<[u8]>>(&mut self, addr: Addr, data: B) {
let bytes = data.as_ref();
self.pages[addr.page_id()]
.get_mut_slice(addr.page_local_addr(), bytes.len())
.copy_from_slice(bytes);
}
/// Returns the `len` bytes starting at `addr`
///
/// # Panics
///
/// Panics if the memory has not been allocated beforehands.
pub fn read_slice(&self, addr: Addr, len: usize) -> &[u8] {
self.pages[addr.page_id()].get_slice(addr.page_local_addr(), len)
}
unsafe fn get_mut_ptr(&mut self, addr: Addr) -> *mut u8 {
self.pages[addr.page_id()].get_mut_ptr(addr.page_local_addr())
}
/// Stores an item's data in the heap
///
/// It allocates the `Item` beforehands.
pub fn store<Item: ArenaStorable>(&mut self, val: Item) -> Addr {
let num_bytes = val.num_bytes();
let addr = self.allocate_space(num_bytes);
unsafe {
self.write(addr, val);
};
addr
}
pub unsafe fn write<Item: ArenaStorable>(&mut self, addr: Addr, val: Item) {
val.write_into(self, addr)
} }
/// Read an item in the heap at the given `address`. /// Read an item in the heap at the given `address`.
@@ -120,21 +174,9 @@ impl MemoryArena {
/// # Panics /// # Panics
/// ///
/// If the address is erroneous /// If the address is erroneous
pub fn read<Item: Copy + 'static>(&self, addr: Addr) -> Item { pub unsafe fn read<Item: Copy>(&self, addr: Addr) -> Item {
load(self.slice(addr, mem::size_of::<Item>())) let ptr = self.pages[addr.page_id()].get_ptr(addr.page_local_addr());
} ptr::read_unaligned(ptr as *const Item)
pub fn slice(&self, addr: Addr, len: usize) -> &[u8] {
self.pages[addr.page_id()].slice(addr.page_local_addr(), len)
}
pub fn slice_from(&self, addr: Addr) -> &[u8] {
self.pages[addr.page_id()].slice_from(addr.page_local_addr())
}
#[inline(always)]
pub fn slice_mut(&mut self, addr: Addr, len: usize) -> &mut [u8] {
self.pages[addr.page_id()].slice_mut(addr.page_local_addr(), len)
} }
/// Allocates `len` bytes and returns the allocated address. /// Allocates `len` bytes and returns the allocated address.
@@ -155,10 +197,14 @@ struct Page {
impl Page { impl Page {
fn new(page_id: usize) -> Page { fn new(page_id: usize) -> Page {
let mut data: Vec<u8> = Vec::with_capacity(PAGE_SIZE);
unsafe {
data.set_len(PAGE_SIZE);
} // avoid initializing page
Page { Page {
page_id, page_id,
len: 0, len: 0,
data: vec![0u8; PAGE_SIZE].into_boxed_slice(), data: data.into_boxed_slice(),
} }
} }
@@ -167,18 +213,14 @@ impl Page {
len + self.len <= PAGE_SIZE len + self.len <= PAGE_SIZE
} }
fn slice(&self, local_addr: usize, len: usize) -> &[u8] { fn get_mut_slice(&mut self, local_addr: usize, len: usize) -> &mut [u8] {
&self.slice_from(local_addr)[..len]
}
fn slice_from(&self, local_addr: usize) -> &[u8] {
&self.data[local_addr..]
}
fn slice_mut(&mut self, local_addr: usize, len: usize) -> &mut [u8] {
&mut self.data[local_addr..][..len] &mut self.data[local_addr..][..len]
} }
fn get_slice(&self, local_addr: usize, len: usize) -> &[u8] {
&self.data[local_addr..][..len]
}
fn allocate_space(&mut self, len: usize) -> Option<Addr> { fn allocate_space(&mut self, len: usize) -> Option<Addr> {
if self.is_available(len) { if self.is_available(len) {
let addr = Addr::new(self.page_id, self.len); let addr = Addr::new(self.page_id, self.len);
@@ -188,6 +230,16 @@ impl Page {
None None
} }
} }
#[inline(always)]
pub(crate) unsafe fn get_ptr(&self, addr: usize) -> *const u8 {
self.data.as_ptr().add(addr)
}
#[inline(always)]
pub(crate) unsafe fn get_mut_ptr(&mut self, addr: usize) -> *mut u8 {
self.data.as_mut_ptr().add(addr)
}
} }
#[cfg(test)] #[cfg(test)]
@@ -202,13 +254,13 @@ mod tests {
let b = b"happy tax payer"; let b = b"happy tax payer";
let addr_a = arena.allocate_space(a.len()); let addr_a = arena.allocate_space(a.len());
arena.slice_mut(addr_a, a.len()).copy_from_slice(a); arena.write_bytes(addr_a, a);
let addr_b = arena.allocate_space(b.len()); let addr_b = arena.allocate_space(b.len());
arena.slice_mut(addr_b, b.len()).copy_from_slice(b); arena.write_bytes(addr_b, b);
assert_eq!(arena.slice(addr_a, a.len()), a); assert_eq!(arena.read_slice(addr_a, a.len()), a);
assert_eq!(arena.slice(addr_b, b.len()), b); assert_eq!(arena.read_slice(addr_b, b.len()), b);
} }
#[derive(Clone, Copy, Debug, Eq, PartialEq)] #[derive(Clone, Copy, Debug, Eq, PartialEq)]
@@ -231,15 +283,9 @@ mod tests {
b: 221, b: 221,
c: 12, c: 12,
}; };
let addr_a = arena.store(a);
let num_bytes = std::mem::size_of::<MyTest>(); let addr_b = arena.store(b);
let addr_a = arena.allocate_space(num_bytes); assert_eq!(unsafe { arena.read::<MyTest>(addr_a) }, a);
arena.write_at(addr_a, a); assert_eq!(unsafe { arena.read::<MyTest>(addr_b) }, b);
let addr_b = arena.allocate_space(num_bytes);
arena.write_at(addr_b, b);
assert_eq!(arena.read::<MyTest>(addr_a), a);
assert_eq!(arena.read::<MyTest>(addr_b), b);
} }
} }

4
src/postings/stacker/mod.rs
View File

@@ -1,7 +1,9 @@
mod expull; mod expull;
mod memory_arena; mod memory_arena;
mod murmurhash2;
mod term_hashmap; mod term_hashmap;
pub use self::expull::ExpUnrolledLinkedList; pub use self::expull::ExpUnrolledLinkedList;
pub use self::memory_arena::{Addr, MemoryArena}; pub use self::memory_arena::{Addr, ArenaStorable, MemoryArena};
use self::murmurhash2::murmurhash2;
pub use self::term_hashmap::{compute_table_size, TermHashMap}; pub use self::term_hashmap::{compute_table_size, TermHashMap};

87
src/postings/stacker/murmurhash2.rs Normal file
View File

@@ -0,0 +1,87 @@
use std::ptr;
const SEED: u32 = 3_242_157_231u32;
const M: u32 = 0x5bd1_e995;
#[inline(always)]
pub fn murmurhash2(key: &[u8]) -> u32 {
#[cfg_attr(feature = "cargo-clippy", allow(clippy::cast_ptr_alignment))]
let mut key_ptr: *const u32 = key.as_ptr() as *const u32;
let len = key.len() as u32;
let mut h: u32 = SEED ^ len;
let num_blocks = len >> 2;
for _ in 0..num_blocks {
let mut k: u32 = unsafe { ptr::read_unaligned(key_ptr) }; // ok because of num_blocks definition
k = k.wrapping_mul(M);
k ^= k >> 24;
k = k.wrapping_mul(M);
h = h.wrapping_mul(M);
h ^= k;
key_ptr = key_ptr.wrapping_offset(1);
}
// Handle the last few bytes of the input array
let remaining: &[u8] = &key[key.len() & !3..];
match remaining.len() {
3 => {
h ^= u32::from(remaining[2]) << 16;
h ^= u32::from(remaining[1]) << 8;
h ^= u32::from(remaining[0]);
h = h.wrapping_mul(M);
}
2 => {
h ^= u32::from(remaining[1]) << 8;
h ^= u32::from(remaining[0]);
h = h.wrapping_mul(M);
}
1 => {
h ^= u32::from(remaining[0]);
h = h.wrapping_mul(M);
}
_ => {}
}
h ^= h >> 13;
h = h.wrapping_mul(M);
h ^ (h >> 15)
}
#[cfg(test)]
mod test {
use super::murmurhash2;
use std::collections::HashSet;
#[test]
fn test_murmur() {
let s1 = "abcdef";
let s2 = "abcdeg";
for i in 0..5 {
assert_eq!(
murmurhash2(&s1[i..5].as_bytes()),
murmurhash2(&s2[i..5].as_bytes())
);
}
}
#[test]
fn test_murmur_against_reference_impl() {
assert_eq!(murmurhash2("".as_bytes()), 3632506080);
assert_eq!(murmurhash2("a".as_bytes()), 455683869);
assert_eq!(murmurhash2("ab".as_bytes()), 2448092234);
assert_eq!(murmurhash2("abc".as_bytes()), 2066295634);
assert_eq!(murmurhash2("abcd".as_bytes()), 2588571162);
assert_eq!(murmurhash2("abcde".as_bytes()), 2988696942);
assert_eq!(murmurhash2("abcdefghijklmnop".as_bytes()), 2350868870);
}
#[test]
fn test_murmur_collisions() {
let mut set: HashSet<u32> = HashSet::default();
for i in 0..10_000 {
let s = format!("hash{}", i);
let hash = murmurhash2(s.as_bytes());
set.insert(hash);
}
assert_eq!(set.len(), 10_000);
}
}

145
src/postings/stacker/term_hashmap.rs
View File

@@ -1,15 +1,37 @@
extern crate murmurhash32; use super::murmurhash2;
use super::{Addr, ArenaStorable, MemoryArena};
use self::murmurhash32::murmurhash2;
use super::{Addr, MemoryArena};
use byteorder::{ByteOrder, NativeEndian};
use postings::stacker::memory_arena::store;
use postings::UnorderedTermId;
use std::iter; use std::iter;
use std::mem; use std::mem;
use std::slice; use std::slice;
pub type BucketId = usize;
struct KeyBytesValue<'a, V> {
key: &'a [u8],
value: V,
}
impl<'a, V> KeyBytesValue<'a, V> {
fn new(key: &'a [u8], value: V) -> KeyBytesValue<'a, V> {
KeyBytesValue { key, value }
}
}
impl<'a, V> ArenaStorable for KeyBytesValue<'a, V>
where
V: ArenaStorable,
{
fn num_bytes(&self) -> usize {
0u16.num_bytes() + self.key.len() + self.value.num_bytes()
}
unsafe fn write_into(self, arena: &mut MemoryArena, addr: Addr) {
arena.write(addr, self.key.len() as u16);
arena.write_bytes(addr.offset(2), self.key);
arena.write(addr.offset(2 + self.key.len() as u32), self.value);
}
}
/// Returns the actual memory size in bytes /// Returns the actual memory size in bytes
/// required to create a table of size $2^num_bits$. /// required to create a table of size $2^num_bits$.
pub fn compute_table_size(num_bits: usize) -> usize { pub fn compute_table_size(num_bits: usize) -> usize {
@@ -27,7 +49,6 @@ pub fn compute_table_size(num_bits: usize) -> usize {
struct KeyValue { struct KeyValue {
key_value_addr: Addr, key_value_addr: Addr,
hash: u32, hash: u32,
unordered_term_id: UnorderedTermId,
} }
impl Default for KeyValue { impl Default for KeyValue {
@@ -35,7 +56,6 @@ impl Default for KeyValue {
KeyValue { KeyValue {
key_value_addr: Addr::null_pointer(), key_value_addr: Addr::null_pointer(),
hash: 0u32, hash: 0u32,
unordered_term_id: UnorderedTermId::default(),
} }
} }
} }
@@ -60,7 +80,6 @@ pub struct TermHashMap {
pub heap: MemoryArena, pub heap: MemoryArena,
mask: usize, mask: usize,
occupied: Vec<usize>, occupied: Vec<usize>,
len: usize,
} }
struct QuadraticProbing { struct QuadraticProbing {
@@ -87,13 +106,14 @@ pub struct Iter<'a> {
} }
impl<'a> Iterator for Iter<'a> { impl<'a> Iterator for Iter<'a> {
type Item = (&'a [u8], Addr, UnorderedTermId); type Item = (&'a [u8], Addr, BucketId);
fn next(&mut self) -> Option<Self::Item> { fn next(&mut self) -> Option<Self::Item> {
self.inner.next().cloned().map(move |bucket: usize| { self.inner.next().cloned().map(move |bucket: usize| {
let kv = self.hashmap.table[bucket]; let kv = self.hashmap.table[bucket];
let (key, offset): (&'a [u8], Addr) = self.hashmap.get_key_value(kv.key_value_addr); let (key, offset): (&'a [u8], Addr) =
(key, offset, kv.unordered_term_id) unsafe { self.hashmap.get_key_value(kv.key_value_addr) };
(key, offset, bucket as BucketId)
}) })
} }
} }
@@ -108,7 +128,6 @@ impl TermHashMap {
heap, heap,
mask: table_size - 1, mask: table_size - 1,
occupied: Vec::with_capacity(table_size / 2), occupied: Vec::with_capacity(table_size / 2),
len: 0,
} }
} }
@@ -124,34 +143,20 @@ impl TermHashMap {
self.table.len() < self.occupied.len() * 3 self.table.len() < self.occupied.len() * 3
} }
#[inline(always)] unsafe fn get_key_value(&self, addr: Addr) -> (&[u8], Addr) {
fn get_key_value(&self, addr: Addr) -> (&[u8], Addr) { let key_bytes_len = self.heap.read::<u16>(addr) as usize;
let data = self.heap.slice_from(addr); let key_addr = addr.offset(2u32);
let key_bytes_len = NativeEndian::read_u16(data) as usize; let key_bytes: &[u8] = self.heap.read_slice(key_addr, key_bytes_len);
let key_bytes: &[u8] = &data[2..][..key_bytes_len]; let val_addr: Addr = key_addr.offset(key_bytes.len() as u32);
(key_bytes, addr.offset(2u32 + key_bytes_len as u32)) (key_bytes, val_addr)
} }
#[inline(always)] pub fn set_bucket(&mut self, hash: u32, key_value_addr: Addr, bucket: usize) {
fn get_value_addr_if_key_match(&self, target_key: &[u8], addr: Addr) -> Option<Addr> {
let (stored_key, value_addr) = self.get_key_value(addr);
if stored_key == target_key {
Some(value_addr)
} else {
None
}
}
fn set_bucket(&mut self, hash: u32, key_value_addr: Addr, bucket: usize) -> UnorderedTermId {
self.occupied.push(bucket); self.occupied.push(bucket);
let unordered_term_id = self.len as UnorderedTermId;
self.len += 1;
self.table[bucket] = KeyValue { self.table[bucket] = KeyValue {
key_value_addr, key_value_addr,
hash, hash,
unordered_term_id,
}; };
unordered_term_id
} }
pub fn iter(&self) -> Iter { pub fn iter(&self) -> Iter {
@@ -191,53 +196,64 @@ impl TermHashMap {
/// will be in charge of returning a default value. /// will be in charge of returning a default value.
/// If the key already as an associated value, then it will be passed /// If the key already as an associated value, then it will be passed
/// `Some(previous_value)`. /// `Some(previous_value)`.
pub fn mutate_or_create<S, V, TMutator>( pub fn mutate_or_create<S, V, TMutator>(&mut self, key: S, mut updater: TMutator) -> BucketId
&mut self,
key: S,
mut updater: TMutator,
) -> UnorderedTermId
where where
S: AsRef<[u8]>, S: AsRef<[u8]>,
V: Copy + 'static, V: Copy,
TMutator: FnMut(Option<V>) -> V, TMutator: FnMut(Option<V>) -> V,
{ {
if self.is_saturated() { if self.is_saturated() {
self.resize(); self.resize();
} }
let key_bytes: &[u8] = key.as_ref(); let key_bytes: &[u8] = key.as_ref();
let hash = murmurhash2(key.as_ref()); let hash = murmurhash2::murmurhash2(key.as_ref());
let mut probe = self.probe(hash); let mut probe = self.probe(hash);
loop { loop {
let bucket = probe.next_probe(); let bucket = probe.next_probe();
let kv: KeyValue = self.table[bucket]; let kv: KeyValue = self.table[bucket];
if kv.is_empty() { if kv.is_empty() {
// The key does not exists yet.
let val = updater(None); let val = updater(None);
let num_bytes = let key_addr = self.heap.store(KeyBytesValue::new(key_bytes, val));
std::mem::size_of::<u16>() + key_bytes.len() + std::mem::size_of::<V>(); self.set_bucket(hash, key_addr, bucket);
let key_addr = self.heap.allocate_space(num_bytes); return bucket as BucketId;
{
let data = self.heap.slice_mut(key_addr, num_bytes);
NativeEndian::write_u16(data, key_bytes.len() as u16);
let stop = 2 + key_bytes.len();
data[2..stop].copy_from_slice(key_bytes);
store(&mut data[stop..], val);
}
return self.set_bucket(hash, key_addr, bucket);
} else if kv.hash == hash { } else if kv.hash == hash {
if let Some(val_addr) = let (key_matches, val_addr) = {
self.get_value_addr_if_key_match(key_bytes, kv.key_value_addr) let (stored_key, val_addr): (&[u8], Addr) =
{ unsafe { self.get_key_value(kv.key_value_addr) };
let v = self.heap.read(val_addr); (stored_key == key_bytes, val_addr)
let new_v = updater(Some(v)); };
self.heap.write_at(val_addr, new_v); if key_matches {
return kv.unordered_term_id; unsafe {
// logic
let v = self.heap.read(val_addr);
let new_v = updater(Some(v));
self.heap.write(val_addr, new_v);
};
return bucket as BucketId;
} }
} }
} }
} }
} }
#[cfg(all(test, feature = "unstable"))]
mod bench {
use super::murmurhash2::murmurhash2;
use test::Bencher;
#[bench]
fn bench_murmurhash2(b: &mut Bencher) {
let keys: [&'static str; 3] = ["wer qwe qwe qwe ", "werbq weqweqwe2 ", "weraq weqweqwe3 "];
b.iter(|| {
let mut s = 0;
for &key in &keys {
s ^= murmurhash2(key.as_bytes());
}
s
});
}
}
#[cfg(test)] #[cfg(test)]
mod tests { mod tests {
@@ -269,7 +285,10 @@ mod tests {
let mut vanilla_hash_map = HashMap::new(); let mut vanilla_hash_map = HashMap::new();
let mut iter_values = hash_map.iter(); let mut iter_values = hash_map.iter();
while let Some((key, addr, _)) = iter_values.next() { while let Some((key, addr, _)) = iter_values.next() {
let val: u32 = hash_map.heap.read(addr); let val: u32 = unsafe {
// test
hash_map.heap.read(addr)
};
vanilla_hash_map.insert(key.to_owned(), val); vanilla_hash_map.insert(key.to_owned(), val);
} }
assert_eq!(vanilla_hash_map.len(), 2); assert_eq!(vanilla_hash_map.len(), 2);

9
src/query/all_query.rs
View File

@@ -86,12 +86,12 @@ mod tests {
use super::AllQuery; use super::AllQuery;
use query::Query; use query::Query;
use schema::{Schema, TEXT}; use schema::{SchemaBuilder, TEXT};
use Index; use Index;
#[test] #[test]
fn test_all_query() { fn test_all_query() {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let field = schema_builder.add_text_field("text", TEXT); let field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
@@ -101,9 +101,8 @@ mod tests {
index_writer.commit().unwrap(); index_writer.commit().unwrap();
index_writer.add_document(doc!(field=>"ccc")); index_writer.add_document(doc!(field=>"ccc"));
index_writer.commit().unwrap(); index_writer.commit().unwrap();
let reader = index.reader().unwrap(); index.load_searchers().unwrap();
reader.reload().unwrap(); let searcher = index.searcher();
let searcher = reader.searcher();
let weight = AllQuery.weight(&searcher, false).unwrap(); let weight = AllQuery.weight(&searcher, false).unwrap();
{ {
let reader = searcher.segment_reader(0); let reader = searcher.segment_reader(0);

8
src/query/automaton_weight.rs
View File

@@ -1,17 +1,17 @@
use common::BitSet; use common::BitSet;
use core::SegmentReader; use core::SegmentReader;
use fst::Automaton;
use query::BitSetDocSet; use query::BitSetDocSet;
use query::ConstScorer; use query::ConstScorer;
use query::{Scorer, Weight}; use query::{Scorer, Weight};
use schema::{Field, IndexRecordOption}; use schema::{Field, IndexRecordOption};
use tantivy_fst::Automaton;
use termdict::{TermDictionary, TermStreamer}; use termdict::{TermDictionary, TermStreamer};
use Result; use Result;
/// A weight struct for Fuzzy Term and Regex Queries /// A weight struct for Fuzzy Term and Regex Queries
pub struct AutomatonWeight<A> pub struct AutomatonWeight<A>
where where
A: Automaton + Send + Sync + 'static, A: Automaton,
{ {
field: Field, field: Field,
automaton: A, automaton: A,
@@ -19,7 +19,7 @@ where
impl<A> AutomatonWeight<A> impl<A> AutomatonWeight<A>
where where
A: Automaton + Send + Sync + 'static, A: Automaton,
{ {
/// Create a new AutomationWeight /// Create a new AutomationWeight
pub fn new(field: Field, automaton: A) -> AutomatonWeight<A> { pub fn new(field: Field, automaton: A) -> AutomatonWeight<A> {
@@ -34,7 +34,7 @@ where
impl<A> Weight for AutomatonWeight<A> impl<A> Weight for AutomatonWeight<A>
where where
A: Automaton + Send + Sync + 'static, A: Automaton,
{ {
fn scorer(&self, reader: &SegmentReader) -> Result<Box<Scorer>> { fn scorer(&self, reader: &SegmentReader) -> Result<Box<Scorer>> {
let max_doc = reader.max_doc(); let max_doc = reader.max_doc();

3
src/query/bm25.rs
View File

@@ -63,8 +63,7 @@ impl BM25Weight {
.map(|term| { .map(|term| {
let term_doc_freq = searcher.doc_freq(term); let term_doc_freq = searcher.doc_freq(term);
idf(term_doc_freq, total_num_docs) idf(term_doc_freq, total_num_docs)
}) }).sum::<f32>();
.sum::<f32>();
BM25Weight::new(idf, average_fieldnorm) BM25Weight::new(idf, average_fieldnorm)
} }

6
src/query/boolean_query/boolean_query.rs
View File

@@ -47,8 +47,7 @@ impl Query for BooleanQuery {
.iter() .iter()
.map(|&(ref occur, ref subquery)| { .map(|&(ref occur, ref subquery)| {
Ok((*occur, subquery.weight(searcher, scoring_enabled)?)) Ok((*occur, subquery.weight(searcher, scoring_enabled)?))
}) }).collect::<Result<_>>()?;
.collect::<Result<_>>()?;
Ok(Box::new(BooleanWeight::new(sub_weights, scoring_enabled))) Ok(Box::new(BooleanWeight::new(sub_weights, scoring_enabled)))
} }
@@ -69,8 +68,7 @@ impl BooleanQuery {
let term_query: Box<Query> = let term_query: Box<Query> =
Box::new(TermQuery::new(term, IndexRecordOption::WithFreqs)); Box::new(TermQuery::new(term, IndexRecordOption::WithFreqs));
(Occur::Should, term_query) (Occur::Should, term_query)
}) }).collect();
.collect();
BooleanQuery::from(occur_term_queries) BooleanQuery::from(occur_term_queries)
} }

9
src/query/boolean_query/boolean_weight.rs
View File

@@ -1,4 +1,5 @@
use core::SegmentReader; use core::SegmentReader;
use downcast::Downcast;
use query::intersect_scorers; use query::intersect_scorers;
use query::score_combiner::{DoNothingCombiner, ScoreCombiner, SumWithCoordsCombiner}; use query::score_combiner::{DoNothingCombiner, ScoreCombiner, SumWithCoordsCombiner};
use query::term_query::TermScorer; use query::term_query::TermScorer;
@@ -9,6 +10,7 @@ use query::RequiredOptionalScorer;
use query::Scorer; use query::Scorer;
use query::Union; use query::Union;
use query::Weight; use query::Weight;
use std::borrow::Borrow;
use std::collections::HashMap; use std::collections::HashMap;
use Result; use Result;
@@ -22,11 +24,14 @@ where
} }
{ {
let is_all_term_queries = scorers.iter().all(|scorer| scorer.is::<TermScorer>()); let is_all_term_queries = scorers.iter().all(|scorer| {
let scorer_ref: &Scorer = scorer.borrow();
Downcast::<TermScorer>::is_type(scorer_ref)
});
if is_all_term_queries { if is_all_term_queries {
let scorers: Vec<TermScorer> = scorers let scorers: Vec<TermScorer> = scorers
.into_iter() .into_iter()
.map(|scorer| *(scorer.downcast::<TermScorer>().map_err(|_| ()).unwrap())) .map(|scorer| *Downcast::<TermScorer>::downcast(scorer).unwrap())
.collect(); .collect();
let scorer: Box<Scorer> = Box::new(Union::<TermScorer, TScoreCombiner>::from(scorers)); let scorer: Box<Scorer> = Box::new(Union::<TermScorer, TScoreCombiner>::from(scorers));
return scorer; return scorer;

56
src/query/boolean_query/mod.rs
View File

@@ -8,6 +8,7 @@ mod tests {
use super::*; use super::*;
use collector::tests::TestCollector; use collector::tests::TestCollector;
use downcast::Downcast;
use query::score_combiner::SumWithCoordsCombiner; use query::score_combiner::SumWithCoordsCombiner;
use query::term_query::TermScorer; use query::term_query::TermScorer;
use query::Intersection; use query::Intersection;
@@ -18,17 +19,16 @@ mod tests {
use query::Scorer; use query::Scorer;
use query::TermQuery; use query::TermQuery;
use schema::*; use schema::*;
use DocId;
use Index; use Index;
fn aux_test_helper() -> (Index, Field) { fn aux_test_helper() -> (Index, Field) {
let mut schema_builder = Schema::builder(); let mut schema_builder = SchemaBuilder::default();
let text_field = schema_builder.add_text_field("text", TEXT); let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build(); let schema = schema_builder.build();
let index = Index::create_in_ram(schema); let index = Index::create_in_ram(schema);
{ {
// writing the segment // writing the segment
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap(); let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
{ {
let doc = doc!(text_field => "a b c"); let doc = doc!(text_field => "a b c");
index_writer.add_document(doc); index_writer.add_document(doc);
@@ -51,6 +51,7 @@ mod tests {
} }
assert!(index_writer.commit().is_ok()); assert!(index_writer.commit().is_ok());
} }
index.load_searchers().unwrap();
(index, text_field) (index, text_field)
} }
@@ -59,8 +60,7 @@ mod tests {
let (index, text_field) = aux_test_helper(); let (index, text_field) = aux_test_helper();
let query_parser = QueryParser::for_index(&index, vec![text_field]); let query_parser = QueryParser::for_index(&index, vec![text_field]);
let query = query_parser.parse_query("(+a +b) d").unwrap(); let query = query_parser.parse_query("(+a +b) d").unwrap();
let searcher = index.reader().unwrap().searcher(); assert_eq!(query.count(&*index.searcher()).unwrap(), 3);
assert_eq!(query.count(&searcher).unwrap(), 3);
} }
#[test] #[test]
@@ -68,28 +68,28 @@ mod tests {
let (index, text_field) = aux_test_helper(); let (index, text_field) = aux_test_helper();
let query_parser = QueryParser::for_index(&index, vec![text_field]); let query_parser = QueryParser::for_index(&index, vec![text_field]);
let query = query_parser.parse_query("+a").unwrap(); let query = query_parser.parse_query("+a").unwrap();
let searcher = index.reader().unwrap().searcher(); let searcher = index.searcher();
let weight = query.weight(&searcher, true).unwrap(); let weight = query.weight(&searcher, true).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap(); let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap();
assert!(scorer.is::<TermScorer>()); assert!(Downcast::<TermScorer>::is_type(&*scorer));
} }
#[test] #[test]
pub fn test_boolean_termonly_intersection() { pub fn test_boolean_termonly_intersection() {
let (index, text_field) = aux_test_helper(); let (index, text_field) = aux_test_helper();
let query_parser = QueryParser::for_index(&index, vec![text_field]); let query_parser = QueryParser::for_index(&index, vec![text_field]);
let searcher = index.reader().unwrap().searcher(); let searcher = index.searcher();
{ {
let query = query_parser.parse_query("+a +b +c").unwrap(); let query = query_parser.parse_query("+a +b +c").unwrap();
let weight = query.weight(&searcher, true).unwrap(); let weight = query.weight(&searcher, true).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap(); let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap();
assert!(scorer.is::<Intersection<TermScorer>>()); assert!(Downcast::<Intersection<TermScorer>>::is_type(&*scorer));
} }
{ {
let query = query_parser.parse_query("+a +(b c)").unwrap(); let query = query_parser.parse_query("+a +(b c)").unwrap();
let weight = query.weight(&searcher, true).unwrap(); let weight = query.weight(&searcher, true).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap(); let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap();
assert!(scorer.is::<Intersection<Box<Scorer>>>()); assert!(Downcast::<Intersection<Box<Scorer>>>::is_type(&*scorer));
} }
} }
@@ -97,19 +97,21 @@ mod tests {
pub fn test_boolean_reqopt() { pub fn test_boolean_reqopt() {
let (index, text_field) = aux_test_helper(); let (index, text_field) = aux_test_helper();
let query_parser = QueryParser::for_index(&index, vec![text_field]); let query_parser = QueryParser::for_index(&index, vec![text_field]);
let searcher = index.reader().unwrap().searcher(); let searcher = index.searcher();
{ {
let query = query_parser.parse_query("+a b").unwrap(); let query = query_parser.parse_query("+a b").unwrap();
let weight = query.weight(&searcher, true).unwrap(); let weight = query.weight(&searcher, true).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap(); let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap();
assert!(scorer assert!(Downcast::<
.is::<RequiredOptionalScorer<Box<Scorer>, Box<Scorer>, SumWithCoordsCombiner>>()); RequiredOptionalScorer<Box<Scorer>, Box<Scorer>, SumWithCoordsCombiner>,
>::is_type(&*scorer));
} }
{ {
let query = query_parser.parse_query("+a b").unwrap(); let query = query_parser.parse_query("+a b").unwrap();
let weight = query.weight(&searcher, false).unwrap(); let weight = query.weight(&searcher, false).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap(); let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap();
assert!(scorer.is::<TermScorer>()); println!("{:?}", scorer.type_name());
assert!(Downcast::<TermScorer>::is_type(&*scorer));
} }
} }
@@ -126,18 +128,11 @@ mod tests {
query query
}; };
let reader = index.reader().unwrap();
let matching_docs = |boolean_query: &Query| { let matching_docs = |boolean_query: &Query| {
reader let searcher = index.searcher();
.searcher() let mut test_collector = TestCollector::default();
.search(boolean_query, &TestCollector) searcher.search(boolean_query, &mut test_collector).unwrap();
.unwrap() test_collector.docs()
.docs()
.iter()
.cloned()
.map(|doc| doc.1)
.collect::<Vec<DocId>>()
}; };
{ {
@@ -188,13 +183,12 @@ mod tests {
let query: Box<Query> = Box::new(term_query); let query: Box<Query> = Box::new(term_query);
query query
}; };
let reader = index.reader().unwrap();
let score_docs = |boolean_query: &Query| { let score_docs = |boolean_query: &Query| {
let fruit = reader let searcher = index.searcher();
.searcher() let mut test_collector = TestCollector::default();
.search(boolean_query, &TestCollector) searcher.search(boolean_query, &mut test_collector).unwrap();
.unwrap(); test_collector.scores()
fruit.scores().to_vec()
}; };
{ {

Some files were not shown because too many files have changed in this diff Show More