rust/tantivy
1
0
Fork 0
mirror of https://github.com/quickwit-oss/tantivy.git synced 2025-12-28 04:52:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

Compare commits

base: rust:slog
Branches Tags
rust:main rust:stuhood.upstream-erased-sort-computer rust:trinity.pointard/ff-query 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:fix_size_hint 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:polling
Branches Tags
rust:stuhood.upstream-erased-sort-computer rust:main rust:trinity.pointard/ff-query 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:fix_size_hint 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
slog ... polling

Author SHA1 Message Date
Halvor Fladsrud Bø
22a702c17f Removed temporary comment 2020-02-04 17:07:06 +01:00
Halvor Fladsrud Bø
14f0c6d01a Changed watcher to use polling 2020-02-04 17:06:29 +01:00
137 changed files with 3770 additions and 6398 deletions
Expand all files Collapse all files

2
.gitignore vendored
View File

@@ -1,5 +1,4 @@
tantivy.iml
proptest-regressions
*.swp
target
target/debug
@@ -12,4 +11,3 @@ cpp/simdcomp/bitpackingbenchmark
*.bk
.idea
trace.dat
cargo-timing*

42
CHANGELOG.md
View File

@@ -1,51 +1,9 @@
Tantivy 0.14.0
=========================
- Remove dependency to atomicwrites #833. Implemented by @pmasurel upon suggestion and research from @asafigan).
- Migrated tantivy error from the now deprecated `failure` crate to `thiserror` #760. (@hirevo)
- Switched to structure logging (via the `slog` crate). (@pmasurel)
Tantivy 0.13.1
===================
Made `Query` and `Collector` `Send + Sync`.
Updated misc dependency versions.
Tantivy 0.13.0
======================
Tantivy 0.13 introduce a change in the index format that will require
you to reindex your index (BlockWAND information are added in the skiplist).
The index size increase is minor as this information is only added for
full blocks.
If you have a massive index for which reindexing is not an option, please contact me
so that we can discuss possible solutions.
- Bugfix in `FuzzyTermQuery` not matching terms by prefix when it should (@Peachball)
- Relaxed constraints on the custom/tweak score functions. At the segment level, they can be mut, and they are not required to be Sync + Send.
- `MMapDirectory::open` does not return a `Result` anymore.
- Change in the DocSet and Scorer API. (@fulmicoton).
A freshly created DocSet point directly to their first doc. A sentinel value called TERMINATED marks the end of a DocSet.
`.advance()` returns the new DocId. `Scorer::skip(target)` has been replaced by `Scorer::seek(target)` and returns the resulting DocId.
As a result, iterating through DocSet now looks as follows
```rust
let mut doc = docset.doc();
while doc != TERMINATED {
// ...
doc = docset.advance();
}
```
The change made it possible to greatly simplify a lot of the docset's code.
- Misc internal optimization and introduction of the `Scorer::for_each_pruning` function. (@fulmicoton)
- Added an offset option to the Top(.*)Collectors. (@robyoung)
- Added Block WAND. Performance on TOP-K on term-unions should be greatly increased. (@fulmicoton, and special thanks
to the PISA team for answering all my questions!)
Tantivy 0.12.0
======================
- Removing static dispatch in tokenizers for simplicity. (#762)
- Added backward iteration for `TermDictionary` stream. (@halvorboe)
- Fixed a performance issue when searching for the posting lists of a missing term (@audunhalland)
- Added a configurable maximum number of docs (10M by default) for a segment to be considered for merge (@hntd187, landed by @halvorboe #713)
- Important Bugfix #777, causing tantivy to retain memory mapping. (diagnosed by @poljar)
- Added support for field boosting. (#547, @fulmicoton)
## How to update?

59
Cargo.toml
View File

@@ -1,11 +1,11 @@
[package]
name = "tantivy"
version = "0.14.0-dev"
version = "0.12.0"
authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT"
categories = ["database-implementations", "data-structures"]
description = """Search engine library"""
documentation = "https://docs.rs/tantivy/"
documentation = "https://tantivy-search.github.io/tantivy/tantivy/index.html"
homepage = "https://github.com/tantivy-search/tantivy"
repository = "https://github.com/tantivy-search/tantivy"
readme = "README.md"
@@ -13,42 +13,44 @@ keywords = ["search", "information", "retrieval"]
edition = "2018"
[dependencies]
base64 = "0.12"
byteorder = "1"
crc32fast = "1"
once_cell = "1"
regex ={version = "1", default-features = false, features = ["std"]}
tantivy-fst = "0.3"
base64 = "0.11.0"
byteorder = "1.0"
crc32fast = "1.2.0"
once_cell = "1.0"
regex ={version = "1.3.0", default-features = false, features = ["std"]}
tantivy-fst = "0.2.1"
memmap = {version = "0.7", optional=true}
lz4 = {version="1", optional=true}
snap = "1"
tempfile = {version="3", optional=true}
slog = "2.5"
slog-stdlog = "4"
serde = {version="1", features=["derive"]}
serde_json = "1"
num_cpus = "1"
lz4 = {version="1.20", optional=true}
snap = {version="0.2"}
atomicwrites = {version="0.2.2", optional=true}
tempfile = "3.0"
log = "0.4"
serde = "1.0"
serde_derive = "1.0"
serde_json = "1.0"
num_cpus = "1.2"
fs2={version="0.4", optional=true}
levenshtein_automata = "0.2"
itertools = "0.8"
levenshtein_automata = {version="0.1", features=["fst_automaton"]}
notify = {version="4", optional=true}
uuid = { version = "0.8", features = ["v4", "serde"] }
crossbeam = "0.7"
futures = {version = "0.3", features=["thread-pool"] }
owning_ref = "0.4"
tantivy-query-grammar = { version="0.14.0-dev", path="./query-grammar" }
stable_deref_trait = "1"
rust-stemmers = "1"
downcast-rs = "1"
stable_deref_trait = "1.0.0"
rust-stemmers = "1.2"
downcast-rs = { version="1.0" }
tantivy-query-grammar = { version="0.11", path="./query-grammar" }
bitpacking = {version="0.8", default-features = false, features=["bitpacker4x"]}
census = "0.4"
fnv = "1"
fnv = "1.0.6"
owned-read = "0.4"
thiserror = "1.0"
htmlescape = "0.3"
fail = "0.4"
failure = "0.1"
htmlescape = "0.3.1"
fail = "0.3"
murmurhash32 = "0.2"
chrono = "0.4"
smallvec = "1"
smallvec = "1.0"
rayon = "1"
[target.'cfg(windows)'.dependencies]
@@ -58,10 +60,9 @@ winapi = "0.3"
rand = "0.7"
maplit = "1"
matches = "0.1.8"
proptest = "0.10"
[dev-dependencies.fail]
version = "0.4"
version = "0.3"
features = ["failpoints"]
[profile.release]
@@ -75,7 +76,7 @@ overflow-checks = true
[features]
default = ["mmap"]
mmap = ["fs2", "tempfile", "memmap", "notify"]
mmap = ["atomicwrites", "fs2", "memmap", "notify"]
lz4-compression = ["lz4"]
failpoints = ["fail/failpoints"]
unstable = [] # useful for benches.

14
README.md
View File

@@ -31,15 +31,16 @@ Tantivy is, in fact, strongly inspired by Lucene's design.
# Benchmark
Tantivy is typically faster than Lucene, but the results 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.
Your mileage WILL vary depending on the nature of queries and their load.
# Features
- Full-text search
- Configurable tokenizer (stemming available for 17 Latin languages with third party support for Chinese ([tantivy-jieba](https://crates.io/crates/tantivy-jieba) and [cang-jie](https://crates.io/crates/cang-jie)), Japanese ([lindera](https://github.com/lindera-morphology/lindera-tantivy) and [tantivy-tokenizer-tiny-segmente](https://crates.io/crates/tantivy-tokenizer-tiny-segmenter)) and Korean ([lindera](https://github.com/lindera-morphology/lindera-tantivy) + [lindera-ko-dic-builder](https://github.com/lindera-morphology/lindera-ko-dic-builder))
- Configurable tokenizer (stemming available for 17 Latin languages with 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:)
- Tiny startup time (<10ms), perfect for command line tools
- BM25 scoring (the same as Lucene)
@@ -58,17 +59,18 @@ Your mileage WILL vary depending on the nature of queries and their load.
- Configurable indexing (optional term frequency and position indexing)
- Cheesy logo with a horse
## Non-features
# Non-features
- Distributed search is out of the scope of Tantivy. That being said, Tantivy is a
library upon which one could build a distributed search. Serializable/mergeable collector state for instance,
are within the scope of Tantivy.
# Getting started
# Supported OS and compiler
Tantivy works on stable Rust (>= 1.27) and supports Linux, MacOS, and Windows.
# Getting started
- [Tantivy's simple search example](https://tantivy-search.github.io/examples/basic_search.html)
- [tantivy-cli and its tutorial](https://github.com/tantivy-search/tantivy-cli) - `tantivy-cli` is an actual command line interface that makes it easy for you to create a search engine,
index documents, and search via the CLI or a small server with a REST API.

2
appveyor.yml
View File

@@ -18,5 +18,5 @@ install:
build: false
test_script:
- REM SET RUST_LOG=tantivy,test & cargo test --all --verbose --no-default-features --features mmap
- REM SET RUST_LOG=tantivy,test & cargo test --verbose --no-default-features --features mmap
- REM SET RUST_BACKTRACE=1 & cargo build --examples

12
examples/basic_search.rs
View File

@@ -112,6 +112,18 @@ fn main() -> tantivy::Result<()> {
limbs and branches that arch over the pool"
));
index_writer.add_document(doc!(
title => "Of Mice and Men",
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 \
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 \
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 \
debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
limbs and branches that arch over the pool"
));
// Multivalued field just need to be repeated.
index_writer.add_document(doc!(
title => "Frankenstein",

4
examples/custom_collector.rs
View File

@@ -14,7 +14,7 @@ use tantivy::fastfield::FastFieldReader;
use tantivy::query::QueryParser;
use tantivy::schema::Field;
use tantivy::schema::{Schema, FAST, INDEXED, TEXT};
use tantivy::{doc, Index, Score, SegmentReader, TantivyError};
use tantivy::{doc, Index, SegmentReader, TantivyError};
#[derive(Default)]
struct Stats {
@@ -114,7 +114,7 @@ struct StatsSegmentCollector {
impl SegmentCollector for StatsSegmentCollector {
type Fruit = Option<Stats>;
fn collect(&mut self, doc: u32, _score: Score) {
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;

98
examples/faceted_search_with_tweaked_score.rs
View File

@@ -1,98 +0,0 @@
use std::collections::HashSet;
use tantivy::collector::TopDocs;
use tantivy::doc;
use tantivy::query::BooleanQuery;
use tantivy::schema::*;
use tantivy::{DocId, Index, Score, SegmentReader};
fn main() -> tantivy::Result<()> {
let mut schema_builder = Schema::builder();
let title = schema_builder.add_text_field("title", STORED);
let ingredient = schema_builder.add_facet_field("ingredient");
let schema = schema_builder.build();
let index = Index::create_in_ram(schema.clone());
let mut index_writer = index.writer(30_000_000)?;
index_writer.add_document(doc!(
title => "Fried egg",
ingredient => Facet::from("/ingredient/egg"),
ingredient => Facet::from("/ingredient/oil"),
));
index_writer.add_document(doc!(
title => "Scrambled egg",
ingredient => Facet::from("/ingredient/egg"),
ingredient => Facet::from("/ingredient/butter"),
ingredient => Facet::from("/ingredient/milk"),
ingredient => Facet::from("/ingredient/salt"),
));
index_writer.add_document(doc!(
title => "Egg rolls",
ingredient => Facet::from("/ingredient/egg"),
ingredient => Facet::from("/ingredient/garlic"),
ingredient => Facet::from("/ingredient/salt"),
ingredient => Facet::from("/ingredient/oil"),
ingredient => Facet::from("/ingredient/tortilla-wrap"),
ingredient => Facet::from("/ingredient/mushroom"),
));
index_writer.commit()?;
let reader = index.reader()?;
let searcher = reader.searcher();
{
let facets = vec![
Facet::from("/ingredient/egg"),
Facet::from("/ingredient/oil"),
Facet::from("/ingredient/garlic"),
Facet::from("/ingredient/mushroom"),
];
let query = BooleanQuery::new_multiterms_query(
facets
.iter()
.map(|key| Term::from_facet(ingredient, &key))
.collect(),
);
let top_docs_by_custom_score =
TopDocs::with_limit(2).tweak_score(move |segment_reader: &SegmentReader| {
let mut ingredient_reader = segment_reader.facet_reader(ingredient).unwrap();
let facet_dict = ingredient_reader.facet_dict();
let query_ords: HashSet<u64> = facets
.iter()
.filter_map(|key| facet_dict.term_ord(key.encoded_str()))
.collect();
let mut facet_ords_buffer: Vec<u64> = Vec::with_capacity(20);
move |doc: DocId, original_score: Score| {
ingredient_reader.facet_ords(doc, &mut facet_ords_buffer);
let missing_ingredients = facet_ords_buffer
.iter()
.filter(|ord| !query_ords.contains(ord))
.count();
let tweak = 1.0 / 4_f32.powi(missing_ingredients as i32);
original_score * tweak
}
});
let top_docs = searcher.search(&query, &top_docs_by_custom_score)?;
let titles: Vec<String> = top_docs
.iter()
.map(|(_, doc_id)| {
searcher
.doc(*doc_id)
.unwrap()
.get_first(title)
.unwrap()
.text()
.unwrap()
.to_owned()
})
.collect();
assert_eq!(titles, vec!["Fried egg", "Egg rolls"]);
}
Ok(())
}

17
examples/iterating_docs_and_positions.rs
View File

@@ -10,7 +10,7 @@
// ---
// Importing tantivy...
use tantivy::schema::*;
use tantivy::{doc, DocSet, Index, Postings, TERMINATED};
use tantivy::{doc, DocId, DocSet, Index, Postings};
fn main() -> tantivy::Result<()> {
// We first create a schema for the sake of the
@@ -62,11 +62,12 @@ fn main() -> tantivy::Result<()> {
{
// this buffer will be used to request for positions
let mut positions: Vec<u32> = Vec::with_capacity(100);
let mut doc_id = segment_postings.doc();
while doc_id != TERMINATED {
while segment_postings.advance() {
// the number of time the term appears in the document.
let doc_id: DocId = segment_postings.doc(); //< do not try to access this before calling advance once.
// This MAY contains deleted documents as well.
if segment_reader.is_deleted(doc_id) {
doc_id = segment_postings.advance();
continue;
}
@@ -85,7 +86,6 @@ fn main() -> tantivy::Result<()> {
// Doc 2: TermFreq 1: [0]
// ```
println!("Doc {}: TermFreq {}: {:?}", doc_id, term_freq, positions);
doc_id = segment_postings.advance();
}
}
}
@@ -117,16 +117,11 @@ fn main() -> tantivy::Result<()> {
if let Some(mut block_segment_postings) =
inverted_index.read_block_postings(&term_the, IndexRecordOption::Basic)
{
loop {
let docs = block_segment_postings.docs();
if docs.is_empty() {
break;
}
while block_segment_postings.advance() {
// Once again these docs MAY contains deleted documents as well.
let docs = block_segment_postings.docs();
// Prints `Docs [0, 2].`
println!("Docs {:?}", docs);
block_segment_postings.advance();
}
}
}

3
examples/pre_tokenized_text.rs
View File

@@ -9,10 +9,11 @@
// - import tokenized text straight from json,
// - perform a search on documents with pre-tokenized text
use tantivy::tokenizer::{PreTokenizedString, SimpleTokenizer, Token, Tokenizer};
use tantivy::collector::{Count, TopDocs};
use tantivy::query::TermQuery;
use tantivy::schema::*;
use tantivy::tokenizer::{PreTokenizedString, SimpleTokenizer, Token, Tokenizer};
use tantivy::{doc, Index, ReloadPolicy};
use tempfile::TempDir;

4
query-grammar/Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy-query-grammar"
version = "0.14.0-dev"
version = "0.11.0"
authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT"
categories = ["database-implementations", "data-structures"]
@@ -13,4 +13,4 @@ keywords = ["search", "information", "retrieval"]
edition = "2018"
[dependencies]
combine = {version="4", default-features=false, features=[] }
combine = ">=3.6.0,<4.0.0"

2
query-grammar/src/lib.rs
View File

@@ -1,3 +1,5 @@
#![recursion_limit = "100"]
mod occur;
mod query_grammar;
mod user_input_ast;

46
query-grammar/src/occur.rs
View File

@@ -31,12 +31,22 @@ impl Occur {
/// Compose two occur values.
pub fn compose(left: Occur, right: Occur) -> Occur {
match (left, right) {
(Occur::Should, _) => right,
(Occur::Must, Occur::MustNot) => Occur::MustNot,
(Occur::Must, _) => Occur::Must,
(Occur::MustNot, Occur::MustNot) => Occur::Must,
(Occur::MustNot, _) => Occur::MustNot,
match left {
Occur::Should => right,
Occur::Must => {
if right == Occur::MustNot {
Occur::MustNot
} else {
Occur::Must
}
}
Occur::MustNot => {
if right == Occur::MustNot {
Occur::Must
} else {
Occur::MustNot
}
}
}
}
}
@@ -46,27 +56,3 @@ impl fmt::Display for Occur {
f.write_char(self.to_char())
}
}
#[cfg(test)]
mod test {
use crate::Occur;
#[test]
fn test_occur_compose() {
assert_eq!(Occur::compose(Occur::Should, Occur::Should), Occur::Should);
assert_eq!(Occur::compose(Occur::Should, Occur::Must), Occur::Must);
assert_eq!(
Occur::compose(Occur::Should, Occur::MustNot),
Occur::MustNot
);
assert_eq!(Occur::compose(Occur::Must, Occur::Should), Occur::Must);
assert_eq!(Occur::compose(Occur::Must, Occur::Must), Occur::Must);
assert_eq!(Occur::compose(Occur::Must, Occur::MustNot), Occur::MustNot);
assert_eq!(
Occur::compose(Occur::MustNot, Occur::Should),
Occur::MustNot
);
assert_eq!(Occur::compose(Occur::MustNot, Occur::Must), Occur::MustNot);
assert_eq!(Occur::compose(Occur::MustNot, Occur::MustNot), Occur::Must);
}
}

562
query-grammar/src/query_grammar.rs
View File

@@ -1,211 +1,171 @@
use super::user_input_ast::{UserInputAST, UserInputBound, UserInputLeaf, UserInputLiteral};
use super::user_input_ast::*;
use crate::Occur;
use combine::error::StringStreamError;
use combine::parser::char::{char, digit, letter, space, spaces, string};
use combine::parser::Parser;
use combine::{
attempt, choice, eof, many, many1, one_of, optional, parser, satisfy, skip_many1, value,
};
use combine::char::*;
use combine::error::StreamError;
use combine::stream::StreamErrorFor;
use combine::*;
fn field<'a>() -> impl Parser<&'a str, Output = String> {
(
(letter().or(char('_'))),
many(satisfy(|c: char| {
c.is_alphanumeric() || c == '_' || c == '-'
})),
)
.skip(char(':'))
parser! {
fn field[I]()(I) -> String
where [I: Stream<Item = char>] {
(
letter(),
many(satisfy(|c: char| c.is_alphanumeric() || c == '_')),
).skip(char(':')).map(|(s1, s2): (char, String)| format!("{}{}", s1, s2))
}
}
parser! {
fn word[I]()(I) -> String
where [I: Stream<Item = char>] {
(
satisfy(|c: char| !c.is_whitespace() && !['-', '`', ':', '{', '}', '"', '[', ']', '(',')'].contains(&c) ),
many(satisfy(|c: char| !c.is_whitespace() && ![':', '{', '}', '"', '[', ']', '(',')'].contains(&c)))
)
.map(|(s1, s2): (char, String)| format!("{}{}", s1, s2))
.and_then(|s: String|
match s.as_str() {
"OR" => Err(StreamErrorFor::<I>::unexpected_static_message("OR")),
"AND" => Err(StreamErrorFor::<I>::unexpected_static_message("AND")),
"NOT" => Err(StreamErrorFor::<I>::unexpected_static_message("NOT")),
_ => Ok(s)
})
}
}
fn word<'a>() -> impl Parser<&'a str, Output = String> {
(
satisfy(|c: char| {
!c.is_whitespace()
&& !['-', '^', '`', ':', '{', '}', '"', '[', ']', '(', ')'].contains(&c)
}),
many(satisfy(|c: char| {
!c.is_whitespace() && ![':', '^', '{', '}', '"', '[', ']', '(', ')'].contains(&c)
})),
)
.map(|(s1, s2): (char, String)| format!("{}{}", s1, s2))
.and_then(|s: String| match s.as_str() {
"OR" | "AND " | "NOT" => Err(StringStreamError::UnexpectedParse),
_ => Ok(s),
parser! {
fn literal[I]()(I) -> UserInputLeaf
where [I: Stream<Item = char>]
{
let term_val = || {
let phrase = char('"').with(many1(satisfy(|c| c != '"'))).skip(char('"'));
phrase.or(word())
};
let term_val_with_field = negative_number().or(term_val());
let term_query =
(field(), term_val_with_field)
.map(|(field_name, phrase)| UserInputLiteral {
field_name: Some(field_name),
phrase,
});
let term_default_field = term_val().map(|phrase| UserInputLiteral {
field_name: None,
phrase,
});
attempt(term_query)
.or(term_default_field)
.map(UserInputLeaf::from)
}
}
parser! {
fn negative_number[I]()(I) -> String
where [I: Stream<Item = char>]
{
(char('-'), many1(satisfy(char::is_numeric)),
optional((char('.'), many1(satisfy(char::is_numeric)))))
.map(|(s1, s2, s3): (char, String, Option<(char, String)>)| {
if let Some(('.', s3)) = s3 {
format!("{}{}.{}", s1, s2, s3)
} else {
format!("{}{}", s1, s2)
}
})
}
}
parser! {
fn spaces1[I]()(I) -> ()
where [I: Stream<Item = char>] {
skip_many1(space())
}
}
parser! {
/// Function that parses a range out of a Stream
/// Supports ranges like:
/// [5 TO 10], {5 TO 10}, [* TO 10], [10 TO *], {10 TO *], >5, <=10
/// [a TO *], [a TO c], [abc TO bcd}
fn range[I]()(I) -> UserInputLeaf
where [I: Stream<Item = char>] {
let range_term_val = || {
word().or(negative_number()).or(char('*').with(value("*".to_string())))
};
// check for unbounded range in the form of <5, <=10, >5, >=5
let elastic_unbounded_range = (choice([attempt(string(">=")),
attempt(string("<=")),
attempt(string("<")),
attempt(string(">"))])
.skip(spaces()),
range_term_val()).
map(|(comparison_sign, bound): (&str, String)|
match comparison_sign {
">=" => (UserInputBound::Inclusive(bound), UserInputBound::Unbounded),
"<=" => (UserInputBound::Unbounded, UserInputBound::Inclusive(bound)),
"<" => (UserInputBound::Unbounded, UserInputBound::Exclusive(bound)),
">" => (UserInputBound::Exclusive(bound), UserInputBound::Unbounded),
// default case
_ => (UserInputBound::Unbounded, UserInputBound::Unbounded)
});
let lower_bound = (one_of("{[".chars()), range_term_val())
.map(|(boundary_char, lower_bound): (char, String)|
if lower_bound == "*" {
UserInputBound::Unbounded
} else if boundary_char == '{' {
UserInputBound::Exclusive(lower_bound)
} else {
UserInputBound::Inclusive(lower_bound)
});
let upper_bound = (range_term_val(), one_of("}]".chars()))
.map(|(higher_bound, boundary_char): (String, char)|
if higher_bound == "*" {
UserInputBound::Unbounded
} else if boundary_char == '}' {
UserInputBound::Exclusive(higher_bound)
} else {
UserInputBound::Inclusive(higher_bound)
});
// return only lower and upper
let lower_to_upper = (lower_bound.
skip((spaces(),
string("TO"),
spaces())),
upper_bound);
(optional(field()).skip(spaces()),
// try elastic first, if it matches, the range is unbounded
attempt(elastic_unbounded_range).or(lower_to_upper))
.map(|(field, (lower, upper))|
// Construct the leaf from extracted field (optional)
// and bounds
UserInputLeaf::Range {
field,
lower,
upper
})
}
fn term_val<'a>() -> impl Parser<&'a str, Output = String> {
let phrase = char('"').with(many1(satisfy(|c| c != '"'))).skip(char('"'));
phrase.or(word())
}
fn term_query<'a>() -> impl Parser<&'a str, Output = UserInputLiteral> {
let term_val_with_field = negative_number().or(term_val());
(field(), term_val_with_field).map(|(field_name, phrase)| UserInputLiteral {
field_name: Some(field_name),
phrase,
})
}
fn literal<'a>() -> impl Parser<&'a str, Output = UserInputLeaf> {
let term_default_field = term_val().map(|phrase| UserInputLiteral {
field_name: None,
phrase,
});
attempt(term_query())
.or(term_default_field)
.map(UserInputLeaf::from)
}
fn negative_number<'a>() -> impl Parser<&'a str, Output = String> {
(
char('-'),
many1(digit()),
optional((char('.'), many1(digit()))),
)
.map(|(s1, s2, s3): (char, String, Option<(char, String)>)| {
if let Some(('.', s3)) = s3 {
format!("{}{}.{}", s1, s2, s3)
} else {
format!("{}{}", s1, s2)
}
})
}
fn spaces1<'a>() -> impl Parser<&'a str, Output = ()> {
skip_many1(space())
}
/// Function that parses a range out of a Stream
/// Supports ranges like:
/// [5 TO 10], {5 TO 10}, [* TO 10], [10 TO *], {10 TO *], >5, <=10
/// [a TO *], [a TO c], [abc TO bcd}
fn range<'a>() -> impl Parser<&'a str, Output = UserInputLeaf> {
let range_term_val = || {
word()
.or(negative_number())
.or(char('*').with(value("*".to_string())))
};
// check for unbounded range in the form of <5, <=10, >5, >=5
let elastic_unbounded_range = (
choice([
attempt(string(">=")),
attempt(string("<=")),
attempt(string("<")),
attempt(string(">")),
])
.skip(spaces()),
range_term_val(),
)
.map(
|(comparison_sign, bound): (&str, String)| match comparison_sign {
">=" => (UserInputBound::Inclusive(bound), UserInputBound::Unbounded),
"<=" => (UserInputBound::Unbounded, UserInputBound::Inclusive(bound)),
"<" => (UserInputBound::Unbounded, UserInputBound::Exclusive(bound)),
">" => (UserInputBound::Exclusive(bound), UserInputBound::Unbounded),
// default case
_ => (UserInputBound::Unbounded, UserInputBound::Unbounded),
},
);
let lower_bound = (one_of("{[".chars()), range_term_val()).map(
|(boundary_char, lower_bound): (char, String)| {
if lower_bound == "*" {
UserInputBound::Unbounded
} else if boundary_char == '{' {
UserInputBound::Exclusive(lower_bound)
} else {
UserInputBound::Inclusive(lower_bound)
}
},
);
let upper_bound = (range_term_val(), one_of("}]".chars())).map(
|(higher_bound, boundary_char): (String, char)| {
if higher_bound == "*" {
UserInputBound::Unbounded
} else if boundary_char == '}' {
UserInputBound::Exclusive(higher_bound)
} else {
UserInputBound::Inclusive(higher_bound)
}
},
);
// return only lower and upper
let lower_to_upper = (
lower_bound.skip((spaces(), string("TO"), spaces())),
upper_bound,
);
(
optional(field()).skip(spaces()),
// try elastic first, if it matches, the range is unbounded
attempt(elastic_unbounded_range).or(lower_to_upper),
)
.map(|(field, (lower, upper))|
// Construct the leaf from extracted field (optional)
// and bounds
UserInputLeaf::Range {
field,
lower,
upper
})
}
}
fn negate(expr: UserInputAST) -> UserInputAST {
expr.unary(Occur::MustNot)
}
fn leaf<'a>() -> impl Parser<&'a str, Output = UserInputAST> {
parser(|input| {
char('(')
.with(ast())
.skip(char(')'))
.or(char('*').map(|_| UserInputAST::from(UserInputLeaf::All)))
.or(attempt(
string("NOT").skip(spaces1()).with(leaf()).map(negate),
))
.or(attempt(range().map(UserInputAST::from)))
.or(literal().map(UserInputAST::from))
.parse_stream(input)
.into_result()
})
fn must(expr: UserInputAST) -> UserInputAST {
expr.unary(Occur::Must)
}
fn occur_symbol<'a>() -> impl Parser<&'a str, Output = Occur> {
char('-')
.map(|_| Occur::MustNot)
.or(char('+').map(|_| Occur::Must))
}
fn occur_leaf<'a>() -> impl Parser<&'a str, Output = (Option<Occur>, UserInputAST)> {
(optional(occur_symbol()), boosted_leaf())
}
fn positive_float_number<'a>() -> impl Parser<&'a str, Output = f64> {
(many1(digit()), optional((char('.'), many1(digit())))).map(
|(int_part, decimal_part_opt): (String, Option<(char, String)>)| {
let mut float_str = int_part;
if let Some((chr, decimal_str)) = decimal_part_opt {
float_str.push(chr);
float_str.push_str(&decimal_str);
}
float_str.parse::<f64>().unwrap()
},
)
}
fn boost<'a>() -> impl Parser<&'a str, Output = f64> {
(char('^'), positive_float_number()).map(|(_, boost)| boost)
}
fn boosted_leaf<'a>() -> impl Parser<&'a str, Output = UserInputAST> {
(leaf(), optional(boost())).map(|(leaf, boost_opt)| match boost_opt {
Some(boost) if (boost - 1.0).abs() > std::f64::EPSILON => {
UserInputAST::Boost(Box::new(leaf), boost)
}
_ => leaf,
})
parser! {
fn leaf[I]()(I) -> UserInputAST
where [I: Stream<Item = char>] {
char('-').with(leaf()).map(negate)
.or(char('+').with(leaf()).map(must))
.or(char('(').with(ast()).skip(char(')')))
.or(char('*').map(|_| UserInputAST::from(UserInputLeaf::All)))
.or(attempt(string("NOT").skip(spaces1()).with(leaf()).map(negate)))
.or(attempt(range().map(UserInputAST::from)))
.or(literal().map(UserInputAST::from))
}
}
#[derive(Clone, Copy)]
@@ -214,10 +174,13 @@ enum BinaryOperand {
And,
}
fn binary_operand<'a>() -> impl Parser<&'a str, Output = BinaryOperand> {
string("AND")
.with(value(BinaryOperand::And))
.or(string("OR").with(value(BinaryOperand::Or)))
parser! {
fn binary_operand[I]()(I) -> BinaryOperand
where [I: Stream<Item = char>]
{
string("AND").with(value(BinaryOperand::And))
.or(string("OR").with(value(BinaryOperand::Or)))
}
}
fn aggregate_binary_expressions(
@@ -245,84 +208,37 @@ fn aggregate_binary_expressions(
}
}
fn operand_leaf<'a>() -> impl Parser<&'a str, Output = (BinaryOperand, UserInputAST)> {
(
binary_operand().skip(spaces()),
boosted_leaf().skip(spaces()),
)
}
pub fn ast<'a>() -> impl Parser<&'a str, Output = UserInputAST> {
let boolean_expr = (boosted_leaf().skip(spaces()), many1(operand_leaf()))
.map(|(left, right)| aggregate_binary_expressions(left, right));
let whitespace_separated_leaves = many1(occur_leaf().skip(spaces().silent())).map(
|subqueries: Vec<(Option<Occur>, UserInputAST)>| {
parser! {
pub fn ast[I]()(I) -> UserInputAST
where [I: Stream<Item = char>]
{
let operand_leaf = (binary_operand().skip(spaces()), leaf().skip(spaces()));
let boolean_expr = (leaf().skip(spaces().silent()), many1(operand_leaf)).map(
|(left, right)| aggregate_binary_expressions(left,right));
let whitespace_separated_leaves = many1(leaf().skip(spaces().silent()))
.map(|subqueries: Vec<UserInputAST>|
if subqueries.len() == 1 {
let (occur_opt, ast) = subqueries.into_iter().next().unwrap();
match occur_opt.unwrap_or(Occur::Should) {
Occur::Must | Occur::Should => ast,
Occur::MustNot => UserInputAST::Clause(vec![(Some(Occur::MustNot), ast)]),
}
subqueries.into_iter().next().unwrap()
} else {
UserInputAST::Clause(subqueries.into_iter().collect())
}
},
);
let expr = attempt(boolean_expr).or(whitespace_separated_leaves);
spaces().with(expr).skip(spaces())
});
let expr = attempt(boolean_expr).or(whitespace_separated_leaves);
spaces().with(expr).skip(spaces())
}
}
pub fn parse_to_ast<'a>() -> impl Parser<&'a str, Output = UserInputAST> {
spaces()
.with(optional(ast()).skip(eof()))
.map(|opt_ast| opt_ast.unwrap_or_else(UserInputAST::empty_query))
parser! {
pub fn parse_to_ast[I]()(I) -> UserInputAST
where [I: Stream<Item = char>]
{
spaces().with(optional(ast()).skip(eof())).map(|opt_ast| opt_ast.unwrap_or_else(UserInputAST::empty_query))
}
}
#[cfg(test)]
mod test {
type TestParseResult = Result<(), StringStreamError>;
use super::*;
use combine::parser::Parser;
pub fn nearly_equals(a: f64, b: f64) -> bool {
(a - b).abs() < 0.0005 * (a + b).abs()
}
fn assert_nearly_equals(expected: f64, val: f64) {
assert!(
nearly_equals(val, expected),
"Got {}, expected {}.",
val,
expected
);
}
#[test]
fn test_occur_symbol() -> TestParseResult {
assert_eq!(super::occur_symbol().parse("-")?, (Occur::MustNot, ""));
assert_eq!(super::occur_symbol().parse("+")?, (Occur::Must, ""));
Ok(())
}
#[test]
fn test_positive_float_number() {
fn valid_parse(float_str: &str, expected_val: f64, expected_remaining: &str) {
let (val, remaining) = positive_float_number().parse(float_str).unwrap();
assert_eq!(remaining, expected_remaining);
assert_nearly_equals(val, expected_val);
}
fn error_parse(float_str: &str) {
assert!(positive_float_number().parse(float_str).is_err());
}
valid_parse("1.0", 1.0, "");
valid_parse("1", 1.0, "");
valid_parse("0.234234 aaa", 0.234234f64, " aaa");
error_parse(".3332");
error_parse("1.");
error_parse("-1.");
}
fn test_parse_query_to_ast_helper(query: &str, expected: &str) {
let query = parse_to_ast().parse(query).unwrap().0;
@@ -353,24 +269,15 @@ mod test {
"Err(UnexpectedParse)"
);
test_parse_query_to_ast_helper("NOTa", "\"NOTa\"");
test_parse_query_to_ast_helper("NOT a", "(-\"a\")");
}
#[test]
fn test_boosting() {
assert!(parse_to_ast().parse("a^2^3").is_err());
assert!(parse_to_ast().parse("a^2^").is_err());
test_parse_query_to_ast_helper("a^3", "(\"a\")^3");
test_parse_query_to_ast_helper("a^3 b^2", "(*(\"a\")^3 *(\"b\")^2)");
test_parse_query_to_ast_helper("a^1", "\"a\"");
test_parse_query_to_ast_helper("NOT a", "-(\"a\")");
}
#[test]
fn test_parse_query_to_ast_binary_op() {
test_parse_query_to_ast_helper("a AND b", "(+\"a\" +\"b\")");
test_parse_query_to_ast_helper("a OR b", "(?\"a\" ?\"b\")");
test_parse_query_to_ast_helper("a OR b AND c", "(?\"a\" ?(+\"b\" +\"c\"))");
test_parse_query_to_ast_helper("a AND b AND c", "(+\"a\" +\"b\" +\"c\")");
test_parse_query_to_ast_helper("a AND b", "(+(\"a\") +(\"b\"))");
test_parse_query_to_ast_helper("a OR b", "(?(\"a\") ?(\"b\"))");
test_parse_query_to_ast_helper("a OR b AND c", "(?(\"a\") ?((+(\"b\") +(\"c\"))))");
test_parse_query_to_ast_helper("a AND b AND c", "(+(\"a\") +(\"b\") +(\"c\"))");
assert_eq!(
format!("{:?}", parse_to_ast().parse("a OR b aaa")),
"Err(UnexpectedParse)"
@@ -408,32 +315,6 @@ mod test {
test_parse_query_to_ast_helper("weight: <= 70.5", "weight:{\"*\" TO \"70.5\"]");
}
#[test]
fn test_occur_leaf() {
let ((occur, ast), _) = super::occur_leaf().parse("+abc").unwrap();
assert_eq!(occur, Some(Occur::Must));
assert_eq!(format!("{:?}", ast), "\"abc\"");
}
#[test]
fn test_field_name() -> TestParseResult {
assert_eq!(
super::field().parse("my-field-name:a")?,
("my-field-name".to_string(), "a")
);
assert_eq!(
super::field().parse("my_field_name:a")?,
("my_field_name".to_string(), "a")
);
assert!(super::field().parse(":a").is_err());
assert!(super::field().parse("-my_field:a").is_err());
assert_eq!(
super::field().parse("_my_field:a")?,
("_my_field".to_string(), "a")
);
Ok(())
}
#[test]
fn test_range_parser() {
// testing the range() parser separately
@@ -462,67 +343,32 @@ mod test {
fn test_parse_query_to_triming_spaces() {
test_parse_query_to_ast_helper(" abc", "\"abc\"");
test_parse_query_to_ast_helper("abc ", "\"abc\"");
test_parse_query_to_ast_helper("( a OR abc)", "(?\"a\" ?\"abc\")");
test_parse_query_to_ast_helper("(a OR abc)", "(?\"a\" ?\"abc\")");
test_parse_query_to_ast_helper("(a OR abc)", "(?\"a\" ?\"abc\")");
test_parse_query_to_ast_helper("a OR abc ", "(?\"a\" ?\"abc\")");
test_parse_query_to_ast_helper("(a OR abc )", "(?\"a\" ?\"abc\")");
test_parse_query_to_ast_helper("(a OR abc) ", "(?\"a\" ?\"abc\")");
test_parse_query_to_ast_helper("( a OR abc)", "(?(\"a\") ?(\"abc\"))");
test_parse_query_to_ast_helper("(a OR abc)", "(?(\"a\") ?(\"abc\"))");
test_parse_query_to_ast_helper("(a OR abc)", "(?(\"a\") ?(\"abc\"))");
test_parse_query_to_ast_helper("a OR abc ", "(?(\"a\") ?(\"abc\"))");
test_parse_query_to_ast_helper("(a OR abc )", "(?(\"a\") ?(\"abc\"))");
test_parse_query_to_ast_helper("(a OR abc) ", "(?(\"a\") ?(\"abc\"))");
}
#[test]
fn test_parse_query_single_term() {
fn test_parse_query_to_ast() {
test_parse_query_to_ast_helper("abc", "\"abc\"");
}
#[test]
fn test_parse_query_default_clause() {
test_parse_query_to_ast_helper("a b", "(*\"a\" *\"b\")");
}
#[test]
fn test_parse_query_must_default_clause() {
test_parse_query_to_ast_helper("+(a b)", "(*\"a\" *\"b\")");
}
#[test]
fn test_parse_query_must_single_term() {
test_parse_query_to_ast_helper("+d", "\"d\"");
}
#[test]
fn test_single_term_with_field() {
test_parse_query_to_ast_helper("a b", "(\"a\" \"b\")");
test_parse_query_to_ast_helper("+(a b)", "+((\"a\" \"b\"))");
test_parse_query_to_ast_helper("+d", "+(\"d\")");
test_parse_query_to_ast_helper("+(a b) +d", "(+((\"a\" \"b\")) +(\"d\"))");
test_parse_query_to_ast_helper("(+a +b) d", "((+(\"a\") +(\"b\")) \"d\")");
test_parse_query_to_ast_helper("(+a)", "+(\"a\")");
test_parse_query_to_ast_helper("(+a +b)", "(+(\"a\") +(\"b\"))");
test_parse_query_to_ast_helper("abc:toto", "abc:\"toto\"");
}
#[test]
fn test_single_term_with_float() {
test_parse_query_to_ast_helper("abc:1.1", "abc:\"1.1\"");
}
#[test]
fn test_must_clause() {
test_parse_query_to_ast_helper("(+a +b)", "(+\"a\" +\"b\")");
}
#[test]
fn test_parse_test_query_plus_a_b_plus_d() {
test_parse_query_to_ast_helper("+(a b) +d", "(+(*\"a\" *\"b\") +\"d\")");
}
#[test]
fn test_parse_test_query_other() {
test_parse_query_to_ast_helper("(+a +b) d", "(*(+\"a\" +\"b\") *\"d\")");
test_parse_query_to_ast_helper("+abc:toto", "abc:\"toto\"");
test_parse_query_to_ast_helper("(+abc:toto -titi)", "(+abc:\"toto\" -\"titi\")");
test_parse_query_to_ast_helper("-abc:toto", "(-abc:\"toto\")");
test_parse_query_to_ast_helper("abc:a b", "(*abc:\"a\" *\"b\")");
test_parse_query_to_ast_helper("+abc:toto", "+(abc:\"toto\")");
test_parse_query_to_ast_helper("(+abc:toto -titi)", "(+(abc:\"toto\") -(\"titi\"))");
test_parse_query_to_ast_helper("-abc:toto", "-(abc:\"toto\")");
test_parse_query_to_ast_helper("abc:a b", "(abc:\"a\" \"b\")");
test_parse_query_to_ast_helper("abc:\"a b\"", "abc:\"a b\"");
test_parse_query_to_ast_helper("foo:[1 TO 5]", "foo:[\"1\" TO \"5\"]");
}
#[test]
fn test_parse_query_with_range() {
test_parse_query_to_ast_helper("[1 TO 5]", "[\"1\" TO \"5\"]");
test_parse_query_to_ast_helper("foo:{a TO z}", "foo:{\"a\" TO \"z\"}");
test_parse_query_to_ast_helper("foo:[1 TO toto}", "foo:[\"1\" TO \"toto\"}");

32
query-grammar/src/user_input_ast.rs
View File

@@ -85,14 +85,14 @@ impl UserInputBound {
}
pub enum UserInputAST {
Clause(Vec<(Option<Occur>, UserInputAST)>),
Clause(Vec<UserInputAST>),
Unary(Occur, Box<UserInputAST>),
Leaf(Box<UserInputLeaf>),
Boost(Box<UserInputAST>, f64),
}
impl UserInputAST {
pub fn unary(self, occur: Occur) -> UserInputAST {
UserInputAST::Clause(vec![(Some(occur), self)])
UserInputAST::Unary(occur, Box::new(self))
}
fn compose(occur: Occur, asts: Vec<UserInputAST>) -> UserInputAST {
@@ -103,7 +103,7 @@ impl UserInputAST {
} else {
UserInputAST::Clause(
asts.into_iter()
.map(|ast: UserInputAST| (Some(occur), ast))
.map(|ast: UserInputAST| ast.unary(occur))
.collect::<Vec<_>>(),
)
}
@@ -134,38 +134,26 @@ impl From<UserInputLeaf> for UserInputAST {
}
}
fn print_occur_ast(
occur_opt: Option<Occur>,
ast: &UserInputAST,
formatter: &mut fmt::Formatter,
) -> fmt::Result {
if let Some(occur) = occur_opt {
write!(formatter, "{}{:?}", occur, ast)?;
} else {
write!(formatter, "*{:?}", ast)?;
}
Ok(())
}
impl fmt::Debug for UserInputAST {
fn fmt(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
match *self {
UserInputAST::Clause(ref subqueries) => {
if subqueries.is_empty() {
write!(formatter, "<emptyclause>")?;
} else {
write!(formatter, "(")?;
print_occur_ast(subqueries[0].0, &subqueries[0].1, formatter)?;
write!(formatter, "{:?}", &subqueries[0])?;
for subquery in &subqueries[1..] {
write!(formatter, " ")?;
print_occur_ast(subquery.0, &subquery.1, formatter)?;
write!(formatter, " {:?}", subquery)?;
}
write!(formatter, ")")?;
}
Ok(())
}
UserInputAST::Unary(ref occur, ref subquery) => {
write!(formatter, "{}({:?})", occur, subquery)
}
UserInputAST::Leaf(ref subquery) => write!(formatter, "{:?}", subquery),
UserInputAST::Boost(ref leaf, boost) => write!(formatter, "({:?})^{}", leaf, boost),
}
}
}

8
src/collector/count_collector.rs
View File

@@ -96,18 +96,18 @@ mod tests {
}
{
let mut count_collector = SegmentCountCollector::default();
count_collector.collect(0u32, 1.0);
count_collector.collect(0u32, 1f32);
assert_eq!(count_collector.harvest(), 1);
}
{
let mut count_collector = SegmentCountCollector::default();
count_collector.collect(0u32, 1.0);
count_collector.collect(0u32, 1f32);
assert_eq!(count_collector.harvest(), 1);
}
{
let mut count_collector = SegmentCountCollector::default();
count_collector.collect(0u32, 1.0);
count_collector.collect(1u32, 1.0);
count_collector.collect(0u32, 1f32);
count_collector.collect(1u32, 1f32);
assert_eq!(count_collector.harvest(), 2);
}
}

14
src/collector/custom_score_top_collector.rs
View File

@@ -11,13 +11,13 @@ impl<TCustomScorer, TScore> CustomScoreTopCollector<TCustomScorer, TScore>
where
TScore: Clone + PartialOrd,
{
pub(crate) fn new(
pub fn new(
custom_scorer: TCustomScorer,
collector: TopCollector<TScore>,
limit: usize,
) -> CustomScoreTopCollector<TCustomScorer, TScore> {
CustomScoreTopCollector {
custom_scorer,
collector,
collector: TopCollector::with_limit(limit),
}
}
}
@@ -28,7 +28,7 @@ where
/// It is the segment local version of the [`CustomScorer`](./trait.CustomScorer.html).
pub trait CustomSegmentScorer<TScore>: 'static {
/// Computes the score of a specific `doc`.
fn score(&mut self, doc: DocId) -> TScore;
fn score(&self, doc: DocId) -> TScore;
}
/// `CustomScorer` makes it possible to define any kind of score.
@@ -46,7 +46,7 @@ pub trait CustomScorer<TScore>: Sync {
impl<TCustomScorer, TScore> Collector for CustomScoreTopCollector<TCustomScorer, TScore>
where
TCustomScorer: CustomScorer<TScore> + Send + Sync,
TCustomScorer: CustomScorer<TScore>,
TScore: 'static + PartialOrd + Clone + Send + Sync,
{
type Fruit = Vec<(TScore, DocAddress)>;
@@ -117,9 +117,9 @@ where
impl<F, TScore> CustomSegmentScorer<TScore> for F
where
F: 'static + FnMut(DocId) -> TScore,
F: 'static + Sync + Send + Fn(DocId) -> TScore,
{
fn score(&mut self, doc: DocId) -> TScore {
fn score(&self, doc: DocId) -> TScore {
(self)(doc)
}
}

50
src/collector/facet_collector.rs
View File

@@ -1,5 +1,6 @@
use crate::collector::Collector;
use crate::collector::SegmentCollector;
use crate::docset::SkipResult;
use crate::fastfield::FacetReader;
use crate::schema::Facet;
use crate::schema::Field;
@@ -187,11 +188,6 @@ pub struct FacetSegmentCollector {
collapse_facet_ords: Vec<u64>,
}
enum SkipResult {
Found,
NotFound,
}
fn skip<'a, I: Iterator<Item = &'a Facet>>(
target: &[u8],
collapse_it: &mut Peekable<I>,
@@ -201,14 +197,14 @@ fn skip<'a, I: Iterator<Item = &'a Facet>>(
Some(facet_bytes) => match facet_bytes.encoded_str().as_bytes().cmp(target) {
Ordering::Less => {}
Ordering::Greater => {
return SkipResult::NotFound;
return SkipResult::OverStep;
}
Ordering::Equal => {
return SkipResult::Found;
return SkipResult::Reached;
}
},
None => {
return SkipResult::NotFound;
return SkipResult::End;
}
}
collapse_it.next();
@@ -285,7 +281,7 @@ impl Collector for FacetCollector {
// is positionned on a term that has not been processed yet.
let skip_result = skip(facet_streamer.key(), &mut collapse_facet_it);
match skip_result {
SkipResult::Found => {
SkipResult::Reached => {
// we reach a facet we decided to collapse.
let collapse_depth = facet_depth(facet_streamer.key());
let mut collapsed_id = 0;
@@ -305,7 +301,7 @@ impl Collector for FacetCollector {
}
break;
}
SkipResult::NotFound => {
SkipResult::End | SkipResult::OverStep => {
collapse_mapping.push(0);
if !facet_streamer.advance() {
break;
@@ -472,7 +468,7 @@ mod tests {
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
let num_facets: usize = 3 * 4 * 5;
let facets: Vec<Facet> = (0..num_facets)
.map(|mut n| {
@@ -531,7 +527,7 @@ mod tests {
let facet_field = schema_builder.add_facet_field("facets");
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(
facet_field => Facet::from_text(&"/subjects/A/a"),
facet_field => Facet::from_text(&"/subjects/B/a"),
@@ -550,12 +546,12 @@ mod tests {
}
#[test]
fn test_doc_search_by_facet() -> crate::Result<()> {
fn test_doc_search_by_facet() {
let mut schema_builder = Schema::builder();
let facet_field = schema_builder.add_facet_field("facet");
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests()?;
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(
facet_field => Facet::from_text(&"/A/A"),
));
@@ -568,8 +564,8 @@ mod tests {
index_writer.add_document(doc!(
facet_field => Facet::from_text(&"/D/C/A"),
));
index_writer.commit()?;
let reader = index.reader()?;
index_writer.commit().unwrap();
let reader = index.reader().unwrap();
let searcher = reader.searcher();
assert_eq!(searcher.num_docs(), 4);
@@ -586,17 +582,17 @@ mod tests {
assert_eq!(count_facet("/A/C"), 1);
assert_eq!(count_facet("/A/C/A"), 1);
assert_eq!(count_facet("/C/A"), 0);
let query_parser = QueryParser::for_index(&index, vec![]);
{
let query = query_parser.parse_query("facet:/A/B")?;
assert_eq!(1, searcher.search(&query, &Count).unwrap());
let query_parser = QueryParser::for_index(&index, vec![]);
{
let query = query_parser.parse_query("facet:/A/B").unwrap();
assert_eq!(1, searcher.search(&query, &Count).unwrap());
}
{
let query = query_parser.parse_query("facet:/A").unwrap();
assert_eq!(3, searcher.search(&query, &Count).unwrap());
}
}
{
let query = query_parser.parse_query("facet:/A")?;
assert_eq!(3, searcher.search(&query, &Count)?);
}
Ok(())
}
#[test]
@@ -631,7 +627,7 @@ mod tests {
.collect();
docs[..].shuffle(&mut thread_rng());
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
for doc in docs {
index_writer.add_document(doc);
}
@@ -684,7 +680,7 @@ mod bench {
// 40425 docs
docs[..].shuffle(&mut thread_rng());
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
for doc in docs {
index_writer.add_document(doc);
}

2
src/collector/int_facet_collector.rs
View File

@@ -89,7 +89,7 @@ mod tests {
let index = Index::create_in_ram(schema.clone());
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
{
for i in 0u64..10u64 {
index_writer.add_document(doc!(

26
src/collector/mod.rs
View File

@@ -109,7 +109,6 @@ pub use self::tweak_score_top_collector::{ScoreSegmentTweaker, ScoreTweaker};
mod facet_collector;
pub use self::facet_collector::FacetCollector;
use crate::query::Weight;
/// `Fruit` is the type for the result of our collection.
/// e.g. `usize` for the `Count` collector.
@@ -133,7 +132,7 @@ impl<T> Fruit for T where T: Send + downcast_rs::Downcast {}
/// The collection logic itself is in the `SegmentCollector`.
///
/// Segments are not guaranteed to be visited in any specific order.
pub trait Collector: Sync + Send {
pub trait Collector: Sync {
/// `Fruit` is the type for the result of our collection.
/// e.g. `usize` for the `Count` collector.
type Fruit: Fruit;
@@ -155,29 +154,6 @@ pub trait Collector: Sync + Send {
/// Combines the fruit associated to the collection of each segments
/// into one fruit.
fn merge_fruits(&self, segment_fruits: Vec<Self::Fruit>) -> crate::Result<Self::Fruit>;
/// Created a segment collector and
fn collect_segment(
&self,
weight: &dyn Weight,
segment_ord: u32,
reader: &SegmentReader,
) -> crate::Result<<Self::Child as SegmentCollector>::Fruit> {
let mut segment_collector = self.for_segment(segment_ord as u32, reader)?;
if let Some(delete_bitset) = reader.delete_bitset() {
weight.for_each(reader, &mut |doc, score| {
if delete_bitset.is_alive(doc) {
segment_collector.collect(doc, score);
}
})?;
} else {
weight.for_each(reader, &mut |doc, score| {
segment_collector.collect(doc, score);
})?;
}
Ok(segment_collector.harvest())
}
}
/// The `SegmentCollector` is the trait in charge of defining the

8
src/collector/multi_collector.rs
View File

@@ -55,7 +55,7 @@ impl<TCollector: Collector> Collector for CollectorWrapper<TCollector> {
impl SegmentCollector for Box<dyn BoxableSegmentCollector> {
type Fruit = Box<dyn Fruit>;
fn collect(&mut self, doc: u32, score: Score) {
fn collect(&mut self, doc: u32, score: f32) {
self.as_mut().collect(doc, score);
}
@@ -65,7 +65,7 @@ impl SegmentCollector for Box<dyn BoxableSegmentCollector> {
}
pub trait BoxableSegmentCollector {
fn collect(&mut self, doc: u32, score: Score);
fn collect(&mut self, doc: u32, score: f32);
fn harvest_from_box(self: Box<Self>) -> Box<dyn Fruit>;
}
@@ -74,7 +74,7 @@ pub struct SegmentCollectorWrapper<TSegmentCollector: SegmentCollector>(TSegment
impl<TSegmentCollector: SegmentCollector> BoxableSegmentCollector
for SegmentCollectorWrapper<TSegmentCollector>
{
fn collect(&mut self, doc: u32, score: Score) {
fn collect(&mut self, doc: u32, score: f32) {
self.0.collect(doc, score);
}
@@ -259,7 +259,7 @@ mod tests {
let index = Index::create_in_ram(schema);
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(text=>"abc"));
index_writer.add_document(doc!(text=>"abc abc abc"));
index_writer.add_document(doc!(text=>"abc abc"));

2
src/collector/tests.rs
View File

@@ -206,7 +206,7 @@ impl Collector for BytesFastFieldTestCollector {
impl SegmentCollector for BytesFastFieldSegmentCollector {
type Fruit = Vec<u8>;
fn collect(&mut self, doc: u32, _score: Score) {
fn collect(&mut self, doc: u32, _score: f32) {
let data = self.reader.get_bytes(doc);
self.vals.extend(data);
}

86
src/collector/top_collector.rs
View File

@@ -18,9 +18,9 @@ use std::collections::BinaryHeap;
/// 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.
pub(crate) struct ComparableDoc<T, D> {
pub feature: T,
pub doc: D,
struct ComparableDoc<T, D> {
feature: T,
doc: D,
}
impl<T: PartialOrd, D: PartialOrd> PartialOrd for ComparableDoc<T, D> {
@@ -56,8 +56,7 @@ impl<T: PartialOrd, D: PartialOrd> PartialEq for ComparableDoc<T, D> {
impl<T: PartialOrd, D: PartialOrd> Eq for ComparableDoc<T, D> {}
pub(crate) struct TopCollector<T> {
pub limit: usize,
pub offset: usize,
limit: usize,
_marker: PhantomData<T>,
}
@@ -73,20 +72,14 @@ where
if limit < 1 {
panic!("Limit must be strictly greater than 0.");
}
Self {
TopCollector {
limit,
offset: 0,
_marker: PhantomData,
}
}
/// Skip the first "offset" documents when collecting.
///
/// This is equivalent to `OFFSET` in MySQL or PostgreSQL and `start` in
/// Lucene's TopDocsCollector.
pub fn and_offset(mut self, offset: usize) -> TopCollector<T> {
self.offset = offset;
self
pub fn limit(&self) -> usize {
self.limit
}
pub fn merge_fruits(
@@ -99,7 +92,7 @@ where
let mut top_collector = BinaryHeap::new();
for child_fruit in children {
for (feature, doc) in child_fruit {
if top_collector.len() < (self.limit + self.offset) {
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 {
@@ -111,7 +104,6 @@ where
Ok(top_collector
.into_sorted_vec()
.into_iter()
.skip(self.offset)
.map(|cdoc| (cdoc.feature, cdoc.doc))
.collect())
}
@@ -121,23 +113,7 @@ where
segment_id: SegmentLocalId,
_: &SegmentReader,
) -> crate::Result<TopSegmentCollector<F>> {
Ok(TopSegmentCollector::new(
segment_id,
self.limit + self.offset,
))
}
/// Create a new TopCollector with the same limit and offset.
///
/// Ideally we would use Into but the blanket implementation seems to cause the Scorer traits
/// to fail.
#[doc(hidden)]
pub(crate) fn into_tscore<TScore: PartialOrd + Clone>(self) -> TopCollector<TScore> {
TopCollector {
limit: self.limit,
offset: self.offset,
_marker: PhantomData,
}
Ok(TopSegmentCollector::new(segment_id, self.limit))
}
}
@@ -211,7 +187,7 @@ impl<T: PartialOrd + Clone> TopSegmentCollector<T> {
#[cfg(test)]
mod tests {
use super::{TopCollector, TopSegmentCollector};
use super::TopSegmentCollector;
use crate::DocAddress;
#[test]
@@ -272,48 +248,6 @@ mod tests {
top_collector_limit_3.harvest()[..2].to_vec(),
);
}
#[test]
fn test_top_collector_with_limit_and_offset() {
let collector = TopCollector::with_limit(2).and_offset(1);
let results = collector
.merge_fruits(vec![vec![
(0.9, DocAddress(0, 1)),
(0.8, DocAddress(0, 2)),
(0.7, DocAddress(0, 3)),
(0.6, DocAddress(0, 4)),
(0.5, DocAddress(0, 5)),
]])
.unwrap();
assert_eq!(
results,
vec![(0.8, DocAddress(0, 2)), (0.7, DocAddress(0, 3)),]
);
}
#[test]
fn test_top_collector_with_limit_larger_than_set_and_offset() {
let collector = TopCollector::with_limit(2).and_offset(1);
let results = collector
.merge_fruits(vec![vec![(0.9, DocAddress(0, 1)), (0.8, DocAddress(0, 2))]])
.unwrap();
assert_eq!(results, vec![(0.8, DocAddress(0, 2)),]);
}
#[test]
fn test_top_collector_with_limit_and_offset_larger_than_set() {
let collector = TopCollector::with_limit(2).and_offset(20);
let results = collector
.merge_fruits(vec![vec![(0.9, DocAddress(0, 1)), (0.8, DocAddress(0, 2))]])
.unwrap();
assert_eq!(results, vec![]);
}
}
#[cfg(all(test, feature = "unstable"))]

261
src/collector/top_score_collector.rs
View File

@@ -1,20 +1,18 @@
use super::Collector;
use crate::collector::custom_score_top_collector::CustomScoreTopCollector;
use crate::collector::top_collector::TopCollector;
use crate::collector::top_collector::TopSegmentCollector;
use crate::collector::top_collector::{ComparableDoc, TopCollector};
use crate::collector::tweak_score_top_collector::TweakedScoreTopCollector;
use crate::collector::{
CustomScorer, CustomSegmentScorer, ScoreSegmentTweaker, ScoreTweaker, SegmentCollector,
};
use crate::fastfield::FastFieldReader;
use crate::query::Weight;
use crate::schema::Field;
use crate::DocAddress;
use crate::DocId;
use crate::Score;
use crate::SegmentLocalId;
use crate::SegmentReader;
use std::collections::BinaryHeap;
use std::fmt;
/// The `TopDocs` collector keeps track of the top `K` documents
@@ -38,7 +36,7 @@ use std::fmt;
/// let schema = schema_builder.build();
/// let index = Index::create_in_ram(schema);
///
/// let mut index_writer = index.writer_with_num_threads(1, 10_000_000).unwrap();
/// let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
/// 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"));
@@ -52,18 +50,14 @@ use std::fmt;
/// let query = query_parser.parse_query("diary").unwrap();
/// let top_docs = searcher.search(&query, &TopDocs::with_limit(2)).unwrap();
///
/// assert_eq!(top_docs[0].1, DocAddress(0, 1));
/// assert_eq!(top_docs[1].1, DocAddress(0, 3));
/// assert_eq!(&top_docs[0], &(0.7261542, DocAddress(0, 1)));
/// assert_eq!(&top_docs[1], &(0.6099695, DocAddress(0, 3)));
/// ```
pub struct TopDocs(TopCollector<Score>);
impl fmt::Debug for TopDocs {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(
f,
"TopDocs(limit={}, offset={})",
self.0.limit, self.0.offset
)
write!(f, "TopDocs({})", self.0.limit())
}
}
@@ -72,7 +66,7 @@ struct ScorerByFastFieldReader {
}
impl CustomSegmentScorer<u64> for ScorerByFastFieldReader {
fn score(&mut self, doc: DocId) -> u64 {
fn score(&self, doc: DocId) -> u64 {
self.ff_reader.get_u64(u64::from(doc))
}
}
@@ -90,8 +84,7 @@ impl CustomScorer<u64> for ScorerByField {
.u64(self.field)
.ok_or_else(|| {
crate::TantivyError::SchemaError(format!(
"Field requested ({:?}) is not a i64/u64 fast field.",
self.field
"Field requested is not a i64/u64 fast field."
))
})?;
Ok(ScorerByFastFieldReader { ff_reader })
@@ -107,45 +100,6 @@ impl TopDocs {
TopDocs(TopCollector::with_limit(limit))
}
/// Skip the first "offset" documents when collecting.
///
/// This is equivalent to `OFFSET` in MySQL or PostgreSQL and `start` in
/// Lucene's TopDocsCollector.
///
/// ```rust
/// use tantivy::collector::TopDocs;
/// use tantivy::query::QueryParser;
/// use tantivy::schema::{Schema, TEXT};
/// use tantivy::{doc, DocAddress, Index};
///
/// 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_with_num_threads(1, 10_000_000).unwrap();
/// 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.add_document(doc!(title => "The Diary of Lena Mukhina"));
/// assert!(index_writer.commit().is_ok());
///
/// let reader = index.reader().unwrap();
/// let searcher = reader.searcher();
///
/// let query_parser = QueryParser::for_index(&index, vec![title]);
/// let query = query_parser.parse_query("diary").unwrap();
/// let top_docs = searcher.search(&query, &TopDocs::with_limit(2).and_offset(1)).unwrap();
///
/// assert_eq!(top_docs.len(), 2);
/// assert_eq!(top_docs[0].1, DocAddress(0, 4));
/// assert_eq!(top_docs[1].1, DocAddress(0, 3));
/// ```
pub fn and_offset(self, offset: usize) -> TopDocs {
TopDocs(self.0.and_offset(offset))
}
/// Set top-K to rank documents by a given fast field.
///
/// ```rust
@@ -163,7 +117,7 @@ impl TopDocs {
/// # let schema = schema_builder.build();
/// #
/// # let index = Index::create_in_ram(schema);
/// # let mut index_writer = index.writer_with_num_threads(1, 10_000_000)?;
/// # let mut index_writer = index.writer_with_num_threads(1, 3_000_000)?;
/// # index_writer.add_document(doc!(title => "The Name of the Wind", rating => 92u64));
/// # index_writer.add_document(doc!(title => "The Diary of Muadib", rating => 97u64));
/// # index_writer.add_document(doc!(title => "A Dairy Cow", rating => 63u64));
@@ -264,7 +218,7 @@ impl TopDocs {
/// fn create_index() -> tantivy::Result<Index> {
/// let schema = create_schema();
/// let index = Index::create_in_ram(schema);
/// let mut index_writer = index.writer_with_num_threads(1, 10_000_000)?;
/// let mut index_writer = index.writer_with_num_threads(1, 3_000_000)?;
/// let product_name = index.schema().get_field("product_name").unwrap();
/// let popularity: Field = index.schema().get_field("popularity").unwrap();
/// index_writer.add_document(doc!(product_name => "The Diary of Muadib", popularity => 1u64));
@@ -303,7 +257,7 @@ impl TopDocs {
/// let popularity: u64 = popularity_reader.get(doc);
/// // Well.. For the sake of the example we use a simple logarithm
/// // function.
/// let popularity_boost_score = ((2u64 + popularity) as Score).log2();
/// let popularity_boost_score = ((2u64 + popularity) as f32).log2();
/// popularity_boost_score * original_score
/// }
/// });
@@ -324,9 +278,9 @@ impl TopDocs {
where
TScore: 'static + Send + Sync + Clone + PartialOrd,
TScoreSegmentTweaker: ScoreSegmentTweaker<TScore> + 'static,
TScoreTweaker: ScoreTweaker<TScore, Child = TScoreSegmentTweaker> + Send + Sync,
TScoreTweaker: ScoreTweaker<TScore, Child = TScoreSegmentTweaker>,
{
TweakedScoreTopCollector::new(score_tweaker, self.0.into_tscore())
TweakedScoreTopCollector::new(score_tweaker, self.0.limit())
}
/// Ranks the documents using a custom score.
@@ -371,7 +325,7 @@ impl TopDocs {
/// # fn main() -> tantivy::Result<()> {
/// # let schema = create_schema();
/// # let index = Index::create_in_ram(schema);
/// # let mut index_writer = index.writer_with_num_threads(1, 10_000_000)?;
/// # let mut index_writer = index.writer_with_num_threads(1, 3_000_000)?;
/// # let product_name = index.schema().get_field("product_name").unwrap();
/// #
/// let popularity: Field = index.schema().get_field("popularity").unwrap();
@@ -438,9 +392,9 @@ impl TopDocs {
where
TScore: 'static + Send + Sync + Clone + PartialOrd,
TCustomSegmentScorer: CustomSegmentScorer<TScore> + 'static,
TCustomScorer: CustomScorer<TScore, Child = TCustomSegmentScorer> + Send + Sync,
TCustomScorer: CustomScorer<TScore, Child = TCustomSegmentScorer>,
{
CustomScoreTopCollector::new(custom_score, self.0.into_tscore())
CustomScoreTopCollector::new(custom_score, self.0.limit())
}
}
@@ -468,64 +422,6 @@ impl Collector for TopDocs {
) -> crate::Result<Self::Fruit> {
self.0.merge_fruits(child_fruits)
}
fn collect_segment(
&self,
weight: &dyn Weight,
segment_ord: u32,
reader: &SegmentReader,
) -> crate::Result<<Self::Child as SegmentCollector>::Fruit> {
let heap_len = self.0.limit + self.0.offset;
let mut heap: BinaryHeap<ComparableDoc<Score, DocId>> = BinaryHeap::with_capacity(heap_len);
if let Some(delete_bitset) = reader.delete_bitset() {
let mut threshold = Score::MIN;
weight.for_each_pruning(threshold, reader, &mut |doc, score| {
if delete_bitset.is_deleted(doc) {
return threshold;
}
let heap_item = ComparableDoc {
feature: score,
doc,
};
if heap.len() < heap_len {
heap.push(heap_item);
if heap.len() == heap_len {
threshold = heap.peek().map(|el| el.feature).unwrap_or(Score::MIN);
}
return threshold;
}
*heap.peek_mut().unwrap() = heap_item;
threshold = heap.peek().map(|el| el.feature).unwrap_or(Score::MIN);
threshold
})?;
} else {
weight.for_each_pruning(Score::MIN, reader, &mut |doc, score| {
let heap_item = ComparableDoc {
feature: score,
doc,
};
if heap.len() < heap_len {
heap.push(heap_item);
// TODO the threshold is suboptimal for heap.len == heap_len
if heap.len() == heap_len {
return heap.peek().map(|el| el.feature).unwrap_or(Score::MIN);
} else {
return Score::MIN;
}
}
*heap.peek_mut().unwrap() = heap_item;
heap.peek().map(|el| el.feature).unwrap_or(Score::MIN)
})?;
}
let fruit = heap
.into_sorted_vec()
.into_iter()
.map(|cid| (cid.feature, DocAddress(segment_ord, cid.doc)))
.collect();
Ok(fruit)
}
}
/// Segment Collector associated to `TopDocs`.
@@ -535,7 +431,7 @@ impl SegmentCollector for TopScoreSegmentCollector {
type Fruit = Vec<(Score, DocAddress)>;
fn collect(&mut self, doc: DocId, score: Score) {
self.0.collect(doc, score);
self.0.collect(doc, score)
}
fn harvest(self) -> Vec<(Score, DocAddress)> {
@@ -549,10 +445,11 @@ mod tests {
use crate::collector::Collector;
use crate::query::{AllQuery, Query, QueryParser};
use crate::schema::{Field, Schema, FAST, STORED, TEXT};
use crate::DocAddress;
use crate::Index;
use crate::IndexWriter;
use crate::Score;
use crate::{DocAddress, DocId, SegmentReader};
use itertools::Itertools;
fn make_index() -> Index {
let mut schema_builder = Schema::builder();
@@ -561,7 +458,7 @@ mod tests {
let index = Index::create_in_ram(schema);
{
// writing the segment
let mut index_writer = index.writer_with_num_threads(1, 10_000_000).unwrap();
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"));
@@ -570,13 +467,6 @@ mod tests {
index
}
fn assert_results_equals(results: &[(Score, DocAddress)], expected: &[(Score, DocAddress)]) {
for (result, expected) in results.iter().zip(expected.iter()) {
assert_eq!(result.1, expected.1);
crate::assert_nearly_equals!(result.0, expected.0);
}
}
#[test]
fn test_top_collector_not_at_capacity() {
let index = make_index();
@@ -589,31 +479,16 @@ mod tests {
.searcher()
.search(&text_query, &TopDocs::with_limit(4))
.unwrap();
assert_results_equals(
&score_docs,
&[
assert_eq!(
score_docs,
vec![
(0.81221175, DocAddress(0u32, 1)),
(0.5376842, DocAddress(0u32, 2)),
(0.48527452, DocAddress(0, 0)),
],
(0.48527452, DocAddress(0, 0))
]
);
}
#[test]
fn test_top_collector_not_at_capacity_with_offset() {
let index = make_index();
let field = index.schema().get_field("text").unwrap();
let query_parser = QueryParser::for_index(&index, vec![field]);
let text_query = query_parser.parse_query("droopy tax").unwrap();
let score_docs: Vec<(Score, DocAddress)> = index
.reader()
.unwrap()
.searcher()
.search(&text_query, &TopDocs::with_limit(4).and_offset(2))
.unwrap();
assert_results_equals(&score_docs[..], &[(0.48527452, DocAddress(0, 0))]);
}
#[test]
fn test_top_collector_at_capacity() {
let index = make_index();
@@ -626,33 +501,12 @@ mod tests {
.searcher()
.search(&text_query, &TopDocs::with_limit(2))
.unwrap();
assert_results_equals(
&score_docs,
&[
assert_eq!(
score_docs,
vec![
(0.81221175, DocAddress(0u32, 1)),
(0.5376842, DocAddress(0u32, 2)),
],
);
}
#[test]
fn test_top_collector_at_capacity_with_offset() {
let index = make_index();
let field = index.schema().get_field("text").unwrap();
let query_parser = QueryParser::for_index(&index, vec![field]);
let text_query = query_parser.parse_query("droopy tax").unwrap();
let score_docs: Vec<(Score, DocAddress)> = index
.reader()
.unwrap()
.searcher()
.search(&text_query, &TopDocs::with_limit(2).and_offset(1))
.unwrap();
assert_results_equals(
&score_docs[..],
&[
(0.5376842, DocAddress(0u32, 2)),
(0.48527452, DocAddress(0, 0)),
],
]
);
}
@@ -669,8 +523,8 @@ mod tests {
// precondition for the test to be meaningful: we did get documents
// with the same score
assert!(page_1.iter().all(|result| result.0 == page_1[0].0));
assert!(page_2.iter().all(|result| result.0 == page_2[0].0));
assert!(page_1.iter().map(|result| result.0).all_equal());
assert!(page_2.iter().map(|result| result.0).all_equal());
// sanity check since we're relying on make_index()
assert_eq!(page_1.len(), 2);
@@ -713,8 +567,8 @@ mod tests {
let top_collector = TopDocs::with_limit(4).order_by_u64_field(size);
let top_docs: Vec<(u64, DocAddress)> = searcher.search(&query, &top_collector).unwrap();
assert_eq!(
&top_docs[..],
&[
top_docs,
vec![
(64, DocAddress(0, 1)),
(16, DocAddress(0, 2)),
(12, DocAddress(0, 0))
@@ -760,59 +614,12 @@ mod tests {
let top_collector = TopDocs::with_limit(4).order_by_u64_field(size);
let err = top_collector.for_segment(0, segment);
if let Err(crate::TantivyError::SchemaError(msg)) = err {
assert_eq!(
msg,
"Field requested (Field(1)) is not a i64/u64 fast field."
);
assert_eq!(msg, "Field requested is not a i64/u64 fast field.");
} else {
assert!(false);
}
}
#[test]
fn test_tweak_score_top_collector_with_offset() {
let index = make_index();
let field = index.schema().get_field("text").unwrap();
let query_parser = QueryParser::for_index(&index, vec![field]);
let text_query = query_parser.parse_query("droopy tax").unwrap();
let collector = TopDocs::with_limit(2).and_offset(1).tweak_score(
move |_segment_reader: &SegmentReader| move |doc: DocId, _original_score: Score| doc,
);
let score_docs: Vec<(u32, DocAddress)> = index
.reader()
.unwrap()
.searcher()
.search(&text_query, &collector)
.unwrap();
assert_eq!(
score_docs,
vec![(1, DocAddress(0, 1)), (0, DocAddress(0, 0)),]
);
}
#[test]
fn test_custom_score_top_collector_with_offset() {
let index = make_index();
let field = index.schema().get_field("text").unwrap();
let query_parser = QueryParser::for_index(&index, vec![field]);
let text_query = query_parser.parse_query("droopy tax").unwrap();
let collector = TopDocs::with_limit(2)
.and_offset(1)
.custom_score(move |_segment_reader: &SegmentReader| move |doc: DocId| doc);
let score_docs: Vec<(u32, DocAddress)> = index
.reader()
.unwrap()
.searcher()
.search(&text_query, &collector)
.unwrap();
assert_eq!(
score_docs,
vec![(1, DocAddress(0, 1)), (0, DocAddress(0, 0)),]
);
}
fn index(
query: &str,
query_field: Field,
@@ -821,7 +628,7 @@ mod tests {
) -> (Index, Box<dyn Query>) {
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_with_num_threads(1, 10_000_000).unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
doc_adder(&mut index_writer);
index_writer.commit().unwrap();
let query_parser = QueryParser::for_index(&index, vec![query_field]);

12
src/collector/tweak_score_top_collector.rs
View File

@@ -14,11 +14,11 @@ where
{
pub fn new(
score_tweaker: TScoreTweaker,
collector: TopCollector<TScore>,
limit: usize,
) -> TweakedScoreTopCollector<TScoreTweaker, TScore> {
TweakedScoreTopCollector {
score_tweaker,
collector,
collector: TopCollector::with_limit(limit),
}
}
}
@@ -29,7 +29,7 @@ where
/// It is the segment local version of the [`ScoreTweaker`](./trait.ScoreTweaker.html).
pub trait ScoreSegmentTweaker<TScore>: 'static {
/// Tweak the given `score` for the document `doc`.
fn score(&mut self, doc: DocId, score: Score) -> TScore;
fn score(&self, doc: DocId, score: Score) -> TScore;
}
/// `ScoreTweaker` makes it possible to tweak the score
@@ -49,7 +49,7 @@ pub trait ScoreTweaker<TScore>: Sync {
impl<TScoreTweaker, TScore> Collector for TweakedScoreTopCollector<TScoreTweaker, TScore>
where
TScoreTweaker: ScoreTweaker<TScore> + Send + Sync,
TScoreTweaker: ScoreTweaker<TScore>,
TScore: 'static + PartialOrd + Clone + Send + Sync,
{
type Fruit = Vec<(TScore, DocAddress)>;
@@ -121,9 +121,9 @@ where
impl<F, TScore> ScoreSegmentTweaker<TScore> for F
where
F: 'static + FnMut(DocId, Score) -> TScore,
F: 'static + Sync + Send + Fn(DocId, Score) -> TScore,
{
fn score(&mut self, doc: DocId, score: Score) -> TScore {
fn score(&self, doc: DocId, score: Score) -> TScore {
(self)(doc, score)
}
}

29
src/common/bitset.rs
View File

@@ -33,10 +33,6 @@ impl TinySet {
TinySet(0u64)
}
pub fn clear(&mut self) {
self.0 = 0u64;
}
/// Returns the complement of the set in `[0, 64[`.
fn complement(self) -> TinySet {
TinySet(!self.0)
@@ -47,11 +43,6 @@ impl TinySet {
!self.intersect(TinySet::singleton(el)).is_empty()
}
/// Returns the number of elements in the TinySet.
pub fn len(self) -> u32 {
self.0.count_ones()
}
/// Returns the intersection of `self` and `other`
pub fn intersect(self, other: TinySet) -> TinySet {
TinySet(self.0 & other.0)
@@ -118,12 +109,22 @@ impl TinySet {
pub fn range_greater_or_equal(from_included: u32) -> TinySet {
TinySet::range_lower(from_included).complement()
}
pub fn clear(&mut self) {
self.0 = 0u64;
}
pub fn len(self) -> u32 {
self.0.count_ones()
}
}
#[derive(Clone)]
pub struct BitSet {
tinysets: Box<[TinySet]>,
len: usize,
len: usize, //< Technically it should be u32, but we
// count multiple inserts.
// `usize` guards us from overflow.
max_value: u32,
}
@@ -203,7 +204,7 @@ mod tests {
use super::BitSet;
use super::TinySet;
use crate::docset::{DocSet, TERMINATED};
use crate::docset::DocSet;
use crate::query::BitSetDocSet;
use crate::tests;
use crate::tests::generate_nonunique_unsorted;
@@ -277,13 +278,11 @@ mod tests {
}
assert_eq!(btreeset.len(), bitset.len());
let mut bitset_docset = BitSetDocSet::from(bitset);
let mut remaining = true;
for el in btreeset.into_iter() {
assert!(remaining);
bitset_docset.advance();
assert_eq!(bitset_docset.doc(), el);
remaining = bitset_docset.advance() != TERMINATED;
}
assert!(!remaining);
assert!(!bitset_docset.advance());
}
#[test]

35
src/common/mod.rs
View File

@@ -10,9 +10,7 @@ pub(crate) use self::bitset::TinySet;
pub(crate) use self::composite_file::{CompositeFile, CompositeWrite};
pub use self::counting_writer::CountingWriter;
pub use self::serialize::{BinarySerializable, FixedSize};
pub use self::vint::{
read_u32_vint, read_u32_vint_no_advance, serialize_vint_u32, write_u32_vint, VInt,
};
pub use self::vint::{read_u32_vint, serialize_vint_u32, write_u32_vint, VInt};
pub use byteorder::LittleEndian as Endianness;
/// Segment's max doc must be `< MAX_DOC_LIMIT`.
@@ -20,19 +18,6 @@ pub use byteorder::LittleEndian as Endianness;
/// We do not allow segments with more than
pub const MAX_DOC_LIMIT: u32 = 1 << 31;
pub fn minmax<I, T>(mut vals: I) -> Option<(T, T)>
where
I: Iterator<Item = T>,
T: Copy + Ord,
{
if let Some(first_el) = vals.next() {
return Some(vals.fold((first_el, first_el), |(min_val, max_val), el| {
(min_val.min(el), max_val.max(el))
}));
}
None
}
/// Computes the number of bits that will be used for bitpacking.
///
/// In general the target is the minimum number of bits
@@ -149,7 +134,6 @@ pub fn u64_to_f64(val: u64) -> f64 {
#[cfg(test)]
pub(crate) mod test {
pub use super::minmax;
pub use super::serialize::test::fixed_size_test;
use super::{compute_num_bits, f64_to_u64, i64_to_u64, u64_to_f64, u64_to_i64};
use std::f64;
@@ -215,21 +199,4 @@ pub(crate) mod test {
assert!(((super::MAX_DOC_LIMIT - 1) as i32) >= 0);
assert!((super::MAX_DOC_LIMIT as i32) < 0);
}
#[test]
fn test_minmax_empty() {
let vals: Vec<u32> = vec![];
assert_eq!(minmax(vals.into_iter()), None);
}
#[test]
fn test_minmax_one() {
assert_eq!(minmax(vec![1].into_iter()), Some((1, 1)));
}
#[test]
fn test_minmax_two() {
assert_eq!(minmax(vec![1, 2].into_iter()), Some((1, 2)));
assert_eq!(minmax(vec![2, 1].into_iter()), Some((1, 2)));
}
}

13
src/common/serialize.rs
View File

@@ -89,19 +89,6 @@ impl FixedSize for u64 {
const SIZE_IN_BYTES: usize = 8;
}
impl BinarySerializable for f32 {
fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
writer.write_f32::<Endianness>(*self)
}
fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
reader.read_f32::<Endianness>()
}
}
impl FixedSize for f32 {
const SIZE_IN_BYTES: usize = 4;
}
impl BinarySerializable for i64 {
fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
writer.write_i64::<Endianness>(*self)

38
src/common/vint.rs
View File

@@ -5,12 +5,12 @@ use std::io::Read;
use std::io::Write;
/// Wrapper over a `u64` that serializes as a variable int.
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
#[derive(Debug, Eq, PartialEq)]
pub struct VInt(pub u64);
const STOP_BIT: u8 = 128;
pub fn serialize_vint_u32(val: u32, buf: &mut [u8; 8]) -> &[u8] {
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;
@@ -29,7 +29,7 @@ pub fn serialize_vint_u32(val: u32, buf: &mut [u8; 8]) -> &[u8] {
let val = u64::from(val);
const STOP_BIT: u64 = 128u64;
let (res, num_bytes) = match val {
match val {
0..=STOP_1 => (val | STOP_BIT, 1),
START_2..=STOP_2 => (
(val & MASK_1) | ((val & MASK_2) << 1) | (STOP_BIT << (8)),
@@ -56,9 +56,7 @@ pub fn serialize_vint_u32(val: u32, buf: &mut [u8; 8]) -> &[u8] {
| (STOP_BIT << (8 * 4)),
5,
),
};
LittleEndian::write_u64(&mut buf[..], res);
&buf[0..num_bytes]
}
}
/// Returns the number of bytes covered by a
@@ -87,26 +85,23 @@ fn vint_len(data: &[u8]) -> usize {
/// If the buffer does not start by a valid
/// vint payload
pub fn read_u32_vint(data: &mut &[u8]) -> u32 {
let (result, vlen) = read_u32_vint_no_advance(*data);
*data = &data[vlen..];
result
}
pub fn read_u32_vint_no_advance(data: &[u8]) -> (u32, usize) {
let vlen = vint_len(data);
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;
}
(result, vlen)
*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 mut buf = [0u8; 8];
let data = serialize_vint_u32(val, &mut buf);
writer.write_all(&data)
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 {
@@ -177,6 +172,7 @@ mod tests {
use super::serialize_vint_u32;
use super::VInt;
use crate::common::BinarySerializable;
use byteorder::{ByteOrder, LittleEndian};
fn aux_test_vint(val: u64) {
let mut v = [14u8; 10];
@@ -212,10 +208,12 @@ mod tests {
fn aux_test_serialize_vint_u32(val: u32) {
let mut buffer = [0u8; 10];
let mut buffer2 = [0u8; 8];
let mut buffer2 = [0u8; 10];
let len_vint = VInt(val as u64).serialize_into(&mut buffer);
let res2 = serialize_vint_u32(val, &mut buffer2);
assert_eq!(&buffer[..len_vint], res2, "array wrong for {}", val);
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]

16
src/core/executor.rs
View File

@@ -1,6 +1,5 @@
use crossbeam::channel;
use rayon::{ThreadPool, ThreadPoolBuilder};
use slog::{error, Logger};
/// Search executor whether search request are single thread or multithread.
///
@@ -44,7 +43,6 @@ impl Executor {
&self,
f: F,
args: AIterator,
logger: Logger,
) -> crate::Result<Vec<R>> {
match self {
Executor::SingleThread => args.map(f).collect::<crate::Result<_>>(),
@@ -59,7 +57,7 @@ impl Executor {
let (idx, arg) = arg_with_idx;
let fruit = f(arg);
if let Err(err) = fruit_sender.send((idx, fruit)) {
error!(logger, "Failed to send search task. It probably means all search threads have panicked. {:?}", err);
error!("Failed to send search task. It probably means all search threads have panicked. {:?}", err);
}
});
}
@@ -89,21 +87,17 @@ impl Executor {
#[cfg(test)]
mod tests {
use slog::{o, Discard, Logger};
use super::Executor;
#[test]
#[should_panic(expected = "panic should propagate")]
fn test_panic_propagates_single_thread() {
let logger = Logger::root(Discard, o!());
let _result: Vec<usize> = Executor::single_thread()
.map(
|_| {
panic!("panic should propagate");
},
vec![0].into_iter(),
logger,
)
.unwrap();
}
@@ -111,7 +105,6 @@ mod tests {
#[test]
#[should_panic] //< unfortunately the panic message is not propagated
fn test_panic_propagates_multi_thread() {
let logger = Logger::root(Discard, o!());
let _result: Vec<usize> = Executor::multi_thread(1, "search-test")
.unwrap()
.map(
@@ -119,16 +112,14 @@ mod tests {
panic!("panic should propagate");
},
vec![0].into_iter(),
logger,
)
.unwrap();
}
#[test]
fn test_map_singlethread() {
let logger = Logger::root(Discard, o!());
let result: Vec<usize> = Executor::single_thread()
.map(|i| Ok(i * 2), 0..1_000, logger)
.map(|i| Ok(i * 2), 0..1_000)
.unwrap();
assert_eq!(result.len(), 1_000);
for i in 0..1_000 {
@@ -138,10 +129,9 @@ mod tests {
#[test]
fn test_map_multithread() {
let logger = Logger::root(Discard, o!());
let result: Vec<usize> = Executor::multi_thread(3, "search-test")
.unwrap()
.map(|i| Ok(i * 2), 0..10, logger)
.map(|i| Ok(i * 2), 0..10)
.unwrap();
assert_eq!(result.len(), 10);
for i in 0..10 {

63
src/core/index.rs
View File

@@ -1,3 +1,4 @@
use super::segment::create_segment;
use super::segment::Segment;
use crate::core::Executor;
use crate::core::IndexMeta;
@@ -21,14 +22,12 @@ use crate::schema::FieldType;
use crate::schema::Schema;
use crate::tokenizer::{TextAnalyzer, TokenizerManager};
use crate::IndexWriter;
use slog::Logger;
use num_cpus;
use std::borrow::BorrowMut;
use std::collections::HashSet;
use std::fmt;
#[cfg(feature = "mmap")]
use std::path::Path;
use std::path::PathBuf;
use std::path::{Path, PathBuf};
use std::sync::Arc;
fn load_metas(
@@ -58,14 +57,7 @@ pub struct Index {
}
impl Index {
pub(crate) fn logger(&self) -> &Logger {
self.directory.logger()
}
/// Examines the directory to see if it contains an index.
///
/// Effectively, it only checks for the presence of the `meta.json` file.
/// Examines the director to see if it contains an index
pub fn exists<Dir: Directory>(dir: &Dir) -> bool {
dir.exists(&META_FILEPATH)
}
@@ -148,18 +140,16 @@ impl Index {
Index::create(mmap_directory, schema)
}
/// Creates a new index given an implementation of the trait `Directory`.
///
/// If a directory previously existed, it will be erased.
/// Creates a new index given an implementation of the trait `Directory`
pub fn create<Dir: Directory>(dir: Dir, schema: Schema) -> crate::Result<Index> {
let directory = ManagedDirectory::wrap(dir)?;
Index::new_from_directory(directory, schema)
Index::from_directory(directory, schema)
}
/// Create a new index from a directory.
///
/// This will overwrite existing meta.json
fn new_from_directory(mut directory: ManagedDirectory, schema: Schema) -> crate::Result<Index> {
fn from_directory(mut directory: ManagedDirectory, schema: Schema) -> crate::Result<Index> {
save_new_metas(schema.clone(), directory.borrow_mut())?;
let metas = IndexMeta::with_schema(schema);
Index::create_from_metas(directory, &metas, SegmentMetaInventory::default())
@@ -250,8 +240,6 @@ impl Index {
/// Open the index using the provided directory
pub fn open<D: Directory>(directory: D) -> crate::Result<Index> {
let logger: &Logger = directory.logger();
slog::info!(logger, "index-open"; "directory" => format!("{:?}", directory));
let directory = ManagedDirectory::wrap(directory)?;
let inventory = SegmentMetaInventory::default();
let metas = load_metas(&directory, &inventory)?;
@@ -295,7 +283,7 @@ impl Index {
TantivyError::LockFailure(
err,
Some(
"Failed to acquire index lock. If you are using \
"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."
@@ -312,15 +300,6 @@ impl Index {
)
}
/// Helper to create an index writer for tests.
///
/// That index writer only simply has a single thread and a heap of 5 MB.
/// Using a single thread gives us a deterministic allocation of DocId.
#[cfg(test)]
pub fn writer_for_tests(&self) -> crate::Result<IndexWriter> {
self.writer_with_num_threads(1, 10_000_000)
}
/// Creates a multithreaded writer
///
/// Tantivy will automatically define the number of threads to use.
@@ -358,7 +337,7 @@ impl Index {
#[doc(hidden)]
pub fn segment(&self, segment_meta: SegmentMeta) -> Segment {
Segment::for_index(self.clone(), segment_meta)
create_segment(self.clone(), segment_meta)
}
/// Creates a new segment.
@@ -523,7 +502,7 @@ mod tests {
let schema = throw_away_schema();
let field = schema.get_field("num_likes").unwrap();
let mut index = Index::create_from_tempdir(schema).unwrap();
let mut writer = index.writer_for_tests().unwrap();
let mut writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
writer.commit().unwrap();
let reader = index
.reader_builder()
@@ -560,33 +539,23 @@ mod tests {
test_index_on_commit_reload_policy_aux(field, &write_index, &reader);
}
}
fn test_index_on_commit_reload_policy_aux(field: Field, index: &Index, reader: &IndexReader) {
let mut reader_index = reader.index();
let (sender, receiver) = crossbeam::channel::unbounded();
let _watch_handle = reader_index.directory_mut().watch(Box::new(move || {
let _ = sender.send(());
}));
let mut writer = index.writer_for_tests().unwrap();
let mut writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
assert_eq!(reader.searcher().num_docs(), 0);
writer.add_document(doc!(field=>1u64));
writer.commit().unwrap();
// We need a loop here because it is possible for notify to send more than
// one modify event. It was observed on CI on MacOS.
loop {
assert!(receiver.recv().is_ok());
if reader.searcher().num_docs() == 1 {
break;
}
}
assert!(receiver.recv().is_ok());
assert_eq!(reader.searcher().num_docs(), 1);
writer.add_document(doc!(field=>2u64));
writer.commit().unwrap();
// ... Same as above
loop {
assert!(receiver.recv().is_ok());
if reader.searcher().num_docs() == 2 {
break;
}
}
assert!(receiver.recv().is_ok());
assert_eq!(reader.searcher().num_docs(), 2);
}
// This test will not pass on windows, because windows

5
src/core/index_meta.rs
View File

@@ -3,7 +3,8 @@ use crate::core::SegmentId;
use crate::schema::Schema;
use crate::Opstamp;
use census::{Inventory, TrackedObject};
use serde::{Deserialize, Serialize};
use serde;
use serde_json;
use std::collections::HashSet;
use std::fmt;
use std::path::PathBuf;
@@ -213,7 +214,7 @@ pub struct IndexMeta {
#[serde(skip_serializing_if = "Option::is_none")]
/// Payload associated to the last commit.
///
/// Upon commit, clients can optionally add a small `String` payload to their commit
/// Upon commit, clients can optionally add a small `Striing` payload to their commit
/// to help identify this commit.
/// This payload is entirely unused by tantivy.
pub payload: Option<String>,

12
src/core/inverted_index_reader.rs
View File

@@ -3,9 +3,11 @@ use crate::directory::ReadOnlySource;
use crate::positions::PositionReader;
use crate::postings::TermInfo;
use crate::postings::{BlockSegmentPostings, SegmentPostings};
use crate::schema::FieldType;
use crate::schema::IndexRecordOption;
use crate::schema::Term;
use crate::termdict::TermDictionary;
use owned_read::OwnedRead;
/// The inverted index reader is in charge of accessing
/// the inverted index associated to a specific field.
@@ -53,7 +55,10 @@ impl InvertedIndexReader {
/// Creates an empty `InvertedIndexReader` object, which
/// contains no terms at all.
pub fn empty(record_option: IndexRecordOption) -> InvertedIndexReader {
pub fn empty(field_type: &FieldType) -> InvertedIndexReader {
let record_option = field_type
.get_index_record_option()
.unwrap_or(IndexRecordOption::Basic);
InvertedIndexReader {
termdict: TermDictionary::empty(),
postings_source: ReadOnlySource::empty(),
@@ -92,7 +97,8 @@ impl InvertedIndexReader {
let offset = term_info.postings_offset as usize;
let end_source = self.postings_source.len();
let postings_slice = self.postings_source.slice(offset, end_source);
block_postings.reset(term_info.doc_freq, postings_slice);
let postings_reader = OwnedRead::new(postings_slice);
block_postings.reset(term_info.doc_freq, postings_reader);
}
/// Returns a block postings given a `Term`.
@@ -121,7 +127,7 @@ impl InvertedIndexReader {
let postings_data = self.postings_source.slice_from(offset);
BlockSegmentPostings::from_data(
term_info.doc_freq,
postings_data,
OwnedRead::new(postings_data),
self.record_option,
requested_option,
)

31
src/core/searcher.rs
View File

@@ -1,8 +1,11 @@
use crate::collector::Collector;
use crate::collector::SegmentCollector;
use crate::core::Executor;
use crate::core::InvertedIndexReader;
use crate::core::SegmentReader;
use crate::query::Query;
use crate::query::Scorer;
use crate::query::Weight;
use crate::schema::Document;
use crate::schema::Schema;
use crate::schema::{Field, Term};
@@ -14,6 +17,26 @@ use crate::Index;
use std::fmt;
use std::sync::Arc;
fn collect_segment<C: Collector>(
collector: &C,
weight: &dyn Weight,
segment_ord: u32,
segment_reader: &SegmentReader,
) -> crate::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_alive(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.
///
/// It guarantees that the `Segment` will not be removed before
@@ -140,10 +163,14 @@ impl Searcher {
let segment_readers = self.segment_readers();
let fruits = executor.map(
|(segment_ord, segment_reader)| {
collector.collect_segment(weight.as_ref(), segment_ord as u32, segment_reader)
collect_segment(
collector,
weight.as_ref(),
segment_ord as u32,
segment_reader,
)
},
segment_readers.iter().enumerate(),
self.index.logger().clone(),
)?;
collector.merge_fruits(fruits)
}

13
src/core/segment.rs
View File

@@ -24,12 +24,15 @@ impl fmt::Debug for Segment {
}
}
impl Segment {
/// Creates a new segment given an `Index` and a `SegmentId`
pub(crate) fn for_index(index: Index, meta: SegmentMeta) -> Segment {
Segment { index, meta }
}
/// Creates a new segment given an `Index` and a `SegmentId`
///
/// The function is here to make it private outside `tantivy`.
/// #[doc(hidden)]
pub fn create_segment(index: Index, meta: SegmentMeta) -> Segment {
Segment { index, meta }
}
impl Segment {
/// Returns the index the segment belongs to.
pub fn index(&self) -> &Index {
&self.index

7
src/core/segment_id.rs
View File

@@ -4,7 +4,6 @@ use uuid::Uuid;
#[cfg(test)]
use once_cell::sync::Lazy;
use serde::{Deserialize, Serialize};
use std::error::Error;
use std::str::FromStr;
#[cfg(test)]
@@ -21,12 +20,6 @@ use std::sync::atomic;
#[derive(Clone, Copy, PartialEq, Eq, Hash, Serialize, Deserialize)]
pub struct SegmentId(Uuid);
impl ToString for SegmentId {
fn to_string(&self) -> String {
self.short_uuid_string()
}
}
#[cfg(test)]
static AUTO_INC_COUNTER: Lazy<atomic::AtomicUsize> = Lazy::new(|| atomic::AtomicUsize::default());

98
src/core/segment_reader.rs
View File

@@ -8,16 +8,15 @@ use crate::directory::ReadOnlySource;
use crate::fastfield::DeleteBitSet;
use crate::fastfield::FacetReader;
use crate::fastfield::FastFieldReaders;
use crate::fieldnorm::{FieldNormReader, FieldNormReaders};
use crate::fieldnorm::FieldNormReader;
use crate::schema::Field;
use crate::schema::FieldType;
use crate::schema::Schema;
use crate::schema::{Field, IndexRecordOption};
use crate::space_usage::SegmentSpaceUsage;
use crate::store::StoreReader;
use crate::termdict::TermDictionary;
use crate::DocId;
use fail::fail_point;
use slog::{warn, Logger};
use std::collections::HashMap;
use std::fmt;
use std::sync::Arc;
@@ -49,12 +48,11 @@ pub struct SegmentReader {
positions_composite: CompositeFile,
positions_idx_composite: CompositeFile,
fast_fields_readers: Arc<FastFieldReaders>,
fieldnorm_readers: FieldNormReaders,
fieldnorms_composite: CompositeFile,
store_source: ReadOnlySource,
delete_bitset_opt: Option<DeleteBitSet>,
schema: Schema,
logger: Logger,
}
impl SegmentReader {
@@ -127,15 +125,17 @@ impl SegmentReader {
///
/// They are simply stored as a fast field, serialized in
/// the `.fieldnorm` file of the segment.
pub fn get_fieldnorms_reader(&self, field: Field) -> crate::Result<FieldNormReader> {
self.fieldnorm_readers.get_field(field).ok_or_else(|| {
pub fn get_fieldnorms_reader(&self, field: Field) -> FieldNormReader {
if let Some(fieldnorm_source) = self.fieldnorms_composite.open_read(field) {
FieldNormReader::open(fieldnorm_source)
} else {
let field_name = self.schema.get_field_name(field);
let err_msg = format!(
"Field norm not found for field {:?}. Was it market as indexed during indexing.",
field_name
);
crate::TantivyError::SchemaError(err_msg)
})
panic!(err_msg);
}
}
/// Accessor to the segment's `StoreReader`.
@@ -178,8 +178,8 @@ impl SegmentReader {
let fast_field_readers =
Arc::new(FastFieldReaders::load_all(&schema, &fast_fields_composite)?);
let fieldnorm_data = segment.open_read(SegmentComponent::FIELDNORMS)?;
let fieldnorm_readers = FieldNormReaders::open(fieldnorm_data)?;
let fieldnorms_data = segment.open_read(SegmentComponent::FIELDNORMS)?;
let fieldnorms_composite = CompositeFile::open(&fieldnorms_data)?;
let delete_bitset_opt = if segment.meta().has_deletes() {
let delete_data = segment.open_read(SegmentComponent::DELETE)?;
@@ -195,14 +195,13 @@ impl SegmentReader {
termdict_composite,
postings_composite,
fast_fields_readers: fast_field_readers,
fieldnorm_readers,
fieldnorms_composite,
segment_id: segment.id(),
store_source,
delete_bitset_opt,
positions_composite,
positions_idx_composite,
schema,
logger: segment.index().logger().clone(),
})
}
@@ -213,11 +212,6 @@ impl SegmentReader {
/// The field reader is in charge of iterating through the
/// term dictionary associated to a specific field,
/// and opening the posting list associated to any term.
///
/// If the field is marked as index, a warn is logged and an empty `InvertedIndexReader`
/// is returned.
/// Similarly if the field is marked as indexed but no term has been indexed for the given
/// index. an empty `InvertedIndexReader` is returned (but no warning is logged).
pub fn inverted_index(&self, field: Field) -> Arc<InvertedIndexReader> {
if let Some(inv_idx_reader) = self
.inv_idx_reader_cache
@@ -232,25 +226,21 @@ impl SegmentReader {
let record_option_opt = field_type.get_index_record_option();
if record_option_opt.is_none() {
warn!(
self.logger,
"Field {:?} does not seem indexed.",
field_entry.name()
);
panic!("Field {:?} does not seem indexed.", field_entry.name());
}
let record_option = record_option_opt.unwrap();
let postings_source_opt = self.postings_composite.open_read(field);
if postings_source_opt.is_none() || record_option_opt.is_none() {
if postings_source_opt.is_none() {
// no documents in the segment contained this field.
// As a result, no data is associated to the inverted index.
//
// Returns an empty inverted index.
let record_option = record_option_opt.unwrap_or(IndexRecordOption::Basic);
return Arc::new(InvertedIndexReader::empty(record_option));
return Arc::new(InvertedIndexReader::empty(field_type));
}
let record_option = record_option_opt.unwrap();
let postings_source = postings_source_opt.unwrap();
let termdict_source = self.termdict_composite.open_read(field).expect(
@@ -305,8 +295,8 @@ impl SegmentReader {
}
/// Returns an iterator that will iterate over the alive document ids
pub fn doc_ids_alive<'a>(&'a self) -> impl Iterator<Item = DocId> + 'a {
(0u32..self.max_doc).filter(move |doc| !self.is_deleted(*doc))
pub fn doc_ids_alive(&self) -> SegmentReaderAliveDocsIterator<'_> {
SegmentReaderAliveDocsIterator::new(&self)
}
/// Summarize total space usage of this segment.
@@ -318,7 +308,7 @@ impl SegmentReader {
self.positions_composite.space_usage(),
self.positions_idx_composite.space_usage(),
self.fast_fields_readers.space_usage(),
self.fieldnorm_readers.space_usage(),
self.fieldnorms_composite.space_usage(),
self.get_store_reader().space_usage(),
self.delete_bitset_opt
.as_ref()
@@ -334,6 +324,52 @@ impl fmt::Debug for SegmentReader {
}
}
/// Implements the iterator trait to allow easy iteration
/// over non-deleted ("alive") DocIds in a SegmentReader
pub struct SegmentReaderAliveDocsIterator<'a> {
reader: &'a SegmentReader,
max_doc: DocId,
current: DocId,
}
impl<'a> SegmentReaderAliveDocsIterator<'a> {
pub fn new(reader: &'a SegmentReader) -> SegmentReaderAliveDocsIterator<'a> {
SegmentReaderAliveDocsIterator {
reader,
max_doc: reader.max_doc(),
current: 0,
}
}
}
impl<'a> Iterator for SegmentReaderAliveDocsIterator<'a> {
type Item = DocId;
fn next(&mut self) -> Option<Self::Item> {
// TODO: Use TinySet (like in BitSetDocSet) to speed this process up
if self.current >= self.max_doc {
return None;
}
// find the next alive doc id
while self.reader.is_deleted(self.current) {
self.current += 1;
if self.current >= self.max_doc {
return None;
}
}
// capture the current alive DocId
let result = Some(self.current);
// move down the chain
self.current += 1;
result
}
}
#[cfg(test)]
mod test {
use crate::core::Index;
@@ -349,7 +385,7 @@ mod test {
let name = schema.get_field("name").unwrap();
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(name => "tantivy"));
index_writer.add_document(doc!(name => "horse"));
index_writer.add_document(doc!(name => "jockey"));

12
src/directory/directory.rs
View File

@@ -1,5 +1,3 @@
use slog::{error, Logger};
use crate::directory::directory_lock::Lock;
use crate::directory::error::LockError;
use crate::directory::error::{DeleteError, OpenReadError, OpenWriteError};
@@ -66,10 +64,7 @@ impl<T: Send + Sync + 'static> From<Box<T>> for DirectoryLock {
impl Drop for DirectoryLockGuard {
fn drop(&mut self) {
if let Err(e) = self.directory.delete(&*self.path) {
error!(
self.directory.logger(),
"Failed to remove the lock file. {:?}", e
);
error!("Failed to remove the lock file. {:?}", e);
}
}
}
@@ -85,7 +80,7 @@ fn try_acquire_lock(
) -> 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),
OpenWriteError::IOError(io_error) => TryAcquireLockError::IOError(io_error.into()),
})?;
write.flush().map_err(TryAcquireLockError::IOError)?;
Ok(DirectoryLock::from(Box::new(DirectoryLockGuard {
@@ -214,9 +209,6 @@ pub trait Directory: DirectoryClone + fmt::Debug + Send + Sync + 'static {
/// `OnCommit` `ReloadPolicy`. Not implementing watch in a `Directory` only prevents the
/// `OnCommit` `ReloadPolicy` to work properly.
fn watch(&self, watch_callback: WatchCallback) -> crate::Result<WatchHandle>;
/// Returns the `slog::Logger` configured for the `Directory`.
fn logger(&self) -> &Logger;
}
/// DirectoryClone

250
src/directory/error.rs
View File

@@ -1,60 +1,160 @@
use crate::Version;
use std::error::Error as StdError;
use std::fmt;
use std::io;
use std::path::PathBuf;
/// Error while trying to acquire a directory lock.
#[derive(Debug, Error)]
#[derive(Debug, Fail)]
pub enum LockError {
/// Failed to acquired a lock as it is already held 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.
#[error("Could not acquire lock as it is already held, possibly by a different process.")]
#[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`
#[error("Failed to acquire the lock due to an io:Error.")]
#[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.
#[derive(Debug)]
pub struct IOError {
path: Option<PathBuf>,
err: io::Error,
}
impl Into<io::Error> for IOError {
fn into(self) -> io::Error {
self.err
}
}
impl fmt::Display for IOError {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match self.path {
Some(ref path) => write!(f, "io error occurred on path '{:?}': '{}'", path, self.err),
None => write!(f, "io error occurred: '{}'", self.err),
}
}
}
impl StdError for IOError {
fn description(&self) -> &str {
"io error occurred"
}
fn cause(&self) -> Option<&dyn StdError> {
Some(&self.err)
}
}
impl IOError {
pub(crate) fn with_path(path: PathBuf, err: io::Error) -> Self {
IOError {
path: Some(path),
err,
}
}
}
impl From<io::Error> for IOError {
fn from(err: io::Error) -> IOError {
IOError { path: None, err }
}
}
/// Error that may occur when opening a directory
#[derive(Debug, Error)]
#[derive(Debug)]
pub enum OpenDirectoryError {
/// The underlying directory does not exists.
#[error("Directory does not exist: '{0}'.")]
DoesNotExist(PathBuf),
/// The path exists but is not a directory.
#[error("Path exists but is not a directory: '{0}'.")]
NotADirectory(PathBuf),
/// Failed to create a temp directory.
#[error("Failed to create a temporary directory: '{0}'.")]
FailedToCreateTempDir(io::Error),
/// IoError
#[error("IOError '{io_error:?}' while create directory in: '{directory_path:?}'.")]
IoError {
/// underlying io Error.
io_error: io::Error,
/// directory we tried to open.
directory_path: PathBuf,
},
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 {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match *self {
OpenDirectoryError::DoesNotExist(ref path) => {
write!(f, "the underlying directory '{:?}' does not exist", path)
}
OpenDirectoryError::NotADirectory(ref 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
),
}
}
}
impl StdError for OpenDirectoryError {
fn description(&self) -> &str {
"error occurred while opening a directory"
}
fn cause(&self) -> Option<&dyn StdError> {
None
}
}
/// Error that may occur when starting to write in a file
#[derive(Debug, Error)]
#[derive(Debug)]
pub enum OpenWriteError {
/// Our directory is WORM, writing an existing file is forbidden.
/// Checkout the `Directory` documentation.
#[error("File already exists: '{0}'")]
FileAlreadyExists(PathBuf),
/// Any kind of IO error that happens when
/// writing in the underlying IO device.
#[error("IOError '{io_error:?}' while opening file for write: '{filepath}'.")]
IOError {
/// The underlying `io::Error`.
io_error: io::Error,
/// File path of the file that tantivy failed to open for write.
filepath: PathBuf,
},
IOError(IOError),
}
impl From<IOError> for OpenWriteError {
fn from(err: IOError) -> OpenWriteError {
OpenWriteError::IOError(err)
}
}
impl fmt::Display for OpenWriteError {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match *self {
OpenWriteError::FileAlreadyExists(ref path) => {
write!(f, "the file '{:?}' already exists", path)
}
OpenWriteError::IOError(ref err) => write!(
f,
"an io error occurred while opening a file for writing: '{}'",
err
),
}
}
}
impl StdError for OpenWriteError {
fn description(&self) -> &str {
"error occurred while opening a file for writing"
}
fn cause(&self) -> Option<&dyn StdError> {
match *self {
OpenWriteError::FileAlreadyExists(_) => None,
OpenWriteError::IOError(ref err) => Some(err),
}
}
}
/// Type of index incompatibility between the library and the index found on disk
@@ -117,41 +217,55 @@ impl fmt::Debug for Incompatibility {
}
/// Error that may occur when accessing a file read
#[derive(Debug, Error)]
#[derive(Debug)]
pub enum OpenReadError {
/// The file does not exists.
#[error("Files does not exists: {0:?}")]
FileDoesNotExist(PathBuf),
/// Any kind of io::Error.
#[error(
"IOError: '{io_error:?}' happened while opening the following file for Read: {filepath}."
)]
IOError {
/// The underlying `io::Error`.
io_error: io::Error,
/// File path of the file that tantivy failed to open for read.
filepath: PathBuf,
},
/// This library does not support the index version found in file footer.
#[error("Index version unsupported: {0:?}")]
IncompatibleIndex(Incompatibility),
}
/// Error that may occur when trying to delete a file
#[derive(Debug, Error)]
pub enum DeleteError {
/// The file does not exists.
#[error("File does not exists: '{0}'.")]
FileDoesNotExist(PathBuf),
/// Any kind of IO error that happens when
/// interacting with the underlying IO device.
#[error("The following IO error happened while deleting file '{filepath}': '{io_error:?}'.")]
IOError {
/// The underlying `io::Error`.
io_error: io::Error,
/// File path of the file that tantivy failed to delete.
filepath: PathBuf,
},
IOError(IOError),
/// This library doesn't support the index version found on disk
IncompatibleIndex(Incompatibility),
}
impl From<IOError> for OpenReadError {
fn from(err: IOError) -> OpenReadError {
OpenReadError::IOError(err)
}
}
impl fmt::Display for OpenReadError {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match *self {
OpenReadError::FileDoesNotExist(ref path) => {
write!(f, "the file '{:?}' does not exist", path)
}
OpenReadError::IOError(ref err) => write!(
f,
"an io error occurred while opening a file for reading: '{}'",
err
),
OpenReadError::IncompatibleIndex(ref footer) => {
write!(f, "Incompatible index format: {:?}", footer)
}
}
}
}
/// Error that may occur when trying to delete a file
#[derive(Debug)]
pub enum DeleteError {
/// The file does not exists.
FileDoesNotExist(PathBuf),
/// Any kind of IO error that happens when
/// interacting with the underlying IO device.
IOError(IOError),
}
impl From<IOError> for DeleteError {
fn from(err: IOError) -> DeleteError {
DeleteError::IOError(err)
}
}
impl From<Incompatibility> for OpenReadError {
@@ -159,3 +273,29 @@ impl From<Incompatibility> for OpenReadError {
OpenReadError::IncompatibleIndex(incompatibility)
}
}
impl fmt::Display for DeleteError {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match *self {
DeleteError::FileDoesNotExist(ref path) => {
write!(f, "the file '{:?}' does not exist", path)
}
DeleteError::IOError(ref err) => {
write!(f, "an io error occurred while deleting a file: '{}'", err)
}
}
}
}
impl StdError for DeleteError {
fn description(&self) -> &str {
"error occurred while deleting a file"
}
fn cause(&self) -> Option<&dyn StdError> {
match *self {
DeleteError::FileDoesNotExist(_) => None,
DeleteError::IOError(ref err) => Some(err),
}
}
}

115
src/directory/footer.rs
View File

@@ -8,8 +8,6 @@ use crc32fast::Hasher;
use std::io;
use std::io::Write;
const FOOTER_MAX_LEN: usize = 10_000;
type CrcHashU32 = u32;
#[derive(Debug, Clone, PartialEq)]
@@ -94,24 +92,12 @@ impl Footer {
match &self.versioned_footer {
VersionedFooter::V1 {
crc32: _crc,
store_compression,
store_compression: compression,
} => {
if &library_version.store_compression != store_compression {
if &library_version.store_compression != compression {
return Err(Incompatibility::CompressionMismatch {
library_compression_format: library_version.store_compression.to_string(),
index_compression_format: store_compression.to_string(),
});
}
Ok(())
}
VersionedFooter::V2 {
crc32: _crc,
store_compression,
} => {
if &library_version.store_compression != store_compression {
return Err(Incompatibility::CompressionMismatch {
library_compression_format: library_version.store_compression.to_string(),
index_compression_format: store_compression.to_string(),
index_compression_format: compression.to_string(),
});
}
Ok(())
@@ -132,29 +118,24 @@ pub enum VersionedFooter {
crc32: CrcHashU32,
store_compression: String,
},
// Introduction of the Block WAND information.
V2 {
crc32: CrcHashU32,
store_compression: String,
},
}
impl BinarySerializable for VersionedFooter {
fn serialize<W: io::Write>(&self, writer: &mut W) -> io::Result<()> {
let mut buf = Vec::new();
match self {
VersionedFooter::V2 {
VersionedFooter::V1 {
crc32,
store_compression: compression,
} => {
// Serializes a valid `VersionedFooter` or panics if the version is unknown
// [ version | crc_hash | compression_mode ]
// [ 0..4 | 4..8 | variable ]
BinarySerializable::serialize(&2u32, &mut buf)?;
BinarySerializable::serialize(&1u32, &mut buf)?;
BinarySerializable::serialize(crc32, &mut buf)?;
BinarySerializable::serialize(compression, &mut buf)?;
}
VersionedFooter::V1 { .. } | VersionedFooter::UnknownVersion => {
VersionedFooter::UnknownVersion => {
return Err(io::Error::new(
io::ErrorKind::InvalidInput,
"Cannot serialize an unknown versioned footer ",
@@ -162,51 +143,32 @@ impl BinarySerializable for VersionedFooter {
}
}
BinarySerializable::serialize(&VInt(buf.len() as u64), writer)?;
assert!(buf.len() <= FOOTER_MAX_LEN);
writer.write_all(&buf[..])?;
Ok(())
}
fn deserialize<R: io::Read>(reader: &mut R) -> io::Result<Self> {
let len = VInt::deserialize(reader)?.0 as usize;
if len > FOOTER_MAX_LEN {
return Err(io::Error::new(
io::ErrorKind::InvalidData,
format!(
"Footer seems invalid as it suggests a footer len of {}. File is corrupted, \
or the index was created with a different & old version of tantivy.",
len
),
));
}
let mut buf = vec![0u8; len];
reader.read_exact(&mut buf[..])?;
let mut cursor = &buf[..];
let version = u32::deserialize(&mut cursor)?;
if version != 1 && version != 2 {
return Ok(VersionedFooter::UnknownVersion);
}
let crc32 = u32::deserialize(&mut cursor)?;
let store_compression = String::deserialize(&mut cursor)?;
Ok(if version == 1 {
VersionedFooter::V1 {
if version == 1 {
let crc32 = u32::deserialize(&mut cursor)?;
let compression = String::deserialize(&mut cursor)?;
Ok(VersionedFooter::V1 {
crc32,
store_compression,
}
store_compression: compression,
})
} else {
assert_eq!(version, 2);
VersionedFooter::V2 {
crc32,
store_compression,
}
})
Ok(VersionedFooter::UnknownVersion)
}
}
}
impl VersionedFooter {
pub fn crc(&self) -> Option<CrcHashU32> {
match self {
VersionedFooter::V2 { crc32, .. } => Some(*crc32),
VersionedFooter::V1 { crc32, .. } => Some(*crc32),
VersionedFooter::UnknownVersion { .. } => None,
}
@@ -244,7 +206,7 @@ impl<W: TerminatingWrite> Write for FooterProxy<W> {
impl<W: TerminatingWrite> TerminatingWrite for FooterProxy<W> {
fn terminate_ref(&mut self, _: AntiCallToken) -> io::Result<()> {
let crc32 = self.hasher.take().unwrap().finalize();
let footer = Footer::new(VersionedFooter::V2 {
let footer = Footer::new(VersionedFooter::V1 {
crc32,
store_compression: crate::store::COMPRESSION.to_string(),
});
@@ -259,29 +221,28 @@ mod tests {
use super::CrcHashU32;
use super::FooterProxy;
use crate::common::{BinarySerializable, VInt};
use crate::common::BinarySerializable;
use crate::directory::footer::{Footer, VersionedFooter};
use crate::directory::TerminatingWrite;
use byteorder::{ByteOrder, LittleEndian};
use regex::Regex;
use std::io;
#[test]
fn test_versioned_footer() {
let mut vec = Vec::new();
let footer_proxy = FooterProxy::new(&mut vec);
assert!(footer_proxy.terminate().is_ok());
if crate::store::COMPRESSION == "lz4" {
assert_eq!(vec.len(), 158);
} else {
assert_eq!(vec.len(), 167);
}
assert_eq!(vec.len(), 167);
let footer = Footer::deserialize(&mut &vec[..]).unwrap();
assert!(matches!(
footer.versioned_footer,
VersionedFooter::V2 { store_compression, .. }
if store_compression == crate::store::COMPRESSION
));
if let VersionedFooter::V1 {
crc32: _,
store_compression,
} = footer.versioned_footer
{
assert_eq!(store_compression, crate::store::COMPRESSION);
} else {
panic!("Versioned footer should be V1.");
}
assert_eq!(&footer.version, crate::version());
}
@@ -289,7 +250,7 @@ mod tests {
fn test_serialize_deserialize_footer() {
let mut buffer = Vec::new();
let crc32 = 123456u32;
let footer: Footer = Footer::new(VersionedFooter::V2 {
let footer: Footer = Footer::new(VersionedFooter::V1 {
crc32,
store_compression: "lz4".to_string(),
});
@@ -301,7 +262,7 @@ mod tests {
#[test]
fn footer_length() {
let crc32 = 1111111u32;
let versioned_footer = VersionedFooter::V2 {
let versioned_footer = VersionedFooter::V1 {
crc32,
store_compression: "lz4".to_string(),
};
@@ -322,7 +283,7 @@ mod tests {
// versionned footer length
12 | 128,
// index format version
2,
1,
0,
0,
0,
@@ -341,7 +302,7 @@ mod tests {
let versioned_footer = VersionedFooter::deserialize(&mut cursor).unwrap();
assert!(cursor.is_empty());
let expected_crc: u32 = LittleEndian::read_u32(&v_footer_bytes[5..9]) as CrcHashU32;
let expected_versioned_footer: VersionedFooter = VersionedFooter::V2 {
let expected_versioned_footer: VersionedFooter = VersionedFooter::V1 {
crc32: expected_crc,
store_compression: "lz4".to_string(),
};
@@ -375,20 +336,4 @@ mod tests {
let res = footer.is_compatible();
assert!(res.is_err());
}
#[test]
fn test_deserialize_too_large_footer() {
let mut buf = vec![];
assert!(FooterProxy::new(&mut buf).terminate().is_ok());
let mut long_len_buf = [0u8; 10];
let num_bytes = VInt(super::FOOTER_MAX_LEN as u64 + 1u64).serialize_into(&mut long_len_buf);
buf[0..num_bytes].copy_from_slice(&long_len_buf[..num_bytes]);
let err = Footer::deserialize(&mut &buf[..]).unwrap_err();
assert_eq!(err.kind(), io::ErrorKind::InvalidData);
assert_eq!(
err.to_string(),
"Footer seems invalid as it suggests a footer len of 10001. File is corrupted, \
or the index was created with a different & old version of tantivy."
);
}
}

78
src/directory/managed_directory.rs
View File

@@ -1,5 +1,5 @@
use crate::core::{MANAGED_FILEPATH, META_FILEPATH};
use crate::directory::error::{DeleteError, LockError, OpenReadError, OpenWriteError};
use crate::core::MANAGED_FILEPATH;
use crate::directory::error::{DeleteError, IOError, LockError, OpenReadError, OpenWriteError};
use crate::directory::footer::{Footer, FooterProxy};
use crate::directory::DirectoryLock;
use crate::directory::GarbageCollectionResult;
@@ -11,9 +11,10 @@ use crate::error::DataCorruption;
use crate::Directory;
use crc32fast::Hasher;
use slog::{debug, error, info};
use serde_json;
use std::collections::HashSet;
use std::io;
use std::io::Write;
use std::path::{Path, PathBuf};
use std::result;
use std::sync::RwLockWriteGuard;
@@ -56,9 +57,9 @@ fn save_managed_paths(
directory: &mut dyn Directory,
wlock: &RwLockWriteGuard<'_, MetaInformation>,
) -> io::Result<()> {
let mut managed_json = serde_json::to_string_pretty(&wlock.managed_paths)?;
managed_json.push_str("\n");
directory.atomic_write(&MANAGED_FILEPATH, managed_json.as_bytes())?;
let mut w = serde_json::to_vec(&wlock.managed_paths)?;
writeln!(&mut w)?;
directory.atomic_write(&MANAGED_FILEPATH, &w[..])?;
Ok(())
}
@@ -86,12 +87,7 @@ impl ManagedDirectory {
directory: Box::new(directory),
meta_informations: Arc::default(),
}),
Err(OpenReadError::IOError { io_error, filepath }) => {
Err(crate::TantivyError::OpenReadError(OpenReadError::IOError {
io_error,
filepath,
}))
}
Err(OpenReadError::IOError(e)) => Err(From::from(e)),
Err(OpenReadError::IncompatibleIndex(incompatibility)) => {
// For the moment, this should never happen `meta.json`
// do not have any footer and cannot detect incompatibility.
@@ -118,7 +114,7 @@ impl ManagedDirectory {
&mut self,
get_living_files: L,
) -> crate::Result<GarbageCollectionResult> {
info!(self.directory.logger(), "gc"; "stage"=>"start");
info!("Garbage collect");
let mut files_to_delete = vec![];
// It is crucial to get the living files after acquiring the
@@ -153,7 +149,7 @@ impl ManagedDirectory {
}
}
Err(err) => {
error!(self.logger(), "Failed to acquire lock for GC");
error!("Failed to acquire lock for GC");
return Err(crate::TantivyError::from(err));
}
}
@@ -165,7 +161,7 @@ impl ManagedDirectory {
for file_to_delete in files_to_delete {
match self.delete(&file_to_delete) {
Ok(_) => {
debug!(self.logger(), "deleted-success"; "file"=>format!("{:?}", file_to_delete));
info!("Deleted {:?}", file_to_delete);
deleted_files.push(file_to_delete);
}
Err(file_error) => {
@@ -173,12 +169,12 @@ impl ManagedDirectory {
DeleteError::FileDoesNotExist(_) => {
deleted_files.push(file_to_delete.clone());
}
DeleteError::IOError { .. } => {
DeleteError::IOError(_) => {
failed_to_delete_files.push(file_to_delete.clone());
if !cfg!(target_os = "windows") {
// On windows, delete is expected to fail if the file
// is mmapped.
error!(self.logger(), "delete-file-fail"; "path"=>file_to_delete.to_str().unwrap_or("<invalid-utf8>"));
error!("Failed to delete {:?}", file_to_delete);
}
}
}
@@ -200,10 +196,6 @@ impl ManagedDirectory {
save_managed_paths(self.directory.as_mut(), &meta_informations_wlock)?;
}
info!(self.directory.logger(), "gc"; "stage"=>"end",
"num-sucess-file-deletes"=>deleted_files.len(),
"num-failed-file-deletes"=>failed_to_delete_files.len());
Ok(GarbageCollectionResult {
deleted_files,
failed_to_delete_files,
@@ -240,11 +232,8 @@ impl ManagedDirectory {
/// Verify checksum of a managed file
pub fn validate_checksum(&self, path: &Path) -> result::Result<bool, OpenReadError> {
let reader = self.directory.open_read(path)?;
let (footer, data) =
Footer::extract_footer(reader).map_err(|io_error| OpenReadError::IOError {
io_error,
filepath: path.to_path_buf(),
})?;
let (footer, data) = Footer::extract_footer(reader)
.map_err(|err| IOError::with_path(path.to_path_buf(), err))?;
let mut hasher = Hasher::new();
hasher.update(data.as_slice());
let crc = hasher.finalize();
@@ -257,46 +246,35 @@ impl ManagedDirectory {
/// List files for which checksum does not match content
pub fn list_damaged(&self) -> result::Result<HashSet<PathBuf>, OpenReadError> {
let mut managed_paths = self
let mut hashset = HashSet::new();
let managed_paths = self
.meta_informations
.read()
.expect("Managed directory rlock poisoned in list damaged.")
.managed_paths
.clone();
managed_paths.remove(*META_FILEPATH);
let mut damaged_files = HashSet::new();
for path in managed_paths {
for path in managed_paths.into_iter() {
if !self.validate_checksum(&path)? {
damaged_files.insert(path);
hashset.insert(path);
}
}
Ok(damaged_files)
Ok(hashset)
}
}
impl Directory for ManagedDirectory {
fn open_read(&self, path: &Path) -> result::Result<ReadOnlySource, OpenReadError> {
slog::debug!(self.logger(), "open-read"; "path" => path.to_str().unwrap_or("<invalid-utf8>"));
let read_only_source = self.directory.open_read(path)?;
let (footer, reader) = Footer::extract_footer(read_only_source).map_err(|io_error| {
OpenReadError::IOError {
io_error,
filepath: path.to_path_buf(),
}
})?;
let (footer, reader) = Footer::extract_footer(read_only_source)
.map_err(|err| IOError::with_path(path.to_path_buf(), err))?;
footer.is_compatible()?;
Ok(reader)
}
fn open_write(&mut self, path: &Path) -> result::Result<WritePtr, OpenWriteError> {
slog::debug!(self.logger(), "open-write"; "path" => path.to_str().unwrap_or("<invalid-utf8>"));
self.register_file_as_managed(path)
.map_err(|io_error| OpenWriteError::IOError {
io_error,
filepath: path.to_path_buf(),
})?;
.map_err(|e| IOError::with_path(path.to_owned(), e))?;
Ok(io::BufWriter::new(Box::new(FooterProxy::new(
self.directory
.open_write(path)?
@@ -306,11 +284,9 @@ impl Directory for ManagedDirectory {
))))
}
fn atomic_write(&mut self, path: &Path, content: &[u8]) -> io::Result<()> {
let content_str = std::str::from_utf8(content).unwrap_or("<content-not-utf-8>");
slog::debug!(self.logger(), "Atomic write"; "path" => format!("{:?}", path), "content_length"=>content_str);
fn atomic_write(&mut self, path: &Path, data: &[u8]) -> io::Result<()> {
self.register_file_as_managed(path)?;
self.directory.atomic_write(path, content)
self.directory.atomic_write(path, data)
}
fn atomic_read(&self, path: &Path) -> result::Result<Vec<u8>, OpenReadError> {
@@ -332,10 +308,6 @@ impl Directory for ManagedDirectory {
fn watch(&self, watch_callback: WatchCallback) -> crate::Result<WatchHandle> {
self.directory.watch(watch_callback)
}
fn logger(&self) -> &slog::Logger {
self.directory.logger()
}
}
impl Clone for ManagedDirectory {

176
src/directory/mmap_directory.rs
View File

@@ -1,6 +1,15 @@
use fs2;
use notify;
use self::fs2::FileExt;
use self::notify::RawEvent;
use self::notify::RecursiveMode;
use self::notify::Watcher;
use crate::core::META_FILEPATH;
use crate::directory::error::LockError;
use crate::directory::error::{DeleteError, OpenDirectoryError, OpenReadError, OpenWriteError};
use crate::directory::error::{
DeleteError, IOError, OpenDirectoryError, OpenReadError, OpenWriteError,
};
use crate::directory::read_only_source::BoxedData;
use crate::directory::AntiCallToken;
use crate::directory::Directory;
@@ -11,14 +20,8 @@ use crate::directory::WatchCallback;
use crate::directory::WatchCallbackList;
use crate::directory::WatchHandle;
use crate::directory::{TerminatingWrite, WritePtr};
use fs2::FileExt;
use atomicwrites;
use memmap::Mmap;
use notify::RawEvent;
use notify::RecursiveMode;
use notify::Watcher;
use serde::{Deserialize, Serialize};
use slog::{debug, o, Drain, Logger};
use slog_stdlog::StdLog;
use std::collections::HashMap;
use std::convert::From;
use std::fmt;
@@ -36,6 +39,11 @@ use std::sync::Weak;
use std::thread;
use tempfile::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
/// cannot be mmapped)
fn open_mmap(full_path: &Path) -> result::Result<Option<Mmap>, OpenReadError> {
@@ -43,17 +51,13 @@ fn open_mmap(full_path: &Path) -> result::Result<Option<Mmap>, OpenReadError> {
if e.kind() == io::ErrorKind::NotFound {
OpenReadError::FileDoesNotExist(full_path.to_owned())
} else {
OpenReadError::IOError {
io_error: e,
filepath: full_path.to_owned(),
}
OpenReadError::IOError(IOError::with_path(full_path.to_owned(), e))
}
})?;
let meta_data = file.metadata().map_err(|e| OpenReadError::IOError {
io_error: e,
filepath: full_path.to_owned(),
})?;
let meta_data = file
.metadata()
.map_err(|e| IOError::with_path(full_path.to_owned(), e))?;
if meta_data.len() == 0 {
// if the file size is 0, it will not be possible
// to mmap the file, so we return None
@@ -63,10 +67,7 @@ fn open_mmap(full_path: &Path) -> result::Result<Option<Mmap>, OpenReadError> {
unsafe {
memmap::Mmap::map(&file)
.map(Some)
.map_err(|e| OpenReadError::IOError {
io_error: e,
filepath: full_path.to_owned(),
})
.map_err(|e| From::from(IOError::with_path(full_path.to_owned(), e)))
}
}
@@ -141,15 +142,15 @@ impl MmapCache {
}
struct WatcherWrapper {
_watcher: Mutex<notify::RecommendedWatcher>,
_watcher: Mutex<notify::PollWatcher>,
watcher_router: Arc<WatchCallbackList>,
}
impl WatcherWrapper {
pub(crate) fn new(path: &Path, logger: Logger) -> Result<Self, OpenDirectoryError> {
pub fn new(path: &Path) -> Result<Self, OpenDirectoryError> {
let (tx, watcher_recv): (Sender<RawEvent>, Receiver<RawEvent>) = channel();
// We need to initialize the
let watcher = notify::raw_watcher(tx)
let watcher = notify::poll::PollWatcher::with_delay_ms(tx, 1)
.and_then(|mut watcher| {
watcher.watch(path, RecursiveMode::Recursive)?;
Ok(watcher)
@@ -160,8 +161,7 @@ impl WatcherWrapper {
panic!("Unknown error while starting watching directory {:?}", path);
}
})?;
let watcher_router: Arc<WatchCallbackList> =
Arc::new(WatchCallbackList::with_logger(logger));
let watcher_router: Arc<WatchCallbackList> = Default::default();
let watcher_router_clone = watcher_router.clone();
thread::Builder::new()
.name("meta-file-watch-thread".to_string())
@@ -186,10 +186,6 @@ impl WatcherWrapper {
}
}
}
})
.map_err(|io_error| OpenDirectoryError::IoError {
io_error,
directory_path: path.to_path_buf(),
})?;
Ok(WatcherWrapper {
_watcher: Mutex::new(watcher),
@@ -224,22 +220,20 @@ struct MmapDirectoryInner {
mmap_cache: RwLock<MmapCache>,
_temp_directory: Option<TempDir>,
watcher: RwLock<Option<WatcherWrapper>>,
logger: Logger,
}
impl MmapDirectoryInner {
fn new(
root_path: PathBuf,
temp_directory: Option<TempDir>,
logger: Logger,
) -> MmapDirectoryInner {
MmapDirectoryInner {
) -> Result<MmapDirectoryInner, OpenDirectoryError> {
let mmap_directory_inner = MmapDirectoryInner {
root_path,
mmap_cache: Default::default(),
_temp_directory: temp_directory,
watcher: RwLock::new(None),
logger,
}
};
Ok(mmap_directory_inner)
}
fn watch(&self, watch_callback: WatchCallback) -> crate::Result<WatchHandle> {
@@ -250,7 +244,7 @@ impl MmapDirectoryInner {
// The downside is that we might create a watch wrapper that is not useful.
let need_initialization = self.watcher.read().unwrap().is_none();
if need_initialization {
let watch_wrapper = WatcherWrapper::new(&self.root_path, self.logger.clone())?;
let watch_wrapper = WatcherWrapper::new(&self.root_path)?;
let mut watch_wlock = self.watcher.write().unwrap();
// the watcher could have been initialized when we released the lock, and
// we do not want to lose the watched files that were set.
@@ -273,11 +267,14 @@ impl fmt::Debug for MmapDirectory {
}
impl MmapDirectory {
fn new(root_path: PathBuf, temp_directory: Option<TempDir>, logger: Logger) -> MmapDirectory {
let inner = MmapDirectoryInner::new(root_path, temp_directory, logger);
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.
@@ -285,19 +282,16 @@ impl MmapDirectory {
/// This is mostly useful to test the MmapDirectory itself.
/// For your unit tests, prefer the RAMDirectory.
pub fn create_from_tempdir() -> Result<MmapDirectory, OpenDirectoryError> {
let tempdir = TempDir::new().map_err(OpenDirectoryError::FailedToCreateTempDir)?;
let logger = Logger::root(StdLog.fuse(), o!());
Ok(MmapDirectory::new(tempdir.path().to_owned(), Some(tempdir), logger))
let tempdir = TempDir::new().map_err(OpenDirectoryError::IoError)?;
let tempdir_path = PathBuf::from(tempdir.path());
MmapDirectory::new(tempdir_path, Some(tempdir))
}
/// Opens a MmapDirectory in a directory.
///
/// Returns an error if the `directory_path` does not
/// exist or if it is not a directory.
pub fn open_with_logger<P: AsRef<Path>>(
directory_path: P,
logger: Logger,
) -> Result<MmapDirectory, OpenDirectoryError> {
pub fn open<P: AsRef<Path>>(directory_path: P) -> Result<MmapDirectory, OpenDirectoryError> {
let directory_path: &Path = directory_path.as_ref();
if !directory_path.exists() {
Err(OpenDirectoryError::DoesNotExist(PathBuf::from(
@@ -308,20 +302,10 @@ impl MmapDirectory {
directory_path,
)))
} else {
Ok(MmapDirectory::new(
PathBuf::from(directory_path),
None,
logger,
))
Ok(MmapDirectory::new(PathBuf::from(directory_path), None)?)
}
}
/// Creates an `MmapDirectory` at the given path.
pub fn open<P: AsRef<Path>>(directory_path: P) -> Result<MmapDirectory, OpenDirectoryError> {
let logger = Logger::root(StdLog.fuse(), o!());
Self::open_with_logger(directory_path, logger)
}
/// Joins a relative_path to the directory `root_path`
/// to create a proper complete `filepath`.
fn resolve_path(&self, relative_path: &Path) -> PathBuf {
@@ -381,12 +365,11 @@ impl MmapDirectory {
struct ReleaseLockFile {
_file: File,
path: PathBuf,
logger: Logger,
}
impl Drop for ReleaseLockFile {
fn drop(&mut self) {
debug!(self.logger, "Releasing lock {:?}", self.path);
debug!("Releasing lock {:?}", self.path);
}
}
@@ -425,18 +408,16 @@ impl TerminatingWrite for SafeFileWriter {
impl Directory for MmapDirectory {
fn open_read(&self, path: &Path) -> result::Result<ReadOnlySource, OpenReadError> {
debug!("Open Read {:?}", path);
let full_path = self.resolve_path(path);
let mut mmap_cache = self.inner.mmap_cache.write().map_err(|_| {
let msg = format!(
"Failed to acquired write lock \
on mmap cache while reading {:?}",
path
);
let io_error = io::Error::new(io::ErrorKind::Other, msg);
OpenReadError::IOError {
io_error,
filepath: path.to_owned(),
}
IOError::with_path(path.to_owned(), make_io_err(msg))
})?;
Ok(mmap_cache
.get_mmap(&full_path)?
@@ -449,18 +430,14 @@ impl Directory for MmapDirectory {
fn delete(&self, path: &Path) -> result::Result<(), DeleteError> {
let full_path = self.resolve_path(path);
match fs::remove_file(&full_path) {
Ok(_) => self.sync_directory().map_err(|e| DeleteError::IOError {
io_error: e,
filepath: path.to_path_buf(),
}),
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(DeleteError::IOError {
io_error: e,
filepath: path.to_path_buf(),
})
Err(IOError::with_path(path.to_owned(), e).into())
}
}
}
@@ -472,7 +449,9 @@ impl Directory for MmapDirectory {
}
fn open_write(&mut self, path: &Path) -> Result<WritePtr, OpenWriteError> {
debug!("Open Write {:?}", path);
let full_path = self.resolve_path(path);
let open_res = OpenOptions::new()
.write(true)
.create_new(true)
@@ -482,25 +461,18 @@ impl Directory for MmapDirectory {
if err.kind() == io::ErrorKind::AlreadyExists {
OpenWriteError::FileAlreadyExists(path.to_owned())
} else {
OpenWriteError::IOError {
io_error: err,
filepath: path.to_owned(),
}
IOError::with_path(path.to_owned(), err).into()
}
})?;
// making sure the file is created.
file.flush().map_err(|io_error| OpenWriteError::IOError {
io_error,
filepath: path.to_owned(),
})?;
file.flush()
.map_err(|e| IOError::with_path(path.to_owned(), e))?;
// Apparetntly, on some filesystem syncing the parent
// directory is required.
self.sync_directory().map_err(|e| OpenWriteError::IOError {
io_error: e,
filepath: path.to_owned(),
})?;
self.sync_directory()
.map_err(|e| IOError::with_path(path.to_owned(), e))?;
let writer = SafeFileWriter::new(file);
Ok(BufWriter::new(Box::new(writer)))
@@ -512,31 +484,24 @@ impl Directory for MmapDirectory {
match File::open(&full_path) {
Ok(mut file) => {
file.read_to_end(&mut buffer)
.map_err(|io_error| OpenReadError::IOError {
io_error,
filepath: path.to_owned(),
})?;
.map_err(|e| IOError::with_path(path.to_owned(), e))?;
Ok(buffer)
}
Err(io_error) => {
if io_error.kind() == io::ErrorKind::NotFound {
Err(e) => {
if e.kind() == io::ErrorKind::NotFound {
Err(OpenReadError::FileDoesNotExist(path.to_owned()))
} else {
Err(OpenReadError::IOError {
io_error,
filepath: path.to_owned(),
})
Err(IOError::with_path(path.to_owned(), e).into())
}
}
}
}
fn atomic_write(&mut self, path: &Path, content: &[u8]) -> io::Result<()> {
let mut tempfile = tempfile::Builder::new().tempfile_in(&self.inner.root_path)?;
tempfile.write_all(content)?;
tempfile.flush()?;
fn atomic_write(&mut self, path: &Path, data: &[u8]) -> io::Result<()> {
debug!("Atomic Write {:?}", path);
let full_path = self.resolve_path(path);
tempfile.into_temp_path().persist(full_path)?;
let meta_file = atomicwrites::AtomicFile::new(full_path, atomicwrites::AllowOverwrite);
meta_file.write(|f| f.write_all(data))?;
Ok(())
}
@@ -553,22 +518,16 @@ impl Directory for MmapDirectory {
} else {
file.try_lock_exclusive().map_err(|_| LockError::LockBusy)?
}
let logger = self.inner.logger.clone();
// dropping the file handle will release the lock.
Ok(DirectoryLock::from(Box::new(ReleaseLockFile {
path: lock.filepath.clone(),
_file: file,
logger,
})))
}
fn watch(&self, watch_callback: WatchCallback) -> crate::Result<WatchHandle> {
self.inner.watch(watch_callback)
}
fn logger(&self) -> &Logger {
&self.inner.logger
}
}
#[cfg(test)]
@@ -678,8 +637,7 @@ mod tests {
let counter_clone = counter.clone();
let tmp_dir = tempfile::TempDir::new().unwrap();
let tmp_dirpath = tmp_dir.path().to_owned();
let logger = Logger::root(slog::Discard, o!());
let mut watch_wrapper = WatcherWrapper::new(&tmp_dirpath, logger).unwrap();
let mut watch_wrapper = WatcherWrapper::new(&tmp_dirpath).unwrap();
let tmp_file = tmp_dirpath.join(*META_FILEPATH);
let _handle = watch_wrapper.watch(Box::new(move || {
counter_clone.fetch_add(1, Ordering::SeqCst);
@@ -704,7 +662,7 @@ mod tests {
{
let index = Index::create(mmap_directory.clone(), schema).unwrap();
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
let mut log_merge_policy = LogMergePolicy::default();
log_merge_policy.set_min_merge_size(3);
index_writer.set_merge_policy(Box::new(log_merge_policy));

3
src/directory/mod.rs
View File

@@ -23,8 +23,7 @@ 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::read_only_source::ReadOnlySource;
pub(crate) use self::watch_event_router::WatchCallbackList;
pub use self::watch_event_router::{WatchCallback, WatchHandle};
pub use self::watch_event_router::{WatchCallback, WatchCallbackList, WatchHandle};
use std::io::{self, BufWriter, Write};
use std::path::PathBuf;
/// Outcome of the Garbage collection

122
src/directory/ram_directory.rs
View File

@@ -5,8 +5,6 @@ use crate::directory::WatchCallbackList;
use crate::directory::{Directory, ReadOnlySource, WatchCallback, WatchHandle};
use crate::directory::{TerminatingWrite, WritePtr};
use fail::fail_point;
use slog::{o, Drain, Logger};
use slog_stdlog::StdLog;
use std::collections::HashMap;
use std::fmt;
use std::io::{self, BufWriter, Cursor, Seek, SeekFrom, Write};
@@ -68,7 +66,7 @@ impl Write for VecWriter {
fn flush(&mut self) -> io::Result<()> {
self.is_flushed = true;
let mut fs = self.shared_directory.fs.inner_directory.write().unwrap();
let mut fs = self.shared_directory.fs.write().unwrap();
fs.write(self.path.clone(), self.data.get_ref());
Ok(())
}
@@ -80,19 +78,13 @@ impl TerminatingWrite for VecWriter {
}
}
#[derive(Default)]
struct InnerDirectory {
fs: HashMap<PathBuf, ReadOnlySource>,
watch_router: WatchCallbackList,
}
impl InnerDirectory {
fn with_logger(logger: Logger) -> Self {
InnerDirectory {
fs: Default::default(),
watch_router: WatchCallbackList::with_logger(logger.clone()),
}
}
fn write(&mut self, path: PathBuf, data: &[u8]) -> bool {
let data = ReadOnlySource::new(Vec::from(data));
self.fs.insert(path, data).is_some()
@@ -125,32 +117,20 @@ impl InnerDirectory {
}
}
impl Default for RAMDirectory {
fn default() -> RAMDirectory {
let logger = Logger::root(StdLog.fuse(), o!());
Self::with_logger(logger)
}
}
impl fmt::Debug for RAMDirectory {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(f, "RAMDirectory")
}
}
struct Inner {
inner_directory: RwLock<InnerDirectory>,
logger: Logger,
}
/// A Directory storing everything in anonymous memory.
///
/// It is mainly meant for unit testing.
/// Writes are only made visible upon flushing.
///
#[derive(Clone)]
#[derive(Clone, Default)]
pub struct RAMDirectory {
fs: Arc<Inner>,
fs: Arc<RwLock<InnerDirectory>>,
}
impl RAMDirectory {
@@ -159,61 +139,33 @@ impl RAMDirectory {
Self::default()
}
/// Create a `RAMDirectory` with a custom logger.
pub fn with_logger(logger: Logger) -> RAMDirectory {
let inner_directory = InnerDirectory::with_logger(logger.clone()).into();
RAMDirectory {
fs: Arc::new(Inner {
inner_directory,
logger,
}),
}
}
/// Returns the sum of the size of the different files
/// in the RAMDirectory.
pub fn total_mem_usage(&self) -> usize {
self.fs.inner_directory.read().unwrap().total_mem_usage()
}
/// Write a copy of all of the files saved in the RAMDirectory in the target `Directory`.
///
/// Files are all written using the `Directory::write` meaning, even if they were
/// written using the `atomic_write` api.
///
/// If an error is encounterred, files may be persisted partially.
pub fn persist(&self, dest: &mut dyn Directory) -> crate::Result<()> {
let wlock = self.fs.inner_directory.write().unwrap();
for (path, source) in wlock.fs.iter() {
let mut dest_wrt = dest.open_write(path)?;
dest_wrt.write_all(source.as_slice())?;
dest_wrt.terminate()?;
}
Ok(())
self.fs.read().unwrap().total_mem_usage()
}
}
impl Directory for RAMDirectory {
fn open_read(&self, path: &Path) -> result::Result<ReadOnlySource, OpenReadError> {
self.fs.inner_directory.read().unwrap().open_read(path)
self.fs.read().unwrap().open_read(path)
}
fn delete(&self, path: &Path) -> result::Result<(), DeleteError> {
fail_point!("RAMDirectory::delete", |_| {
Err(DeleteError::IOError {
io_error: io::Error::from(io::ErrorKind::Other),
filepath: path.to_path_buf(),
})
use crate::directory::error::IOError;
let io_error = IOError::from(io::Error::from(io::ErrorKind::Other));
Err(DeleteError::from(io_error))
});
self.fs.inner_directory.write().unwrap().delete(path)
self.fs.write().unwrap().delete(path)
}
fn exists(&self, path: &Path) -> bool {
self.fs.inner_directory.read().unwrap().exists(path)
self.fs.read().unwrap().exists(path)
}
fn open_write(&mut self, path: &Path) -> Result<WritePtr, OpenWriteError> {
let mut fs = self.fs.inner_directory.write().unwrap();
let mut fs = self.fs.write().unwrap();
let path_buf = PathBuf::from(path);
let vec_writer = VecWriter::new(path_buf.clone(), self.clone());
let exists = fs.write(path_buf.clone(), &[]);
@@ -237,62 +189,18 @@ impl Directory for RAMDirectory {
let path_buf = PathBuf::from(path);
// Reserve the path to prevent calls to .write() to succeed.
self.fs
.inner_directory
.write()
.unwrap()
.write(path_buf.clone(), &[]);
self.fs.write().unwrap().write(path_buf.clone(), &[]);
let mut vec_writer = VecWriter::new(path_buf, self.clone());
vec_writer.write_all(data)?;
vec_writer.flush()?;
if path == Path::new(&*META_FILEPATH) {
let _ = self
.fs
.inner_directory
.write()
.unwrap()
.watch_router
.broadcast();
let _ = self.fs.write().unwrap().watch_router.broadcast();
}
Ok(())
}
fn watch(&self, watch_callback: WatchCallback) -> crate::Result<WatchHandle> {
Ok(self
.fs
.inner_directory
.write()
.unwrap()
.watch(watch_callback))
}
fn logger(&self) -> &Logger {
&self.fs.logger
}
}
#[cfg(test)]
mod tests {
use super::RAMDirectory;
use crate::Directory;
use std::io::Write;
use std::path::Path;
#[test]
fn test_persist() {
let msg_atomic: &'static [u8] = b"atomic is the way";
let msg_seq: &'static [u8] = b"sequential is the way";
let path_atomic: &'static Path = Path::new("atomic");
let path_seq: &'static Path = Path::new("seq");
let mut directory = RAMDirectory::create();
assert!(directory.atomic_write(path_atomic, msg_atomic).is_ok());
let mut wrt = directory.open_write(path_seq).unwrap();
assert!(wrt.write_all(msg_seq).is_ok());
assert!(wrt.flush().is_ok());
let mut directory_copy = RAMDirectory::create();
assert!(directory.persist(&mut directory_copy).is_ok());
assert_eq!(directory_copy.atomic_read(path_atomic).unwrap(), msg_atomic);
assert_eq!(directory_copy.atomic_read(path_seq).unwrap(), msg_seq);
Ok(self.fs.write().unwrap().watch(watch_callback))
}
}

6
src/directory/tests.rs
View File

@@ -211,19 +211,19 @@ fn test_watch(directory: &mut dyn Directory) {
.unwrap();
for i in 0..10 {
assert!(i <= counter.load(SeqCst));
assert_eq!(i, counter.load(SeqCst));
assert!(directory
.atomic_write(Path::new("meta.json"), b"random_test_data_2")
.is_ok());
assert_eq!(receiver.recv_timeout(Duration::from_millis(500)), Ok(i));
assert!(i + 1 <= counter.load(SeqCst)); // notify can trigger more than once.
assert_eq!(i + 1, counter.load(SeqCst));
}
mem::drop(watch_handle);
assert!(directory
.atomic_write(Path::new("meta.json"), b"random_test_data")
.is_ok());
assert!(receiver.recv_timeout(Duration::from_millis(500)).is_ok());
assert!(10 <= counter.load(SeqCst));
assert_eq!(10, counter.load(SeqCst));
}
fn test_lock_non_blocking(directory: &mut dyn Directory) {

31
src/directory/watch_event_router.rs
View File

@@ -1,20 +1,19 @@
use futures::channel::oneshot;
use futures::{Future, TryFutureExt};
use slog::{error, Logger};
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<dyn Fn() + Sync + Send>;
pub type WatchCallback = Box<dyn 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(...)`.
pub(crate) struct WatchCallbackList {
#[derive(Default)]
pub struct WatchCallbackList {
router: RwLock<Vec<Weak<WatchCallback>>>,
logger: Logger,
}
/// Controls how long a directory should watch for a file change.
@@ -33,14 +32,7 @@ impl WatchHandle {
}
impl WatchCallbackList {
pub fn with_logger(logger: Logger) -> Self {
WatchCallbackList {
logger,
router: Default::default(),
}
}
/// Subscribes a new callback and returns a handle that controls the lifetime of the callback.
/// 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);
@@ -82,8 +74,8 @@ impl WatchCallbackList {
});
if let Err(err) = spawn_res {
error!(
self.logger,
"Failed to spawn thread to call watch callbacks. Cause: {:?}", err
"Failed to spawn thread to call watch callbacks. Cause: {:?}",
err
);
}
result
@@ -94,18 +86,13 @@ impl WatchCallbackList {
mod tests {
use crate::directory::WatchCallbackList;
use futures::executor::block_on;
use slog::{o, Discard, Logger};
use std::mem;
use std::sync::atomic::{AtomicUsize, Ordering};
use std::sync::Arc;
fn default_watch_callback_list() -> WatchCallbackList {
WatchCallbackList::with_logger(Logger::root(Discard, o!()))
}
#[test]
fn test_watch_event_router_simple() {
let watch_event_router = default_watch_callback_list();
let watch_event_router = WatchCallbackList::default();
let counter: Arc<AtomicUsize> = Default::default();
let counter_clone = counter.clone();
let inc_callback = Box::new(move || {
@@ -132,7 +119,7 @@ mod tests {
#[test]
fn test_watch_event_router_multiple_callback_same_key() {
let watch_event_router = default_watch_callback_list();
let watch_event_router = WatchCallbackList::default();
let counter: Arc<AtomicUsize> = Default::default();
let inc_callback = |inc: usize| {
let counter_clone = counter.clone();
@@ -161,7 +148,7 @@ mod tests {
#[test]
fn test_watch_event_router_multiple_callback_different_key() {
let watch_event_router = default_watch_callback_list();
let watch_event_router = WatchCallbackList::default();
let counter: Arc<AtomicUsize> = Default::default();
let counter_clone = counter.clone();
let inc_callback = Box::new(move || {

131
src/docset.rs
View File

@@ -1,48 +1,58 @@
use crate::common::BitSet;
use crate::fastfield::DeleteBitSet;
use crate::DocId;
use std::borrow::Borrow;
use std::borrow::BorrowMut;
use std::cmp::Ordering;
/// Sentinel value returned when a DocSet has been entirely consumed.
///
/// This is not u32::MAX as one would have expected, due to the lack of SSE2 instructions
/// to compare [u32; 4].
pub const TERMINATED: DocId = std::i32::MAX as u32;
/// Expresses the outcome of a call to `DocSet`'s `.skip_next(...)`.
#[derive(PartialEq, Eq, Debug)]
pub enum SkipResult {
/// target was in the docset
Reached,
/// target was not in the docset, skipping stopped as a greater element was found
OverStep,
/// the docset was entirely consumed without finding the target, nor any
/// element greater than the target.
End,
}
/// Represents an iterable set of sorted doc ids.
pub trait DocSet {
/// Goes to the next element.
///
/// The DocId of the next element is returned.
/// In other words we should always have :
/// ```ignore
/// let doc = docset.advance();
/// assert_eq!(doc, docset.doc());
/// ```
///
/// If we reached the end of the DocSet, TERMINATED should be returned.
///
/// Calling `.advance()` on a terminated DocSet should be supported, and TERMINATED should
/// be returned.
/// TODO Test existing docsets.
fn advance(&mut self) -> DocId;
/// `.advance(...)` needs to be called a first time to point to the correct
/// element.
fn advance(&mut self) -> bool;
/// Advances the DocSet forward until reaching the target, or going to the
/// lowest DocId greater than the target.
/// After skipping, position the iterator in such a way that `.doc()`
/// will return a value greater than or equal to target.
///
/// If the end of the DocSet is reached, TERMINATED is returned.
/// SkipResult expresses whether the `target value` was reached, overstepped,
/// or if the `DocSet` was entirely consumed without finding any value
/// greater or equal to the `target`.
///
/// Calling `.seek(target)` on a terminated DocSet is legal. Implementation
/// of DocSet should support it.
/// WARNING: Calling skip always advances the docset.
/// More specifically, if the docset is already positionned on the target
/// skipping will advance to the next position and return SkipResult::Overstep.
///
/// Calling `seek(TERMINATED)` is also legal and is the normal way to consume a DocSet.
fn seek(&mut self, target: DocId) -> DocId {
let mut doc = self.doc();
debug_assert!(doc <= target);
while doc < target {
doc = self.advance();
/// If `.skip_next()` oversteps, then the docset must be positionned correctly
/// on an existing document. In other words, `.doc()` should return the first document
/// greater than `DocId`.
fn skip_next(&mut self, target: DocId) -> SkipResult {
if !self.advance() {
return SkipResult::End;
}
loop {
match self.doc().cmp(&target) {
Ordering::Less => {
if !self.advance() {
return SkipResult::End;
}
}
Ordering::Equal => return SkipResult::Reached,
Ordering::Greater => return SkipResult::OverStep,
}
}
doc
}
/// Fills a given mutable buffer with the next doc ids from the
@@ -61,38 +71,38 @@ pub trait DocSet {
/// use case where batching. The normal way to
/// go through the `DocId`'s is to call `.advance()`.
fn fill_buffer(&mut self, buffer: &mut [DocId]) -> usize {
if self.doc() == TERMINATED {
return 0;
}
for (i, buffer_val) in buffer.iter_mut().enumerate() {
*buffer_val = self.doc();
if self.advance() == TERMINATED {
return i + 1;
if self.advance() {
*buffer_val = self.doc();
} else {
return i;
}
}
buffer.len()
}
/// Returns the current document
/// Right after creating a new DocSet, the docset points to the first document.
///
/// If the DocSet is empty, .doc() should return `TERMINATED`.
fn doc(&self) -> DocId;
/// Returns a best-effort hint of the
/// length of the docset.
fn size_hint(&self) -> u32;
/// Appends all docs to a `bitset`.
fn append_to_bitset(&mut self, bitset: &mut BitSet) {
while self.advance() {
bitset.insert(self.doc());
}
}
/// Returns the number documents matching.
/// Calling this method consumes the `DocSet`.
fn count(&mut self, delete_bitset: &DeleteBitSet) -> u32 {
let mut count = 0u32;
let mut doc = self.doc();
while doc != TERMINATED {
if !delete_bitset.is_deleted(doc) {
while self.advance() {
if !delete_bitset.is_deleted(self.doc()) {
count += 1u32;
}
doc = self.advance();
}
count
}
@@ -104,42 +114,22 @@ pub trait DocSet {
/// given by `count()`.
fn count_including_deleted(&mut self) -> u32 {
let mut count = 0u32;
let mut doc = self.doc();
while doc != TERMINATED {
while self.advance() {
count += 1u32;
doc = self.advance();
}
count
}
}
impl<'a> DocSet for &'a mut dyn DocSet {
fn advance(&mut self) -> u32 {
(**self).advance()
}
fn seek(&mut self, target: DocId) -> DocId {
(**self).seek(target)
}
fn doc(&self) -> u32 {
(**self).doc()
}
fn size_hint(&self) -> u32 {
(**self).size_hint()
}
}
impl<TDocSet: DocSet + ?Sized> DocSet for Box<TDocSet> {
fn advance(&mut self) -> DocId {
fn advance(&mut self) -> bool {
let unboxed: &mut TDocSet = self.borrow_mut();
unboxed.advance()
}
fn seek(&mut self, target: DocId) -> DocId {
fn skip_next(&mut self, target: DocId) -> SkipResult {
let unboxed: &mut TDocSet = self.borrow_mut();
unboxed.seek(target)
unboxed.skip_next(target)
}
fn doc(&self) -> DocId {
@@ -161,4 +151,9 @@ impl<TDocSet: DocSet + ?Sized> DocSet for Box<TDocSet> {
let unboxed: &mut TDocSet = self.borrow_mut();
unboxed.count_including_deleted()
}
fn append_to_bitset(&mut self, bitset: &mut BitSet) {
let unboxed: &mut TDocSet = self.borrow_mut();
unboxed.append_to_bitset(bitset);
}
}

99
src/error.rs
View File

@@ -2,13 +2,12 @@
use std::io;
use crate::directory::error::{IOError, OpenDirectoryError, OpenReadError, OpenWriteError};
use crate::directory::error::{Incompatibility, LockError};
use crate::fastfield::FastFieldNotAvailableError;
use crate::query;
use crate::{
directory::error::{OpenDirectoryError, OpenReadError, OpenWriteError},
schema,
};
use crate::schema;
use serde_json;
use std::fmt;
use std::path::PathBuf;
use std::sync::PoisonError;
@@ -45,47 +44,44 @@ impl fmt::Debug for DataCorruption {
}
}
/// The library's error enum
#[derive(Debug, Error)]
/// The library's failure based error enum
#[derive(Debug, Fail)]
pub enum TantivyError {
/// Failed to open the directory.
#[error("Failed to open the directory: '{0:?}'")]
OpenDirectoryError(#[from] OpenDirectoryError),
/// Failed to open a file for read.
#[error("Failed to open file for read: '{0:?}'")]
OpenReadError(#[from] OpenReadError),
/// Failed to open a file for write.
#[error("Failed to open file for write: '{0:?}'")]
OpenWriteError(#[from] OpenWriteError),
/// Path does not exist.
#[fail(display = "Path does not exist: '{:?}'", _0)]
PathDoesNotExist(PathBuf),
/// File already exists, this is a problem when we try to write into a new file.
#[fail(display = "File already exists: '{:?}'", _0)]
FileAlreadyExists(PathBuf),
/// Index already exists in this directory
#[error("Index already exists")]
#[fail(display = "Index already exists")]
IndexAlreadyExists,
/// Failed to acquire file lock
#[error("Failed to acquire Lockfile: {0:?}. {1:?}")]
#[fail(display = "Failed to acquire Lockfile: {:?}. {:?}", _0, _1)]
LockFailure(LockError, Option<String>),
/// IO Error.
#[error("An IO error occurred: '{0}'")]
IOError(#[from] io::Error),
#[fail(display = "An IO error occurred: '{}'", _0)]
IOError(#[cause] IOError),
/// Data corruption.
#[error("Data corrupted: '{0:?}'")]
#[fail(display = "{:?}", _0)]
DataCorruption(DataCorruption),
/// A thread holding the locked panicked and poisoned the lock.
#[error("A thread holding the locked panicked and poisoned the lock")]
#[fail(display = "A thread holding the locked panicked and poisoned the lock")]
Poisoned,
/// Invalid argument was passed by the user.
#[error("An invalid argument was passed: '{0}'")]
#[fail(display = "An invalid argument was passed: '{}'", _0)]
InvalidArgument(String),
/// An Error happened in one of the thread.
#[error("An error occurred in a thread: '{0}'")]
#[fail(display = "An error occurred in a thread: '{}'", _0)]
ErrorInThread(String),
/// An Error appeared related to the schema.
#[error("Schema error: '{0}'")]
#[fail(display = "Schema error: '{}'", _0)]
SchemaError(String),
/// System error. (e.g.: We failed spawning a new thread)
#[error("System error.'{0}'")]
#[fail(display = "System error.'{}'", _0)]
SystemError(String),
/// Index incompatible with current version of tantivy
#[error("{0:?}")]
#[fail(display = "{:?}", _0)]
IncompatibleIndex(Incompatibility),
}
@@ -94,17 +90,31 @@ impl From<DataCorruption> for TantivyError {
TantivyError::DataCorruption(data_corruption)
}
}
impl From<FastFieldNotAvailableError> for TantivyError {
fn from(fastfield_error: FastFieldNotAvailableError) -> TantivyError {
TantivyError::SchemaError(format!("{}", fastfield_error))
}
}
impl From<LockError> for TantivyError {
fn from(lock_error: LockError) -> TantivyError {
TantivyError::LockFailure(lock_error, None)
}
}
impl From<IOError> for TantivyError {
fn from(io_error: IOError) -> TantivyError {
TantivyError::IOError(io_error)
}
}
impl From<io::Error> for TantivyError {
fn from(io_error: io::Error) -> TantivyError {
TantivyError::IOError(io_error.into())
}
}
impl From<query::QueryParserError> for TantivyError {
fn from(parsing_error: query::QueryParserError) -> TantivyError {
TantivyError::InvalidArgument(format!("Query is invalid. {:?}", parsing_error))
@@ -117,12 +127,49 @@ impl<Guard> From<PoisonError<Guard>> for TantivyError {
}
}
impl From<OpenReadError> for TantivyError {
fn from(error: OpenReadError) -> TantivyError {
match error {
OpenReadError::FileDoesNotExist(filepath) => TantivyError::PathDoesNotExist(filepath),
OpenReadError::IOError(io_error) => TantivyError::IOError(io_error),
OpenReadError::IncompatibleIndex(incompatibility) => {
TantivyError::IncompatibleIndex(incompatibility)
}
}
}
}
impl From<schema::DocParsingError> for TantivyError {
fn from(error: schema::DocParsingError) -> TantivyError {
TantivyError::InvalidArgument(format!("Failed to parse document {:?}", error))
}
}
impl From<OpenWriteError> for TantivyError {
fn from(error: OpenWriteError) -> TantivyError {
match error {
OpenWriteError::FileAlreadyExists(filepath) => {
TantivyError::FileAlreadyExists(filepath)
}
OpenWriteError::IOError(io_error) => TantivyError::IOError(io_error),
}
}
}
impl From<OpenDirectoryError> for TantivyError {
fn from(error: OpenDirectoryError) -> TantivyError {
match error {
OpenDirectoryError::DoesNotExist(directory_path) => {
TantivyError::PathDoesNotExist(directory_path)
}
OpenDirectoryError::NotADirectory(directory_path) => {
TantivyError::InvalidArgument(format!("{:?} is not a directory", directory_path))
}
OpenDirectoryError::IoError(err) => TantivyError::IOError(IOError::from(err)),
}
}
}
impl From<serde_json::Error> for TantivyError {
fn from(error: serde_json::Error) -> TantivyError {
let io_err = io::Error::from(error);

2
src/fastfield/bytes/mod.rs
View File

@@ -15,7 +15,7 @@ mod tests {
let field = schema_builder.add_bytes_field("bytesfield");
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(field=>vec![0u8, 1, 2, 3]));
index_writer.add_document(doc!(field=>vec![]));
index_writer.add_document(doc!(field=>vec![255u8]));

73
src/fastfield/delete.rs
View File

@@ -9,8 +9,6 @@ use std::io::Write;
/// Write a delete `BitSet`
///
/// where `delete_bitset` is the set of deleted `DocId`.
/// Warning: this function does not call terminate. The caller is in charge of
/// closing the writer properly.
pub fn write_delete_bitset(
delete_bitset: &BitSet,
max_doc: u32,
@@ -44,24 +42,6 @@ pub struct DeleteBitSet {
}
impl DeleteBitSet {
#[cfg(test)]
pub(crate) fn for_test(docs: &[DocId], max_doc: u32) -> DeleteBitSet {
use crate::directory::{Directory, RAMDirectory, TerminatingWrite};
use std::path::Path;
assert!(docs.iter().all(|&doc| doc < max_doc));
let mut bitset = BitSet::with_max_value(max_doc);
for &doc in docs {
bitset.insert(doc);
}
let mut directory = RAMDirectory::create();
let path = Path::new("dummydeletebitset");
let mut wrt = directory.open_write(path).unwrap();
write_delete_bitset(&bitset, max_doc, &mut wrt).unwrap();
wrt.terminate().unwrap();
let source = directory.open_read(path).unwrap();
Self::open(source)
}
/// Opens a delete bitset given its data source.
pub fn open(data: ReadOnlySource) -> DeleteBitSet {
let num_deleted: usize = data
@@ -103,35 +83,42 @@ impl HasLen for DeleteBitSet {
#[cfg(test)]
mod tests {
use super::DeleteBitSet;
use crate::common::HasLen;
use super::*;
use crate::directory::*;
use std::path::PathBuf;
#[test]
fn test_delete_bitset_empty() {
let delete_bitset = DeleteBitSet::for_test(&[], 10);
for doc in 0..10 {
assert_eq!(delete_bitset.is_deleted(doc), !delete_bitset.is_alive(doc));
fn test_delete_bitset_helper(bitset: &BitSet, max_doc: u32) {
let test_path = PathBuf::from("test");
let mut directory = RAMDirectory::create();
{
let mut writer = directory.open_write(&*test_path).unwrap();
write_delete_bitset(bitset, max_doc, &mut writer).unwrap();
writer.terminate().unwrap();
}
assert_eq!(delete_bitset.len(), 0);
let source = directory.open_read(&test_path).unwrap();
let delete_bitset = DeleteBitSet::open(source);
for doc in 0..max_doc {
assert_eq!(bitset.contains(doc), delete_bitset.is_deleted(doc as DocId));
}
assert_eq!(delete_bitset.len(), bitset.len());
}
#[test]
fn test_delete_bitset() {
let delete_bitset = DeleteBitSet::for_test(&[1, 9], 10);
assert!(delete_bitset.is_alive(0));
assert!(delete_bitset.is_deleted(1));
assert!(delete_bitset.is_alive(2));
assert!(delete_bitset.is_alive(3));
assert!(delete_bitset.is_alive(4));
assert!(delete_bitset.is_alive(5));
assert!(delete_bitset.is_alive(6));
assert!(delete_bitset.is_alive(6));
assert!(delete_bitset.is_alive(7));
assert!(delete_bitset.is_alive(8));
assert!(delete_bitset.is_deleted(9));
for doc in 0..10 {
assert_eq!(delete_bitset.is_deleted(doc), !delete_bitset.is_alive(doc));
{
let mut bitset = BitSet::with_max_value(10);
bitset.insert(1);
bitset.insert(9);
test_delete_bitset_helper(&bitset, 10);
}
{
let mut bitset = BitSet::with_max_value(8);
bitset.insert(1);
bitset.insert(2);
bitset.insert(3);
bitset.insert(5);
bitset.insert(7);
test_delete_bitset_helper(&bitset, 8);
}
assert_eq!(delete_bitset.len(), 2);
}
}

4
src/fastfield/error.rs
View File

@@ -4,8 +4,8 @@ use std::result;
/// `FastFieldNotAvailableError` is returned when the
/// user requested for a fast field reader, and the field was not
/// defined in the schema as a fast field.
#[derive(Debug, Error)]
#[error("Fast field not available: '{field_name:?}'")]
#[derive(Debug, Fail)]
#[fail(display = "Fast field not available: '{:?}'", field_name)]
pub struct FastFieldNotAvailableError {
field_name: String,
}

4
src/fastfield/mod.rs
View File

@@ -474,7 +474,7 @@ mod tests {
let date_field = schema_builder.add_date_field("date", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.set_merge_policy(Box::new(NoMergePolicy));
index_writer.add_document(doc!(date_field =>crate::chrono::prelude::Utc::now()));
index_writer.commit().unwrap();
@@ -511,7 +511,7 @@ mod tests {
);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.set_merge_policy(Box::new(NoMergePolicy));
index_writer.add_document(doc!(
date_field => crate::DateTime::from_u64(1i64.to_u64()),

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

@@ -25,7 +25,7 @@ mod tests {
);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(field=>1u64, field=>3u64));
index_writer.add_document(doc!());
index_writer.add_document(doc!(field=>4u64));
@@ -64,7 +64,7 @@ mod tests {
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_for_tests().unwrap();
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),
@@ -186,7 +186,7 @@ mod tests {
);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(field=> 1i64, field => 3i64));
index_writer.add_document(doc!());
index_writer.add_document(doc!(field=> -4i64));
@@ -221,7 +221,7 @@ mod tests {
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_for_tests().unwrap();
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())));
}

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

@@ -74,7 +74,7 @@ mod tests {
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index
.writer_for_tests()
.writer_with_num_threads(1, 30_000_000)
.expect("Failed to create index writer.");
index_writer.add_document(doc!(
facet_field => Facet::from("/category/cat2"),

5
src/fastfield/multivalued/writer.rs
View File

@@ -6,6 +6,7 @@ use crate::schema::{Document, Field};
use crate::termdict::TermOrdinal;
use crate::DocId;
use fnv::FnvHashMap;
use itertools::Itertools;
use std::io;
/// Writer for multi-valued (as in, more than one value per document)
@@ -150,8 +151,8 @@ impl MultiValueIntFastFieldWriter {
}
}
None => {
let val_min_max = crate::common::minmax(self.vals.iter().cloned());
let (val_min, val_max) = val_min_max.unwrap_or((0u64, 0u64));
let val_min_max = self.vals.iter().cloned().minmax();
let (val_min, val_max) = val_min_max.into_option().unwrap_or((0u64, 0u64));
value_serializer =
serializer.new_u64_fast_field_with_idx(self.field, val_min, val_max, 1)?;
for &val in &self.vals {

2
src/fieldnorm/mod.rs
View File

@@ -21,7 +21,7 @@ mod reader;
mod serializer;
mod writer;
pub use self::reader::{FieldNormReader, FieldNormReaders};
pub use self::reader::FieldNormReader;
pub use self::serializer::FieldNormsSerializer;
pub use self::writer::FieldNormsWriter;

67
src/fieldnorm/reader.rs
View File

@@ -1,41 +1,6 @@
use super::{fieldnorm_to_id, id_to_fieldnorm};
use crate::common::CompositeFile;
use crate::directory::ReadOnlySource;
use crate::schema::Field;
use crate::space_usage::PerFieldSpaceUsage;
use crate::DocId;
use std::sync::Arc;
/// Reader for the fieldnorm (for each document, the number of tokens indexed in the
/// field) of all indexed fields in the index.
///
/// Each fieldnorm is approximately compressed over one byte. We refer to this byte as
/// `fieldnorm_id`.
/// The mapping from `fieldnorm` to `fieldnorm_id` is given by monotonic.
#[derive(Clone)]
pub struct FieldNormReaders {
data: Arc<CompositeFile>,
}
impl FieldNormReaders {
/// Creates a field norm reader.
pub fn open(source: ReadOnlySource) -> crate::Result<FieldNormReaders> {
let data = CompositeFile::open(&source)?;
Ok(FieldNormReaders {
data: Arc::new(data),
})
}
/// Returns the FieldNormReader for a specific field.
pub fn get_field(&self, field: Field) -> Option<FieldNormReader> {
self.data.open_read(field).map(FieldNormReader::open)
}
/// Return a break down of the space usage per field.
pub fn space_usage(&self) -> PerFieldSpaceUsage {
self.data.space_usage()
}
}
/// Reads the fieldnorm associated to a document.
/// The fieldnorm represents the length associated to
@@ -54,7 +19,6 @@ impl FieldNormReaders {
/// Apart from compression, this scale also makes it possible to
/// precompute computationally expensive functions of the fieldnorm
/// in a very short array.
#[derive(Clone)]
pub struct FieldNormReader {
data: ReadOnlySource,
}
@@ -65,11 +29,6 @@ impl FieldNormReader {
FieldNormReader { data }
}
/// Returns the number of documents in this segment.
pub fn num_docs(&self) -> u32 {
self.data.len() as u32
}
/// Returns the `fieldnorm` associated to a doc id.
/// The fieldnorm is a value approximating the number
/// of tokens in a given field of the `doc_id`.
@@ -103,12 +62,13 @@ impl FieldNormReader {
pub fn fieldnorm_to_id(fieldnorm: u32) -> u8 {
fieldnorm_to_id(fieldnorm)
}
}
#[cfg(test)]
pub fn for_test(field_norms: &[u32]) -> FieldNormReader {
#[cfg(test)]
impl From<Vec<u32>> for FieldNormReader {
fn from(field_norms: Vec<u32>) -> FieldNormReader {
let field_norms_id = field_norms
.iter()
.cloned()
.into_iter()
.map(FieldNormReader::fieldnorm_to_id)
.collect::<Vec<u8>>();
let field_norms_data = ReadOnlySource::from(field_norms_id);
@@ -117,20 +77,3 @@ impl FieldNormReader {
}
}
}
#[cfg(test)]
mod tests {
use crate::fieldnorm::FieldNormReader;
#[test]
fn test_from_fieldnorms_array() {
let fieldnorms = &[1, 2, 3, 4, 1_000_000];
let fieldnorm_reader = FieldNormReader::for_test(fieldnorms);
assert_eq!(fieldnorm_reader.num_docs(), 5);
assert_eq!(fieldnorm_reader.fieldnorm(0), 1);
assert_eq!(fieldnorm_reader.fieldnorm(1), 2);
assert_eq!(fieldnorm_reader.fieldnorm(2), 3);
assert_eq!(fieldnorm_reader.fieldnorm(3), 4);
assert_eq!(fieldnorm_reader.fieldnorm(4), 983_064);
}
}

3
src/fieldnorm/writer.rs
View File

@@ -78,12 +78,11 @@ impl FieldNormsWriter {
}
/// Serialize the seen fieldnorm values to the serializer for all fields.
pub fn serialize(&self, mut fieldnorms_serializer: FieldNormsSerializer) -> io::Result<()> {
pub fn serialize(&self, fieldnorms_serializer: &mut FieldNormsSerializer) -> io::Result<()> {
for &field in self.fields.iter() {
let fieldnorm_values: &[u8] = &self.fieldnorms_buffer[field.field_id() as usize][..];
fieldnorms_serializer.serialize_field(field, fieldnorm_values)?;
}
fieldnorms_serializer.close()?;
Ok(())
}
}

130
src/indexer/index_writer.rs
View File

@@ -10,7 +10,7 @@ use crate::core::SegmentMeta;
use crate::core::SegmentReader;
use crate::directory::TerminatingWrite;
use crate::directory::{DirectoryLock, GarbageCollectionResult};
use crate::docset::{DocSet, TERMINATED};
use crate::docset::DocSet;
use crate::error::TantivyError;
use crate::fastfield::write_delete_bitset;
use crate::indexer::delete_queue::{DeleteCursor, DeleteQueue};
@@ -27,7 +27,6 @@ use crate::Opstamp;
use crossbeam::channel;
use futures::executor::block_on;
use futures::future::Future;
use slog::{error, info, Logger};
use smallvec::smallvec;
use smallvec::SmallVec;
use std::mem;
@@ -113,15 +112,15 @@ fn compute_deleted_bitset(
if let Some(mut docset) =
inverted_index.read_postings(&delete_op.term, IndexRecordOption::Basic)
{
let mut deleted_doc = docset.doc();
while deleted_doc != TERMINATED {
while docset.advance() {
let deleted_doc = docset.doc();
if deleted_doc < limit_doc {
delete_bitset.insert(deleted_doc);
might_have_changed = true;
}
deleted_doc = docset.advance();
}
}
delete_cursor.advance();
}
Ok(might_have_changed)
@@ -156,8 +155,6 @@ pub(crate) fn advance_deletes(
None => BitSet::with_max_value(max_doc),
};
let num_deleted_docs_before = segment.meta().num_deleted_docs();
compute_deleted_bitset(
&mut delete_bitset,
&segment_reader,
@@ -167,8 +164,6 @@ pub(crate) fn advance_deletes(
)?;
// TODO optimize
// It should be possible to do something smarter by manipulation bitsets directly
// to compute this union.
if let Some(seg_delete_bitset) = segment_reader.delete_bitset() {
for doc in 0u32..max_doc {
if seg_delete_bitset.is_deleted(doc) {
@@ -177,9 +172,8 @@ pub(crate) fn advance_deletes(
}
}
let num_deleted_docs: u32 = delete_bitset.len() as u32;
if num_deleted_docs > num_deleted_docs_before {
// There are new deletes. We need to write a new delete file.
let num_deleted_docs = delete_bitset.len();
if num_deleted_docs > 0 {
segment = segment.with_delete_meta(num_deleted_docs as u32, target_opstamp);
let mut delete_file = segment.open_write(SegmentComponent::DELETE)?;
write_delete_bitset(&delete_bitset, max_doc, &mut delete_file)?;
@@ -196,21 +190,20 @@ fn index_documents(
grouped_document_iterator: &mut dyn Iterator<Item = OperationGroup>,
segment_updater: &mut SegmentUpdater,
mut delete_cursor: DeleteCursor,
logger: &Logger,
) -> crate::Result<bool> {
let schema = segment.schema();
info!(logger, "segment-index"; "stage"=>"start");
let mut segment_writer = SegmentWriter::for_segment(memory_budget, segment.clone(), &schema)?;
let mut buffer_limit_reached = false;
for document_group in grouped_document_iterator {
for doc in document_group {
segment_writer.add_document(doc, &schema)?;
}
let mem_usage = segment_writer.mem_usage();
if mem_usage >= memory_budget - MARGIN_IN_BYTES {
buffer_limit_reached = true;
info!(
"Buffer limit reached, flushing segment with maxdoc={}.",
segment_writer.max_doc()
);
break;
}
}
@@ -230,14 +223,6 @@ fn index_documents(
let segment_with_max_doc = segment.with_max_doc(max_doc);
let last_docstamp: Opstamp = *(doc_opstamps.last().unwrap());
info!(
logger,
"segment-index";
"stage" => "serialize",
"cause" => if buffer_limit_reached { "buffer-limit" } else { "commit" },
"maxdoc" => max_doc,
"last_docstamp" => last_docstamp
);
let delete_bitset_opt = apply_deletes(
&segment_with_max_doc,
@@ -251,18 +236,7 @@ fn index_documents(
delete_cursor,
delete_bitset_opt,
);
info!(
logger,
"segment-index";
"stage" => "publish",
);
block_on(segment_updater.schedule_add_segment(segment_entry))?;
info!(
logger,
"segment-index";
"stage" => "end",
);
Ok(true)
}
@@ -365,20 +339,14 @@ impl IndexWriter {
Ok(index_writer)
}
pub(crate) fn logger(&self) -> &Logger {
self.index.logger()
}
fn drop_sender(&mut self) {
let (sender, _receiver) = channel::bounded(1);
self.operation_sender = sender;
mem::replace(&mut self.operation_sender, sender);
}
/// If there are some merging threads, blocks until they all finish their work and
/// then drop the `IndexWriter`.
pub fn wait_merging_threads(mut self) -> crate::Result<()> {
info!(self.logger(), "wait-merge-threads"; "stage"=>"start");
// this will stop the indexing thread,
// dropping the last reference to the segment_updater.
self.drop_sender();
@@ -399,9 +367,9 @@ impl IndexWriter {
.map_err(|_| TantivyError::ErrorInThread("Failed to join merging thread.".into()));
if let Err(ref e) = result {
error!(self.logger(), "some merge thread failed"; "cause"=>e.to_string());
error!("Some merging thread failed {:?}", e);
}
info!(self.logger(), "wait-merge-threads"; "stage"=>"stop");
result
}
@@ -461,16 +429,12 @@ impl IndexWriter {
return Ok(());
}
let segment = index.new_segment();
let segment_id = segment.id();
index_documents(
mem_budget,
segment,
&mut document_iterator,
&mut segment_updater,
delete_cursor.clone(),
&index
.logger()
.new(slog::o!("segment"=>segment_id.to_string())),
)?;
}
})?;
@@ -567,7 +531,6 @@ impl IndexWriter {
/// when no documents are remaining.
///
/// Returns the former segment_ready channel.
#[allow(unused_must_use)]
fn recreate_document_channel(&mut self) -> OperationReceiver {
let (document_sender, document_receiver): (OperationSender, OperationReceiver) =
channel::bounded(PIPELINE_MAX_SIZE_IN_DOCS);
@@ -584,10 +547,7 @@ impl IndexWriter {
///
/// The opstamp at the last commit is returned.
pub fn rollback(&mut self) -> crate::Result<Opstamp> {
info!(
self.logger(),
"Rolling back to opstamp {}", self.committed_opstamp
);
info!("Rolling back to opstamp {}", self.committed_opstamp);
// marks the segment updater as killed. From now on, all
// segment updates will be ignored.
self.segment_updater.kill();
@@ -610,7 +570,7 @@ impl IndexWriter {
//
// This will drop the document queue, and the thread
// should terminate.
*self = new_index_writer;
mem::replace(self, new_index_writer);
// Drains the document receiver pipeline :
// Workers don't need to index the pending documents.
@@ -644,8 +604,6 @@ impl IndexWriter {
/// using this API.
/// See [`PreparedCommit::set_payload()`](PreparedCommit.html)
pub fn prepare_commit(&mut self) -> crate::Result<PreparedCommit> {
let logger = self.logger().clone();
// Here, because we join all of the worker threads,
// all of the segment update for this commit have been
// sent.
@@ -656,10 +614,7 @@ impl IndexWriter {
//
// This will move uncommitted segments to the state of
// committed segments.
let commit_opstamp = self.stamper.stamp();
info!(logger, "prepare-commit"; "opstamp" => commit_opstamp);
info!("Preparing commit");
// this will drop the current document channel
// and recreate a new one.
@@ -675,8 +630,9 @@ impl IndexWriter {
self.add_indexing_worker()?;
}
let commit_opstamp = self.stamper.stamp();
let prepared_commit = PreparedCommit::new(self, commit_opstamp);
info!(logger, "Prepared commit {}", commit_opstamp);
info!("Prepared commit {}", commit_opstamp);
Ok(prepared_commit)
}
@@ -838,7 +794,7 @@ mod tests {
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 index_writer = index.writer_for_tests().unwrap();
let 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")),
@@ -847,46 +803,6 @@ mod tests {
assert_eq!(batch_opstamp1, 2u64);
}
#[test]
fn test_no_need_to_rewrite_delete_file_if_no_new_deletes() {
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_for_tests().unwrap();
index_writer.add_document(doc!(text_field => "hello1"));
index_writer.add_document(doc!(text_field => "hello2"));
assert!(index_writer.commit().is_ok());
let reader = index.reader().unwrap();
let searcher = reader.searcher();
assert_eq!(searcher.segment_readers().len(), 1);
assert_eq!(searcher.segment_reader(0u32).num_deleted_docs(), 0);
index_writer.delete_term(Term::from_field_text(text_field, "hello1"));
assert!(index_writer.commit().is_ok());
assert!(reader.reload().is_ok());
let searcher = reader.searcher();
assert_eq!(searcher.segment_readers().len(), 1);
assert_eq!(searcher.segment_reader(0u32).num_deleted_docs(), 1);
let previous_delete_opstamp = index.load_metas().unwrap().segments[0].delete_opstamp();
// All docs containing hello1 have been already removed.
// We should not update the delete meta.
index_writer.delete_term(Term::from_field_text(text_field, "hello1"));
assert!(index_writer.commit().is_ok());
assert!(reader.reload().is_ok());
let searcher = reader.searcher();
assert_eq!(searcher.segment_readers().len(), 1);
assert_eq!(searcher.segment_reader(0u32).num_deleted_docs(), 1);
let after_delete_opstamp = index.load_metas().unwrap().segments[0].delete_opstamp();
assert_eq!(after_delete_opstamp, previous_delete_opstamp);
}
#[test]
fn test_ordered_batched_operations() {
// * one delete for `doc!(field=>"a")`
@@ -902,7 +818,7 @@ mod tests {
.reload_policy(ReloadPolicy::Manual)
.try_into()
.unwrap();
let mut index_writer = index.writer_for_tests().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![
@@ -964,8 +880,8 @@ mod tests {
fn test_lockfile_already_exists_error_msg() {
let schema_builder = schema::Schema::builder();
let index = Index::create_in_ram(schema_builder.build());
let _index_writer = index.writer_for_tests().unwrap();
match index.writer_for_tests() {
let _index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
match index.writer_with_num_threads(1, 3_000_000) {
Err(err) => {
let err_msg = err.to_string();
assert!(err_msg.contains("already an `IndexWriter`"));
@@ -1299,7 +1215,7 @@ mod tests {
let idfield = schema_builder.add_text_field("id", STRING);
schema_builder.add_text_field("optfield", STRING);
let index = Index::create_in_ram(schema_builder.build());
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(idfield=>"myid"));
let commit = index_writer.commit();
assert!(commit.is_ok());

52
src/indexer/log_merge_policy.rs
View File

@@ -54,6 +54,10 @@ impl LogMergePolicy {
impl MergePolicy for LogMergePolicy {
fn compute_merge_candidates(&self, segments: &[SegmentMeta]) -> Vec<MergeCandidate> {
if segments.is_empty() {
return Vec::new();
}
let mut size_sorted_tuples = segments
.iter()
.map(SegmentMeta::num_docs)
@@ -63,35 +67,27 @@ impl MergePolicy for LogMergePolicy {
size_sorted_tuples.sort_by(|x, y| y.1.cmp(&(x.1)));
if size_sorted_tuples.len() <= 1 {
return Vec::new();
}
let size_sorted_log_tuples: Vec<_> = size_sorted_tuples
.into_iter()
.map(|(ind, num_docs)| (ind, f64::from(self.clip_min_size(num_docs)).log2()))
.collect();
if let Some(&(first_ind, first_score)) = size_sorted_log_tuples.first() {
let mut current_max_log_size = first_score;
let mut levels = vec![vec![first_ind]];
for &(ind, score) in (&size_sorted_log_tuples).iter().skip(1) {
if score < (current_max_log_size - self.level_log_size) {
current_max_log_size = score;
levels.push(Vec::new());
}
levels.last_mut().unwrap().push(ind);
let (first_ind, first_score) = size_sorted_log_tuples[0];
let mut current_max_log_size = first_score;
let mut levels = vec![vec![first_ind]];
for &(ind, score) in (&size_sorted_log_tuples).iter().skip(1) {
if score < (current_max_log_size - self.level_log_size) {
current_max_log_size = score;
levels.push(Vec::new());
}
levels
.iter()
.filter(|level| level.len() >= self.min_merge_size)
.map(|ind_vec| {
MergeCandidate(ind_vec.iter().map(|&ind| segments[ind].id()).collect())
})
.collect()
} else {
return vec![];
levels.last_mut().unwrap().push(ind);
}
levels
.iter()
.filter(|level| level.len() >= self.min_merge_size)
.map(|ind_vec| MergeCandidate(ind_vec.iter().map(|&ind| segments[ind].id()).collect()))
.collect()
}
}
@@ -183,7 +179,6 @@ mod tests {
let result_list = test_merge_policy().compute_merge_candidates(&test_input);
assert_eq!(result_list.len(), 2);
}
#[test]
fn test_log_merge_policy_small_segments() {
// segments under min_layer_size are merged together
@@ -199,17 +194,6 @@ mod tests {
assert_eq!(result_list.len(), 1);
}
#[test]
fn test_log_merge_policy_all_segments_too_large_to_merge() {
let eight_large_segments: Vec<SegmentMeta> =
std::iter::repeat_with(|| create_random_segment_meta(100_001))
.take(8)
.collect();
assert!(test_merge_policy()
.compute_merge_candidates(&eight_large_segments)
.is_empty());
}
#[test]
fn test_large_merge_segments() {
let test_input = vec![

284
src/indexer/merger.rs
View File

@@ -2,37 +2,37 @@ use crate::common::MAX_DOC_LIMIT;
use crate::core::Segment;
use crate::core::SegmentReader;
use crate::core::SerializableSegment;
use crate::docset::{DocSet, TERMINATED};
use crate::docset::DocSet;
use crate::fastfield::BytesFastFieldReader;
use crate::fastfield::DeleteBitSet;
use crate::fastfield::FastFieldReader;
use crate::fastfield::FastFieldSerializer;
use crate::fastfield::MultiValueIntFastFieldReader;
use crate::fieldnorm::FieldNormReader;
use crate::fieldnorm::FieldNormsSerializer;
use crate::fieldnorm::FieldNormsWriter;
use crate::fieldnorm::{FieldNormReader, FieldNormReaders};
use crate::indexer::SegmentSerializer;
use crate::postings::InvertedIndexSerializer;
use crate::postings::Postings;
use crate::postings::{InvertedIndexSerializer, SegmentPostings};
use crate::schema::Cardinality;
use crate::schema::FieldType;
use crate::schema::{Field, Schema};
use crate::store::StoreWriter;
use crate::termdict::TermMerger;
use crate::termdict::TermOrdinal;
use crate::{DocId, InvertedIndexReader, SegmentComponent};
use crate::DocId;
use itertools::Itertools;
use std::cmp;
use std::collections::HashMap;
use std::sync::Arc;
fn compute_total_num_tokens(readers: &[SegmentReader], field: Field) -> crate::Result<u64> {
fn compute_total_num_tokens(readers: &[SegmentReader], field: Field) -> u64 {
let mut total_tokens = 0u64;
let mut count: [usize; 256] = [0; 256];
for reader in readers {
if reader.has_deletes() {
// if there are deletes, then we use an approximation
// using the fieldnorm
let fieldnorms_reader = reader.get_fieldnorms_reader(field)?;
let fieldnorms_reader = reader.get_fieldnorms_reader(field);
for doc in reader.doc_ids_alive() {
let fieldnorm_id = fieldnorms_reader.fieldnorm_id(doc);
count[fieldnorm_id as usize] += 1;
@@ -41,7 +41,7 @@ fn compute_total_num_tokens(readers: &[SegmentReader], field: Field) -> crate::R
total_tokens += reader.inverted_index(field).total_num_tokens();
}
}
Ok(total_tokens
total_tokens
+ count
.iter()
.cloned()
@@ -49,7 +49,7 @@ fn compute_total_num_tokens(readers: &[SegmentReader], field: Field) -> crate::R
.map(|(fieldnorm_ord, count)| {
count as u64 * u64::from(FieldNormReader::id_to_fieldnorm(fieldnorm_ord as u8))
})
.sum::<u64>())
.sum::<u64>()
}
pub struct IndexMerger {
@@ -70,11 +70,11 @@ fn compute_min_max_val(
Some(delete_bitset) => {
// some deleted documents,
// we need to recompute the max / min
crate::common::minmax(
(0..max_doc)
.filter(|doc_id| delete_bitset.is_alive(*doc_id))
.map(|doc_id| u64_reader.get(doc_id)),
)
(0..max_doc)
.filter(|doc_id| delete_bitset.is_alive(*doc_id))
.map(|doc_id| u64_reader.get(doc_id))
.minmax()
.into_option()
}
None => {
// no deleted documents,
@@ -168,14 +168,14 @@ impl IndexMerger {
fn write_fieldnorms(
&self,
mut fieldnorms_serializer: FieldNormsSerializer,
fieldnorms_serializer: &mut FieldNormsSerializer,
) -> crate::Result<()> {
let fields = FieldNormsWriter::fields_with_fieldnorm(&self.schema);
let mut fieldnorms_data = Vec::with_capacity(self.max_doc as usize);
for field in fields {
fieldnorms_data.clear();
for reader in &self.readers {
let fieldnorms_reader = reader.get_fieldnorms_reader(field)?;
let fieldnorms_reader = reader.get_fieldnorms_reader(field);
for doc_id in reader.doc_ids_alive() {
let fieldnorm_id = fieldnorms_reader.fieldnorm_id(doc_id);
fieldnorms_data.push(fieldnorm_id);
@@ -183,7 +183,6 @@ impl IndexMerger {
}
fieldnorms_serializer.serialize_field(field, &fieldnorms_data[..])?;
}
fieldnorms_serializer.close()?;
Ok(())
}
@@ -494,11 +493,10 @@ impl IndexMerger {
indexed_field: Field,
field_type: &FieldType,
serializer: &mut InvertedIndexSerializer,
fieldnorm_reader: Option<FieldNormReader>,
) -> crate::Result<Option<TermOrdinalMapping>> {
let mut positions_buffer: Vec<u32> = Vec::with_capacity(1_000);
let mut delta_computer = DeltaComputer::new();
let field_readers: Vec<Arc<InvertedIndexReader>> = self
let field_readers = self
.readers
.iter()
.map(|reader| reader.inverted_index(indexed_field))
@@ -541,7 +539,7 @@ impl IndexMerger {
// The total number of tokens will only be exact when there has been no deletes.
//
// Otherwise, we approximate by removing deleted documents proportionally.
let total_num_tokens: u64 = compute_total_num_tokens(&self.readers, indexed_field)?;
let total_num_tokens: u64 = compute_total_num_tokens(&self.readers, indexed_field);
// Create the total list of doc ids
// by stacking the doc ids from the different segment.
@@ -553,8 +551,7 @@ impl IndexMerger {
// - Segment 2's doc ids become [seg0.max_doc + seg1.max_doc,
// seg0.max_doc + seg1.max_doc + seg2.max_doc]
// ...
let mut field_serializer =
serializer.new_field(indexed_field, total_num_tokens, fieldnorm_reader)?;
let mut field_serializer = serializer.new_field(indexed_field, total_num_tokens)?;
let field_entry = self.schema.get_field_entry(indexed_field);
@@ -564,78 +561,84 @@ impl IndexMerger {
indexed. Have you modified the schema?",
);
let mut segment_postings_containing_the_term: Vec<(usize, SegmentPostings)> = vec![];
while merged_terms.advance() {
segment_postings_containing_the_term.clear();
let term_bytes: &[u8] = merged_terms.key();
let mut total_doc_freq = 0;
// Let's compute the list of non-empty posting lists
for heap_item in merged_terms.current_kvs() {
let segment_ord = heap_item.segment_ord;
let term_info = heap_item.streamer.value();
let segment_reader = &self.readers[heap_item.segment_ord];
let inverted_index: &InvertedIndexReader = &*field_readers[segment_ord];
let segment_postings =
inverted_index.read_postings_from_terminfo(term_info, segment_postings_option);
let delete_bitset_opt = segment_reader.delete_bitset();
let doc_freq = if let Some(delete_bitset) = delete_bitset_opt {
segment_postings.doc_freq_given_deletes(delete_bitset)
} else {
segment_postings.doc_freq()
};
if doc_freq > 0u32 {
total_doc_freq += doc_freq;
segment_postings_containing_the_term.push((segment_ord, segment_postings));
}
}
let segment_postings: Vec<_> = merged_terms
.current_kvs()
.iter()
.flat_map(|heap_item| {
let segment_ord = heap_item.segment_ord;
let term_info = heap_item.streamer.value();
let segment_reader = &self.readers[heap_item.segment_ord];
let inverted_index = segment_reader.inverted_index(indexed_field);
let mut segment_postings = inverted_index
.read_postings_from_terminfo(term_info, segment_postings_option);
while segment_postings.advance() {
if !segment_reader.is_deleted(segment_postings.doc()) {
return Some((segment_ord, segment_postings));
}
}
None
})
.collect();
// At this point, `segment_postings` contains the posting list
// of all of the segments containing the given term (and that are non-empty)
// of all of the segments containing the given term.
//
// These segments are non-empty and advance has already been called.
if total_doc_freq == 0u32 {
// All docs that used to contain the term have been deleted. The `term` will be
// entirely removed.
continue;
}
if !segment_postings.is_empty() {
// If not, the `term` will be entirely removed.
let to_term_ord = field_serializer.new_term(term_bytes, total_doc_freq)?;
// We know that there is at least one document containing
// the term, so we add it.
let to_term_ord = field_serializer.new_term(term_bytes)?;
if let Some(ref mut term_ord_mapping) = term_ord_mapping_opt {
for (segment_ord, from_term_ord) in merged_terms.matching_segments() {
term_ord_mapping.register_from_to(segment_ord, from_term_ord, to_term_ord);
}
}
// We can now serialize this postings, by pushing each document to the
// postings serializer.
for (segment_ord, mut segment_postings) in
segment_postings_containing_the_term.drain(..)
{
let old_to_new_doc_id = &merged_doc_id_map[segment_ord];
let mut doc = segment_postings.doc();
while doc != TERMINATED {
// deleted doc are skipped as they do not have a `remapped_doc_id`.
if let Some(remapped_doc_id) = old_to_new_doc_id[doc as usize] {
// we make sure to only write the term iff
// there is at least one document.
let term_freq = segment_postings.term_freq();
segment_postings.positions(&mut positions_buffer);
let delta_positions = delta_computer.compute_delta(&positions_buffer);
field_serializer.write_doc(remapped_doc_id, term_freq, delta_positions)?;
if let Some(ref mut term_ord_mapping) = term_ord_mapping_opt {
for (segment_ord, from_term_ord) in merged_terms.matching_segments() {
term_ord_mapping.register_from_to(segment_ord, from_term_ord, to_term_ord);
}
doc = segment_postings.advance();
}
}
// closing the term.
field_serializer.close_term()?;
// We can now serialize this postings, by pushing each document to the
// postings serializer.
for (segment_ord, mut segment_postings) in segment_postings {
let old_to_new_doc_id = &merged_doc_id_map[segment_ord];
loop {
let doc = segment_postings.doc();
// `.advance()` has been called once before the loop.
//
// It was required to make sure we only consider segments
// that effectively contain at least one non-deleted document
// and remove terms that do not have documents associated.
//
// For this reason, we cannot use a `while segment_postings.advance()` loop.
// deleted doc are skipped as they do not have a `remapped_doc_id`.
if let Some(remapped_doc_id) = old_to_new_doc_id[doc as usize] {
// we make sure to only write the term iff
// there is at least one document.
let term_freq = segment_postings.term_freq();
segment_postings.positions(&mut positions_buffer);
let delta_positions = delta_computer.compute_delta(&positions_buffer);
field_serializer.write_doc(
remapped_doc_id,
term_freq,
delta_positions,
)?;
}
if !segment_postings.advance() {
break;
}
}
}
// closing the term.
field_serializer.close_term()?;
}
}
field_serializer.close()?;
Ok(term_ord_mapping_opt)
@@ -644,18 +647,13 @@ impl IndexMerger {
fn write_postings(
&self,
serializer: &mut InvertedIndexSerializer,
fieldnorm_readers: FieldNormReaders,
) -> crate::Result<HashMap<Field, TermOrdinalMapping>> {
let mut term_ordinal_mappings = HashMap::new();
for (field, field_entry) in self.schema.fields() {
let fieldnorm_reader = fieldnorm_readers.get_field(field);
if field_entry.is_indexed() {
if let Some(term_ordinal_mapping) = self.write_postings_for_field(
field,
field_entry.field_type(),
serializer,
fieldnorm_reader,
)? {
if let Some(term_ordinal_mapping) =
self.write_postings_for_field(field, field_entry.field_type(), serializer)?
{
term_ordinal_mappings.insert(field, term_ordinal_mapping);
}
}
@@ -681,15 +679,8 @@ impl IndexMerger {
impl SerializableSegment for IndexMerger {
fn write(&self, mut serializer: SegmentSerializer) -> crate::Result<u32> {
if let Some(fieldnorms_serializer) = serializer.extract_fieldnorms_serializer() {
self.write_fieldnorms(fieldnorms_serializer)?;
}
let fieldnorm_data = serializer
.segment()
.open_read(SegmentComponent::FIELDNORMS)?;
let fieldnorm_readers = FieldNormReaders::open(fieldnorm_data)?;
let term_ord_mappings =
self.write_postings(serializer.get_postings_serializer(), fieldnorm_readers)?;
let term_ord_mappings = self.write_postings(serializer.get_postings_serializer())?;
self.write_fieldnorms(serializer.get_fieldnorms_serializer())?;
self.write_fast_fields(serializer.get_fast_field_serializer(), term_ord_mappings)?;
self.write_storable_fields(serializer.get_store_writer())?;
serializer.close()?;
@@ -699,15 +690,15 @@ impl SerializableSegment for IndexMerger {
#[cfg(test)]
mod tests {
use crate::assert_nearly_equals;
use crate::collector::tests::TEST_COLLECTOR_WITH_SCORE;
use crate::collector::tests::{BytesFastFieldTestCollector, FastFieldTestCollector};
use crate::collector::{Count, FacetCollector};
use crate::core::Index;
use crate::query::AllQuery;
use crate::query::BooleanQuery;
use crate::query::Scorer;
use crate::query::TermQuery;
use crate::schema;
use crate::schema::Cardinality;
use crate::schema::Document;
use crate::schema::Facet;
use crate::schema::IndexRecordOption;
@@ -715,11 +706,9 @@ mod tests {
use crate::schema::Term;
use crate::schema::TextFieldIndexing;
use crate::schema::INDEXED;
use crate::schema::{Cardinality, TEXT};
use crate::DocAddress;
use crate::IndexWriter;
use crate::Searcher;
use crate::{schema, DocSet, SegmentId};
use byteorder::{BigEndian, ReadBytesExt, WriteBytesExt};
use futures::executor::block_on;
use std::io::Cursor;
@@ -751,7 +740,7 @@ mod tests {
};
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
{
// writing the segment
{
@@ -803,7 +792,7 @@ mod tests {
let segment_ids = index
.searchable_segment_ids()
.expect("Searchable segments failed.");
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
block_on(index_writer.merge(&segment_ids)).expect("Merging failed");
index_writer.wait_merging_threads().unwrap();
}
@@ -904,7 +893,7 @@ mod tests {
let score_field = schema_builder.add_u64_field("score", score_fieldtype);
let bytes_score_field = schema_builder.add_bytes_field("score_bytes");
let index = Index::create_in_ram(schema_builder.build());
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
let reader = index.reader().unwrap();
let search_term = |searcher: &Searcher, term: Term| {
let collector = FastFieldTestCollector::for_field(score_field);
@@ -1211,7 +1200,7 @@ mod tests {
let index = Index::create_in_ram(schema_builder.build());
let reader = index.reader().unwrap();
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
let index_doc = |index_writer: &mut IndexWriter, doc_facets: &[&str]| {
let mut doc = Document::default();
for facet in doc_facets {
@@ -1276,7 +1265,7 @@ mod tests {
let segment_ids = index
.searchable_segment_ids()
.expect("Searchable segments failed.");
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
block_on(index_writer.merge(&segment_ids)).expect("Merging failed");
index_writer.wait_merging_threads().unwrap();
reader.reload().unwrap();
@@ -1295,7 +1284,7 @@ mod tests {
// Deleting one term
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
let facet = Facet::from_path(vec!["top", "a", "firstdoc"]);
let facet_term = Term::from_facet(facet_field, &facet);
index_writer.delete_term(facet_term);
@@ -1320,7 +1309,7 @@ mod tests {
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_for_tests().unwrap();
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));
@@ -1349,7 +1338,7 @@ mod tests {
let index = Index::create_in_ram(schema_builder.build());
let reader = index.reader().unwrap();
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
let mut doc = Document::default();
doc.add_u64(int_field, 1);
index_writer.add_document(doc.clone());
@@ -1388,7 +1377,7 @@ mod tests {
let index = Index::create_in_ram(schema_builder.build());
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
let index_doc = |index_writer: &mut IndexWriter, int_vals: &[u64]| {
let mut doc = Document::default();
for &val in int_vals {
@@ -1462,7 +1451,7 @@ mod tests {
let segment_ids = index
.searchable_segment_ids()
.expect("Searchable segments failed.");
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
assert!(block_on(index_writer.merge(&segment_ids)).is_ok());
assert!(index_writer.wait_merging_threads().is_ok());
}
@@ -1516,7 +1505,7 @@ mod tests {
let index = Index::create_in_ram(builder.build());
let mut writer = index.writer_for_tests()?;
let mut writer = index.writer_with_num_threads(1, 3_000_000)?;
// Make sure we'll attempt to merge every created segment
let mut policy = crate::indexer::LogMergePolicy::default();
@@ -1526,9 +1515,12 @@ mod tests {
for i in 0..100 {
let mut doc = Document::new();
doc.add_f64(field, 42.0);
doc.add_f64(multi_field, 0.24);
doc.add_f64(multi_field, 0.27);
writer.add_document(doc);
if i % 5 == 0 {
writer.commit()?;
}
@@ -1540,72 +1532,6 @@ mod tests {
// If a merging thread fails, we should end up with more
// than one segment here
assert_eq!(1, index.searchable_segments()?.len());
Ok(())
}
#[test]
fn test_merged_index_has_blockwand() -> crate::Result<()> {
let mut builder = schema::SchemaBuilder::new();
let text = builder.add_text_field("text", TEXT);
let index = Index::create_in_ram(builder.build());
let mut writer = index.writer_for_tests()?;
let happy_term = Term::from_field_text(text, "happy");
let term_query = TermQuery::new(happy_term, IndexRecordOption::WithFreqs);
for _ in 0..62 {
writer.add_document(doc!(text=>"hello happy tax payer"));
}
writer.commit()?;
let reader = index.reader()?;
let searcher = reader.searcher();
let mut term_scorer = term_query
.specialized_weight(&searcher, true)
.specialized_scorer(searcher.segment_reader(0u32), 1.0)?;
assert_eq!(term_scorer.doc(), 0);
assert_nearly_equals!(term_scorer.block_max_score(), 0.0079681855);
assert_nearly_equals!(term_scorer.score(), 0.0079681855);
for _ in 0..81 {
writer.add_document(doc!(text=>"hello happy tax payer"));
}
writer.commit()?;
reader.reload()?;
let searcher = reader.searcher();
assert_eq!(searcher.segment_readers().len(), 2);
for segment_reader in searcher.segment_readers() {
let mut term_scorer = term_query
.specialized_weight(&searcher, true)
.specialized_scorer(segment_reader, 1.0)?;
// the difference compared to before is instrinsic to the bm25 formula. no worries there.
for doc in segment_reader.doc_ids_alive() {
assert_eq!(term_scorer.doc(), doc);
assert_nearly_equals!(term_scorer.block_max_score(), 0.003478312);
assert_nearly_equals!(term_scorer.score(), 0.003478312);
term_scorer.advance();
}
}
let segment_ids: Vec<SegmentId> = searcher
.segment_readers()
.iter()
.map(|reader| reader.segment_id())
.collect();
block_on(writer.merge(&segment_ids[..]))?;
reader.reload()?;
let searcher = reader.searcher();
assert_eq!(searcher.segment_readers().len(), 1);
let segment_reader = searcher.segment_reader(0u32);
let mut term_scorer = term_query
.specialized_weight(&searcher, true)
.specialized_scorer(segment_reader, 1.0)?;
// the difference compared to before is instrinsic to the bm25 formula. no worries there.
for doc in segment_reader.doc_ids_alive() {
assert_eq!(term_scorer.doc(), doc);
assert_nearly_equals!(term_scorer.block_max_score(), 0.003478312);
assert_nearly_equals!(term_scorer.score(), 0.003478312);
term_scorer.advance();
}
Ok(())
}

5
src/indexer/mod.rs
View File

@@ -29,9 +29,8 @@ pub use self::segment_writer::SegmentWriter;
/// Alias for the default merge policy, which is the `LogMergePolicy`.
pub type DefaultMergePolicy = LogMergePolicy;
#[cfg(feature = "mmap")]
#[cfg(test)]
mod tests_mmap {
mod tests {
use crate::schema::{self, Schema};
use crate::{Index, Term};
@@ -40,7 +39,7 @@ mod tests_mmap {
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("text", schema::TEXT);
let index = Index::create_from_tempdir(schema_builder.build()).unwrap();
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
// there must be one deleted document in the segment
index_writer.add_document(doc!(text_field=>"b"));
index_writer.delete_term(Term::from_field_text(text_field, "b"));

3
src/indexer/prepared_commit.rs
View File

@@ -1,7 +1,6 @@
use super::IndexWriter;
use crate::Opstamp;
use futures::executor::block_on;
use slog::info;
/// A prepared commit
pub struct PreparedCommit<'a> {
@@ -32,7 +31,7 @@ impl<'a> PreparedCommit<'a> {
}
pub fn commit(self) -> crate::Result<Opstamp> {
info!(self.index_writer.logger(), "committing {}", self.opstamp);
info!("committing {}", self.opstamp);
let _ = block_on(
self.index_writer
.segment_updater()

8
src/indexer/segment_manager.rs
View File

@@ -1,5 +1,3 @@
use slog::{warn, Logger};
use super::segment_register::SegmentRegister;
use crate::core::SegmentId;
use crate::core::SegmentMeta;
@@ -44,9 +42,9 @@ impl SegmentRegisters {
///
/// It guarantees the atomicity of the
/// changes (merges especially)
#[derive(Default)]
pub struct SegmentManager {
registers: RwLock<SegmentRegisters>,
logger: Logger,
}
impl Debug for SegmentManager {
@@ -79,14 +77,12 @@ impl SegmentManager {
pub fn from_segments(
segment_metas: Vec<SegmentMeta>,
delete_cursor: &DeleteCursor,
logger: Logger,
) -> SegmentManager {
SegmentManager {
registers: RwLock::new(SegmentRegisters {
uncommitted: SegmentRegister::default(),
committed: SegmentRegister::new(segment_metas, delete_cursor),
}),
logger,
}
}
@@ -190,7 +186,7 @@ impl SegmentManager {
let segments_status = registers_lock
.segments_status(before_merge_segment_ids)
.ok_or_else(|| {
warn!(self.logger, "couldn't find segment in SegmentManager");
warn!("couldn't find segment in SegmentManager");
crate::TantivyError::InvalidArgument(
"The segments that were merged could not be found in the SegmentManager. \
This is not necessarily a bug, and can happen after a rollback for instance."

28
src/indexer/segment_serializer.rs
View File

@@ -8,16 +8,15 @@ use crate::store::StoreWriter;
/// Segment serializer is in charge of laying out on disk
/// the data accumulated and sorted by the `SegmentWriter`.
pub struct SegmentSerializer {
segment: Segment,
store_writer: StoreWriter,
fast_field_serializer: FastFieldSerializer,
fieldnorms_serializer: Option<FieldNormsSerializer>,
fieldnorms_serializer: FieldNormsSerializer,
postings_serializer: InvertedIndexSerializer,
}
impl SegmentSerializer {
/// Creates a new `SegmentSerializer`.
pub fn for_segment(mut segment: Segment) -> crate::Result<SegmentSerializer> {
pub fn for_segment(segment: &mut Segment) -> crate::Result<SegmentSerializer> {
let store_write = segment.open_write(SegmentComponent::STORE)?;
let fast_field_write = segment.open_write(SegmentComponent::FASTFIELDS)?;
@@ -26,20 +25,15 @@ impl SegmentSerializer {
let fieldnorms_write = segment.open_write(SegmentComponent::FIELDNORMS)?;
let fieldnorms_serializer = FieldNormsSerializer::from_write(fieldnorms_write)?;
let postings_serializer = InvertedIndexSerializer::open(&mut segment)?;
let postings_serializer = InvertedIndexSerializer::open(segment)?;
Ok(SegmentSerializer {
segment,
store_writer: StoreWriter::new(store_write),
fast_field_serializer,
fieldnorms_serializer: Some(fieldnorms_serializer),
fieldnorms_serializer,
postings_serializer,
})
}
pub fn segment(&self) -> &Segment {
&self.segment
}
/// Accessor to the `PostingsSerializer`.
pub fn get_postings_serializer(&mut self) -> &mut InvertedIndexSerializer {
&mut self.postings_serializer
@@ -50,11 +44,9 @@ impl SegmentSerializer {
&mut self.fast_field_serializer
}
/// Extract the field norm serializer.
///
/// Note the fieldnorms serializer can only be extracted once.
pub fn extract_fieldnorms_serializer(&mut self) -> Option<FieldNormsSerializer> {
self.fieldnorms_serializer.take()
/// Accessor to the field norm serializer.
pub fn get_fieldnorms_serializer(&mut self) -> &mut FieldNormsSerializer {
&mut self.fieldnorms_serializer
}
/// Accessor to the `StoreWriter`.
@@ -63,13 +55,11 @@ impl SegmentSerializer {
}
/// Finalize the segment serialization.
pub fn close(mut self) -> crate::Result<()> {
if let Some(fieldnorms_serializer) = self.extract_fieldnorms_serializer() {
fieldnorms_serializer.close()?;
}
pub fn close(self) -> crate::Result<()> {
self.fast_field_serializer.close()?;
self.postings_serializer.close()?;
self.store_writer.close()?;
self.fieldnorms_serializer.close()?;
Ok(())
}
}

67
src/indexer/segment_updater.rs
View File

@@ -23,9 +23,10 @@ use futures::channel::oneshot;
use futures::executor::{ThreadPool, ThreadPoolBuilder};
use futures::future::Future;
use futures::future::TryFutureExt;
use slog::{debug, error, info, warn};
use serde_json;
use std::borrow::BorrowMut;
use std::collections::HashSet;
use std::io::Write;
use std::ops::Deref;
use std::path::PathBuf;
use std::sync::atomic::{AtomicBool, Ordering};
@@ -65,11 +66,12 @@ pub fn save_new_metas(schema: Schema, directory: &mut dyn Directory) -> crate::R
///
/// This method is not part of tantivy's public API
fn save_metas(metas: &IndexMeta, directory: &mut dyn Directory) -> crate::Result<()> {
let mut meta_json = serde_json::to_string_pretty(metas)?;
info!("save metas");
let mut buffer = serde_json::to_vec_pretty(metas)?;
// Just adding a new line at the end of the buffer.
meta_json.push_str("\n");
debug!(directory.logger(), "save meta"; "content"=>&meta_json);
directory.atomic_write(&META_FILEPATH, meta_json.as_bytes())?;
writeln!(&mut buffer)?;
directory.atomic_write(&META_FILEPATH, &buffer[..])?;
debug!("Saved metas {:?}", serde_json::to_string_pretty(&metas));
Ok(())
}
@@ -96,6 +98,7 @@ impl Deref for SegmentUpdater {
async fn garbage_collect_files(
segment_updater: SegmentUpdater,
) -> crate::Result<GarbageCollectionResult> {
info!("Running garbage collection");
let mut index = segment_updater.index.clone();
index
.directory_mut()
@@ -105,12 +108,14 @@ async fn garbage_collect_files(
/// Merges a list of segments the list of segment givens in the `segment_entries`.
/// This function happens in the calling thread and is computationally expensive.
fn merge(
merged_segment: Segment,
index: &Index,
mut segment_entries: Vec<SegmentEntry>,
target_opstamp: Opstamp,
) -> crate::Result<SegmentEntry> {
// First we apply all of the delete to the merged segment, up to the target opstamp.
// first we need to apply deletes to our segment.
let mut merged_segment = index.new_segment();
// First we apply all of the delet to the merged segment, up to the target opstamp.
for segment_entry in &mut segment_entries {
let segment = index.segment(segment_entry.meta().clone());
advance_deletes(segment, segment_entry, target_opstamp)?;
@@ -127,13 +132,12 @@ fn merge(
let merger: IndexMerger = IndexMerger::open(index.schema(), &segments[..])?;
// ... we just serialize this index merger in our new segment to merge the two segments.
let segment_serializer = SegmentSerializer::for_segment(merged_segment.clone())?;
let segment_serializer = SegmentSerializer::for_segment(&mut merged_segment)?;
let num_docs = merger.write(segment_serializer)?;
let merged_segment_id = merged_segment.id();
let segment_meta = index.new_segment_meta(merged_segment.id(), num_docs);
let segment_meta = index.new_segment_meta(merged_segment_id, num_docs);
Ok(SegmentEntry::new(segment_meta, delete_cursor, None))
}
@@ -163,8 +167,7 @@ impl SegmentUpdater {
delete_cursor: &DeleteCursor,
) -> crate::Result<SegmentUpdater> {
let segments = index.searchable_segment_metas()?;
let segment_manager =
SegmentManager::from_segments(segments, delete_cursor, index.logger().clone());
let segment_manager = SegmentManager::from_segments(segments, delete_cursor);
let pool = ThreadPoolBuilder::new()
.name_prefix("segment_updater")
.pool_size(1)
@@ -384,18 +387,7 @@ impl SegmentUpdater {
.segment_manager
.start_merge(merge_operation.segment_ids())?;
let segment_ids_str: String = merge_operation
.segment_ids()
.iter()
.map(|segment_id| segment_id.to_string())
.collect::<Vec<String>>()
.join(",");
let merged_segment = self.index.new_segment();
let logger = self.index.logger().new(slog::o!("segments"=>segment_ids_str, "merged-segment"=>merged_segment.id().to_string()));
let num_merges: usize = self.merge_operations.list().len();
slog::info!(&logger, "merge"; "stage"=>"start", "num-merges" => num_merges);
info!("Starting merge - {:?}", merge_operation.segment_ids());
let (merging_future_send, merging_future_recv) =
oneshot::channel::<crate::Result<SegmentMeta>>();
@@ -406,20 +398,22 @@ impl SegmentUpdater {
// as well as which segment is currently in merge and therefore should not be
// candidate for another merge.
match merge(
merged_segment,
&segment_updater.index,
segment_entries,
merge_operation.target_opstamp(),
) {
Ok(after_merge_segment_entry) => {
info!(&logger, "merge"; "stage" => "end");
let segment_meta = segment_updater
.end_merge(merge_operation, after_merge_segment_entry)
.await;
let _send_result = merging_future_send.send(segment_meta);
}
Err(e) => {
error!(&logger, "merge"; "stage" => "fail", "cause"=>e.to_string());
warn!(
"Merge of {:?} was cancelled: {:?}",
merge_operation.segment_ids().to_vec(),
e
);
// ... cancel merge
if cfg!(test) {
panic!("Merge failed.");
@@ -460,12 +454,11 @@ impl SegmentUpdater {
.collect::<Vec<_>>();
merge_candidates.extend(committed_merge_candidates.into_iter());
let logger = self.index.logger();
for merge_operation in merge_candidates {
if let Err(err) = self.start_merge(merge_operation) {
warn!(
logger,
"merge-start-fail (not fatal, not necessarily a problem)"; "reason" => format!("{}", err),
"Starting the merge failed for the following reason. This is not fatal. {}",
err
);
}
}
@@ -478,11 +471,8 @@ impl SegmentUpdater {
) -> impl Future<Output = crate::Result<SegmentMeta>> {
let segment_updater = self.clone();
let after_merge_segment_meta = after_merge_segment_entry.meta().clone();
let logger = self.index.logger().new(
slog::o!("segment"=>after_merge_segment_meta.id().to_string(),
"delete-opstamp"=>after_merge_segment_meta.delete_opstamp()),
);
let end_merge_future = self.schedule_future(async move {
info!("End merge {:?}", after_merge_segment_entry.meta());
{
let mut delete_cursor = after_merge_segment_entry.delete_cursor().clone();
if let Some(delete_operation) = delete_cursor.get() {
@@ -496,7 +486,6 @@ impl SegmentUpdater {
committed_opstamp,
) {
error!(
logger,
"Merge of {:?} was cancelled (advancing deletes failed): {:?}",
merge_operation.segment_ids(),
e
@@ -533,7 +522,7 @@ impl SegmentUpdater {
///
/// Upon termination of the current merging threads,
/// merge opportunity may appear.
///
//
/// We keep waiting until the merge policy judges that
/// no opportunity is available.
///
@@ -566,7 +555,7 @@ mod tests {
let index = Index::create_in_ram(schema);
// writing the segment
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.set_merge_policy(Box::new(MergeWheneverPossible));
{
@@ -619,7 +608,7 @@ mod tests {
let index = Index::create_in_ram(schema);
// writing the segment
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
{
for _ in 0..100 {
@@ -690,7 +679,7 @@ mod tests {
let index = Index::create_in_ram(schema);
// writing the segment
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
{
for _ in 0..100 {

98
src/indexer/segment_writer.rs
View File

@@ -1,7 +1,8 @@
use super::operation::AddOperation;
use crate::core::Segment;
use crate::core::SerializableSegment;
use crate::fastfield::FastFieldsWriter;
use crate::fieldnorm::{FieldNormReaders, FieldNormsWriter};
use crate::fieldnorm::FieldNormsWriter;
use crate::indexer::segment_serializer::SegmentSerializer;
use crate::postings::compute_table_size;
use crate::postings::MultiFieldPostingsWriter;
@@ -13,10 +14,10 @@ use crate::schema::{Field, FieldEntry};
use crate::tokenizer::{BoxTokenStream, PreTokenizedStream};
use crate::tokenizer::{FacetTokenizer, TextAnalyzer};
use crate::tokenizer::{TokenStreamChain, Tokenizer};
use crate::DocId;
use crate::Opstamp;
use crate::{core::Segment, tokenizer::MAX_TOKEN_LEN};
use crate::{DocId, SegmentComponent};
use std::io;
use std::str;
/// Computes the initial size of the hash table.
///
@@ -47,7 +48,6 @@ pub struct SegmentWriter {
fieldnorms_writer: FieldNormsWriter,
doc_opstamps: Vec<Opstamp>,
tokenizers: Vec<Option<TextAnalyzer>>,
term_buffer: Term,
}
impl SegmentWriter {
@@ -62,12 +62,11 @@ impl SegmentWriter {
/// - schema
pub fn for_segment(
memory_budget: usize,
segment: Segment,
mut segment: Segment,
schema: &Schema,
) -> crate::Result<SegmentWriter> {
let tokenizer_manager = segment.index().tokenizers().clone();
let table_num_bits = initial_table_size(memory_budget)?;
let segment_serializer = SegmentSerializer::for_segment(segment)?;
let segment_serializer = SegmentSerializer::for_segment(&mut segment)?;
let multifield_postings = MultiFieldPostingsWriter::new(schema, table_num_bits);
let tokenizers = schema
.fields()
@@ -77,7 +76,7 @@ impl SegmentWriter {
.get_indexing_options()
.and_then(|text_index_option| {
let tokenizer_name = &text_index_option.tokenizer();
tokenizer_manager.get(tokenizer_name)
segment.index().tokenizers().get(tokenizer_name)
}),
_ => None,
},
@@ -91,7 +90,6 @@ impl SegmentWriter {
fast_field_writers: FastFieldsWriter::from_schema(schema),
doc_opstamps: Vec::with_capacity(1_000),
tokenizers,
term_buffer: Term::new(),
})
}
@@ -129,29 +127,24 @@ impl SegmentWriter {
if !field_options.is_indexed() {
continue;
}
let (term_buffer, multifield_postings) =
(&mut self.term_buffer, &mut self.multifield_postings);
match *field_options.field_type() {
FieldType::HierarchicalFacet => {
term_buffer.set_field(field);
let facets =
field_values
.iter()
.flat_map(|field_value| match *field_value.value() {
Value::Facet(ref facet) => Some(facet.encoded_str()),
_ => {
panic!("Expected hierarchical facet");
}
});
let facets: Vec<&str> = field_values
.iter()
.flat_map(|field_value| match *field_value.value() {
Value::Facet(ref facet) => Some(facet.encoded_str()),
_ => {
panic!("Expected hierarchical facet");
}
})
.collect();
let mut term = Term::for_field(field); // we set the Term
for fake_str in facets {
let mut unordered_term_id_opt = None;
FacetTokenizer.token_stream(fake_str).process(&mut |token| {
if token.text.len() > MAX_TOKEN_LEN {
return;
}
term_buffer.set_text(&token.text);
term.set_text(&token.text);
let unordered_term_id =
multifield_postings.subscribe(doc_id, &term_buffer);
self.multifield_postings.subscribe(doc_id, &term);
unordered_term_id_opt = Some(unordered_term_id);
});
if let Some(unordered_term_id) = unordered_term_id_opt {
@@ -174,6 +167,7 @@ impl SegmentWriter {
if let Some(last_token) = tok_str.tokens.last() {
total_offset += last_token.offset_to;
}
token_streams
.push(PreTokenizedStream::from(tok_str.clone()).into());
}
@@ -183,6 +177,7 @@ impl SegmentWriter {
{
offsets.push(total_offset);
total_offset += text.len();
token_streams.push(tokenizer.token_stream(text));
}
}
@@ -194,12 +189,8 @@ impl SegmentWriter {
0
} else {
let mut token_stream = TokenStreamChain::new(offsets, token_streams);
multifield_postings.index_text(
doc_id,
field,
&mut token_stream,
term_buffer,
)
self.multifield_postings
.index_text(doc_id, field, &mut token_stream)
};
self.fieldnorms_writer.record(doc_id, field, num_tokens);
@@ -207,36 +198,44 @@ impl SegmentWriter {
FieldType::U64(ref int_option) => {
if int_option.is_indexed() {
for field_value in field_values {
term_buffer.set_field(field_value.field());
term_buffer.set_u64(field_value.value().u64_value());
multifield_postings.subscribe(doc_id, &term_buffer);
let term = Term::from_field_u64(
field_value.field(),
field_value.value().u64_value(),
);
self.multifield_postings.subscribe(doc_id, &term);
}
}
}
FieldType::Date(ref int_option) => {
if int_option.is_indexed() {
for field_value in field_values {
term_buffer.set_field(field_value.field());
term_buffer.set_i64(field_value.value().date_value().timestamp());
multifield_postings.subscribe(doc_id, &term_buffer);
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) => {
if int_option.is_indexed() {
for field_value in field_values {
term_buffer.set_field(field_value.field());
term_buffer.set_i64(field_value.value().i64_value());
multifield_postings.subscribe(doc_id, &term_buffer);
let term = Term::from_field_i64(
field_value.field(),
field_value.value().i64_value(),
);
self.multifield_postings.subscribe(doc_id, &term);
}
}
}
FieldType::F64(ref int_option) => {
if int_option.is_indexed() {
for field_value in field_values {
term_buffer.set_field(field_value.field());
term_buffer.set_f64(field_value.value().f64_value());
multifield_postings.subscribe(doc_id, &term_buffer);
let term = Term::from_field_f64(
field_value.field(),
field_value.value().f64_value(),
);
self.multifield_postings.subscribe(doc_id, &term);
}
}
}
@@ -281,16 +280,9 @@ fn write(
fieldnorms_writer: &FieldNormsWriter,
mut serializer: SegmentSerializer,
) -> crate::Result<()> {
if let Some(fieldnorms_serializer) = serializer.extract_fieldnorms_serializer() {
fieldnorms_writer.serialize(fieldnorms_serializer)?;
}
let fieldnorm_data = serializer
.segment()
.open_read(SegmentComponent::FIELDNORMS)?;
let fieldnorm_readers = FieldNormReaders::open(fieldnorm_data)?;
let term_ord_map =
multifield_postings.serialize(serializer.get_postings_serializer(), fieldnorm_readers)?;
let term_ord_map = multifield_postings.serialize(serializer.get_postings_serializer())?;
fast_field_writers.serialize(serializer.get_fast_field_serializer(), &term_ord_map)?;
fieldnorms_writer.serialize(serializer.get_fieldnorms_serializer())?;
serializer.close()?;
Ok(())
}

456
src/lib.rs
View File

@@ -98,11 +98,17 @@
//! [literate programming](https://tantivy-search.github.io/examples/basic_search.html) /
//! [source code](https://github.com/tantivy-search/tantivy/blob/master/examples/basic_search.rs))
#[macro_use]
extern crate serde_derive;
#[cfg_attr(test, macro_use)]
extern crate serde_json;
#[macro_use]
extern crate thiserror;
extern crate log;
#[macro_use]
extern crate failure;
#[cfg(all(test, feature = "unstable"))]
extern crate test;
@@ -145,7 +151,6 @@ pub mod schema;
pub mod space_usage;
pub mod store;
pub mod termdict;
pub use slog;
mod reader;
@@ -154,7 +159,7 @@ mod snippet;
pub use self::snippet::{Snippet, SnippetGenerator};
mod docset;
pub use self::docset::{DocSet, TERMINATED};
pub use self::docset::{DocSet, SkipResult};
pub use crate::common::{f64_to_u64, i64_to_u64, u64_to_f64, u64_to_i64};
pub use crate::core::{Executor, SegmentComponent};
pub use crate::core::{Index, IndexMeta, Searcher, Segment, SegmentId, SegmentMeta};
@@ -168,10 +173,9 @@ pub use crate::schema::{Document, Term};
use std::fmt;
use once_cell::sync::Lazy;
use serde::{Deserialize, Serialize};
/// Index format version.
const INDEX_FORMAT_VERSION: u32 = 2;
const INDEX_FORMAT_VERSION: u32 = 1;
/// Structure version for the index.
#[derive(Clone, PartialEq, Eq, Serialize, Deserialize)]
@@ -243,10 +247,11 @@ pub type DocId = u32;
/// with opstamp `n+1`.
pub type Opstamp = u64;
/// A Score that represents the relevance of the document to the query
/// A f32 that represents the relevance of the document to the query
///
/// This is modelled internally as a `f32`. The larger the number, the more relevant
/// the document to the search query.
/// This is modelled internally as a `f32`. The
/// larger the number, the more relevant the document
/// to the search
pub type Score = f32;
/// A `SegmentLocalId` identifies a segment.
@@ -279,13 +284,15 @@ pub struct DocAddress(pub SegmentLocalId, pub DocId);
#[cfg(test)]
mod tests {
use crate::collector::tests::TEST_COLLECTOR_WITH_SCORE;
use crate::core::SegmentReader;
use crate::docset::{DocSet, TERMINATED};
use crate::docset::DocSet;
use crate::query::BooleanQuery;
use crate::schema::*;
use crate::DocAddress;
use crate::Index;
use crate::IndexWriter;
use crate::Postings;
use crate::ReloadPolicy;
use rand::distributions::Bernoulli;
@@ -293,26 +300,17 @@ mod tests {
use rand::rngs::StdRng;
use rand::{Rng, SeedableRng};
/// Checks if left and right are close one to each other.
/// Panics if the two values are more than 0.5% apart.
#[macro_export]
macro_rules! assert_nearly_equals {
($left:expr, $right:expr) => {{
match (&$left, &$right) {
(left_val, right_val) => {
let diff = (left_val - right_val).abs();
let add = left_val.abs() + right_val.abs();
if diff > 0.0005 * add {
panic!(
r#"assertion failed: `(left ~= right)`
left: `{:?}`,
right: `{:?}`"#,
&*left_val, &*right_val
)
}
}
}
}};
pub fn assert_nearly_equals(expected: f32, val: f32) {
assert!(
nearly_equals(val, expected),
"Got {}, expected {}.",
val,
expected
);
}
pub fn nearly_equals(a: f32, b: f32) -> bool {
(a - b).abs() < 0.0005 * (a + b).abs()
}
pub fn generate_nonunique_unsorted(max_value: u32, n_elems: usize) -> Vec<u32> {
@@ -350,14 +348,14 @@ mod tests {
#[test]
#[cfg(feature = "mmap")]
fn test_indexing() -> crate::Result<()> {
fn test_indexing() {
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_from_tempdir(schema).unwrap();
{
// writing the segment
let mut index_writer = index.writer_for_tests()?;
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
{
let doc = doc!(text_field=>"af b");
index_writer.add_document(doc);
@@ -372,30 +370,36 @@ mod tests {
}
assert!(index_writer.commit().is_ok());
}
Ok(())
}
#[test]
fn test_docfreq1() -> crate::Result<()> {
fn test_docfreq1() {
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("text", TEXT);
let index = Index::create_in_ram(schema_builder.build());
let mut index_writer = index.writer_for_tests()?;
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
{
index_writer.add_document(doc!(text_field=>"a b c"));
index_writer.commit()?;
index_writer.commit().unwrap();
}
{
index_writer.add_document(doc!(text_field=>"a"));
index_writer.add_document(doc!(text_field=>"a a"));
index_writer.commit()?;
{
let doc = doc!(text_field=>"a");
index_writer.add_document(doc);
}
{
let doc = doc!(text_field=>"a a");
index_writer.add_document(doc);
}
index_writer.commit().unwrap();
}
{
index_writer.add_document(doc!(text_field=>"c"));
index_writer.commit()?;
let doc = doc!(text_field=>"c");
index_writer.add_document(doc);
index_writer.commit().unwrap();
}
{
let reader = index.reader()?;
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let term_a = Term::from_field_text(text_field, "a");
assert_eq!(searcher.doc_freq(&term_a), 3);
@@ -406,65 +410,80 @@ mod tests {
let term_d = Term::from_field_text(text_field, "d");
assert_eq!(searcher.doc_freq(&term_d), 0);
}
Ok(())
}
#[test]
fn test_fieldnorm_no_docs_with_field() -> crate::Result<()> {
fn test_fieldnorm_no_docs_with_field() {
let mut schema_builder = Schema::builder();
let title_field = schema_builder.add_text_field("title", TEXT);
let text_field = schema_builder.add_text_field("text", TEXT);
let index = Index::create_in_ram(schema_builder.build());
let mut index_writer = index.writer_for_tests()?;
index_writer.add_document(doc!(text_field=>"a b c"));
index_writer.commit()?;
let index_reader = index.reader()?;
let searcher = index_reader.searcher();
let reader = searcher.segment_reader(0);
{
let fieldnorm_reader = reader.get_fieldnorms_reader(text_field)?;
assert_eq!(fieldnorm_reader.fieldnorm(0), 3);
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
{
let doc = doc!(text_field=>"a b c");
index_writer.add_document(doc);
}
index_writer.commit().unwrap();
}
{
let fieldnorm_reader = reader.get_fieldnorms_reader(title_field)?;
assert_eq!(fieldnorm_reader.fieldnorm_id(0), 0);
let index_reader = index.reader().unwrap();
let searcher = index_reader.searcher();
let reader = searcher.segment_reader(0);
{
let fieldnorm_reader = reader.get_fieldnorms_reader(text_field);
assert_eq!(fieldnorm_reader.fieldnorm(0), 3);
}
{
let fieldnorm_reader = reader.get_fieldnorms_reader(title_field);
assert_eq!(fieldnorm_reader.fieldnorm_id(0), 0);
}
}
Ok(())
}
#[test]
fn test_fieldnorm() -> crate::Result<()> {
fn test_fieldnorm() {
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("text", TEXT);
let index = Index::create_in_ram(schema_builder.build());
let mut index_writer = index.writer_for_tests()?;
index_writer.add_document(doc!(text_field=>"a b c"));
index_writer.add_document(doc!());
index_writer.add_document(doc!(text_field=>"a b"));
index_writer.commit()?;
let reader = index.reader()?;
let searcher = reader.searcher();
let segment_reader: &SegmentReader = searcher.segment_reader(0);
let fieldnorms_reader = segment_reader.get_fieldnorms_reader(text_field)?;
assert_eq!(fieldnorms_reader.fieldnorm(0), 3);
assert_eq!(fieldnorms_reader.fieldnorm(1), 0);
assert_eq!(fieldnorms_reader.fieldnorm(2), 2);
Ok(())
{
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
{
let doc = doc!(text_field=>"a b c");
index_writer.add_document(doc);
}
{
let doc = doc!();
index_writer.add_document(doc);
}
{
let doc = doc!(text_field=>"a b");
index_writer.add_document(doc);
}
index_writer.commit().unwrap();
}
{
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let segment_reader: &SegmentReader = searcher.segment_reader(0);
let fieldnorms_reader = segment_reader.get_fieldnorms_reader(text_field);
assert_eq!(fieldnorms_reader.fieldnorm(0), 3);
assert_eq!(fieldnorms_reader.fieldnorm(1), 0);
assert_eq!(fieldnorms_reader.fieldnorm(2), 2);
}
}
fn advance_undeleted(docset: &mut dyn DocSet, reader: &SegmentReader) -> bool {
let mut doc = docset.advance();
while doc != TERMINATED {
if !reader.is_deleted(doc) {
while docset.advance() {
if !reader.is_deleted(docset.doc()) {
return true;
}
doc = docset.advance();
}
false
}
#[test]
fn test_delete_postings1() -> crate::Result<()> {
fn test_delete_postings1() {
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("text", TEXT);
let term_abcd = Term::from_field_text(text_field, "abcd");
@@ -480,7 +499,7 @@ mod tests {
.unwrap();
{
// writing the segment
let mut index_writer = index.writer_for_tests()?;
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
// 0
index_writer.add_document(doc!(text_field=>"a b"));
// 1
@@ -496,10 +515,10 @@ mod tests {
index_writer.add_document(doc!(text_field=>" b c"));
// 5
index_writer.add_document(doc!(text_field=>" a"));
index_writer.commit()?;
index_writer.commit().unwrap();
}
{
reader.reload()?;
reader.reload().unwrap();
let searcher = reader.searcher();
let segment_reader = searcher.segment_reader(0);
let inverted_index = segment_reader.inverted_index(text_field);
@@ -527,15 +546,15 @@ mod tests {
}
{
// writing the segment
let mut index_writer = index.writer_for_tests()?;
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
// 0
index_writer.add_document(doc!(text_field=>"a b"));
// 1
index_writer.delete_term(Term::from_field_text(text_field, "c"));
index_writer.rollback()?;
index_writer.rollback().unwrap();
}
{
reader.reload()?;
reader.reload().unwrap();
let searcher = reader.searcher();
let seg_reader = searcher.segment_reader(0);
let inverted_index = seg_reader.inverted_index(term_abcd.field());
@@ -564,15 +583,15 @@ mod tests {
}
{
// writing the segment
let mut index_writer = index.writer_for_tests()?;
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"a b"));
index_writer.delete_term(Term::from_field_text(text_field, "c"));
index_writer.rollback()?;
index_writer.rollback().unwrap();
index_writer.delete_term(Term::from_field_text(text_field, "a"));
index_writer.commit()?;
index_writer.commit().unwrap();
}
{
reader.reload()?;
reader.reload().unwrap();
let searcher = reader.searcher();
let segment_reader = searcher.segment_reader(0);
let inverted_index = segment_reader.inverted_index(term_abcd.field());
@@ -604,20 +623,19 @@ mod tests {
assert!(!advance_undeleted(&mut postings, segment_reader));
}
}
Ok(())
}
#[test]
fn test_indexed_u64() -> crate::Result<()> {
fn test_indexed_u64() {
let mut schema_builder = Schema::builder();
let field = schema_builder.add_u64_field("value", INDEXED);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests()?;
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(field=>1u64));
index_writer.commit()?;
let reader = index.reader()?;
index_writer.commit().unwrap();
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let term = Term::from_field_u64(field, 1u64);
let mut postings = searcher
@@ -625,23 +643,23 @@ mod tests {
.inverted_index(term.field())
.read_postings(&term, IndexRecordOption::Basic)
.unwrap();
assert!(postings.advance());
assert_eq!(postings.doc(), 0);
assert_eq!(postings.advance(), TERMINATED);
Ok(())
assert!(!postings.advance());
}
#[test]
fn test_indexed_i64() -> crate::Result<()> {
fn test_indexed_i64() {
let mut schema_builder = Schema::builder();
let value_field = schema_builder.add_i64_field("value", INDEXED);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests()?;
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
let negative_val = -1i64;
index_writer.add_document(doc!(value_field => negative_val));
index_writer.commit()?;
let reader = index.reader()?;
index_writer.commit().unwrap();
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let term = Term::from_field_i64(value_field, negative_val);
let mut postings = searcher
@@ -649,23 +667,23 @@ mod tests {
.inverted_index(term.field())
.read_postings(&term, IndexRecordOption::Basic)
.unwrap();
assert!(postings.advance());
assert_eq!(postings.doc(), 0);
assert_eq!(postings.advance(), TERMINATED);
Ok(())
assert!(!postings.advance());
}
#[test]
fn test_indexed_f64() -> crate::Result<()> {
fn test_indexed_f64() {
let mut schema_builder = Schema::builder();
let value_field = schema_builder.add_f64_field("value", INDEXED);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests()?;
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
let val = std::f64::consts::PI;
index_writer.add_document(doc!(value_field => val));
index_writer.commit()?;
let reader = index.reader()?;
index_writer.commit().unwrap();
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let term = Term::from_field_f64(value_field, val);
let mut postings = searcher
@@ -673,31 +691,29 @@ mod tests {
.inverted_index(term.field())
.read_postings(&term, IndexRecordOption::Basic)
.unwrap();
assert!(postings.advance());
assert_eq!(postings.doc(), 0);
assert_eq!(postings.advance(), TERMINATED);
Ok(())
assert!(!postings.advance());
}
#[test]
fn test_indexedfield_not_in_documents() -> crate::Result<()> {
fn test_indexedfield_not_in_documents() {
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("text", TEXT);
let absent_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests()?;
let mut index_writer = index.writer_with_num_threads(2, 6_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"a"));
assert!(index_writer.commit().is_ok());
let reader = index.reader()?;
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let segment_reader = searcher.segment_reader(0);
let inverted_index = segment_reader.inverted_index(absent_field); //< should not panic
assert_eq!(inverted_index.terms().num_terms(), 0);
Ok(())
segment_reader.inverted_index(absent_field); //< should not panic
}
#[test]
fn test_delete_postings2() -> crate::Result<()> {
fn test_delete_postings2() {
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build();
@@ -705,40 +721,55 @@ mod tests {
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::Manual)
.try_into()?;
.try_into()
.unwrap();
// writing the segment
let mut index_writer = index.writer_for_tests()?;
index_writer.add_document(doc!(text_field=>"63"));
index_writer.add_document(doc!(text_field=>"70"));
index_writer.add_document(doc!(text_field=>"34"));
index_writer.add_document(doc!(text_field=>"1"));
index_writer.add_document(doc!(text_field=>"38"));
index_writer.add_document(doc!(text_field=>"33"));
index_writer.add_document(doc!(text_field=>"40"));
index_writer.add_document(doc!(text_field=>"17"));
index_writer.delete_term(Term::from_field_text(text_field, "38"));
index_writer.delete_term(Term::from_field_text(text_field, "34"));
index_writer.commit()?;
reader.reload()?;
assert_eq!(reader.searcher().num_docs(), 6);
Ok(())
let mut index_writer = index.writer_with_num_threads(2, 6_000_000).unwrap();
let add_document = |index_writer: &mut IndexWriter, val: &'static str| {
let doc = doc!(text_field=>val);
index_writer.add_document(doc);
};
let remove_document = |index_writer: &mut IndexWriter, val: &'static str| {
let delterm = Term::from_field_text(text_field, val);
index_writer.delete_term(delterm);
};
add_document(&mut index_writer, "63");
add_document(&mut index_writer, "70");
add_document(&mut index_writer, "34");
add_document(&mut index_writer, "1");
add_document(&mut index_writer, "38");
add_document(&mut index_writer, "33");
add_document(&mut index_writer, "40");
add_document(&mut index_writer, "17");
remove_document(&mut index_writer, "38");
remove_document(&mut index_writer, "34");
index_writer.commit().unwrap();
reader.reload().unwrap();
let searcher = reader.searcher();
assert_eq!(searcher.num_docs(), 6);
}
#[test]
fn test_termfreq() -> crate::Result<()> {
fn test_termfreq() {
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_for_tests()?;
index_writer.add_document(doc!(text_field=>"af af af bc bc"));
index_writer.commit()?;
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
{
let doc = doc!(text_field=>"af af af bc bc");
index_writer.add_document(doc);
}
index_writer.commit().unwrap();
}
{
let index_reader = index.reader()?;
let index_reader = index.reader().unwrap();
let searcher = index_reader.searcher();
let reader = searcher.segment_reader(0);
let inverted_index = reader.inverted_index(text_field);
@@ -750,67 +781,68 @@ mod tests {
let mut postings = inverted_index
.read_postings(&term_af, IndexRecordOption::WithFreqsAndPositions)
.unwrap();
assert!(postings.advance());
assert_eq!(postings.doc(), 0);
assert_eq!(postings.term_freq(), 3);
assert_eq!(postings.advance(), TERMINATED);
assert!(!postings.advance());
}
Ok(())
}
#[test]
fn test_searcher_1() -> crate::Result<()> {
fn test_searcher_1() {
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);
let reader = index.reader()?;
// writing the segment
let mut index_writer = index.writer_for_tests()?;
index_writer.add_document(doc!(text_field=>"af af af b"));
index_writer.add_document(doc!(text_field=>"a b c"));
index_writer.add_document(doc!(text_field=>"a b c d"));
index_writer.commit()?;
reader.reload()?;
let searcher = reader.searcher();
let get_doc_ids = |terms: Vec<Term>| {
let query = BooleanQuery::new_multiterms_query(terms);
searcher
.search(&query, &TEST_COLLECTOR_WITH_SCORE)
.map(|topdocs| topdocs.docs().to_vec())
};
assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "a")])?,
vec![DocAddress(0, 1), DocAddress(0, 2)]
);
assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "af")])?,
vec![DocAddress(0, 0)]
);
assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "b")])?,
vec![DocAddress(0, 0), DocAddress(0, 1), DocAddress(0, 2)]
);
assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "c")])?,
vec![DocAddress(0, 1), DocAddress(0, 2)]
);
assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "d")])?,
vec![DocAddress(0, 2)]
);
assert_eq!(
get_doc_ids(vec![
Term::from_field_text(text_field, "b"),
Term::from_field_text(text_field, "a"),
])?,
vec![DocAddress(0, 0), DocAddress(0, 1), DocAddress(0, 2)]
);
Ok(())
let reader = index.reader().unwrap();
{
// writing the segment
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"af af af b"));
index_writer.add_document(doc!(text_field=>"a b c"));
index_writer.add_document(doc!(text_field=>"a b c d"));
index_writer.commit().unwrap();
}
{
reader.reload().unwrap();
let searcher = reader.searcher();
let get_doc_ids = |terms: Vec<Term>| {
let query = BooleanQuery::new_multiterms_query(terms);
let topdocs = searcher.search(&query, &TEST_COLLECTOR_WITH_SCORE).unwrap();
topdocs.docs().to_vec()
};
assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "a")]),
vec![DocAddress(0, 1), DocAddress(0, 2)]
);
assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "af")]),
vec![DocAddress(0, 0)]
);
assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "b")]),
vec![DocAddress(0, 0), DocAddress(0, 1), DocAddress(0, 2)]
);
assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "c")]),
vec![DocAddress(0, 1), DocAddress(0, 2)]
);
assert_eq!(
get_doc_ids(vec![Term::from_field_text(text_field, "d")]),
vec![DocAddress(0, 2)]
);
assert_eq!(
get_doc_ids(vec![
Term::from_field_text(text_field, "b"),
Term::from_field_text(text_field, "a"),
]),
vec![DocAddress(0, 0), DocAddress(0, 1), DocAddress(0, 2)]
);
}
}
#[test]
fn test_searcher_2() -> crate::Result<()> {
fn test_searcher_2() {
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build();
@@ -818,17 +850,19 @@ mod tests {
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::Manual)
.try_into()?;
.try_into()
.unwrap();
assert_eq!(reader.searcher().num_docs(), 0u64);
// writing the segment
let mut index_writer = index.writer_for_tests()?;
index_writer.add_document(doc!(text_field=>"af b"));
index_writer.add_document(doc!(text_field=>"a b c"));
index_writer.add_document(doc!(text_field=>"a b c d"));
index_writer.commit()?;
reader.reload()?;
{
// writing the segment
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"af b"));
index_writer.add_document(doc!(text_field=>"a b c"));
index_writer.add_document(doc!(text_field=>"a b c d"));
index_writer.commit().unwrap();
}
reader.reload().unwrap();
assert_eq!(reader.searcher().num_docs(), 3u64);
Ok(())
}
#[test]
@@ -850,7 +884,7 @@ mod tests {
}
#[test]
fn test_wrong_fast_field_type() -> crate::Result<()> {
fn test_wrong_fast_field_type() {
let mut schema_builder = Schema::builder();
let fast_field_unsigned = schema_builder.add_u64_field("unsigned", FAST);
let fast_field_signed = schema_builder.add_i64_field("signed", FAST);
@@ -860,14 +894,14 @@ mod tests {
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests()?;
let mut index_writer = index.writer_with_num_threads(1, 50_000_000).unwrap();
{
let document =
doc!(fast_field_unsigned => 4u64, fast_field_signed=>4i64, fast_field_float=>4f64);
index_writer.add_document(document);
index_writer.commit()?;
index_writer.commit().unwrap();
}
let reader = index.reader()?;
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let segment_reader: &SegmentReader = searcher.segment_reader(0);
{
@@ -906,12 +940,11 @@ mod tests {
let fast_field_reader = fast_field_reader_opt.unwrap();
assert_eq!(fast_field_reader.get(0), 4f64)
}
Ok(())
}
// motivated by #729
#[test]
fn test_update_via_delete_insert() -> crate::Result<()> {
fn test_update_via_delete_insert() {
use crate::collector::Count;
use crate::indexer::NoMergePolicy;
use crate::query::AllQuery;
@@ -925,17 +958,17 @@ mod tests {
let schema = schema_builder.build();
let index = Index::create_in_ram(schema.clone());
let index_reader = index.reader()?;
let index_reader = index.reader().unwrap();
let mut index_writer = index.writer_for_tests()?;
let mut index_writer = index.writer(3_000_000).unwrap();
index_writer.set_merge_policy(Box::new(NoMergePolicy));
for doc_id in 0u64..DOC_COUNT {
index_writer.add_document(doc!(id => doc_id));
}
index_writer.commit()?;
index_writer.commit().unwrap();
index_reader.reload()?;
index_reader.reload().unwrap();
let searcher = index_reader.searcher();
assert_eq!(
@@ -946,11 +979,12 @@ mod tests {
// update the 10 elements by deleting and re-adding
for doc_id in 0u64..DOC_COUNT {
index_writer.delete_term(Term::from_field_u64(id, doc_id));
index_writer.commit()?;
index_reader.reload()?;
index_writer.add_document(doc!(id => doc_id));
index_writer.commit()?;
index_reader.reload()?;
index_writer.commit().unwrap();
index_reader.reload().unwrap();
let doc = doc!(id => doc_id);
index_writer.add_document(doc);
index_writer.commit().unwrap();
index_reader.reload().unwrap();
let searcher = index_reader.searcher();
// The number of document should be stable.
assert_eq!(
@@ -959,7 +993,7 @@ mod tests {
);
}
index_reader.reload()?;
index_reader.reload().unwrap();
let searcher = index_reader.searcher();
let segment_ids: Vec<SegmentId> = searcher
.segment_readers()
@@ -968,18 +1002,12 @@ mod tests {
.collect();
block_on(index_writer.merge(&segment_ids)).unwrap();
index_reader.reload()?;
index_reader.reload().unwrap();
let searcher = index_reader.searcher();
assert_eq!(searcher.search(&AllQuery, &Count)?, DOC_COUNT as usize);
Ok(())
}
#[test]
fn test_validate_checksum() -> crate::Result<()> {
let index_path = tempfile::tempdir().expect("dir");
let schema = Schema::builder().build();
let index = Index::create_in_dir(&index_path, schema)?;
assert!(index.validate_checksum()?.is_empty());
Ok(())
assert_eq!(
searcher.search(&AllQuery, &Count).unwrap(),
DOC_COUNT as usize
);
}
}

89
src/positions/mod.rs
View File

@@ -37,9 +37,9 @@ const LONG_SKIP_INTERVAL: u64 = (LONG_SKIP_IN_BLOCKS * COMPRESSION_BLOCK_SIZE) a
#[cfg(test)]
pub mod tests {
use super::PositionSerializer;
use super::{PositionReader, PositionSerializer};
use crate::directory::ReadOnlySource;
use crate::positions::reader::PositionReader;
use crate::positions::COMPRESSION_BLOCK_SIZE;
use std::iter;
fn create_stream_buffer(vals: &[u32]) -> (ReadOnlySource, ReadOnlySource) {
@@ -68,7 +68,7 @@ pub mod tests {
let mut position_reader = PositionReader::new(stream, skip, 0u64);
for &n in &[1, 10, 127, 128, 130, 312] {
let mut v = vec![0u32; n];
position_reader.read(0, &mut v[..]);
position_reader.read(&mut v[..n]);
for i in 0..n {
assert_eq!(v[i], i as u32);
}
@@ -76,19 +76,19 @@ pub mod tests {
}
#[test]
fn test_position_read_with_offset() {
let v: Vec<u32> = (0..1000).collect();
fn test_position_skip() {
let v: Vec<u32> = (0..1_000).collect();
let (stream, skip) = create_stream_buffer(&v[..]);
assert_eq!(skip.len(), 12);
assert_eq!(stream.len(), 1168);
let mut position_reader = PositionReader::new(stream, skip, 0u64);
for &offset in &[1u64, 10u64, 127u64, 128u64, 130u64, 312u64] {
for &len in &[1, 10, 130, 500] {
let mut v = vec![0u32; len];
position_reader.read(offset, &mut v[..]);
for i in 0..len {
assert_eq!(v[i], i as u32 + offset as u32);
}
position_reader.skip(10);
for &n in &[10, 127, COMPRESSION_BLOCK_SIZE, 130, 312] {
let mut v = vec![0u32; n];
position_reader.read(&mut v[..n]);
for i in 0..n {
assert_eq!(v[i], 10u32 + i as u32);
}
}
}
@@ -103,12 +103,11 @@ pub mod tests {
let mut position_reader = PositionReader::new(stream, skip, 0u64);
let mut buf = [0u32; 7];
let mut c = 0;
let mut offset = 0;
for _ in 0..100 {
position_reader.read(offset, &mut buf);
position_reader.read(offset, &mut buf);
offset += 7;
position_reader.read(&mut buf);
position_reader.read(&mut buf);
position_reader.skip(4);
position_reader.skip(3);
for &el in &buf {
assert_eq!(c, el);
c += 1;
@@ -116,58 +115,6 @@ pub mod tests {
}
}
#[test]
fn test_position_reread_anchor_different_than_block() {
let v: Vec<u32> = (0..2_000_000).collect();
let (stream, skip) = create_stream_buffer(&v[..]);
assert_eq!(skip.len(), 15_749);
assert_eq!(stream.len(), 4_987_872);
let mut position_reader = PositionReader::new(stream.clone(), skip.clone(), 0);
let mut buf = [0u32; 256];
position_reader.read(128, &mut buf);
for i in 0..256 {
assert_eq!(buf[i], (128 + i) as u32);
}
position_reader.read(128, &mut buf);
for i in 0..256 {
assert_eq!(buf[i], (128 + i) as u32);
}
}
#[test]
#[should_panic(expected = "offset arguments should be increasing.")]
fn test_position_panic_if_called_previous_anchor() {
let v: Vec<u32> = (0..2_000_000).collect();
let (stream, skip) = create_stream_buffer(&v[..]);
assert_eq!(skip.len(), 15_749);
assert_eq!(stream.len(), 4_987_872);
let mut buf = [0u32; 1];
let mut position_reader = PositionReader::new(stream.clone(), skip.clone(), 200_000);
position_reader.read(230, &mut buf);
position_reader.read(9, &mut buf);
}
#[test]
fn test_positions_bug() {
let mut v: Vec<u32> = vec![];
for i in 1..200 {
for j in 0..i {
v.push(j);
}
}
let (stream, skip) = create_stream_buffer(&v[..]);
let mut buf = Vec::new();
let mut position_reader = PositionReader::new(stream.clone(), skip.clone(), 0);
let mut offset = 0;
for i in 1..24 {
buf.resize(i, 0);
position_reader.read(offset, &mut buf[..]);
offset += i as u64;
let r: Vec<u32> = (0..i).map(|el| el as u32).collect();
assert_eq!(buf, &r[..]);
}
}
#[test]
fn test_position_long_skip_const() {
const CONST_VAL: u32 = 9u32;
@@ -177,7 +124,7 @@ pub mod tests {
assert_eq!(stream.len(), 1_000_000);
let mut position_reader = PositionReader::new(stream, skip, 128 * 1024);
let mut buf = [0u32; 1];
position_reader.read(0, &mut buf);
position_reader.read(&mut buf);
assert_eq!(buf[0], CONST_VAL);
}
@@ -196,7 +143,7 @@ pub mod tests {
] {
let mut position_reader = PositionReader::new(stream.clone(), skip.clone(), offset);
let mut buf = [0u32; 1];
position_reader.read(0, &mut buf);
position_reader.read(&mut buf);
assert_eq!(buf[0], offset as u32);
}
}

169
src/positions/reader.rs
View File

@@ -3,6 +3,7 @@ use crate::directory::ReadOnlySource;
use crate::positions::COMPRESSION_BLOCK_SIZE;
use crate::positions::LONG_SKIP_INTERVAL;
use crate::positions::LONG_SKIP_IN_BLOCKS;
use crate::postings::compression::compressed_block_size;
/// Positions works as a long sequence of compressed block.
/// All terms are chained one after the other.
///
@@ -61,34 +62,74 @@ impl Positions {
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);
PositionReader {
let mut position_reader = PositionReader {
bit_packer: self.bit_packer,
skip_read,
position_read,
inner_offset: 0,
buffer: Box::new([0u32; 128]),
block_offset: std::i64::MAX as u64,
anchor_offset: (long_skip_id as u64) * LONG_SKIP_INTERVAL,
abs_offset: offset,
}
ahead: None,
};
position_reader.skip(small_skip);
position_reader
}
}
#[derive(Clone)]
pub struct PositionReader {
skip_read: OwnedRead,
position_read: OwnedRead,
bit_packer: BitPacker4x,
buffer: Box<[u32; COMPRESSION_BLOCK_SIZE]>,
inner_offset: usize,
buffer: Box<[u32; 128]>,
ahead: Option<usize>, // if None, no block is loaded.
// if Some(num_blocks), the block currently loaded is num_blocks ahead
// of the block of the next int to read.
}
block_offset: u64,
anchor_offset: u64,
abs_offset: u64,
// `ahead` represents the offset of the block currently loaded
// compared to the cursor of the actual stream.
//
// By contract, when this function is called, the current block has to be
// decompressed.
//
// If the requested number of els ends exactly at a given block, the next
// block is not decompressed.
fn read_impl(
bit_packer: BitPacker4x,
mut position: &[u8],
buffer: &mut [u32; 128],
mut inner_offset: usize,
num_bits: &[u8],
output: &mut [u32],
) -> usize {
let mut output_start = 0;
let mut output_len = output.len();
let mut ahead = 0;
loop {
let available_len = COMPRESSION_BLOCK_SIZE - 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 {
output[output_start..].copy_from_slice(&buffer[inner_offset..][..output_len]);
return ahead;
}
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;
}
}
impl PositionReader {
@@ -100,65 +141,57 @@ impl PositionReader {
Positions::new(position_source, skip_source).reader(offset)
}
fn advance_num_blocks(&mut self, num_blocks: usize) {
let num_bits: usize = self.skip_read.as_ref()[..num_blocks]
.iter()
.cloned()
.map(|num_bits| num_bits as usize)
.sum();
let num_bytes_to_skip = num_bits * COMPRESSION_BLOCK_SIZE / 8;
self.skip_read.advance(num_blocks as usize);
self.position_read.advance(num_bytes_to_skip);
}
/// Fills a buffer with the positions `[offset..offset+output.len())` integers.
///
/// `offset` is required to have a value >= to the offsets given in previous calls
/// for the given `PositionReaderAbsolute` instance.
pub fn read(&mut self, mut offset: u64, mut output: &mut [u32]) {
offset += self.abs_offset;
assert!(
offset >= self.anchor_offset,
"offset arguments should be increasing."
);
let delta_to_block_offset = offset as i64 - self.block_offset as i64;
if delta_to_block_offset < 0 || delta_to_block_offset >= 128 {
// The first position is not within the first block.
// We need to decompress the first block.
let delta_to_anchor_offset = offset - self.anchor_offset;
let num_blocks_to_skip =
(delta_to_anchor_offset / (COMPRESSION_BLOCK_SIZE as u64)) as usize;
self.advance_num_blocks(num_blocks_to_skip);
self.anchor_offset = offset - (offset % COMPRESSION_BLOCK_SIZE as u64);
self.block_offset = self.anchor_offset;
let num_bits = self.skip_read.get(0);
self.bit_packer
.decompress(self.position_read.as_ref(), self.buffer.as_mut(), num_bits);
} else {
let num_blocks_to_skip =
((self.block_offset - self.anchor_offset) / COMPRESSION_BLOCK_SIZE as u64) as usize;
self.advance_num_blocks(num_blocks_to_skip);
self.anchor_offset = self.block_offset;
}
let mut num_bits = self.skip_read.get(0);
let mut position_data = self.position_read.as_ref();
for i in 1.. {
let offset_in_block = (offset as usize) % COMPRESSION_BLOCK_SIZE;
let remaining_in_block = COMPRESSION_BLOCK_SIZE - offset_in_block;
if remaining_in_block >= output.len() {
output.copy_from_slice(&self.buffer[offset_in_block..][..output.len()]);
break;
}
output[..remaining_in_block].copy_from_slice(&self.buffer[offset_in_block..]);
output = &mut output[remaining_in_block..];
offset += remaining_in_block as u64;
position_data = &position_data[(num_bits as usize * COMPRESSION_BLOCK_SIZE / 8)..];
num_bits = self.skip_read.get(i);
/// Fills a buffer with the next `output.len()` integers.
/// This does not consume / advance the stream.
pub fn read(&mut self, output: &mut [u32]) {
let skip_data = self.skip_read.as_ref();
let position_data = self.position_read.as_ref();
let num_bits = self.skip_read.get(0);
if self.ahead != Some(0) {
// the block currently available is not the block
// for the current position
self.bit_packer
.decompress(position_data, self.buffer.as_mut(), num_bits);
self.block_offset += COMPRESSION_BLOCK_SIZE as u64;
self.ahead = Some(0);
}
let block_len = compressed_block_size(num_bits);
self.ahead = Some(read_impl(
self.bit_packer,
&position_data[block_len..],
self.buffer.as_mut(),
self.inner_offset,
&skip_data[1..],
output,
));
}
/// Skip the next `skip_len` integer.
///
/// If a full block is skipped, calling
/// `.skip(...)` will avoid decompressing it.
///
/// May panic if the end of the stream is reached.
pub fn skip(&mut self, skip_len: usize) {
let skip_len_plus_inner_offset = skip_len + self.inner_offset;
let num_blocks_to_advance = skip_len_plus_inner_offset / COMPRESSION_BLOCK_SIZE;
self.inner_offset = skip_len_plus_inner_offset % COMPRESSION_BLOCK_SIZE;
self.ahead = self.ahead.and_then(|num_blocks| {
if num_blocks >= num_blocks_to_advance {
Some(num_blocks - num_blocks_to_advance)
} else {
None
}
});
let skip_len_in_bits = self.skip_read.as_ref()[..num_blocks_to_advance]
.iter()
.map(|num_bits| *num_bits as usize)
.sum::<usize>()
* COMPRESSION_BLOCK_SIZE;
let skip_len_in_bytes = skip_len_in_bits / 8;
self.skip_read.advance(num_blocks_to_advance);
self.position_read.advance(skip_len_in_bytes);
}
}

40
src/postings/block_search.rs
View File

@@ -87,7 +87,6 @@ fn exponential_search(arr: &[u32], target: u32) -> (usize, usize) {
(begin, end)
}
#[inline(never)]
fn galloping(block_docs: &[u32], target: u32) -> usize {
let (start, end) = exponential_search(&block_docs, target);
start + linear_search(&block_docs[start..end], target)
@@ -107,7 +106,7 @@ impl BlockSearcher {
/// the target.
///
/// The results should be equivalent to
/// ```compile_fail
/// ```ignore
/// block[..]
// .iter()
// .take_while(|&&val| val < target)
@@ -130,18 +129,23 @@ impl BlockSearcher {
///
/// If SSE2 instructions are available in the `(platform, running CPU)`,
/// then we use a different implementation that does an exhaustive linear search over
/// the block regardless of whether the block is full or not.
///
/// Indeed, if the block is not full, the remaining items are TERMINATED.
/// It is surprisingly faster, most likely because of the lack of branch misprediction.
pub(crate) fn search_in_block(self, block_docs: &AlignedBuffer, target: u32) -> usize {
/// the full block whenever the block is full (`len == 128`). It is surprisingly faster, most likely because of the lack
/// of branch.
pub(crate) fn search_in_block(
self,
block_docs: &AlignedBuffer,
len: usize,
start: usize,
target: u32,
) -> usize {
#[cfg(target_arch = "x86_64")]
{
if self == BlockSearcher::SSE2 {
use crate::postings::compression::COMPRESSION_BLOCK_SIZE;
if self == BlockSearcher::SSE2 && len == COMPRESSION_BLOCK_SIZE {
return sse2::linear_search_sse2_128(block_docs, target);
}
}
galloping(&block_docs.0[..], target)
start + galloping(&block_docs.0[start..len], target)
}
}
@@ -162,7 +166,6 @@ mod tests {
use super::exponential_search;
use super::linear_search;
use super::BlockSearcher;
use crate::docset::TERMINATED;
use crate::postings::compression::{AlignedBuffer, COMPRESSION_BLOCK_SIZE};
#[test]
@@ -193,12 +196,19 @@ mod tests {
fn util_test_search_in_block(block_searcher: BlockSearcher, block: &[u32], target: u32) {
let cursor = search_in_block_trivial_but_slow(block, target);
assert!(block.len() < COMPRESSION_BLOCK_SIZE);
let mut output_buffer = [TERMINATED; COMPRESSION_BLOCK_SIZE];
let mut output_buffer = [u32::max_value(); COMPRESSION_BLOCK_SIZE];
output_buffer[..block.len()].copy_from_slice(block);
assert_eq!(
block_searcher.search_in_block(&AlignedBuffer(output_buffer), target),
cursor
);
for i in 0..cursor {
assert_eq!(
block_searcher.search_in_block(
&AlignedBuffer(output_buffer),
block.len(),
i,
target
),
cursor
);
}
}
fn util_test_search_in_block_all(block_searcher: BlockSearcher, block: &[u32]) {

527
src/postings/block_segment_postings.rs
View File

@@ -1,527 +0,0 @@
use crate::common::{BinarySerializable, VInt};
use crate::directory::ReadOnlySource;
use crate::fieldnorm::FieldNormReader;
use crate::postings::compression::{
AlignedBuffer, BlockDecoder, VIntDecoder, COMPRESSION_BLOCK_SIZE,
};
use crate::postings::{BlockInfo, FreqReadingOption, SkipReader};
use crate::query::BM25Weight;
use crate::schema::IndexRecordOption;
use crate::{DocId, Score, TERMINATED};
fn max_score<I: Iterator<Item = Score>>(mut it: I) -> Option<Score> {
if let Some(first) = it.next() {
Some(it.fold(first, Score::max))
} else {
None
}
}
/// `BlockSegmentPostings` is a cursor iterating over blocks
/// of documents.
///
/// # Warning
///
/// While it is useful for some very specific high-performance
/// use cases, you should prefer using `SegmentPostings` for most usage.
#[derive(Clone)]
pub struct BlockSegmentPostings {
pub(crate) doc_decoder: BlockDecoder,
loaded_offset: usize,
freq_decoder: BlockDecoder,
freq_reading_option: FreqReadingOption,
block_max_score_cache: Option<Score>,
doc_freq: u32,
data: ReadOnlySource,
pub(crate) skip_reader: SkipReader,
}
fn decode_bitpacked_block(
doc_decoder: &mut BlockDecoder,
freq_decoder_opt: Option<&mut BlockDecoder>,
data: &[u8],
doc_offset: DocId,
doc_num_bits: u8,
tf_num_bits: u8,
) {
let num_consumed_bytes = doc_decoder.uncompress_block_sorted(data, doc_offset, doc_num_bits);
if let Some(freq_decoder) = freq_decoder_opt {
freq_decoder.uncompress_block_unsorted(&data[num_consumed_bytes..], tf_num_bits);
}
}
fn decode_vint_block(
doc_decoder: &mut BlockDecoder,
freq_decoder_opt: Option<&mut BlockDecoder>,
data: &[u8],
doc_offset: DocId,
num_vint_docs: usize,
) {
let num_consumed_bytes =
doc_decoder.uncompress_vint_sorted(data, doc_offset, num_vint_docs, TERMINATED);
if let Some(freq_decoder) = freq_decoder_opt {
freq_decoder.uncompress_vint_unsorted(
&data[num_consumed_bytes..],
num_vint_docs,
TERMINATED,
);
}
}
fn split_into_skips_and_postings(
doc_freq: u32,
data: ReadOnlySource,
) -> (Option<ReadOnlySource>, ReadOnlySource) {
if doc_freq < COMPRESSION_BLOCK_SIZE as u32 {
return (None, data);
}
let mut data_byte_arr = data.as_slice();
let skip_len = VInt::deserialize(&mut data_byte_arr)
.expect("Data corrupted")
.0 as usize;
let vint_len = data.len() - data_byte_arr.len();
let (skip_data, postings_data) = data.slice_from(vint_len).split(skip_len);
(Some(skip_data), postings_data)
}
impl BlockSegmentPostings {
pub(crate) fn from_data(
doc_freq: u32,
data: ReadOnlySource,
record_option: IndexRecordOption,
requested_option: IndexRecordOption,
) -> BlockSegmentPostings {
let freq_reading_option = match (record_option, requested_option) {
(IndexRecordOption::Basic, _) => FreqReadingOption::NoFreq,
(_, IndexRecordOption::Basic) => FreqReadingOption::SkipFreq,
(_, _) => FreqReadingOption::ReadFreq,
};
let (skip_data_opt, postings_data) = split_into_skips_and_postings(doc_freq, data);
let skip_reader = match skip_data_opt {
Some(skip_data) => SkipReader::new(skip_data, doc_freq, record_option),
None => SkipReader::new(ReadOnlySource::empty(), doc_freq, record_option),
};
let mut block_segment_postings = BlockSegmentPostings {
doc_decoder: BlockDecoder::with_val(TERMINATED),
loaded_offset: std::usize::MAX,
freq_decoder: BlockDecoder::with_val(1),
freq_reading_option,
block_max_score_cache: None,
doc_freq,
data: postings_data,
skip_reader,
};
block_segment_postings.load_block();
block_segment_postings
}
/// Returns the block_max_score for the current block.
/// It does not require the block to be loaded. For instance, it is ok to call this method
/// after having called `.shallow_advance(..)`.
///
/// See `TermScorer::block_max_score(..)` for more information.
pub fn block_max_score(
&mut self,
fieldnorm_reader: &FieldNormReader,
bm25_weight: &BM25Weight,
) -> Score {
if let Some(score) = self.block_max_score_cache {
return score;
}
if let Some(skip_reader_max_score) = self.skip_reader.block_max_score(bm25_weight) {
// if we are on a full block, the skip reader should have the block max information
// for us
self.block_max_score_cache = Some(skip_reader_max_score);
return skip_reader_max_score;
}
// this is the last block of the segment posting list.
// If it is actually loaded, we can compute block max manually.
if self.block_is_loaded() {
let docs = self.doc_decoder.output_array().iter().cloned();
let freqs = self.freq_decoder.output_array().iter().cloned();
let bm25_scores = docs.zip(freqs).map(|(doc, term_freq)| {
let fieldnorm_id = fieldnorm_reader.fieldnorm_id(doc);
bm25_weight.score(fieldnorm_id, term_freq)
});
let block_max_score = max_score(bm25_scores).unwrap_or(0.0);
self.block_max_score_cache = Some(block_max_score);
return block_max_score;
}
// We do not have access to any good block max value. We return bm25_weight.max_score()
// as it is a valid upperbound.
//
// We do not cache it however, so that it gets computed when once block is loaded.
bm25_weight.max_score()
}
pub(crate) fn freq_reading_option(&self) -> FreqReadingOption {
self.freq_reading_option
}
// Resets the block segment postings on another position
// in the postings file.
//
// This is useful for enumerating through a list of terms,
// and consuming the associated posting lists while avoiding
// reallocating a `BlockSegmentPostings`.
//
// # Warning
//
// This does not reset the positions list.
pub(crate) fn reset(&mut self, doc_freq: u32, postings_data: ReadOnlySource) {
let (skip_data_opt, postings_data) = split_into_skips_and_postings(doc_freq, postings_data);
self.data = ReadOnlySource::new(postings_data);
self.block_max_score_cache = None;
self.loaded_offset = std::usize::MAX;
if let Some(skip_data) = skip_data_opt {
self.skip_reader.reset(skip_data, doc_freq);
} else {
self.skip_reader.reset(ReadOnlySource::empty(), doc_freq);
}
self.doc_freq = doc_freq;
self.load_block();
}
/// Returns the overall number of documents in the block postings.
/// It does not take in account whether documents are deleted or not.
///
/// This `doc_freq` is simply the sum of the length of all of the blocks
/// length, and it does not take in account deleted documents.
pub fn doc_freq(&self) -> u32 {
self.doc_freq
}
/// Returns the array of docs in the current block.
///
/// Before the first call to `.advance()`, the block
/// returned by `.docs()` is empty.
#[inline]
pub fn docs(&self) -> &[DocId] {
debug_assert!(self.block_is_loaded());
self.doc_decoder.output_array()
}
/// Returns a full block, regardless of whetehr the block is complete or incomplete (
/// as it happens for the last block of the posting list).
///
/// In the latter case, the block is guaranteed to be padded with the sentinel value:
/// `TERMINATED`. The array is also guaranteed to be aligned on 16 bytes = 128 bits.
///
/// This method is useful to run SSE2 linear search.
#[inline(always)]
pub(crate) fn docs_aligned(&self) -> &AlignedBuffer {
debug_assert!(self.block_is_loaded());
self.doc_decoder.output_aligned()
}
/// Return the document at index `idx` of the block.
#[inline(always)]
pub fn doc(&self, idx: usize) -> u32 {
self.doc_decoder.output(idx)
}
/// Return the array of `term freq` in the block.
#[inline]
pub fn freqs(&self) -> &[u32] {
debug_assert!(self.block_is_loaded());
self.freq_decoder.output_array()
}
/// Return the frequency at index `idx` of the block.
#[inline]
pub fn freq(&self, idx: usize) -> u32 {
debug_assert!(self.block_is_loaded());
self.freq_decoder.output(idx)
}
/// Returns the length of the current block.
///
/// All blocks have a length of `NUM_DOCS_PER_BLOCK`,
/// except the last block that may have a length
/// of any number between 1 and `NUM_DOCS_PER_BLOCK - 1`
#[inline]
pub fn block_len(&self) -> usize {
debug_assert!(self.block_is_loaded());
self.doc_decoder.output_len
}
/// Position on a block that may contains `target_doc`.
///
/// If all docs are smaller than target, the block loaded may be empty,
/// or be the last an incomplete VInt block.
pub fn seek(&mut self, target_doc: DocId) {
self.shallow_seek(target_doc);
self.load_block();
}
pub(crate) fn position_offset(&self) -> u64 {
self.skip_reader.position_offset()
}
/// Dangerous API! This calls seek on the skip list,
/// but does not `.load_block()` afterwards.
///
/// `.load_block()` needs to be called manually afterwards.
/// If all docs are smaller than target, the block loaded may be empty,
/// or be the last an incomplete VInt block.
pub(crate) fn shallow_seek(&mut self, target_doc: DocId) {
if self.skip_reader.seek(target_doc) {
self.block_max_score_cache = None;
}
}
pub(crate) fn block_is_loaded(&self) -> bool {
self.loaded_offset == self.skip_reader.byte_offset()
}
pub(crate) fn load_block(&mut self) {
let offset = self.skip_reader.byte_offset();
if self.loaded_offset == offset {
return;
}
self.loaded_offset = offset;
match self.skip_reader.block_info() {
BlockInfo::BitPacked {
doc_num_bits,
tf_num_bits,
..
} => {
decode_bitpacked_block(
&mut self.doc_decoder,
if let FreqReadingOption::ReadFreq = self.freq_reading_option {
Some(&mut self.freq_decoder)
} else {
None
},
&self.data.as_slice()[offset..],
self.skip_reader.last_doc_in_previous_block,
doc_num_bits,
tf_num_bits,
);
}
BlockInfo::VInt { num_docs } => {
let data = {
if num_docs == 0 {
&[]
} else {
&self.data.as_slice()[offset..]
}
};
decode_vint_block(
&mut self.doc_decoder,
if let FreqReadingOption::ReadFreq = self.freq_reading_option {
Some(&mut self.freq_decoder)
} else {
None
},
data,
self.skip_reader.last_doc_in_previous_block,
num_docs as usize,
);
}
}
}
/// Advance to the next block.
///
/// Returns false iff there was no remaining blocks.
pub fn advance(&mut self) {
self.skip_reader.advance();
self.block_max_score_cache = None;
self.load_block();
}
/// Returns an empty segment postings object
pub fn empty() -> BlockSegmentPostings {
BlockSegmentPostings {
doc_decoder: BlockDecoder::with_val(TERMINATED),
loaded_offset: 0,
freq_decoder: BlockDecoder::with_val(1),
freq_reading_option: FreqReadingOption::NoFreq,
block_max_score_cache: None,
doc_freq: 0,
data: ReadOnlySource::new(vec![]),
skip_reader: SkipReader::new(ReadOnlySource::new(vec![]), 0, IndexRecordOption::Basic),
}
}
}
#[cfg(test)]
mod tests {
use super::BlockSegmentPostings;
use crate::common::HasLen;
use crate::core::Index;
use crate::docset::{DocSet, TERMINATED};
use crate::postings::compression::COMPRESSION_BLOCK_SIZE;
use crate::postings::postings::Postings;
use crate::postings::SegmentPostings;
use crate::schema::IndexRecordOption;
use crate::schema::Schema;
use crate::schema::Term;
use crate::schema::INDEXED;
use crate::DocId;
#[test]
fn test_empty_segment_postings() {
let mut postings = SegmentPostings::empty();
assert_eq!(postings.doc(), TERMINATED);
assert_eq!(postings.advance(), TERMINATED);
assert_eq!(postings.advance(), TERMINATED);
assert_eq!(postings.doc_freq(), 0);
assert_eq!(postings.len(), 0);
}
#[test]
fn test_empty_postings_doc_returns_terminated() {
let mut postings = SegmentPostings::empty();
assert_eq!(postings.doc(), TERMINATED);
assert_eq!(postings.advance(), TERMINATED);
}
#[test]
fn test_empty_postings_doc_term_freq_returns_0() {
let postings = SegmentPostings::empty();
assert_eq!(postings.term_freq(), 1);
}
#[test]
fn test_empty_block_segment_postings() {
let mut postings = BlockSegmentPostings::empty();
assert!(postings.docs().is_empty());
assert_eq!(postings.doc_freq(), 0);
postings.advance();
assert!(postings.docs().is_empty());
assert_eq!(postings.doc_freq(), 0);
}
#[test]
fn test_block_segment_postings() {
let mut block_segments = build_block_postings(&(0..100_000).collect::<Vec<u32>>());
let mut offset: u32 = 0u32;
// checking that the `doc_freq` is correct
assert_eq!(block_segments.doc_freq(), 100_000);
loop {
let block = block_segments.docs();
if block.is_empty() {
break;
}
for (i, doc) in block.iter().cloned().enumerate() {
assert_eq!(offset + (i as u32), doc);
}
offset += block.len() as u32;
block_segments.advance();
}
}
#[test]
fn test_skip_right_at_new_block() {
let mut doc_ids = (0..128).collect::<Vec<u32>>();
// 128 is missing
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.seek(128), 129);
assert_eq!(docset.doc(), 129);
assert_eq!(docset.advance(), 130);
assert_eq!(docset.doc(), 130);
assert_eq!(docset.advance(), TERMINATED);
}
{
let block_segments = build_block_postings(&doc_ids);
let mut docset = SegmentPostings::from_block_postings(block_segments, None);
assert_eq!(docset.seek(129), 129);
assert_eq!(docset.doc(), 129);
assert_eq!(docset.advance(), 130);
assert_eq!(docset.doc(), 130);
assert_eq!(docset.advance(), TERMINATED);
}
{
let block_segments = build_block_postings(&doc_ids);
let mut docset = SegmentPostings::from_block_postings(block_segments, None);
assert_eq!(docset.doc(), 0);
assert_eq!(docset.seek(131), TERMINATED);
assert_eq!(docset.doc(), TERMINATED);
}
}
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 index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
let mut last_doc = 0u32;
for &doc in docs {
for _ in last_doc..doc {
index_writer.add_document(doc!(int_field=>1u64));
}
index_writer.add_document(doc!(int_field=>0u64));
last_doc = doc + 1;
}
index_writer.commit().unwrap();
let searcher = index.reader().unwrap().searcher();
let segment_reader = searcher.segment_reader(0);
let inverted_index = segment_reader.inverted_index(int_field);
let term = Term::from_field_u64(int_field, 0u64);
let term_info = inverted_index.get_term_info(&term).unwrap();
inverted_index.read_block_postings_from_terminfo(&term_info, IndexRecordOption::Basic)
}
#[test]
fn test_block_segment_postings_seek() {
let mut docs = vec![0];
for i in 0..1300 {
docs.push((i * i / 100) + i);
}
let mut block_postings = build_block_postings(&docs[..]);
for i in vec![0, 424, 10000] {
block_postings.seek(i);
let docs = block_postings.docs();
assert!(docs[0] <= i);
assert!(docs.last().cloned().unwrap_or(0u32) >= i);
}
block_postings.seek(100_000);
assert_eq!(block_postings.doc(COMPRESSION_BLOCK_SIZE - 1), TERMINATED);
}
#[test]
fn test_reset_block_segment_postings() {
let mut schema_builder = Schema::builder();
let int_field = schema_builder.add_u64_field("id", INDEXED);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
// create two postings list, one containg even number,
// the other containing odd numbers.
for i in 0..6 {
let doc = doc!(int_field=> (i % 2) as u64);
index_writer.add_document(doc);
}
index_writer.commit().unwrap();
let searcher = index.reader().unwrap().searcher();
let segment_reader = searcher.segment_reader(0);
let mut block_segments;
{
let term = Term::from_field_u64(int_field, 0u64);
let inverted_index = segment_reader.inverted_index(int_field);
let term_info = inverted_index.get_term_info(&term).unwrap();
block_segments = inverted_index
.read_block_postings_from_terminfo(&term_info, IndexRecordOption::Basic);
}
assert_eq!(block_segments.docs(), &[0, 2, 4]);
{
let term = Term::from_field_u64(int_field, 1u64);
let inverted_index = segment_reader.inverted_index(int_field);
let term_info = inverted_index.get_term_info(&term).unwrap();
inverted_index.reset_block_postings_from_terminfo(&term_info, &mut block_segments);
}
assert_eq!(block_segments.docs(), &[1, 3, 5]);
}
}

101
src/postings/compression/mod.rs
View File

@@ -17,12 +17,6 @@ pub struct BlockEncoder {
pub output_len: usize,
}
impl Default for BlockEncoder {
fn default() -> Self {
BlockEncoder::new()
}
}
impl BlockEncoder {
pub fn new() -> BlockEncoder {
BlockEncoder {
@@ -52,23 +46,19 @@ impl BlockEncoder {
/// We ensure that the OutputBuffer is align on 128 bits
/// in order to run SSE2 linear search on it.
#[repr(align(128))]
#[derive(Clone)]
pub(crate) struct AlignedBuffer(pub [u32; COMPRESSION_BLOCK_SIZE]);
#[derive(Clone)]
pub struct BlockDecoder {
bitpacker: BitPacker4x,
output: AlignedBuffer,
pub output_len: usize,
}
impl Default for BlockDecoder {
fn default() -> Self {
impl BlockDecoder {
pub fn new() -> BlockDecoder {
BlockDecoder::with_val(0u32)
}
}
impl BlockDecoder {
pub fn with_val(val: u32) -> BlockDecoder {
BlockDecoder {
bitpacker: BitPacker4x::new(),
@@ -100,8 +90,8 @@ impl BlockDecoder {
}
#[inline]
pub(crate) fn output_aligned(&self) -> &AlignedBuffer {
&self.output
pub(crate) fn output_aligned(&self) -> (&AlignedBuffer, usize) {
(&self.output, self.output_len)
}
#[inline]
@@ -144,14 +134,11 @@ pub trait VIntDecoder {
/// For instance, if delta encoded are `1, 3, 9`, and the
/// `offset` is 5, then the output will be:
/// `5 + 1 = 6, 6 + 3= 9, 9 + 9 = 18`
///
/// The value given in `padding` will be used to fill the remaining `128 - num_els` values.
fn uncompress_vint_sorted(
fn uncompress_vint_sorted<'a>(
&mut self,
compressed_data: &[u8],
compressed_data: &'a [u8],
offset: u32,
num_els: usize,
padding: u32,
) -> usize;
/// Uncompress an array of `u32s`, compressed using variable
@@ -159,14 +146,7 @@ pub trait VIntDecoder {
///
/// The method takes a number of int to decompress, and returns
/// the amount of bytes that were read to decompress them.
///
/// The value given in `padding` will be used to fill the remaining `128 - num_els` values.
fn uncompress_vint_unsorted(
&mut self,
compressed_data: &[u8],
num_els: usize,
padding: u32,
) -> usize;
fn uncompress_vint_unsorted<'a>(&mut self, compressed_data: &'a [u8], num_els: usize) -> usize;
}
impl VIntEncoder for BlockEncoder {
@@ -180,26 +160,18 @@ impl VIntEncoder for BlockEncoder {
}
impl VIntDecoder for BlockDecoder {
fn uncompress_vint_sorted(
fn uncompress_vint_sorted<'a>(
&mut self,
compressed_data: &[u8],
compressed_data: &'a [u8],
offset: u32,
num_els: usize,
padding: u32,
) -> usize {
self.output_len = num_els;
self.output.0.iter_mut().for_each(|el| *el = padding);
vint::uncompress_sorted(compressed_data, &mut self.output.0[..num_els], offset)
}
fn uncompress_vint_unsorted(
&mut self,
compressed_data: &[u8],
num_els: usize,
padding: u32,
) -> usize {
fn uncompress_vint_unsorted<'a>(&mut self, compressed_data: &'a [u8], num_els: usize) -> usize {
self.output_len = num_els;
self.output.0.iter_mut().for_each(|el| *el = padding);
vint::uncompress_unsorted(compressed_data, &mut self.output.0[..num_els])
}
}
@@ -208,14 +180,13 @@ impl VIntDecoder for BlockDecoder {
pub mod tests {
use super::*;
use crate::TERMINATED;
#[test]
fn test_encode_sorted_block() {
let vals: Vec<u32> = (0u32..128u32).map(|i| i * 7).collect();
let mut encoder = BlockEncoder::new();
let (num_bits, compressed_data) = encoder.compress_block_sorted(&vals, 0);
let mut decoder = BlockDecoder::default();
let mut decoder = BlockDecoder::new();
{
let consumed_num_bytes = decoder.uncompress_block_sorted(compressed_data, 0, num_bits);
assert_eq!(consumed_num_bytes, compressed_data.len());
@@ -228,9 +199,9 @@ pub mod tests {
#[test]
fn test_encode_sorted_block_with_offset() {
let vals: Vec<u32> = (0u32..128u32).map(|i| 11 + i * 7).collect();
let mut encoder = BlockEncoder::default();
let mut encoder = BlockEncoder::new();
let (num_bits, compressed_data) = encoder.compress_block_sorted(&vals, 10);
let mut decoder = BlockDecoder::default();
let mut decoder = BlockDecoder::new();
{
let consumed_num_bytes = decoder.uncompress_block_sorted(compressed_data, 10, num_bits);
assert_eq!(consumed_num_bytes, compressed_data.len());
@@ -245,11 +216,11 @@ pub mod tests {
let mut compressed: Vec<u8> = Vec::new();
let n = 128;
let vals: Vec<u32> = (0..n).map(|i| 11u32 + (i as u32) * 7u32).collect();
let mut encoder = BlockEncoder::default();
let mut encoder = BlockEncoder::new();
let (num_bits, compressed_data) = encoder.compress_block_sorted(&vals, 10);
compressed.extend_from_slice(compressed_data);
compressed.push(173u8);
let mut decoder = BlockDecoder::default();
let mut decoder = BlockDecoder::new();
{
let consumed_num_bytes = decoder.uncompress_block_sorted(&compressed, 10, num_bits);
assert_eq!(consumed_num_bytes, compressed.len() - 1);
@@ -265,11 +236,11 @@ pub mod tests {
let mut compressed: Vec<u8> = Vec::new();
let n = 128;
let vals: Vec<u32> = (0..n).map(|i| 11u32 + (i as u32) * 7u32 % 12).collect();
let mut encoder = BlockEncoder::default();
let mut encoder = BlockEncoder::new();
let (num_bits, compressed_data) = encoder.compress_block_unsorted(&vals);
compressed.extend_from_slice(compressed_data);
compressed.push(173u8);
let mut decoder = BlockDecoder::default();
let mut decoder = BlockDecoder::new();
{
let consumed_num_bytes = decoder.uncompress_block_unsorted(&compressed, num_bits);
assert_eq!(consumed_num_bytes + 1, compressed.len());
@@ -280,27 +251,20 @@ pub mod tests {
}
}
#[test]
fn test_block_decoder_initialization() {
let block = BlockDecoder::with_val(TERMINATED);
assert_eq!(block.output(0), TERMINATED);
}
#[test]
fn test_encode_vint() {
const PADDING_VALUE: u32 = 234_234_345u32;
let expected_length = 154;
let mut encoder = BlockEncoder::new();
let input: Vec<u32> = (0u32..123u32).map(|i| 4 + i * 7 / 2).into_iter().collect();
for offset in &[0u32, 1u32, 2u32] {
let encoded_data = encoder.compress_vint_sorted(&input, *offset);
assert!(encoded_data.len() <= expected_length);
let mut decoder = BlockDecoder::default();
let consumed_num_bytes =
decoder.uncompress_vint_sorted(&encoded_data, *offset, input.len(), PADDING_VALUE);
assert_eq!(consumed_num_bytes, encoded_data.len());
assert_eq!(input, decoder.output_array());
for i in input.len()..COMPRESSION_BLOCK_SIZE {
assert_eq!(decoder.output(i), PADDING_VALUE);
{
let expected_length = 154;
let mut encoder = BlockEncoder::new();
let input: Vec<u32> = (0u32..123u32).map(|i| 4 + i * 7 / 2).into_iter().collect();
for offset in &[0u32, 1u32, 2u32] {
let encoded_data = encoder.compress_vint_sorted(&input, *offset);
assert!(encoded_data.len() <= expected_length);
let mut decoder = BlockDecoder::new();
let consumed_num_bytes =
decoder.uncompress_vint_sorted(&encoded_data, *offset, input.len());
assert_eq!(consumed_num_bytes, encoded_data.len());
assert_eq!(input, decoder.output_array());
}
}
}
@@ -310,7 +274,6 @@ pub mod tests {
mod bench {
use super::*;
use crate::TERMINATED;
use rand::rngs::StdRng;
use rand::Rng;
use rand::SeedableRng;
@@ -341,7 +304,7 @@ mod bench {
let mut encoder = BlockEncoder::new();
let data = generate_array(COMPRESSION_BLOCK_SIZE, 0.1);
let (num_bits, compressed) = encoder.compress_block_sorted(&data, 0u32);
let mut decoder = BlockDecoder::default();
let mut decoder = BlockDecoder::new();
b.iter(|| {
decoder.uncompress_block_sorted(compressed, 0u32, num_bits);
});
@@ -376,9 +339,9 @@ mod bench {
let mut encoder = BlockEncoder::new();
let data = generate_array(NUM_INTS_BENCH_VINT, 0.001);
let compressed = encoder.compress_vint_sorted(&data, 0u32);
let mut decoder = BlockDecoder::default();
let mut decoder = BlockDecoder::new();
b.iter(|| {
decoder.uncompress_vint_sorted(compressed, 0u32, NUM_INTS_BENCH_VINT, TERMINATED);
decoder.uncompress_vint_sorted(compressed, 0u32, NUM_INTS_BENCH_VINT);
});
}
}

2
src/postings/compression/vint.rs
View File

@@ -42,7 +42,7 @@ pub(crate) fn compress_unsorted<'a>(input: &[u32], output: &'a mut [u8]) -> &'a
}
#[inline(always)]
pub fn uncompress_sorted(compressed_data: &[u8], output: &mut [u32], offset: u32) -> usize {
pub fn uncompress_sorted<'a>(compressed_data: &'a [u8], output: &mut [u32], offset: u32) -> usize {
let mut read_byte = 0;
let mut result = offset;
for output_mut in output.iter_mut() {

247
src/postings/mod.rs
View File

@@ -3,8 +3,11 @@ Postings module (also called inverted index)
*/
mod block_search;
mod block_segment_postings;
pub(crate) mod compression;
/// Postings module
///
/// Postings, also called inverted lists, is the key datastructure
/// to full-text search.
mod postings;
mod postings_writer;
mod recorder;
@@ -19,17 +22,18 @@ pub(crate) use self::block_search::BlockSearcher;
pub(crate) use self::postings_writer::MultiFieldPostingsWriter;
pub use self::serializer::{FieldSerializer, InvertedIndexSerializer};
use self::compression::COMPRESSION_BLOCK_SIZE;
pub use self::postings::Postings;
pub(crate) use self::skip::{BlockInfo, SkipReader};
pub(crate) use self::skip::SkipReader;
pub use self::term_info::TermInfo;
pub use self::block_segment_postings::BlockSegmentPostings;
pub use self::segment_postings::SegmentPostings;
pub use self::segment_postings::{BlockSegmentPostings, SegmentPostings};
pub(crate) use self::stacker::compute_table_size;
pub use crate::common::HasLen;
pub(crate) const USE_SKIP_INFO_LIMIT: u32 = COMPRESSION_BLOCK_SIZE as u32;
pub(crate) type UnorderedTermId = u64;
#[cfg_attr(feature = "cargo-clippy", allow(clippy::enum_variant_names))]
@@ -47,7 +51,7 @@ pub mod tests {
use crate::core::Index;
use crate::core::SegmentComponent;
use crate::core::SegmentReader;
use crate::docset::{DocSet, TERMINATED};
use crate::docset::{DocSet, SkipResult};
use crate::fieldnorm::FieldNormReader;
use crate::indexer::operation::AddOperation;
use crate::indexer::SegmentWriter;
@@ -65,42 +69,45 @@ pub mod tests {
use std::iter;
#[test]
pub fn test_position_write() -> crate::Result<()> {
pub fn test_position_write() {
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);
let mut segment = index.new_segment();
let mut posting_serializer = InvertedIndexSerializer::open(&mut segment)?;
let mut field_serializer = posting_serializer.new_field(text_field, 120 * 4, None)?;
field_serializer.new_term("abc".as_bytes(), 12u32)?;
for doc_id in 0u32..120u32 {
let delta_positions = vec![1, 2, 3, 2];
field_serializer.write_doc(doc_id, 4, &delta_positions)?;
let mut posting_serializer = InvertedIndexSerializer::open(&mut segment).unwrap();
{
let mut field_serializer = posting_serializer.new_field(text_field, 120 * 4).unwrap();
field_serializer.new_term("abc".as_bytes()).unwrap();
for doc_id in 0u32..120u32 {
let delta_positions = vec![1, 2, 3, 2];
field_serializer
.write_doc(doc_id, 4, &delta_positions)
.unwrap();
}
field_serializer.close_term().unwrap();
}
field_serializer.close_term()?;
posting_serializer.close()?;
let read = segment.open_read(SegmentComponent::POSITIONS)?;
posting_serializer.close().unwrap();
let read = segment.open_read(SegmentComponent::POSITIONS).unwrap();
assert!(read.len() <= 140);
Ok(())
}
#[test]
pub fn test_skip_positions() -> crate::Result<()> {
pub fn test_skip_positions() {
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_for_tests()?;
let mut index_writer = index.writer_with_num_threads(1, 30_000_000).unwrap();
index_writer.add_document(doc!(title => r#"abc abc abc"#));
index_writer.add_document(doc!(title => r#"abc be be be be abc"#));
for _ in 0..1_000 {
index_writer.add_document(doc!(title => r#"abc abc abc"#));
}
index_writer.add_document(doc!(title => r#"abc be be be be abc"#));
index_writer.commit()?;
index_writer.commit().unwrap();
let searcher = index.reader()?.searcher();
let searcher = index.reader().unwrap().searcher();
let inverted_index = searcher.segment_reader(0u32).inverted_index(title);
let term = Term::from_field_text(title, "abc");
let mut positions = Vec::new();
@@ -108,12 +115,29 @@ pub mod tests {
let mut postings = inverted_index
.read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
.unwrap();
assert_eq!(postings.doc(), 0);
postings.advance();
postings.positions(&mut positions);
assert_eq!(&[0, 1, 2], &positions[..]);
postings.positions(&mut positions);
assert_eq!(&[0, 1, 2], &positions[..]);
assert_eq!(postings.advance(), 1);
postings.advance();
postings.positions(&mut positions);
assert_eq!(&[0, 5], &positions[..]);
}
{
let mut postings = inverted_index
.read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
.unwrap();
postings.advance();
postings.advance();
postings.positions(&mut positions);
assert_eq!(&[0, 5], &positions[..]);
}
{
let mut postings = inverted_index
.read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
.unwrap();
assert_eq!(postings.skip_next(1), SkipResult::Reached);
assert_eq!(postings.doc(), 1);
postings.positions(&mut positions);
assert_eq!(&[0, 5], &positions[..]);
@@ -122,25 +146,7 @@ pub mod tests {
let mut postings = inverted_index
.read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
.unwrap();
assert_eq!(postings.doc(), 0);
assert_eq!(postings.advance(), 1);
postings.positions(&mut positions);
assert_eq!(&[0, 5], &positions[..]);
}
{
let mut postings = inverted_index
.read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
.unwrap();
assert_eq!(postings.seek(1), 1);
assert_eq!(postings.doc(), 1);
postings.positions(&mut positions);
assert_eq!(&[0, 5], &positions[..]);
}
{
let mut postings = inverted_index
.read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
.unwrap();
assert_eq!(postings.seek(1002), 1002);
assert_eq!(postings.skip_next(1002), SkipResult::Reached);
assert_eq!(postings.doc(), 1002);
postings.positions(&mut positions);
assert_eq!(&[0, 5], &positions[..]);
@@ -149,13 +155,12 @@ pub mod tests {
let mut postings = inverted_index
.read_postings(&term, IndexRecordOption::WithFreqsAndPositions)
.unwrap();
assert_eq!(postings.seek(100), 100);
assert_eq!(postings.seek(1002), 1002);
assert_eq!(postings.skip_next(100), SkipResult::Reached);
assert_eq!(postings.skip_next(1002), SkipResult::Reached);
assert_eq!(postings.doc(), 1002);
postings.positions(&mut positions);
assert_eq!(&[0, 5], &positions[..]);
}
Ok(())
}
#[test]
@@ -176,7 +181,7 @@ pub mod tests {
.tokenizers()
.register("simple_no_truncation", SimpleTokenizer);
let reader = index.reader().unwrap();
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.set_merge_policy(Box::new(NoMergePolicy));
{
index_writer.add_document(doc!(text_field=>exceeding_token_text));
@@ -205,7 +210,7 @@ pub mod tests {
}
#[test]
pub fn test_position_and_fieldnorm1() -> crate::Result<()> {
pub fn test_position_and_fieldnorm1() {
let mut positions = Vec::new();
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("text", TEXT);
@@ -217,38 +222,42 @@ pub mod tests {
let mut segment_writer =
SegmentWriter::for_segment(3_000_000, segment.clone(), &schema).unwrap();
{
let mut doc = Document::default();
// checking that position works if the field has two values
doc.add_text(text_field, "a b a c a d a a.");
doc.add_text(text_field, "d d d d a");
let op = AddOperation {
opstamp: 0u64,
document: doc!(
text_field => "a b a c a d a a.",
text_field => "d d d d a"
),
document: doc,
};
segment_writer.add_document(op, &schema)?;
segment_writer.add_document(op, &schema).unwrap();
}
{
let mut doc = Document::default();
doc.add_text(text_field, "b a");
let op = AddOperation {
opstamp: 1u64,
document: doc!(text_field => "b a"),
document: doc,
};
segment_writer.add_document(op, &schema).unwrap();
}
for i in 2..1000 {
let mut text: String = iter::repeat("e ").take(i).collect();
let mut doc = Document::default();
let mut text = iter::repeat("e ").take(i).collect::<String>();
text.push_str(" a");
doc.add_text(text_field, &text);
let op = AddOperation {
opstamp: 2u64,
document: doc!(text_field => text),
document: doc,
};
segment_writer.add_document(op, &schema).unwrap();
}
segment_writer.finalize()?;
segment_writer.finalize().unwrap();
}
{
let segment_reader = SegmentReader::open(&segment)?;
let segment_reader = SegmentReader::open(&segment).unwrap();
{
let fieldnorm_reader = segment_reader.get_fieldnorms_reader(text_field)?;
let fieldnorm_reader = segment_reader.get_fieldnorms_reader(text_field);
assert_eq!(fieldnorm_reader.fieldnorm(0), 8 + 5);
assert_eq!(fieldnorm_reader.fieldnorm(1), 2);
for i in 2..1000 {
@@ -272,21 +281,22 @@ pub mod tests {
.read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions)
.unwrap();
assert_eq!(postings_a.len(), 1000);
assert!(postings_a.advance());
assert_eq!(postings_a.doc(), 0);
assert_eq!(postings_a.term_freq(), 6);
postings_a.positions(&mut positions);
assert_eq!(&positions[..], [0, 2, 4, 6, 7, 13]);
assert_eq!(postings_a.advance(), 1u32);
assert!(postings_a.advance());
assert_eq!(postings_a.doc(), 1u32);
assert_eq!(postings_a.term_freq(), 1);
for i in 2u32..1000u32 {
assert_eq!(postings_a.advance(), i);
assert!(postings_a.advance());
assert_eq!(postings_a.term_freq(), 1);
postings_a.positions(&mut positions);
assert_eq!(&positions[..], [i]);
assert_eq!(postings_a.doc(), i);
}
assert_eq!(postings_a.advance(), TERMINATED);
assert!(!postings_a.advance());
}
{
let term_e = Term::from_field_text(text_field, "e");
@@ -296,6 +306,7 @@ pub mod tests {
.unwrap();
assert_eq!(postings_e.len(), 1000 - 2);
for i in 2u32..1000u32 {
assert!(postings_e.advance());
assert_eq!(postings_e.term_freq(), i);
postings_e.positions(&mut positions);
assert_eq!(positions.len(), i as usize);
@@ -303,12 +314,10 @@ pub mod tests {
assert_eq!(positions[j], (j as u32));
}
assert_eq!(postings_e.doc(), i);
postings_e.advance();
}
assert_eq!(postings_e.doc(), TERMINATED);
assert!(!postings_e.advance());
}
}
Ok(())
}
#[test]
@@ -319,9 +328,17 @@ pub mod tests {
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{
let mut index_writer = index.writer_for_tests().unwrap();
index_writer.add_document(doc!(text_field => "g b b d c g c"));
index_writer.add_document(doc!(text_field => "g a b b a d c g c"));
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
{
let mut doc = Document::default();
doc.add_text(text_field, "g b b d c g c");
index_writer.add_document(doc);
}
{
let mut doc = Document::default();
doc.add_text(text_field, "g a b b a d c g c");
index_writer.add_document(doc);
}
assert!(index_writer.commit().is_ok());
}
let term_a = Term::from_field_text(text_field, "a");
@@ -331,6 +348,7 @@ pub mod tests {
.inverted_index(text_field)
.read_postings(&term_a, IndexRecordOption::WithFreqsAndPositions)
.unwrap();
assert!(postings.advance());
assert_eq!(postings.doc(), 1u32);
postings.positions(&mut positions);
assert_eq!(&positions[..], &[1u32, 4]);
@@ -351,9 +369,12 @@ pub mod tests {
let index = Index::create_in_ram(schema);
{
let mut index_writer = index.writer_for_tests().unwrap();
for i in 0u64..num_docs as u64 {
let doc = doc!(value_field => 2u64, value_field => i % 2u64);
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
for i in 0..num_docs {
let mut doc = Document::default();
doc.add_u64(value_field, 2);
doc.add_u64(value_field, (i % 2) as u64);
index_writer.add_document(doc);
}
assert!(index_writer.commit().is_ok());
@@ -370,10 +391,11 @@ pub mod tests {
.inverted_index(term_2.field())
.read_postings(&term_2, IndexRecordOption::Basic)
.unwrap();
assert_eq!(segment_postings.seek(i), i);
assert_eq!(segment_postings.skip_next(i), SkipResult::Reached);
assert_eq!(segment_postings.doc(), i);
assert_eq!(segment_postings.seek(j), j);
assert_eq!(segment_postings.skip_next(j), SkipResult::Reached);
assert_eq!(segment_postings.doc(), j);
}
}
@@ -385,16 +407,17 @@ pub mod tests {
.unwrap();
// check that `skip_next` advances the iterator
assert!(segment_postings.advance());
assert_eq!(segment_postings.doc(), 0);
assert_eq!(segment_postings.seek(1), 1);
assert_eq!(segment_postings.skip_next(1), SkipResult::Reached);
assert_eq!(segment_postings.doc(), 1);
assert_eq!(segment_postings.seek(1), 1);
assert_eq!(segment_postings.doc(), 1);
assert_eq!(segment_postings.skip_next(1), SkipResult::OverStep);
assert_eq!(segment_postings.doc(), 2);
// check that going beyond the end is handled
assert_eq!(segment_postings.seek(num_docs), TERMINATED);
assert_eq!(segment_postings.skip_next(num_docs), SkipResult::End);
}
// check that filtering works
@@ -405,7 +428,7 @@ pub mod tests {
.unwrap();
for i in 0..num_docs / 2 {
assert_eq!(segment_postings.seek(i * 2), i * 2);
assert_eq!(segment_postings.skip_next(i * 2), SkipResult::Reached);
assert_eq!(segment_postings.doc(), i * 2);
}
@@ -415,19 +438,18 @@ pub mod tests {
.unwrap();
for i in 0..num_docs / 2 - 1 {
assert!(segment_postings.seek(i * 2 + 1) > (i * 1) * 2);
assert_eq!(segment_postings.skip_next(i * 2 + 1), SkipResult::OverStep);
assert_eq!(segment_postings.doc(), (i + 1) * 2);
}
}
// delete some of the documents
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.delete_term(term_0);
assert!(index_writer.commit().is_ok());
}
let searcher = index.reader().unwrap().searcher();
assert_eq!(searcher.segment_readers().len(), 1);
let segment_reader = searcher.segment_reader(0);
// make sure seeking still works
@@ -438,11 +460,11 @@ pub mod tests {
.unwrap();
if i % 2 == 0 {
assert_eq!(segment_postings.seek(i), i);
assert_eq!(segment_postings.skip_next(i), SkipResult::Reached);
assert_eq!(segment_postings.doc(), i);
assert!(segment_reader.is_deleted(i));
} else {
assert_eq!(segment_postings.seek(i), i);
assert_eq!(segment_postings.skip_next(i), SkipResult::Reached);
assert_eq!(segment_postings.doc(), i);
}
}
@@ -457,16 +479,12 @@ pub mod tests {
let mut last = 2; // start from 5 to avoid seeking to 3 twice
let mut cur = 3;
loop {
let seek = segment_postings.seek(cur);
if seek == TERMINATED {
break;
}
assert_eq!(seek, segment_postings.doc());
if seek == cur {
assert_eq!(segment_postings.doc(), cur);
} else {
assert_eq!(segment_postings.doc(), cur + 1);
match segment_postings.skip_next(cur) {
SkipResult::End => break,
SkipResult::Reached => assert_eq!(segment_postings.doc(), cur),
SkipResult::OverStep => assert_eq!(segment_postings.doc(), cur + 1),
}
let next = cur + last;
last = cur;
cur = next;
@@ -476,7 +494,7 @@ pub mod tests {
// delete everything else
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.delete_term(term_1);
assert!(index_writer.commit().is_ok());
}
@@ -519,7 +537,7 @@ pub mod tests {
let index = Index::create_in_ram(schema);
let posting_list_size = 1_000_000;
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
for _ in 0..posting_list_size {
let mut doc = Document::default();
if rng.gen_bool(1f64 / 15f64) {
@@ -552,7 +570,7 @@ pub mod tests {
}
impl<TDocSet: DocSet> DocSet for UnoptimizedDocSet<TDocSet> {
fn advance(&mut self) -> DocId {
fn advance(&mut self) -> bool {
self.0.advance()
}
@@ -577,26 +595,31 @@ pub mod tests {
) {
for target in targets {
let mut postings_opt = postings_factory();
if target < postings_opt.doc() {
continue;
}
let mut postings_unopt = UnoptimizedDocSet::wrap(postings_factory());
let skip_result_opt = postings_opt.seek(target);
let skip_result_unopt = postings_unopt.seek(target);
let skip_result_opt = postings_opt.skip_next(target);
let skip_result_unopt = postings_unopt.skip_next(target);
assert_eq!(
skip_result_unopt, skip_result_opt,
"Failed while skipping to {}",
target
);
assert!(skip_result_opt >= target);
assert_eq!(skip_result_opt, postings_opt.doc());
if skip_result_opt == TERMINATED {
return;
match skip_result_opt {
SkipResult::Reached => assert_eq!(postings_opt.doc(), target),
SkipResult::OverStep => assert!(postings_opt.doc() > target),
SkipResult::End => {
return;
}
}
while postings_opt.doc() != TERMINATED {
assert_eq!(postings_opt.doc(), postings_unopt.doc());
assert_eq!(postings_opt.advance(), postings_unopt.advance());
while postings_opt.advance() {
assert!(postings_unopt.advance());
assert_eq!(
postings_opt.doc(),
postings_unopt.doc(),
"Failed while skipping to {}",
target
);
}
assert!(!postings_unopt.advance());
}
}
}
@@ -605,7 +628,7 @@ pub mod tests {
mod bench {
use super::tests::*;
use crate::docset::TERMINATED;
use crate::docset::SkipResult;
use crate::query::Intersection;
use crate::schema::IndexRecordOption;
use crate::tests;
@@ -623,7 +646,7 @@ mod bench {
.inverted_index(TERM_A.field())
.read_postings(&*TERM_A, IndexRecordOption::Basic)
.unwrap();
while segment_postings.advance() != TERMINATED {}
while segment_postings.advance() {}
});
}
@@ -655,7 +678,7 @@ mod bench {
segment_postings_c,
segment_postings_d,
]);
while intersection.advance() != TERMINATED {}
while intersection.advance() {}
});
}
@@ -671,10 +694,11 @@ mod bench {
.unwrap();
let mut existing_docs = Vec::new();
segment_postings.advance();
for doc in &docs {
if *doc >= segment_postings.doc() {
existing_docs.push(*doc);
if segment_postings.seek(*doc) == TERMINATED {
if segment_postings.skip_next(*doc) == SkipResult::End {
break;
}
}
@@ -686,7 +710,7 @@ mod bench {
.read_postings(&*TERM_A, IndexRecordOption::Basic)
.unwrap();
for doc in &existing_docs {
if segment_postings.seek(*doc) == TERMINATED {
if segment_postings.skip_next(*doc) == SkipResult::End {
break;
}
}
@@ -725,9 +749,8 @@ mod bench {
.read_postings(&*TERM_A, IndexRecordOption::Basic)
.unwrap();
let mut s = 0u32;
while segment_postings.doc() != TERMINATED {
while segment_postings.advance() {
s += (segment_postings.doc() & n) % 1024;
segment_postings.advance();
}
s
});

38
src/postings/postings_writer.rs
View File

@@ -1,6 +1,5 @@
use super::stacker::{Addr, MemoryArena, TermHashMap};
use crate::fieldnorm::FieldNormReaders;
use crate::postings::recorder::{
BufferLender, NothingRecorder, Recorder, TFAndPositionRecorder, TermFrequencyRecorder,
};
@@ -105,7 +104,6 @@ impl MultiFieldPostingsWriter {
doc: DocId,
field: Field,
token_stream: &mut dyn TokenStream,
term_buffer: &mut Term,
) -> u32 {
let postings_writer =
self.per_field_postings_writers[field.field_id() as usize].deref_mut();
@@ -115,7 +113,6 @@ impl MultiFieldPostingsWriter {
field,
token_stream,
&mut self.heap,
term_buffer,
)
}
@@ -131,7 +128,6 @@ impl MultiFieldPostingsWriter {
pub fn serialize(
&self,
serializer: &mut InvertedIndexSerializer,
fieldnorm_readers: FieldNormReaders,
) -> crate::Result<HashMap<Field, FnvHashMap<UnorderedTermId, TermOrdinal>>> {
let mut term_offsets: Vec<(&[u8], Addr, UnorderedTermId)> =
self.term_index.iter().collect();
@@ -165,12 +161,8 @@ impl MultiFieldPostingsWriter {
}
let postings_writer = &self.per_field_postings_writers[field.field_id() as usize];
let fieldnorm_reader = fieldnorm_readers.get_field(field);
let mut field_serializer = serializer.new_field(
field,
postings_writer.total_num_tokens(),
fieldnorm_reader,
)?;
let mut field_serializer =
serializer.new_field(field, postings_writer.total_num_tokens())?;
postings_writer.serialize(
&term_offsets[start..stop],
&mut field_serializer,
@@ -222,22 +214,21 @@ pub trait PostingsWriter {
field: Field,
token_stream: &mut dyn TokenStream,
heap: &mut MemoryArena,
term_buffer: &mut Term,
) -> u32 {
term_buffer.set_field(field);
let mut term = Term::for_field(field);
let mut sink = |token: &Token| {
// We skip all tokens with a len greater than u16.
if token.text.len() > MAX_TOKEN_LEN {
return;
if token.text.len() <= MAX_TOKEN_LEN {
term.set_text(token.text.as_str());
self.subscribe(term_index, doc_id, token.position as u32, &term, heap);
} else {
info!(
"A token exceeding MAX_TOKEN_LEN ({}>{}) was dropped. Search for \
MAX_TOKEN_LEN in the documentation for more information.",
token.text.len(),
MAX_TOKEN_LEN
);
}
term_buffer.set_text(token.text.as_str());
self.subscribe(
term_index,
doc_id,
token.position as u32,
&term_buffer,
heap,
);
};
token_stream.process(&mut sink)
}
@@ -306,8 +297,7 @@ impl<Rec: Recorder + 'static> PostingsWriter for SpecializedPostingsWriter<Rec>
let mut buffer_lender = BufferLender::default();
for &(term_bytes, addr, _) in term_addrs {
let recorder: Rec = termdict_heap.read(addr);
let term_doc_freq = recorder.term_doc_freq().unwrap_or(0u32);
serializer.new_term(&term_bytes[4..], term_doc_freq)?;
serializer.new_term(&term_bytes[4..])?;
recorder.serialize(&mut buffer_lender, serializer, heap)?;
serializer.close_term()?;
}

23
src/postings/recorder.rs
View File

@@ -75,10 +75,6 @@ pub(crate) trait Recorder: Copy + 'static {
serializer: &mut FieldSerializer<'_>,
heap: &MemoryArena,
) -> io::Result<()>;
/// Returns the number of document containing this term.
///
/// Returns `None` if not available.
fn term_doc_freq(&self) -> Option<u32>;
}
/// Only records the doc ids
@@ -117,16 +113,11 @@ impl Recorder for NothingRecorder {
) -> io::Result<()> {
let buffer = buffer_lender.lend_u8();
self.stack.read_to_end(heap, buffer);
// TODO avoid reading twice.
for doc in VInt32Reader::new(&buffer[..]) {
serializer.write_doc(doc as u32, 0u32, &[][..])?;
}
Ok(())
}
fn term_doc_freq(&self) -> Option<u32> {
None
}
}
/// Recorder encoding document ids, and term frequencies
@@ -135,7 +126,6 @@ pub struct TermFrequencyRecorder {
stack: ExpUnrolledLinkedList,
current_doc: DocId,
current_tf: u32,
term_doc_freq: u32,
}
impl Recorder for TermFrequencyRecorder {
@@ -144,7 +134,6 @@ impl Recorder for TermFrequencyRecorder {
stack: ExpUnrolledLinkedList::new(),
current_doc: u32::max_value(),
current_tf: 0u32,
term_doc_freq: 0u32,
}
}
@@ -153,7 +142,6 @@ impl Recorder for TermFrequencyRecorder {
}
fn new_doc(&mut self, doc: DocId, heap: &mut MemoryArena) {
self.term_doc_freq += 1;
self.current_doc = doc;
let _ = write_u32_vint(doc, &mut self.stack.writer(heap));
}
@@ -184,10 +172,6 @@ impl Recorder for TermFrequencyRecorder {
Ok(())
}
fn term_doc_freq(&self) -> Option<u32> {
Some(self.term_doc_freq)
}
}
/// Recorder encoding term frequencies as well as positions.
@@ -195,14 +179,12 @@ impl Recorder for TermFrequencyRecorder {
pub struct TFAndPositionRecorder {
stack: ExpUnrolledLinkedList,
current_doc: DocId,
term_doc_freq: u32,
}
impl Recorder for TFAndPositionRecorder {
fn new() -> Self {
TFAndPositionRecorder {
stack: ExpUnrolledLinkedList::new(),
current_doc: u32::max_value(),
term_doc_freq: 0u32,
}
}
@@ -212,7 +194,6 @@ impl Recorder for TFAndPositionRecorder {
fn new_doc(&mut self, doc: DocId, heap: &mut MemoryArena) {
self.current_doc = doc;
self.term_doc_freq += 1u32;
let _ = write_u32_vint(doc, &mut self.stack.writer(heap));
}
@@ -252,10 +233,6 @@ impl Recorder for TFAndPositionRecorder {
}
Ok(())
}
fn term_doc_freq(&self) -> Option<u32> {
Some(self.term_doc_freq)
}
}
#[cfg(test)]

807
src/postings/segment_postings.rs
View File

@@ -1,70 +1,81 @@
use crate::common::BitSet;
use crate::common::HasLen;
use crate::docset::DocSet;
use crate::common::{BinarySerializable, VInt};
use crate::docset::{DocSet, SkipResult};
use crate::positions::PositionReader;
use crate::postings::compression::COMPRESSION_BLOCK_SIZE;
use crate::postings::compression::{compressed_block_size, AlignedBuffer};
use crate::postings::compression::{BlockDecoder, VIntDecoder, COMPRESSION_BLOCK_SIZE};
use crate::postings::serializer::PostingsSerializer;
use crate::postings::BlockSearcher;
use crate::postings::FreqReadingOption;
use crate::postings::Postings;
use crate::postings::SkipReader;
use crate::postings::USE_SKIP_INFO_LIMIT;
use crate::schema::IndexRecordOption;
use crate::{DocId, TERMINATED};
use crate::DocId;
use owned_read::OwnedRead;
use std::cmp::Ordering;
use tantivy_fst::Streamer;
use crate::directory::ReadOnlySource;
use crate::fastfield::DeleteBitSet;
use crate::postings::BlockSegmentPostings;
struct PositionComputer {
// store the amount of position int
// before reading positions.
//
// if none, position are already loaded in
// the positions vec.
position_to_skip: usize,
position_reader: PositionReader,
}
impl PositionComputer {
pub fn new(position_reader: PositionReader) -> PositionComputer {
PositionComputer {
position_to_skip: 0,
position_reader,
}
}
pub fn add_skip(&mut self, num_skip: usize) {
self.position_to_skip += num_skip;
}
// Positions can only be read once.
pub fn positions_with_offset(&mut self, offset: u32, output: &mut [u32]) {
self.position_reader.skip(self.position_to_skip);
self.position_to_skip = 0;
self.position_reader.read(output);
let mut cum = offset;
for output_mut in output.iter_mut() {
cum += *output_mut;
*output_mut = cum;
}
}
}
/// `SegmentPostings` represents the inverted list or postings associated to
/// a term in a `Segment`.
///
/// As we iterate through the `SegmentPostings`, the frequencies are optionally decoded.
/// Positions on the other hand, are optionally entirely decoded upfront.
#[derive(Clone)]
pub struct SegmentPostings {
pub(crate) block_cursor: BlockSegmentPostings,
block_cursor: BlockSegmentPostings,
cur: usize,
position_reader: Option<PositionReader>,
position_computer: Option<PositionComputer>,
block_searcher: BlockSearcher,
}
impl SegmentPostings {
/// Returns an empty segment postings object
pub fn empty() -> Self {
let empty_block_cursor = BlockSegmentPostings::empty();
SegmentPostings {
block_cursor: BlockSegmentPostings::empty(),
cur: 0,
position_reader: None,
block_cursor: empty_block_cursor,
cur: COMPRESSION_BLOCK_SIZE,
position_computer: None,
block_searcher: BlockSearcher::default(),
}
}
/// Compute the number of non-deleted documents.
///
/// This method will clone and scan through the posting lists.
/// (this is a rather expensive operation).
pub fn doc_freq_given_deletes(&self, delete_bitset: &DeleteBitSet) -> u32 {
let mut docset = self.clone();
let mut doc_freq = 0;
loop {
let doc = docset.doc();
if doc == TERMINATED {
return doc_freq;
}
if delete_bitset.is_alive(doc) {
doc_freq += 1u32;
}
docset.advance();
}
}
/// Returns the overall number of documents in the block postings.
/// It does not take in account whether documents are deleted or not.
pub fn doc_freq(&self) -> u32 {
self.block_cursor.doc_freq()
}
/// Creates a segment postings object with the given documents
/// and no frequency encoded.
///
@@ -76,9 +87,7 @@ impl SegmentPostings {
pub fn create_from_docs(docs: &[u32]) -> SegmentPostings {
let mut buffer = Vec::new();
{
let mut postings_serializer =
PostingsSerializer::new(&mut buffer, 0.0, false, false, None);
postings_serializer.new_term(docs.len() as u32);
let mut postings_serializer = PostingsSerializer::new(&mut buffer, false, false);
for &doc in docs {
postings_serializer.write_doc(doc, 1u32);
}
@@ -88,58 +97,15 @@ impl SegmentPostings {
}
let block_segment_postings = BlockSegmentPostings::from_data(
docs.len() as u32,
ReadOnlySource::from(buffer),
OwnedRead::new(buffer),
IndexRecordOption::Basic,
IndexRecordOption::Basic,
);
SegmentPostings::from_block_postings(block_segment_postings, None)
}
}
/// Helper functions to create `SegmentPostings` for tests.
#[cfg(test)]
pub fn create_from_docs_and_tfs(
doc_and_tfs: &[(u32, u32)],
fieldnorms: Option<&[u32]>,
) -> SegmentPostings {
use crate::fieldnorm::FieldNormReader;
use crate::Score;
let mut buffer: Vec<u8> = Vec::new();
let fieldnorm_reader = fieldnorms.map(FieldNormReader::for_test);
let average_field_norm = fieldnorms
.map(|fieldnorms| {
if fieldnorms.len() == 0 {
return 0.0;
}
let total_num_tokens: u64 = fieldnorms
.iter()
.map(|&fieldnorm| fieldnorm as u64)
.sum::<u64>();
total_num_tokens as Score / fieldnorms.len() as Score
})
.unwrap_or(0.0);
let mut postings_serializer = PostingsSerializer::new(
&mut buffer,
average_field_norm,
true,
false,
fieldnorm_reader,
);
postings_serializer.new_term(doc_and_tfs.len() as u32);
for &(doc, tf) in doc_and_tfs {
postings_serializer.write_doc(doc, tf);
}
postings_serializer
.close_term(doc_and_tfs.len() as u32)
.unwrap();
let block_segment_postings = BlockSegmentPostings::from_data(
doc_and_tfs.len() as u32,
ReadOnlySource::from(buffer),
IndexRecordOption::WithFreqs,
IndexRecordOption::WithFreqs,
);
SegmentPostings::from_block_postings(block_segment_postings, None)
}
impl SegmentPostings {
/// Reads a Segment postings from an &[u8]
///
/// * `len` - number of document in the posting lists.
@@ -148,12 +114,12 @@ impl SegmentPostings {
/// frequencies and/or positions
pub(crate) fn from_block_postings(
segment_block_postings: BlockSegmentPostings,
position_reader: Option<PositionReader>,
positions_stream_opt: Option<PositionReader>,
) -> SegmentPostings {
SegmentPostings {
block_cursor: segment_block_postings,
cur: 0, // cursor within the block
position_reader,
cur: COMPRESSION_BLOCK_SIZE, // cursor within the block
position_computer: positions_stream_opt.map(PositionComputer::new),
block_searcher: BlockSearcher::default(),
}
}
@@ -163,60 +129,139 @@ 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) -> DocId {
debug_assert!(self.block_cursor.block_is_loaded());
if self.cur == COMPRESSION_BLOCK_SIZE - 1 {
self.cur = 0;
self.block_cursor.advance();
} else {
self.cur += 1;
fn advance(&mut self) -> bool {
if self.position_computer.is_some() && self.cur < COMPRESSION_BLOCK_SIZE {
let term_freq = self.term_freq() as usize;
if let Some(position_computer) = self.position_computer.as_mut() {
position_computer.add_skip(term_freq);
}
}
self.doc()
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 seek(&mut self, target: DocId) -> DocId {
debug_assert!(self.doc() <= target);
if self.doc() >= target {
return self.doc();
fn skip_next(&mut self, target: DocId) -> SkipResult {
if !self.advance() {
return SkipResult::End;
}
match self.doc().cmp(&target) {
Ordering::Equal => {
return SkipResult::Reached;
}
Ordering::Greater => {
return SkipResult::OverStep;
}
_ => {
// ...
}
}
self.block_cursor.seek(target);
// In the following, thanks to the call to advance above,
// we know that the position is not loaded and we need
// to skip every doc_freq we cross.
// At this point we are on the block, that might contain our document.
let output = self.block_cursor.docs_aligned();
self.cur = self.block_searcher.search_in_block(&output, target);
// skip blocks until one that might contain the target
// check if we need to go to the next block
let mut sum_freqs_skipped: u32 = 0;
if !self
.block_cursor
.docs()
.last()
.map(|doc| *doc >= target)
.unwrap_or(false)
// there should always be at least a document in the block
// since advance returned.
{
// we are not in the right block.
//
// First compute all of the freqs skipped from the current block.
if self.position_computer.is_some() {
sum_freqs_skipped = self.block_cursor.freqs()[self.cur..].iter().sum();
match self.block_cursor.skip_to(target) {
BlockSegmentPostingsSkipResult::Success(block_skip_freqs) => {
sum_freqs_skipped += block_skip_freqs;
}
BlockSegmentPostingsSkipResult::Terminated => {
return SkipResult::End;
}
}
} else if self.block_cursor.skip_to(target)
== BlockSegmentPostingsSkipResult::Terminated
{
// no positions needed. no need to sum freqs.
return SkipResult::End;
}
self.cur = 0;
}
// The last block is not full and padded with the value TERMINATED,
// so that we are guaranteed to have at least doc in the block (a real one or the padding)
// that is greater or equal to the target.
debug_assert!(self.cur < COMPRESSION_BLOCK_SIZE);
let cur = self.cur;
// we're in the right block now, start with an exponential search
let (output, len) = self.block_cursor.docs_aligned();
let new_cur = self
.block_searcher
.search_in_block(&output, len, cur, target);
if let Some(position_computer) = self.position_computer.as_mut() {
sum_freqs_skipped += self.block_cursor.freqs()[cur..new_cur].iter().sum::<u32>();
position_computer.add_skip(sum_freqs_skipped as usize);
}
self.cur = new_cur;
// `doc` is now the first element >= `target`
// If all docs are smaller than target the current block should be incomplemented and padded
// with the value `TERMINATED`.
//
// After the search, the cursor should point to the first value of TERMINATED.
let doc = output.0[self.cur];
let doc = output.0[new_cur];
debug_assert!(doc >= target);
debug_assert_eq!(doc, self.doc());
doc
if doc == target {
SkipResult::Reached
} else {
SkipResult::OverStep
}
}
/// Return the current document's `DocId`.
///
/// # Panics
///
/// Will panics if called without having called advance before.
#[inline]
fn doc(&self) -> DocId {
self.block_cursor.doc(self.cur)
let docs = self.block_cursor.docs();
debug_assert!(
self.cur < docs.len(),
"Have you forgotten to call `.advance()` at least once before calling `.doc()` ."
);
docs[self.cur]
}
fn size_hint(&self) -> u32 {
self.len() as u32
}
fn append_to_bitset(&mut self, bitset: &mut BitSet) {
// finish the current block
if self.advance() {
for &doc in &self.block_cursor.docs()[self.cur..] {
bitset.insert(doc);
}
// ... iterate through the remaining blocks.
while self.block_cursor.advance() {
for &doc in self.block_cursor.docs() {
bitset.insert(doc);
}
}
}
}
}
impl HasLen for SegmentPostings {
fn len(&self) -> usize {
self.block_cursor.doc_freq() as usize
self.block_cursor.doc_freq()
}
}
@@ -245,63 +290,515 @@ impl Postings for SegmentPostings {
fn positions_with_offset(&mut self, offset: u32, output: &mut Vec<u32>) {
let term_freq = self.term_freq() as usize;
if let Some(position_reader) = self.position_reader.as_mut() {
let read_offset = self.block_cursor.position_offset()
+ (self.block_cursor.freqs()[..self.cur]
.iter()
.cloned()
.sum::<u32>() as u64);
if let Some(position_comp) = self.position_computer.as_mut() {
output.resize(term_freq, 0u32);
position_reader.read(read_offset, &mut output[..]);
let mut cum = offset;
for output_mut in output.iter_mut() {
cum += *output_mut;
*output_mut = cum;
}
position_comp.positions_with_offset(offset, &mut output[..]);
} else {
output.clear();
}
}
}
/// `BlockSegmentPostings` is a cursor iterating over blocks
/// of documents.
///
/// # Warning
///
/// While it is useful for some very specific high-performance
/// use cases, you should prefer using `SegmentPostings` for most usage.
pub struct BlockSegmentPostings {
doc_decoder: BlockDecoder,
freq_decoder: BlockDecoder,
freq_reading_option: FreqReadingOption,
doc_freq: usize,
doc_offset: DocId,
num_vint_docs: usize,
remaining_data: OwnedRead,
skip_reader: SkipReader,
}
fn split_into_skips_and_postings(
doc_freq: u32,
mut data: OwnedRead,
) -> (Option<OwnedRead>, OwnedRead) {
if doc_freq >= USE_SKIP_INFO_LIMIT {
let skip_len = VInt::deserialize(&mut data).expect("Data corrupted").0 as usize;
let mut postings_data = data.clone();
postings_data.advance(skip_len);
data.clip(skip_len);
(Some(data), postings_data)
} else {
(None, data)
}
}
#[derive(Debug, Eq, PartialEq)]
pub enum BlockSegmentPostingsSkipResult {
Terminated,
Success(u32), //< number of term freqs to skip
}
impl BlockSegmentPostings {
pub(crate) fn from_data(
doc_freq: u32,
data: OwnedRead,
record_option: IndexRecordOption,
requested_option: IndexRecordOption,
) -> BlockSegmentPostings {
let freq_reading_option = match (record_option, requested_option) {
(IndexRecordOption::Basic, _) => FreqReadingOption::NoFreq,
(_, IndexRecordOption::Basic) => FreqReadingOption::SkipFreq,
(_, _) => FreqReadingOption::ReadFreq,
};
let (skip_data_opt, postings_data) = split_into_skips_and_postings(doc_freq, data);
let skip_reader = match skip_data_opt {
Some(skip_data) => SkipReader::new(skip_data, record_option),
None => SkipReader::new(OwnedRead::new(&[][..]), record_option),
};
let doc_freq = doc_freq as usize;
let num_vint_docs = doc_freq % COMPRESSION_BLOCK_SIZE;
BlockSegmentPostings {
num_vint_docs,
doc_decoder: BlockDecoder::new(),
freq_decoder: BlockDecoder::with_val(1),
freq_reading_option,
doc_offset: 0,
doc_freq,
remaining_data: postings_data,
skip_reader,
}
}
// Resets the block segment postings on another position
// in the postings file.
//
// This is useful for enumerating through a list of terms,
// and consuming the associated posting lists while avoiding
// reallocating a `BlockSegmentPostings`.
//
// # Warning
//
// This does not reset the positions list.
pub(crate) fn reset(&mut self, doc_freq: u32, postings_data: OwnedRead) {
let (skip_data_opt, postings_data) = split_into_skips_and_postings(doc_freq, postings_data);
let num_vint_docs = (doc_freq as usize) & (COMPRESSION_BLOCK_SIZE - 1);
self.num_vint_docs = num_vint_docs;
self.remaining_data = postings_data;
if let Some(skip_data) = skip_data_opt {
self.skip_reader.reset(skip_data);
} else {
self.skip_reader.reset(OwnedRead::new(&[][..]))
}
self.doc_offset = 0;
self.doc_freq = doc_freq as usize;
}
/// Returns the document frequency associated to this block postings.
///
/// This `doc_freq` is simply the sum of the length of all of the blocks
/// length, and it does not take in account deleted documents.
pub fn doc_freq(&self) -> usize {
self.doc_freq
}
/// Returns the array of docs in the current block.
///
/// Before the first call to `.advance()`, the block
/// returned by `.docs()` is empty.
#[inline]
pub fn docs(&self) -> &[DocId] {
self.doc_decoder.output_array()
}
pub(crate) fn docs_aligned(&self) -> (&AlignedBuffer, usize) {
self.doc_decoder.output_aligned()
}
/// Return the document at index `idx` of the block.
#[inline]
pub fn doc(&self, idx: usize) -> u32 {
self.doc_decoder.output(idx)
}
/// Return the array of `term freq` in the block.
#[inline]
pub fn freqs(&self) -> &[u32] {
self.freq_decoder.output_array()
}
/// Return the frequency at index `idx` of the block.
#[inline]
pub fn freq(&self, idx: usize) -> u32 {
self.freq_decoder.output(idx)
}
/// Returns the length of the current block.
///
/// All blocks have a length of `NUM_DOCS_PER_BLOCK`,
/// except the last block that may have a length
/// of any number between 1 and `NUM_DOCS_PER_BLOCK - 1`
#[inline]
fn block_len(&self) -> usize {
self.doc_decoder.output_len
}
/// position on a block that may contains `doc_id`.
/// Always advance the current block.
///
/// Returns true if a block that has an element greater or equal to the target is found.
/// Returning true does not guarantee that the smallest element of the block is smaller
/// than the target. It only guarantees that the last element is greater or equal.
///
/// Returns false iff all of the document remaining are smaller than
/// `doc_id`. In that case, all of these document are consumed.
///
pub fn skip_to(&mut self, target_doc: DocId) -> BlockSegmentPostingsSkipResult {
let mut skip_freqs = 0u32;
while self.skip_reader.advance() {
if self.skip_reader.doc() >= target_doc {
// the last document of the current block is larger
// than the target.
//
// We found our block!
let num_bits = self.skip_reader.doc_num_bits();
let num_consumed_bytes = self.doc_decoder.uncompress_block_sorted(
self.remaining_data.as_ref(),
self.doc_offset,
num_bits,
);
self.remaining_data.advance(num_consumed_bytes);
let tf_num_bits = self.skip_reader.tf_num_bits();
match self.freq_reading_option {
FreqReadingOption::NoFreq => {}
FreqReadingOption::SkipFreq => {
let num_bytes_to_skip = compressed_block_size(tf_num_bits);
self.remaining_data.advance(num_bytes_to_skip);
}
FreqReadingOption::ReadFreq => {
let num_consumed_bytes = self
.freq_decoder
.uncompress_block_unsorted(self.remaining_data.as_ref(), tf_num_bits);
self.remaining_data.advance(num_consumed_bytes);
}
}
self.doc_offset = self.skip_reader.doc();
return BlockSegmentPostingsSkipResult::Success(skip_freqs);
} else {
skip_freqs += self.skip_reader.tf_sum();
let advance_len = self.skip_reader.total_block_len();
self.doc_offset = self.skip_reader.doc();
self.remaining_data.advance(advance_len);
}
}
// we are now on the last, incomplete, variable encoded block.
if self.num_vint_docs > 0 {
let num_compressed_bytes = self.doc_decoder.uncompress_vint_sorted(
self.remaining_data.as_ref(),
self.doc_offset,
self.num_vint_docs,
);
self.remaining_data.advance(num_compressed_bytes);
match self.freq_reading_option {
FreqReadingOption::NoFreq | FreqReadingOption::SkipFreq => {}
FreqReadingOption::ReadFreq => {
self.freq_decoder
.uncompress_vint_unsorted(self.remaining_data.as_ref(), self.num_vint_docs);
}
}
self.num_vint_docs = 0;
return self
.docs()
.last()
.map(|last_doc| {
if *last_doc >= target_doc {
BlockSegmentPostingsSkipResult::Success(skip_freqs)
} else {
BlockSegmentPostingsSkipResult::Terminated
}
})
.unwrap_or(BlockSegmentPostingsSkipResult::Terminated);
}
BlockSegmentPostingsSkipResult::Terminated
}
/// Advance to the next block.
///
/// Returns false iff there was no remaining blocks.
pub fn advance(&mut self) -> bool {
if self.skip_reader.advance() {
let num_bits = self.skip_reader.doc_num_bits();
let num_consumed_bytes = self.doc_decoder.uncompress_block_sorted(
self.remaining_data.as_ref(),
self.doc_offset,
num_bits,
);
self.remaining_data.advance(num_consumed_bytes);
let tf_num_bits = self.skip_reader.tf_num_bits();
match self.freq_reading_option {
FreqReadingOption::NoFreq => {}
FreqReadingOption::SkipFreq => {
let num_bytes_to_skip = compressed_block_size(tf_num_bits);
self.remaining_data.advance(num_bytes_to_skip);
}
FreqReadingOption::ReadFreq => {
let num_consumed_bytes = self
.freq_decoder
.uncompress_block_unsorted(self.remaining_data.as_ref(), tf_num_bits);
self.remaining_data.advance(num_consumed_bytes);
}
}
// it will be used as the next offset.
self.doc_offset = self.doc_decoder.output(COMPRESSION_BLOCK_SIZE - 1);
true
} else if self.num_vint_docs > 0 {
let num_compressed_bytes = self.doc_decoder.uncompress_vint_sorted(
self.remaining_data.as_ref(),
self.doc_offset,
self.num_vint_docs,
);
self.remaining_data.advance(num_compressed_bytes);
match self.freq_reading_option {
FreqReadingOption::NoFreq | FreqReadingOption::SkipFreq => {}
FreqReadingOption::ReadFreq => {
self.freq_decoder
.uncompress_vint_unsorted(self.remaining_data.as_ref(), self.num_vint_docs);
}
}
self.num_vint_docs = 0;
true
} else {
false
}
}
/// Returns an empty segment postings object
pub fn empty() -> BlockSegmentPostings {
BlockSegmentPostings {
num_vint_docs: 0,
doc_decoder: BlockDecoder::new(),
freq_decoder: BlockDecoder::with_val(1),
freq_reading_option: FreqReadingOption::NoFreq,
doc_offset: 0,
doc_freq: 0,
remaining_data: OwnedRead::new(vec![]),
skip_reader: SkipReader::new(OwnedRead::new(vec![]), IndexRecordOption::Basic),
}
}
}
impl<'b> Streamer<'b> for BlockSegmentPostings {
type Item = &'b [DocId];
fn next(&'b mut self) -> Option<&'b [DocId]> {
if self.advance() {
Some(self.docs())
} else {
None
}
}
}
#[cfg(test)]
mod tests {
use super::BlockSegmentPostings;
use super::BlockSegmentPostingsSkipResult;
use super::SegmentPostings;
use crate::common::HasLen;
use crate::docset::{DocSet, TERMINATED};
use crate::fastfield::DeleteBitSet;
use crate::core::Index;
use crate::docset::DocSet;
use crate::postings::postings::Postings;
use crate::schema::IndexRecordOption;
use crate::schema::Schema;
use crate::schema::Term;
use crate::schema::INDEXED;
use crate::DocId;
use crate::SkipResult;
use tantivy_fst::Streamer;
#[test]
fn test_empty_segment_postings() {
let mut postings = SegmentPostings::empty();
assert_eq!(postings.advance(), TERMINATED);
assert_eq!(postings.advance(), TERMINATED);
assert!(!postings.advance());
assert!(!postings.advance());
assert_eq!(postings.len(), 0);
}
#[test]
fn test_empty_postings_doc_returns_terminated() {
let mut postings = SegmentPostings::empty();
assert_eq!(postings.doc(), TERMINATED);
assert_eq!(postings.advance(), TERMINATED);
#[should_panic(expected = "Have you forgotten to call `.advance()`")]
fn test_panic_if_doc_called_before_advance() {
SegmentPostings::empty().doc();
}
#[test]
fn test_empty_postings_doc_term_freq_returns_0() {
let postings = SegmentPostings::empty();
assert_eq!(postings.term_freq(), 1);
#[should_panic(expected = "Have you forgotten to call `.advance()`")]
fn test_panic_if_freq_called_before_advance() {
SegmentPostings::empty().term_freq();
}
#[test]
fn test_doc_freq() {
let docs = SegmentPostings::create_from_docs(&[0, 2, 10]);
assert_eq!(docs.doc_freq(), 3);
let delete_bitset = DeleteBitSet::for_test(&[2], 12);
assert_eq!(docs.doc_freq_given_deletes(&delete_bitset), 2);
let all_deleted = DeleteBitSet::for_test(&[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11], 12);
assert_eq!(docs.doc_freq_given_deletes(&all_deleted), 0);
fn test_empty_block_segment_postings() {
let mut postings = BlockSegmentPostings::empty();
assert!(!postings.advance());
assert_eq!(postings.doc_freq(), 0);
}
#[test]
fn test_block_segment_postings() {
let mut block_segments = build_block_postings(&(0..100_000).collect::<Vec<u32>>());
let mut offset: u32 = 0u32;
// checking that the block before calling advance is empty
assert!(block_segments.docs().is_empty());
// checking that the `doc_freq` is correct
assert_eq!(block_segments.doc_freq(), 100_000);
while let Some(block) = block_segments.next() {
for (i, doc) in block.iter().cloned().enumerate() {
assert_eq!(offset + (i as u32), doc);
}
offset += block.len() as u32;
}
}
#[test]
fn test_skip_right_at_new_block() {
let mut doc_ids = (0..128).collect::<Vec<u32>>();
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 index = Index::create_in_ram(schema);
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
let mut last_doc = 0u32;
for &doc in docs {
for _ in last_doc..doc {
index_writer.add_document(doc!(int_field=>1u64));
}
index_writer.add_document(doc!(int_field=>0u64));
last_doc = doc + 1;
}
index_writer.commit().unwrap();
let searcher = index.reader().unwrap().searcher();
let segment_reader = searcher.segment_reader(0);
let inverted_index = segment_reader.inverted_index(int_field);
let term = Term::from_field_u64(int_field, 0u64);
let term_info = inverted_index.get_term_info(&term).unwrap();
inverted_index.read_block_postings_from_terminfo(&term_info, IndexRecordOption::Basic)
}
#[test]
fn test_block_segment_postings_skip() {
for i in 0..4 {
let mut block_postings = build_block_postings(&[3]);
assert_eq!(
block_postings.skip_to(i),
BlockSegmentPostingsSkipResult::Success(0u32)
);
assert_eq!(
block_postings.skip_to(i),
BlockSegmentPostingsSkipResult::Terminated
);
}
let mut block_postings = build_block_postings(&[3]);
assert_eq!(
block_postings.skip_to(4u32),
BlockSegmentPostingsSkipResult::Terminated
);
}
#[test]
fn test_block_segment_postings_skip2() {
let mut docs = vec![0];
for i in 0..1300 {
docs.push((i * i / 100) + i);
}
let mut block_postings = build_block_postings(&docs[..]);
for i in vec![0, 424, 10000] {
assert_eq!(
block_postings.skip_to(i),
BlockSegmentPostingsSkipResult::Success(0u32)
);
let docs = block_postings.docs();
assert!(docs[0] <= i);
assert!(docs.last().cloned().unwrap_or(0u32) >= i);
}
assert_eq!(
block_postings.skip_to(100_000),
BlockSegmentPostingsSkipResult::Terminated
);
assert_eq!(
block_postings.skip_to(101_000),
BlockSegmentPostingsSkipResult::Terminated
);
}
#[test]
fn test_reset_block_segment_postings() {
let mut schema_builder = Schema::builder();
let int_field = schema_builder.add_u64_field("id", INDEXED);
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();
// create two postings list, one containg even number,
// the other containing odd numbers.
for i in 0..6 {
let doc = doc!(int_field=> (i % 2) as u64);
index_writer.add_document(doc);
}
index_writer.commit().unwrap();
let searcher = index.reader().unwrap().searcher();
let segment_reader = searcher.segment_reader(0);
let mut block_segments;
{
let term = Term::from_field_u64(int_field, 0u64);
let inverted_index = segment_reader.inverted_index(int_field);
let term_info = inverted_index.get_term_info(&term).unwrap();
block_segments = inverted_index
.read_block_postings_from_terminfo(&term_info, IndexRecordOption::Basic);
}
assert!(block_segments.advance());
assert_eq!(block_segments.docs(), &[0, 2, 4]);
{
let term = Term::from_field_u64(int_field, 1u64);
let inverted_index = segment_reader.inverted_index(int_field);
let term_info = inverted_index.get_term_info(&term).unwrap();
inverted_index.reset_block_postings_from_terminfo(&term_info, &mut block_segments);
}
assert!(block_segments.advance());
assert_eq!(block_segments.docs(), &[1, 3, 5]);
}
}

91
src/postings/serializer.rs
View File

@@ -3,16 +3,14 @@ use crate::common::{BinarySerializable, VInt};
use crate::common::{CompositeWrite, CountingWriter};
use crate::core::Segment;
use crate::directory::WritePtr;
use crate::fieldnorm::FieldNormReader;
use crate::positions::PositionSerializer;
use crate::postings::compression::{BlockEncoder, VIntEncoder, COMPRESSION_BLOCK_SIZE};
use crate::postings::skip::SkipSerializer;
use crate::query::BM25Weight;
use crate::postings::USE_SKIP_INFO_LIMIT;
use crate::schema::Schema;
use crate::schema::{Field, FieldEntry, FieldType};
use crate::termdict::{TermDictionaryBuilder, TermOrdinal};
use crate::{DocId, Score};
use std::cmp::Ordering;
use crate::DocId;
use std::io::{self, Write};
/// `InvertedIndexSerializer` is in charge of serializing
@@ -92,22 +90,20 @@ impl InvertedIndexSerializer {
&mut self,
field: Field,
total_num_tokens: u64,
fieldnorm_reader: Option<FieldNormReader>,
) -> io::Result<FieldSerializer<'_>> {
let field_entry: &FieldEntry = self.schema.get_field_entry(field);
let term_dictionary_write = self.terms_write.for_field(field);
let postings_write = self.postings_write.for_field(field);
total_num_tokens.serialize(postings_write)?;
let positions_write = self.positions_write.for_field(field);
let positionsidx_write = self.positionsidx_write.for_field(field);
let field_type: FieldType = (*field_entry.field_type()).clone();
FieldSerializer::create(
&field_type,
total_num_tokens,
term_dictionary_write,
postings_write,
positions_write,
positionsidx_write,
fieldnorm_reader,
)
}
@@ -135,14 +131,11 @@ pub struct FieldSerializer<'a> {
impl<'a> FieldSerializer<'a> {
fn create(
field_type: &FieldType,
total_num_tokens: u64,
term_dictionary_write: &'a mut CountingWriter<WritePtr>,
postings_write: &'a mut CountingWriter<WritePtr>,
positions_write: &'a mut CountingWriter<WritePtr>,
positionsidx_write: &'a mut CountingWriter<WritePtr>,
fieldnorm_reader: Option<FieldNormReader>,
) -> io::Result<FieldSerializer<'a>> {
total_num_tokens.serialize(postings_write)?;
let (term_freq_enabled, position_enabled): (bool, bool) = match field_type {
FieldType::Str(ref text_options) => {
if let Some(text_indexing_options) = text_options.get_indexing_options() {
@@ -155,17 +148,8 @@ impl<'a> FieldSerializer<'a> {
_ => (false, false),
};
let term_dictionary_builder = TermDictionaryBuilder::create(term_dictionary_write)?;
let average_fieldnorm = fieldnorm_reader
.as_ref()
.map(|ff_reader| (total_num_tokens as Score / ff_reader.num_docs() as Score))
.unwrap_or(0.0);
let postings_serializer = PostingsSerializer::new(
postings_write,
average_fieldnorm,
term_freq_enabled,
position_enabled,
fieldnorm_reader,
);
let postings_serializer =
PostingsSerializer::new(postings_write, term_freq_enabled, position_enabled);
let positions_serializer_opt = if position_enabled {
Some(PositionSerializer::new(positions_write, positionsidx_write))
} else {
@@ -198,20 +182,18 @@ impl<'a> FieldSerializer<'a> {
/// Starts the postings for a new term.
/// * term - the term. It needs to come after the previous term according
/// to the lexicographical order.
/// * term_doc_freq - return the number of document containing the term.
pub fn new_term(&mut self, term: &[u8], term_doc_freq: u32) -> io::Result<TermOrdinal> {
/// * doc_freq - return the number of document containing the term.
pub fn new_term(&mut self, term: &[u8]) -> io::Result<TermOrdinal> {
assert!(
!self.term_open,
"Called new_term, while the previous term was not closed."
);
self.term_open = true;
self.postings_serializer.clear();
self.current_term_info = self.current_term_info();
self.term_dictionary_builder.insert_key(term)?;
let term_ordinal = self.num_terms;
self.num_terms += 1;
self.postings_serializer.new_term(term_doc_freq);
Ok(term_ordinal)
}
@@ -325,27 +307,14 @@ pub struct PostingsSerializer<W: Write> {
termfreq_enabled: bool,
termfreq_sum_enabled: bool,
fieldnorm_reader: Option<FieldNormReader>,
bm25_weight: Option<BM25Weight>,
num_docs: u32, // Number of docs in the segment
avg_fieldnorm: Score, // Average number of term in the field for that segment.
// this value is used to compute the block wand information.
}
impl<W: Write> PostingsSerializer<W> {
pub fn new(
write: W,
avg_fieldnorm: Score,
termfreq_enabled: bool,
termfreq_sum_enabled: bool,
fieldnorm_reader: Option<FieldNormReader>,
) -> PostingsSerializer<W> {
let num_docs = fieldnorm_reader
.as_ref()
.map(|fieldnorm_reader| fieldnorm_reader.num_docs())
.unwrap_or(0u32);
PostingsSerializer {
output_write: CountingWriter::wrap(write),
@@ -358,23 +327,6 @@ impl<W: Write> PostingsSerializer<W> {
last_doc_id_encoded: 0u32,
termfreq_enabled,
termfreq_sum_enabled,
fieldnorm_reader,
bm25_weight: None,
num_docs,
avg_fieldnorm,
}
}
pub fn new_term(&mut self, term_doc_freq: u32) {
if self.termfreq_enabled && self.num_docs > 0 {
let bm25_weight = BM25Weight::for_one_term(
term_doc_freq as u64,
self.num_docs as u64,
self.avg_fieldnorm,
);
self.bm25_weight = Some(bm25_weight);
}
}
@@ -391,6 +343,7 @@ impl<W: Write> PostingsSerializer<W> {
self.postings_write.extend(block_encoded);
}
if self.termfreq_enabled {
// encode the term_freqs
let (num_bits, block_encoded): (u8, &[u8]) = self
.block_encoder
.compress_block_unsorted(&self.block.term_freqs());
@@ -400,31 +353,6 @@ impl<W: Write> PostingsSerializer<W> {
let sum_freq = self.block.term_freqs().iter().cloned().sum();
self.skip_write.write_total_term_freq(sum_freq);
}
let mut blockwand_params = (0u8, 0u32);
if let Some(bm25_weight) = self.bm25_weight.as_ref() {
if let Some(fieldnorm_reader) = self.fieldnorm_reader.as_ref() {
let docs = self.block.doc_ids().iter().cloned();
let term_freqs = self.block.term_freqs().iter().cloned();
let fieldnorms = docs.map(|doc| fieldnorm_reader.fieldnorm_id(doc));
blockwand_params = fieldnorms
.zip(term_freqs)
.max_by(
|(left_fieldnorm_id, left_term_freq),
(right_fieldnorm_id, right_term_freq)| {
let left_score =
bm25_weight.tf_factor(*left_fieldnorm_id, *left_term_freq);
let right_score =
bm25_weight.tf_factor(*right_fieldnorm_id, *right_term_freq);
left_score
.partial_cmp(&right_score)
.unwrap_or(Ordering::Equal)
},
)
.unwrap();
}
}
let (fieldnorm_id, term_freq) = blockwand_params;
self.skip_write.write_blockwand_max(fieldnorm_id, term_freq);
}
self.block.clear();
}
@@ -463,7 +391,7 @@ impl<W: Write> PostingsSerializer<W> {
}
self.block.clear();
}
if doc_freq >= COMPRESSION_BLOCK_SIZE as u32 {
if doc_freq >= USE_SKIP_INFO_LIMIT {
let skip_data = self.skip_write.data();
VInt(skip_data.len() as u64).serialize(&mut self.output_write)?;
self.output_write.write_all(skip_data)?;
@@ -473,7 +401,6 @@ impl<W: Write> PostingsSerializer<W> {
}
self.skip_write.clear();
self.postings_write.clear();
self.bm25_weight = None;
Ok(())
}

362
src/postings/skip.rs
View File

@@ -1,9 +1,7 @@
use crate::common::{read_u32_vint_no_advance, serialize_vint_u32, BinarySerializable, VInt};
use crate::directory::ReadOnlySource;
use crate::postings::compression::{compressed_block_size, COMPRESSION_BLOCK_SIZE};
use crate::query::BM25Weight;
use crate::common::BinarySerializable;
use crate::postings::compression::COMPRESSION_BLOCK_SIZE;
use crate::schema::IndexRecordOption;
use crate::{DocId, Score, TERMINATED};
use crate::DocId;
use owned_read::OwnedRead;
pub struct SkipSerializer {
@@ -41,13 +39,6 @@ impl SkipSerializer {
.expect("Should never fail");
}
pub fn write_blockwand_max(&mut self, fieldnorm_id: u8, term_freq: u32) {
self.buffer.push(fieldnorm_id);
let mut buf = [0u8; 8];
let bytes = serialize_vint_u32(term_freq, &mut buf);
self.buffer.extend_from_slice(bytes);
}
pub fn data(&self) -> &[u8] {
&self.buffer[..]
}
@@ -58,200 +49,81 @@ impl SkipSerializer {
}
}
#[derive(Clone)]
pub(crate) struct SkipReader {
last_doc_in_block: DocId,
pub(crate) last_doc_in_previous_block: DocId,
doc: DocId,
owned_read: OwnedRead,
doc_num_bits: u8,
tf_num_bits: u8,
tf_sum: u32,
skip_info: IndexRecordOption,
byte_offset: usize,
remaining_docs: u32, // number of docs remaining, including the
// documents in the current block.
block_info: BlockInfo,
position_offset: u64,
}
#[derive(Clone, Eq, PartialEq, Copy, Debug)]
pub(crate) enum BlockInfo {
BitPacked {
doc_num_bits: u8,
tf_num_bits: u8,
tf_sum: u32,
block_wand_fieldnorm_id: u8,
block_wand_term_freq: u32,
},
VInt {
num_docs: u32,
},
}
impl Default for BlockInfo {
fn default() -> Self {
BlockInfo::VInt { num_docs: 0u32 }
}
}
impl SkipReader {
pub fn new(data: ReadOnlySource, doc_freq: u32, skip_info: IndexRecordOption) -> SkipReader {
let mut skip_reader = SkipReader {
last_doc_in_block: if doc_freq >= COMPRESSION_BLOCK_SIZE as u32 {
0
} else {
TERMINATED
},
last_doc_in_previous_block: 0u32,
owned_read: OwnedRead::new(data),
pub fn new(data: OwnedRead, skip_info: IndexRecordOption) -> SkipReader {
SkipReader {
doc: 0u32,
owned_read: data,
skip_info,
block_info: BlockInfo::VInt { num_docs: doc_freq },
byte_offset: 0,
remaining_docs: doc_freq,
position_offset: 0u64,
};
if doc_freq >= COMPRESSION_BLOCK_SIZE as u32 {
skip_reader.read_block_info();
}
skip_reader
}
pub fn reset(&mut self, data: ReadOnlySource, doc_freq: u32) {
self.last_doc_in_block = if doc_freq >= COMPRESSION_BLOCK_SIZE as u32 {
0
} else {
TERMINATED
};
self.last_doc_in_previous_block = 0u32;
self.owned_read = OwnedRead::new(data);
self.block_info = BlockInfo::VInt { num_docs: doc_freq };
self.byte_offset = 0;
self.remaining_docs = doc_freq;
self.position_offset = 0u64;
if doc_freq >= COMPRESSION_BLOCK_SIZE as u32 {
self.read_block_info();
doc_num_bits: 0u8,
tf_num_bits: 0u8,
tf_sum: 0u32,
}
}
// Returns the block max score for this block if available.
//
// The block max score is available for all full bitpacked block,
// but no available for the last VInt encoded incomplete block.
pub fn block_max_score(&self, bm25_weight: &BM25Weight) -> Option<Score> {
match self.block_info {
BlockInfo::BitPacked {
block_wand_fieldnorm_id,
block_wand_term_freq,
..
} => Some(bm25_weight.score(block_wand_fieldnorm_id, block_wand_term_freq)),
BlockInfo::VInt { .. } => None,
}
pub fn reset(&mut self, data: OwnedRead) {
self.doc = 0u32;
self.owned_read = data;
self.doc_num_bits = 0u8;
self.tf_num_bits = 0u8;
self.tf_sum = 0u32;
}
pub(crate) fn last_doc_in_block(&self) -> DocId {
self.last_doc_in_block
pub fn total_block_len(&self) -> usize {
(self.doc_num_bits + self.tf_num_bits) as usize * COMPRESSION_BLOCK_SIZE / 8
}
pub fn position_offset(&self) -> u64 {
self.position_offset
pub fn doc(&self) -> DocId {
self.doc
}
pub fn byte_offset(&self) -> usize {
self.byte_offset
pub fn doc_num_bits(&self) -> u8 {
self.doc_num_bits
}
fn read_block_info(&mut self) {
let doc_delta = u32::deserialize(&mut self.owned_read).expect("Skip data corrupted");
self.last_doc_in_block += doc_delta as DocId;
let doc_num_bits = self.owned_read.get(0);
match self.skip_info {
IndexRecordOption::Basic => {
self.owned_read.advance(1);
self.block_info = BlockInfo::BitPacked {
doc_num_bits,
tf_num_bits: 0,
tf_sum: 0,
block_wand_fieldnorm_id: 0,
block_wand_term_freq: 0,
};
}
IndexRecordOption::WithFreqs => {
let tf_num_bits = self.owned_read.get(1);
let block_wand_fieldnorm_id = self.owned_read.get(2);
let data = &self.owned_read.as_ref()[3..];
let (block_wand_term_freq, num_bytes) = read_u32_vint_no_advance(data);
self.owned_read.advance(3 + num_bytes);
self.block_info = BlockInfo::BitPacked {
doc_num_bits,
tf_num_bits,
tf_sum: 0,
block_wand_fieldnorm_id,
block_wand_term_freq,
};
}
IndexRecordOption::WithFreqsAndPositions => {
let tf_num_bits = self.owned_read.get(1);
self.owned_read.advance(2);
let tf_sum = u32::deserialize(&mut self.owned_read).expect("Failed reading tf_sum");
let block_wand_fieldnorm_id = self.owned_read.get(0);
self.owned_read.advance(1);
let block_wand_term_freq =
VInt::deserialize_u64(&mut self.owned_read).unwrap() as u32;
self.block_info = BlockInfo::BitPacked {
doc_num_bits,
tf_num_bits,
tf_sum,
block_wand_fieldnorm_id,
block_wand_term_freq,
};
}
}
}
pub fn block_info(&self) -> BlockInfo {
self.block_info
}
/// Advance the skip reader to the block that may contain the target.
/// Number of bits used to encode term frequencies
///
/// If the target is larger than all documents, the skip_reader
/// then advance to the last Variable In block.
pub fn seek(&mut self, target: DocId) -> bool {
if self.last_doc_in_block() >= target {
return false;
}
loop {
self.advance();
if self.last_doc_in_block() >= target {
return true;
}
}
/// 0 if term frequencies are not enabled.
pub fn tf_num_bits(&self) -> u8 {
self.tf_num_bits
}
pub fn advance(&mut self) {
match self.block_info {
BlockInfo::BitPacked {
doc_num_bits,
tf_num_bits,
tf_sum,
..
} => {
self.remaining_docs -= COMPRESSION_BLOCK_SIZE as u32;
self.byte_offset += compressed_block_size(doc_num_bits + tf_num_bits);
self.position_offset += tf_sum as u64;
}
BlockInfo::VInt { num_docs } => {
debug_assert_eq!(num_docs, self.remaining_docs);
self.remaining_docs = 0;
self.byte_offset = std::usize::MAX;
}
}
self.last_doc_in_previous_block = self.last_doc_in_block;
if self.remaining_docs >= COMPRESSION_BLOCK_SIZE as u32 {
self.read_block_info();
pub fn tf_sum(&self) -> u32 {
self.tf_sum
}
pub fn advance(&mut self) -> bool {
if self.owned_read.as_ref().is_empty() {
false
} else {
self.last_doc_in_block = TERMINATED;
self.block_info = BlockInfo::VInt {
num_docs: self.remaining_docs,
};
let doc_delta = u32::deserialize(&mut self.owned_read).expect("Skip data corrupted");
self.doc += doc_delta as DocId;
self.doc_num_bits = self.owned_read.get(0);
match self.skip_info {
IndexRecordOption::Basic => {
self.owned_read.advance(1);
}
IndexRecordOption::WithFreqs => {
self.tf_num_bits = self.owned_read.get(1);
self.owned_read.advance(2);
}
IndexRecordOption::WithFreqsAndPositions => {
self.tf_num_bits = self.owned_read.get(1);
self.owned_read.advance(2);
self.tf_sum =
u32::deserialize(&mut self.owned_read).expect("Failed reading tf_sum");
}
}
true
}
}
}
@@ -259,11 +131,9 @@ impl SkipReader {
#[cfg(test)]
mod tests {
use super::BlockInfo;
use super::IndexRecordOption;
use super::{SkipReader, SkipSerializer};
use crate::directory::ReadOnlySource;
use crate::postings::compression::COMPRESSION_BLOCK_SIZE;
use owned_read::OwnedRead;
#[test]
fn test_skip_with_freq() {
@@ -271,47 +141,20 @@ mod tests {
let mut skip_serializer = SkipSerializer::new();
skip_serializer.write_doc(1u32, 2u8);
skip_serializer.write_term_freq(3u8);
skip_serializer.write_blockwand_max(13u8, 3u32);
skip_serializer.write_doc(5u32, 5u8);
skip_serializer.write_term_freq(2u8);
skip_serializer.write_blockwand_max(8u8, 2u32);
skip_serializer.data().to_owned()
};
let doc_freq = 3u32 + (COMPRESSION_BLOCK_SIZE * 2) as u32;
let mut skip_reader = SkipReader::new(
ReadOnlySource::new(buf),
doc_freq,
IndexRecordOption::WithFreqs,
);
assert_eq!(skip_reader.last_doc_in_block(), 1u32);
assert_eq!(
skip_reader.block_info,
BlockInfo::BitPacked {
doc_num_bits: 2u8,
tf_num_bits: 3u8,
tf_sum: 0,
block_wand_fieldnorm_id: 13,
block_wand_term_freq: 3
}
);
skip_reader.advance();
assert_eq!(skip_reader.last_doc_in_block(), 5u32);
assert_eq!(
skip_reader.block_info(),
BlockInfo::BitPacked {
doc_num_bits: 5u8,
tf_num_bits: 2u8,
tf_sum: 0,
block_wand_fieldnorm_id: 8,
block_wand_term_freq: 2
}
);
skip_reader.advance();
assert_eq!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 3u32 });
skip_reader.advance();
assert_eq!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 0u32 });
skip_reader.advance();
assert_eq!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 0u32 });
let mut skip_reader = SkipReader::new(OwnedRead::new(buf), IndexRecordOption::WithFreqs);
assert!(skip_reader.advance());
assert_eq!(skip_reader.doc(), 1u32);
assert_eq!(skip_reader.doc_num_bits(), 2u8);
assert_eq!(skip_reader.tf_num_bits(), 3u8);
assert!(skip_reader.advance());
assert_eq!(skip_reader.doc(), 5u32);
assert_eq!(skip_reader.doc_num_bits(), 5u8);
assert_eq!(skip_reader.tf_num_bits(), 2u8);
assert!(!skip_reader.advance());
}
#[test]
@@ -322,68 +165,13 @@ mod tests {
skip_serializer.write_doc(5u32, 5u8);
skip_serializer.data().to_owned()
};
let doc_freq = 3u32 + (COMPRESSION_BLOCK_SIZE * 2) as u32;
let mut skip_reader = SkipReader::new(
ReadOnlySource::from(buf),
doc_freq,
IndexRecordOption::Basic,
);
assert_eq!(skip_reader.last_doc_in_block(), 1u32);
assert_eq!(
skip_reader.block_info(),
BlockInfo::BitPacked {
doc_num_bits: 2u8,
tf_num_bits: 0,
tf_sum: 0u32,
block_wand_fieldnorm_id: 0,
block_wand_term_freq: 0
}
);
skip_reader.advance();
assert_eq!(skip_reader.last_doc_in_block(), 5u32);
assert_eq!(
skip_reader.block_info(),
BlockInfo::BitPacked {
doc_num_bits: 5u8,
tf_num_bits: 0,
tf_sum: 0u32,
block_wand_fieldnorm_id: 0,
block_wand_term_freq: 0
}
);
skip_reader.advance();
assert_eq!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 3u32 });
skip_reader.advance();
assert_eq!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 0u32 });
skip_reader.advance();
assert_eq!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 0u32 });
}
#[test]
fn test_skip_multiple_of_block_size() {
let buf = {
let mut skip_serializer = SkipSerializer::new();
skip_serializer.write_doc(1u32, 2u8);
skip_serializer.data().to_owned()
};
let doc_freq = COMPRESSION_BLOCK_SIZE as u32;
let mut skip_reader = SkipReader::new(
ReadOnlySource::from(buf),
doc_freq,
IndexRecordOption::Basic,
);
assert_eq!(skip_reader.last_doc_in_block(), 1u32);
assert_eq!(
skip_reader.block_info(),
BlockInfo::BitPacked {
doc_num_bits: 2u8,
tf_num_bits: 0,
tf_sum: 0u32,
block_wand_fieldnorm_id: 0,
block_wand_term_freq: 0
}
);
skip_reader.advance();
assert_eq!(skip_reader.block_info(), BlockInfo::VInt { num_docs: 0u32 });
let mut skip_reader = SkipReader::new(OwnedRead::new(buf), IndexRecordOption::Basic);
assert!(skip_reader.advance());
assert_eq!(skip_reader.doc(), 1u32);
assert_eq!(skip_reader.doc_num_bits(), 2u8);
assert!(skip_reader.advance());
assert_eq!(skip_reader.doc(), 5u32);
assert_eq!(skip_reader.doc_num_bits(), 5u8);
assert!(!skip_reader.advance());
}
}

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

@@ -1,4 +1,6 @@
use murmurhash32::murmurhash2;
use murmurhash32;
use self::murmurhash32::murmurhash2;
use super::{Addr, MemoryArena};
use crate::postings::stacker::memory_arena::store;

93
src/query/all_query.rs
View File

@@ -1,7 +1,6 @@
use crate::core::Searcher;
use crate::core::SegmentReader;
use crate::docset::{DocSet, TERMINATED};
use crate::query::boost_query::BoostScorer;
use crate::docset::DocSet;
use crate::query::explanation::does_not_match;
use crate::query::{Explanation, Query, Scorer, Weight};
use crate::DocId;
@@ -9,7 +8,7 @@ use crate::Score;
/// Query that matches all of the documents.
///
/// All of the document get the score 1.0.
/// All of the document get the score 1f32.
#[derive(Clone, Debug)]
pub struct AllQuery;
@@ -23,36 +22,55 @@ impl Query for AllQuery {
pub struct AllWeight;
impl Weight for AllWeight {
fn scorer(&self, reader: &SegmentReader, boost: Score) -> crate::Result<Box<dyn Scorer>> {
let all_scorer = AllScorer {
fn scorer(&self, reader: &SegmentReader) -> crate::Result<Box<dyn Scorer>> {
Ok(Box::new(AllScorer {
state: State::NotStarted,
doc: 0u32,
max_doc: reader.max_doc(),
};
Ok(Box::new(BoostScorer::new(all_scorer, boost)))
}))
}
fn explain(&self, reader: &SegmentReader, doc: DocId) -> crate::Result<Explanation> {
if doc >= reader.max_doc() {
return Err(does_not_match(doc));
}
Ok(Explanation::new("AllQuery", 1.0))
Ok(Explanation::new("AllQuery", 1f32))
}
}
enum State {
NotStarted,
Started,
Finished,
}
/// Scorer associated to the `AllQuery` query.
pub struct AllScorer {
state: State,
doc: DocId,
max_doc: DocId,
}
impl DocSet for AllScorer {
fn advance(&mut self) -> DocId {
if self.doc + 1 >= self.max_doc {
self.doc = TERMINATED;
return TERMINATED;
fn advance(&mut self) -> bool {
match self.state {
State::NotStarted => {
self.state = State::Started;
self.doc = 0;
}
State::Started => {
self.doc += 1u32;
}
State::Finished => {
return false;
}
}
if self.doc < self.max_doc {
true
} else {
self.state = State::Finished;
false
}
self.doc += 1;
self.doc
}
fn doc(&self) -> DocId {
@@ -66,70 +84,49 @@ impl DocSet for AllScorer {
impl Scorer for AllScorer {
fn score(&mut self) -> Score {
1.0
1f32
}
}
#[cfg(test)]
mod tests {
use super::AllQuery;
use crate::docset::TERMINATED;
use crate::query::Query;
use crate::schema::{Schema, TEXT};
use crate::Index;
fn create_test_index() -> Index {
#[test]
fn test_all_query() {
let mut schema_builder = Schema::builder();
let field = schema_builder.add_text_field("text", TEXT);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 10_000_000).unwrap();
index_writer.add_document(doc!(field=>"aaa"));
index_writer.add_document(doc!(field=>"bbb"));
index_writer.commit().unwrap();
index_writer.add_document(doc!(field=>"ccc"));
index_writer.commit().unwrap();
index
}
#[test]
fn test_all_query() {
let index = create_test_index();
let reader = index.reader().unwrap();
reader.reload().unwrap();
let searcher = reader.searcher();
let weight = AllQuery.weight(&searcher, false).unwrap();
{
let reader = searcher.segment_reader(0);
let mut scorer = weight.scorer(reader, 1.0).unwrap();
let mut scorer = weight.scorer(reader).unwrap();
assert!(scorer.advance());
assert_eq!(scorer.doc(), 0u32);
assert_eq!(scorer.advance(), 1u32);
assert!(scorer.advance());
assert_eq!(scorer.doc(), 1u32);
assert_eq!(scorer.advance(), TERMINATED);
assert!(!scorer.advance());
}
{
let reader = searcher.segment_reader(1);
let mut scorer = weight.scorer(reader, 1.0).unwrap();
let mut scorer = weight.scorer(reader).unwrap();
assert!(scorer.advance());
assert_eq!(scorer.doc(), 0u32);
assert_eq!(scorer.advance(), TERMINATED);
}
}
#[test]
fn test_all_query_with_boost() {
let index = create_test_index();
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let weight = AllQuery.weight(&searcher, false).unwrap();
let reader = searcher.segment_reader(0);
{
let mut scorer = weight.scorer(reader, 2.0).unwrap();
assert_eq!(scorer.doc(), 0u32);
assert_eq!(scorer.score(), 2.0);
}
{
let mut scorer = weight.scorer(reader, 1.5).unwrap();
assert_eq!(scorer.doc(), 0u32);
assert_eq!(scorer.score(), 1.5);
assert!(!scorer.advance());
}
}
}

117
src/query/automaton_weight.rs
View File

@@ -5,8 +5,9 @@ use crate::query::{BitSetDocSet, Explanation};
use crate::query::{Scorer, Weight};
use crate::schema::{Field, IndexRecordOption};
use crate::termdict::{TermDictionary, TermStreamer};
use crate::DocId;
use crate::TantivyError;
use crate::{DocId, Score};
use crate::{Result, SkipResult};
use std::sync::Arc;
use tantivy_fst::Automaton;
@@ -39,9 +40,10 @@ impl<A> Weight for AutomatonWeight<A>
where
A: Automaton + Send + Sync + 'static,
{
fn scorer(&self, reader: &SegmentReader, boost: Score) -> crate::Result<Box<dyn Scorer>> {
fn scorer(&self, reader: &SegmentReader) -> Result<Box<dyn Scorer>> {
let max_doc = reader.max_doc();
let mut doc_bitset = BitSet::with_max_value(max_doc);
let inverted_index = reader.inverted_index(self.field);
let term_dict = inverted_index.terms();
let mut term_stream = self.automaton_stream(term_dict);
@@ -49,26 +51,20 @@ where
let term_info = term_stream.value();
let mut block_segment_postings = inverted_index
.read_block_postings_from_terminfo(term_info, IndexRecordOption::Basic);
loop {
let docs = block_segment_postings.docs();
if docs.is_empty() {
break;
}
for &doc in docs {
while block_segment_postings.advance() {
for &doc in block_segment_postings.docs() {
doc_bitset.insert(doc);
}
block_segment_postings.advance();
}
}
let doc_bitset = BitSetDocSet::from(doc_bitset);
let const_scorer = ConstScorer::new(doc_bitset, boost);
Ok(Box::new(const_scorer))
Ok(Box::new(ConstScorer::new(doc_bitset)))
}
fn explain(&self, reader: &SegmentReader, doc: DocId) -> crate::Result<Explanation> {
let mut scorer = self.scorer(reader, 1.0)?;
if scorer.seek(doc) == doc {
Ok(Explanation::new("AutomatonScorer", 1.0))
fn explain(&self, reader: &SegmentReader, doc: DocId) -> Result<Explanation> {
let mut scorer = self.scorer(reader)?;
if scorer.skip_next(doc) == SkipResult::Reached {
Ok(Explanation::new("AutomatonScorer", 1.0f32))
} else {
Err(TantivyError::InvalidArgument(
"Document does not exist".to_string(),
@@ -76,94 +72,3 @@ where
}
}
}
#[cfg(test)]
mod tests {
use super::AutomatonWeight;
use crate::docset::TERMINATED;
use crate::query::Weight;
use crate::schema::{Schema, STRING};
use crate::Index;
use tantivy_fst::Automaton;
fn create_index() -> Index {
let mut schema = Schema::builder();
let title = schema.add_text_field("title", STRING);
let index = Index::create_in_ram(schema.build());
let mut index_writer = index.writer_for_tests().unwrap();
index_writer.add_document(doc!(title=>"abc"));
index_writer.add_document(doc!(title=>"bcd"));
index_writer.add_document(doc!(title=>"abcd"));
assert!(index_writer.commit().is_ok());
index
}
enum State {
Start,
NotMatching,
AfterA,
}
struct PrefixedByA;
impl Automaton for PrefixedByA {
type State = State;
fn start(&self) -> Self::State {
State::Start
}
fn is_match(&self, state: &Self::State) -> bool {
match *state {
State::AfterA => true,
_ => false,
}
}
fn accept(&self, state: &Self::State, byte: u8) -> Self::State {
match *state {
State::Start => {
if byte == b'a' {
State::AfterA
} else {
State::NotMatching
}
}
State::AfterA => State::AfterA,
State::NotMatching => State::NotMatching,
}
}
}
#[test]
fn test_automaton_weight() {
let index = create_index();
let field = index.schema().get_field("title").unwrap();
let automaton_weight = AutomatonWeight::new(field, PrefixedByA);
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let mut scorer = automaton_weight
.scorer(searcher.segment_reader(0u32), 1.0)
.unwrap();
assert_eq!(scorer.doc(), 0u32);
assert_eq!(scorer.score(), 1.0);
assert_eq!(scorer.advance(), 2u32);
assert_eq!(scorer.doc(), 2u32);
assert_eq!(scorer.score(), 1.0);
assert_eq!(scorer.advance(), TERMINATED);
}
#[test]
fn test_automaton_weight_boost() {
let index = create_index();
let field = index.schema().get_field("title").unwrap();
let automaton_weight = AutomatonWeight::new(field, PrefixedByA);
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let mut scorer = automaton_weight
.scorer(searcher.segment_reader(0u32), 1.32)
.unwrap();
assert_eq!(scorer.doc(), 0u32);
assert_eq!(scorer.score(), 1.32);
}
}

143
src/query/bitset/mod.rs
View File

@@ -1,6 +1,7 @@
use crate::common::{BitSet, TinySet};
use crate::docset::{DocSet, TERMINATED};
use crate::docset::{DocSet, SkipResult};
use crate::DocId;
use std::cmp::Ordering;
/// A `BitSetDocSet` makes it possible to iterate through a bitset as if it was a `DocSet`.
///
@@ -32,51 +33,74 @@ impl From<BitSet> for BitSetDocSet {
} else {
docs.tinyset(0)
};
let mut docset = BitSetDocSet {
BitSetDocSet {
docs,
cursor_bucket: 0,
cursor_tinybitset: first_tiny_bitset,
doc: 0u32,
};
docset.advance();
docset
}
}
}
impl DocSet for BitSetDocSet {
fn advance(&mut self) -> DocId {
fn advance(&mut self) -> bool {
if let Some(lower) = self.cursor_tinybitset.pop_lowest() {
self.doc = (self.cursor_bucket as u32 * 64u32) | lower;
return self.doc;
return true;
}
if let Some(cursor_bucket) = self.docs.first_non_empty_bucket(self.cursor_bucket + 1) {
self.go_to_bucket(cursor_bucket);
let lower = self.cursor_tinybitset.pop_lowest().unwrap();
self.doc = (cursor_bucket * 64u32) | lower;
self.doc
true
} else {
self.doc = TERMINATED;
TERMINATED
false
}
}
fn seek(&mut self, target: DocId) -> DocId {
if target >= self.docs.max_value() {
self.doc = TERMINATED;
return TERMINATED;
fn skip_next(&mut self, target: DocId) -> SkipResult {
// skip is required to advance.
if !self.advance() {
return SkipResult::End;
}
let target_bucket = target / 64u32;
if target_bucket > self.cursor_bucket {
self.go_to_bucket(target_bucket);
let greater_filter: TinySet = TinySet::range_greater_or_equal(target);
self.cursor_tinybitset = self.cursor_tinybitset.intersect(greater_filter);
self.advance()
} else {
let mut doc = self.doc();
while doc < target {
doc = self.advance();
// Mask for all of the bits greater or equal
// to our target document.
match target_bucket.cmp(&self.cursor_bucket) {
Ordering::Greater => {
self.go_to_bucket(target_bucket);
let greater_filter: TinySet = TinySet::range_greater_or_equal(target);
self.cursor_tinybitset = self.cursor_tinybitset.intersect(greater_filter);
if !self.advance() {
SkipResult::End
} else if self.doc() == target {
SkipResult::Reached
} else {
debug_assert!(self.doc() > target);
SkipResult::OverStep
}
}
Ordering::Equal => loop {
match self.doc().cmp(&target) {
Ordering::Less => {
if !self.advance() {
return SkipResult::End;
}
}
Ordering::Equal => {
return SkipResult::Reached;
}
Ordering::Greater => {
debug_assert!(self.doc() > target);
return SkipResult::OverStep;
}
}
},
Ordering::Less => {
debug_assert!(self.doc() > target);
SkipResult::OverStep
}
doc
}
}
@@ -98,7 +122,7 @@ impl DocSet for BitSetDocSet {
mod tests {
use super::BitSetDocSet;
use crate::common::BitSet;
use crate::docset::{DocSet, TERMINATED};
use crate::docset::{DocSet, SkipResult};
use crate::DocId;
fn create_docbitset(docs: &[DocId], max_doc: DocId) -> BitSetDocSet {
@@ -109,31 +133,19 @@ mod tests {
BitSetDocSet::from(docset)
}
#[test]
fn test_empty() {
let bitset = BitSet::with_max_value(1000);
let mut empty = BitSetDocSet::from(bitset);
assert_eq!(empty.advance(), TERMINATED)
}
#[test]
fn test_seek_terminated() {
let bitset = BitSet::with_max_value(1000);
let mut empty = BitSetDocSet::from(bitset);
assert_eq!(empty.seek(TERMINATED), TERMINATED)
}
fn test_go_through_sequential(docs: &[DocId]) {
let mut docset = create_docbitset(docs, 1_000u32);
for &doc in docs {
assert!(docset.advance());
assert_eq!(doc, docset.doc());
docset.advance();
}
assert_eq!(docset.advance(), TERMINATED);
assert!(!docset.advance());
assert!(!docset.advance());
}
#[test]
fn test_docbitset_sequential() {
test_go_through_sequential(&[]);
test_go_through_sequential(&[1, 2, 3]);
test_go_through_sequential(&[1, 2, 3, 4, 5, 63, 64, 65]);
test_go_through_sequential(&[63, 64, 65]);
@@ -144,64 +156,64 @@ mod tests {
fn test_docbitset_skip() {
{
let mut docset = create_docbitset(&[1, 5, 6, 7, 5112], 10_000);
assert_eq!(docset.seek(7), 7);
assert_eq!(docset.skip_next(7), SkipResult::Reached);
assert_eq!(docset.doc(), 7);
assert_eq!(docset.advance(), 5112);
assert!(docset.advance(), 7);
assert_eq!(docset.doc(), 5112);
assert_eq!(docset.advance(), TERMINATED);
assert!(!docset.advance());
}
{
let mut docset = create_docbitset(&[1, 5, 6, 7, 5112], 10_000);
assert_eq!(docset.seek(3), 5);
assert_eq!(docset.skip_next(3), SkipResult::OverStep);
assert_eq!(docset.doc(), 5);
assert_eq!(docset.advance(), 6);
assert!(docset.advance());
}
{
let mut docset = create_docbitset(&[5112], 10_000);
assert_eq!(docset.seek(5112), 5112);
assert_eq!(docset.skip_next(5112), SkipResult::Reached);
assert_eq!(docset.doc(), 5112);
assert_eq!(docset.advance(), TERMINATED);
assert!(!docset.advance());
}
{
let mut docset = create_docbitset(&[5112], 10_000);
assert_eq!(docset.seek(5113), TERMINATED);
assert_eq!(docset.advance(), TERMINATED);
assert_eq!(docset.skip_next(5113), SkipResult::End);
assert!(!docset.advance());
}
{
let mut docset = create_docbitset(&[5112], 10_000);
assert_eq!(docset.seek(5111), 5112);
assert_eq!(docset.skip_next(5111), SkipResult::OverStep);
assert_eq!(docset.doc(), 5112);
assert_eq!(docset.advance(), TERMINATED);
assert!(!docset.advance());
}
{
let mut docset = create_docbitset(&[1, 5, 6, 7, 5112, 5500, 6666], 10_000);
assert_eq!(docset.seek(5112), 5112);
assert_eq!(docset.skip_next(5112), SkipResult::Reached);
assert_eq!(docset.doc(), 5112);
assert_eq!(docset.advance(), 5500);
assert!(docset.advance());
assert_eq!(docset.doc(), 5500);
assert_eq!(docset.advance(), 6666);
assert!(docset.advance());
assert_eq!(docset.doc(), 6666);
assert_eq!(docset.advance(), TERMINATED);
assert!(!docset.advance());
}
{
let mut docset = create_docbitset(&[1, 5, 6, 7, 5112, 5500, 6666], 10_000);
assert_eq!(docset.seek(5111), 5112);
assert_eq!(docset.skip_next(5111), SkipResult::OverStep);
assert_eq!(docset.doc(), 5112);
assert_eq!(docset.advance(), 5500);
assert!(docset.advance());
assert_eq!(docset.doc(), 5500);
assert_eq!(docset.advance(), 6666);
assert!(docset.advance());
assert_eq!(docset.doc(), 6666);
assert_eq!(docset.advance(), TERMINATED);
assert!(!docset.advance());
}
{
let mut docset = create_docbitset(&[1, 5, 6, 7, 5112, 5513, 6666], 10_000);
assert_eq!(docset.seek(5111), 5112);
assert_eq!(docset.skip_next(5111), SkipResult::OverStep);
assert_eq!(docset.doc(), 5112);
assert_eq!(docset.advance(), 5513);
assert!(docset.advance());
assert_eq!(docset.doc(), 5513);
assert_eq!(docset.advance(), 6666);
assert!(docset.advance());
assert_eq!(docset.doc(), 6666);
assert_eq!(docset.advance(), TERMINATED);
assert!(!docset.advance());
}
}
}
@@ -211,7 +223,6 @@ mod bench {
use super::BitSet;
use super::BitSetDocSet;
use crate::docset::TERMINATED;
use crate::test;
use crate::tests;
use crate::DocSet;
@@ -246,7 +257,7 @@ mod bench {
}
b.iter(|| {
let mut docset = BitSetDocSet::from(bitset.clone());
while docset.advance() != TERMINATED {}
while docset.advance() {}
});
}
}

105
src/query/bm25.rs
View File

@@ -3,24 +3,21 @@ use crate::query::Explanation;
use crate::Score;
use crate::Searcher;
use crate::Term;
use serde::Deserialize;
use serde::Serialize;
const K1: Score = 1.2;
const B: Score = 0.75;
const K1: f32 = 1.2;
const B: f32 = 0.75;
fn idf(doc_freq: u64, doc_count: u64) -> Score {
assert!(doc_count >= doc_freq, "{} >= {}", doc_count, doc_freq);
let x = ((doc_count - doc_freq) as Score + 0.5) / (doc_freq as Score + 0.5);
(1.0 + x).ln()
fn idf(doc_freq: u64, doc_count: u64) -> f32 {
let x = ((doc_count - doc_freq) as f32 + 0.5) / (doc_freq as f32 + 0.5);
(1f32 + x).ln()
}
fn cached_tf_component(fieldnorm: u32, average_fieldnorm: Score) -> Score {
K1 * (1.0 - B + B * fieldnorm as Score / average_fieldnorm)
fn cached_tf_component(fieldnorm: u32, average_fieldnorm: f32) -> f32 {
K1 * (1f32 - B + B * fieldnorm as f32 / average_fieldnorm)
}
fn compute_tf_cache(average_fieldnorm: Score) -> [Score; 256] {
let mut cache: [Score; 256] = [0.0; 256];
fn compute_tf_cache(average_fieldnorm: f32) -> [f32; 256] {
let mut cache = [0f32; 256];
for (fieldnorm_id, cache_mut) in cache.iter_mut().enumerate() {
let fieldnorm = FieldNormReader::id_to_fieldnorm(fieldnorm_id as u8);
*cache_mut = cached_tf_component(fieldnorm, average_fieldnorm);
@@ -28,30 +25,15 @@ fn compute_tf_cache(average_fieldnorm: Score) -> [Score; 256] {
cache
}
#[derive(Clone, PartialEq, Debug, Serialize, Deserialize)]
pub struct BM25Params {
pub idf: Score,
pub avg_fieldnorm: Score,
}
#[derive(Clone)]
pub struct BM25Weight {
idf_explain: Explanation,
weight: Score,
cache: [Score; 256],
average_fieldnorm: Score,
weight: f32,
cache: [f32; 256],
average_fieldnorm: f32,
}
impl BM25Weight {
pub fn boost_by(&self, boost: Score) -> BM25Weight {
BM25Weight {
idf_explain: self.idf_explain.clone(),
weight: self.weight * boost,
cache: self.cache,
average_fieldnorm: self.average_fieldnorm,
}
}
pub fn for_terms(searcher: &Searcher, terms: &[Term]) -> BM25Weight {
assert!(!terms.is_empty(), "BM25 requires at least one term");
let field = terms[0].field();
@@ -70,11 +52,19 @@ impl BM25Weight {
total_num_tokens += inverted_index.total_num_tokens();
total_num_docs += u64::from(segment_reader.max_doc());
}
let average_fieldnorm = total_num_tokens as Score / total_num_docs as Score;
let average_fieldnorm = total_num_tokens as f32 / total_num_docs as f32;
let mut idf_explain: Explanation;
if terms.len() == 1 {
let term_doc_freq = searcher.doc_freq(&terms[0]);
BM25Weight::for_one_term(term_doc_freq, total_num_docs, average_fieldnorm)
let idf = idf(term_doc_freq, total_num_docs);
idf_explain =
Explanation::new("idf, computed as log(1 + (N - n + 0.5) / (n + 0.5))", idf);
idf_explain.add_const(
"n, number of docs containing this term",
term_doc_freq as f32,
);
idf_explain.add_const("N, total number of docs", total_num_docs as f32);
} else {
let idf = terms
.iter()
@@ -82,30 +72,14 @@ impl BM25Weight {
let term_doc_freq = searcher.doc_freq(term);
idf(term_doc_freq, total_num_docs)
})
.sum::<Score>();
let idf_explain = Explanation::new("idf", idf);
BM25Weight::new(idf_explain, average_fieldnorm)
.sum::<f32>();
idf_explain = Explanation::new("idf", idf);
}
BM25Weight::new(idf_explain, average_fieldnorm)
}
pub fn for_one_term(
term_doc_freq: u64,
total_num_docs: u64,
avg_fieldnorm: Score,
) -> BM25Weight {
let idf = idf(term_doc_freq, total_num_docs);
let mut idf_explain =
Explanation::new("idf, computed as log(1 + (N - n + 0.5) / (n + 0.5))", idf);
idf_explain.add_const(
"n, number of docs containing this term",
term_doc_freq as Score,
);
idf_explain.add_const("N, total number of docs", total_num_docs as Score);
BM25Weight::new(idf_explain, avg_fieldnorm)
}
fn new(idf_explain: Explanation, average_fieldnorm: Score) -> BM25Weight {
let weight = idf_explain.value() * (1.0 + K1);
fn new(idf_explain: Explanation, average_fieldnorm: f32) -> BM25Weight {
let weight = idf_explain.value() * (1f32 + K1);
BM25Weight {
idf_explain,
weight,
@@ -116,27 +90,19 @@ impl BM25Weight {
#[inline(always)]
pub fn score(&self, fieldnorm_id: u8, term_freq: u32) -> Score {
self.weight * self.tf_factor(fieldnorm_id, term_freq)
}
pub fn max_score(&self) -> Score {
self.score(255u8, 2_013_265_944)
}
#[inline(always)]
pub(crate) fn tf_factor(&self, fieldnorm_id: u8, term_freq: u32) -> Score {
let term_freq = term_freq as Score;
let norm = self.cache[fieldnorm_id as usize];
term_freq / (term_freq + norm)
let term_freq = term_freq as f32;
self.weight * term_freq / (term_freq + norm)
}
pub fn explain(&self, fieldnorm_id: u8, term_freq: u32) -> Explanation {
// The explain format is directly copied from Lucene's.
// (So, Kudos to Lucene)
let score = self.score(fieldnorm_id, term_freq);
let norm = self.cache[fieldnorm_id as usize];
let term_freq = term_freq as Score;
let term_freq = term_freq as f32;
let right_factor = term_freq / (term_freq + norm);
let mut tf_explanation = Explanation::new(
@@ -149,12 +115,12 @@ impl BM25Weight {
tf_explanation.add_const("b, length normalization parameter", B);
tf_explanation.add_const(
"dl, length of field",
FieldNormReader::id_to_fieldnorm(fieldnorm_id) as Score,
FieldNormReader::id_to_fieldnorm(fieldnorm_id) as f32,
);
tf_explanation.add_const("avgdl, average length of field", self.average_fieldnorm);
let mut explanation = Explanation::new("TermQuery, product of...", score);
explanation.add_detail(Explanation::new("(K1+1)", K1 + 1.0));
explanation.add_detail(Explanation::new("(K1+1)", K1 + 1f32));
explanation.add_detail(self.idf_explain.clone());
explanation.add_detail(tf_explanation);
explanation
@@ -165,11 +131,10 @@ impl BM25Weight {
mod tests {
use super::idf;
use crate::{assert_nearly_equals, Score};
use crate::tests::assert_nearly_equals;
#[test]
fn test_idf() {
let score: Score = 2.0;
assert_nearly_equals!(idf(1, 2), score.ln());
assert_nearly_equals(idf(1, 2), 0.6931472);
}
}

434
src/query/boolean_query/block_wand.rs
View File

@@ -1,434 +0,0 @@
use crate::query::term_query::TermScorer;
use crate::query::Scorer;
use crate::{DocId, DocSet, Score, TERMINATED};
use std::ops::Deref;
use std::ops::DerefMut;
/// Takes a term_scorers sorted by their current doc() and a threshold and returns
/// Returns (pivot_len, pivot_ord) defined as follows:
/// - `pivot_doc` lowest document that has a chance of exceeding (>) the threshold score.
/// - `before_pivot_len` number of term_scorers such that term_scorer.doc() < pivot.
/// - `pivot_len` number of term_scorers such that term_scorer.doc() <= pivot.
///
/// We always have `before_pivot_len` < `pivot_len`.
///
/// None is returned if we establish that no document can exceed the threshold.
fn find_pivot_doc(
term_scorers: &[TermScorerWithMaxScore],
threshold: Score,
) -> Option<(usize, usize, DocId)> {
let mut max_score = 0.0;
let mut before_pivot_len = 0;
let mut pivot_doc = TERMINATED;
while before_pivot_len < term_scorers.len() {
let term_scorer = &term_scorers[before_pivot_len];
max_score += term_scorer.max_score;
if max_score > threshold {
pivot_doc = term_scorer.doc();
break;
}
before_pivot_len += 1;
}
if pivot_doc == TERMINATED {
return None;
}
// Right now i is an ordinal, we want a len.
let mut pivot_len = before_pivot_len + 1;
// Some other term_scorer may be positioned on the same document.
pivot_len += term_scorers[pivot_len..]
.iter()
.take_while(|term_scorer| term_scorer.doc() == pivot_doc)
.count();
Some((before_pivot_len, pivot_len, pivot_doc))
}
// Before and after calling this method, scorers need to be sorted by their `.doc()`.
fn block_max_was_too_low_advance_one_scorer(
scorers: &mut Vec<TermScorerWithMaxScore>,
pivot_len: usize,
) {
debug_assert!(is_sorted(scorers.iter().map(|scorer| scorer.doc())));
let mut scorer_to_seek = pivot_len - 1;
let mut doc_to_seek_after = scorers[scorer_to_seek].doc();
for scorer_ord in (0..pivot_len - 1).rev() {
let scorer = &scorers[scorer_ord];
if scorer.last_doc_in_block() <= doc_to_seek_after {
doc_to_seek_after = scorer.last_doc_in_block();
scorer_to_seek = scorer_ord;
}
}
for scorer in &scorers[pivot_len..] {
if scorer.doc() <= doc_to_seek_after {
doc_to_seek_after = scorer.doc();
}
}
scorers[scorer_to_seek].seek(doc_to_seek_after + 1);
restore_ordering(scorers, scorer_to_seek);
debug_assert!(is_sorted(scorers.iter().map(|scorer| scorer.doc())));
}
// Given a list of term_scorers and a `ord` and assuming that `term_scorers[ord]` is sorted
// except term_scorers[ord] that might be in advance compared to its ranks,
// bubble up term_scorers[ord] in order to restore the ordering.
fn restore_ordering(term_scorers: &mut Vec<TermScorerWithMaxScore>, ord: usize) {
let doc = term_scorers[ord].doc();
for i in ord + 1..term_scorers.len() {
if term_scorers[i].doc() >= doc {
break;
}
term_scorers.swap(i, i - 1);
}
debug_assert!(is_sorted(term_scorers.iter().map(|scorer| scorer.doc())));
}
// Attempts to advance all term_scorers between `&term_scorers[0..before_len]` to the pivot.
// If this works, return true.
// If this fails (ie: one of the term_scorer does not contain `pivot_doc` and seek goes past the
// pivot), reorder the term_scorers to ensure the list is still sorted and returns `false`.
// If a term_scorer reach TERMINATED in the process return false remove the term_scorer and return.
fn align_scorers(
term_scorers: &mut Vec<TermScorerWithMaxScore>,
pivot_doc: DocId,
before_pivot_len: usize,
) -> bool {
debug_assert_ne!(pivot_doc, TERMINATED);
for i in (0..before_pivot_len).rev() {
let new_doc = term_scorers[i].seek(pivot_doc);
if new_doc != pivot_doc {
if new_doc == TERMINATED {
term_scorers.swap_remove(i);
}
// We went past the pivot.
// We just go through the outer loop mechanic (Note that pivot is
// still a possible candidate).
//
// Termination is still guaranteed since we can only consider the same
// pivot at most term_scorers.len() - 1 times.
restore_ordering(term_scorers, i);
return false;
}
}
true
}
// Assumes terms_scorers[..pivot_len] are positioned on the same doc (pivot_doc).
// Advance term_scorers[..pivot_len] and out of these removes the terminated scores.
// Restores the ordering of term_scorers.
fn advance_all_scorers_on_pivot(term_scorers: &mut Vec<TermScorerWithMaxScore>, pivot_len: usize) {
for term_scorer in &mut term_scorers[..pivot_len] {
term_scorer.advance();
}
// TODO use drain_filter when available.
let mut i = 0;
while i != term_scorers.len() {
if term_scorers[i].doc() == TERMINATED {
term_scorers.swap_remove(i);
} else {
i += 1;
}
}
term_scorers.sort_by_key(|scorer| scorer.doc());
}
pub fn block_wand(
mut scorers: Vec<TermScorer>,
mut threshold: Score,
callback: &mut dyn FnMut(u32, Score) -> Score,
) {
let mut scorers: Vec<TermScorerWithMaxScore> = scorers
.iter_mut()
.map(TermScorerWithMaxScore::from)
.collect();
scorers.sort_by_key(|scorer| scorer.doc());
// At this point we need to ensure that the scorers are sorted!
debug_assert!(is_sorted(scorers.iter().map(|scorer| scorer.doc())));
while let Some((before_pivot_len, pivot_len, pivot_doc)) =
find_pivot_doc(&scorers[..], threshold)
{
debug_assert!(is_sorted(scorers.iter().map(|scorer| scorer.doc())));
debug_assert_ne!(pivot_doc, TERMINATED);
debug_assert!(before_pivot_len < pivot_len);
let block_max_score_upperbound: Score = scorers[..pivot_len]
.iter_mut()
.map(|scorer| {
scorer.shallow_seek(pivot_doc);
scorer.block_max_score()
})
.sum();
// Beware after shallow advance, skip readers can be in advance compared to
// the segment posting lists.
//
// `block_segment_postings.load_block()` need to be called separately.
if block_max_score_upperbound <= threshold {
// Block max condition was not reached
// We could get away by simply advancing the scorers to DocId + 1 but it would
// be inefficient. The optimization requires proper explanation and was
// isolated in a different function.
block_max_was_too_low_advance_one_scorer(&mut scorers, pivot_len);
continue;
}
// Block max condition is observed.
//
// Let's try and advance all scorers before the pivot to the pivot.
if !align_scorers(&mut scorers, pivot_doc, before_pivot_len) {
// At least of the scorer does not contain the pivot.
//
// Let's stop scoring this pivot and go through the pivot selection again.
// Note that the current pivot is not necessarily a bad candidate and it
// may be picked again.
continue;
}
// At this point, all scorers are positioned on the doc.
let score = scorers[..pivot_len]
.iter_mut()
.map(|scorer| scorer.score())
.sum();
if score > threshold {
threshold = callback(pivot_doc, score);
}
// let's advance all of the scorers that are currently positioned on the pivot.
advance_all_scorers_on_pivot(&mut scorers, pivot_len);
}
}
struct TermScorerWithMaxScore<'a> {
scorer: &'a mut TermScorer,
max_score: Score,
}
impl<'a> From<&'a mut TermScorer> for TermScorerWithMaxScore<'a> {
fn from(scorer: &'a mut TermScorer) -> Self {
let max_score = scorer.max_score();
TermScorerWithMaxScore { scorer, max_score }
}
}
impl<'a> Deref for TermScorerWithMaxScore<'a> {
type Target = TermScorer;
fn deref(&self) -> &Self::Target {
self.scorer
}
}
impl<'a> DerefMut for TermScorerWithMaxScore<'a> {
fn deref_mut(&mut self) -> &mut Self::Target {
self.scorer
}
}
fn is_sorted<I: Iterator<Item = DocId>>(mut it: I) -> bool {
if let Some(first) = it.next() {
let mut prev = first;
for doc in it {
if doc < prev {
return false;
}
prev = doc;
}
}
true
}
#[cfg(test)]
mod tests {
use crate::query::score_combiner::SumCombiner;
use crate::query::term_query::TermScorer;
use crate::query::Union;
use crate::query::{BM25Weight, Scorer};
use crate::{DocId, DocSet, Score, TERMINATED};
use proptest::prelude::*;
use std::cmp::Ordering;
use std::collections::BinaryHeap;
use std::iter;
struct Float(Score);
impl Eq for Float {}
impl PartialEq for Float {
fn eq(&self, other: &Self) -> bool {
self.cmp(&other) == Ordering::Equal
}
}
impl PartialOrd for Float {
fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
Some(self.cmp(other))
}
}
impl Ord for Float {
fn cmp(&self, other: &Self) -> Ordering {
other.0.partial_cmp(&self.0).unwrap_or(Ordering::Equal)
}
}
fn nearly_equals(left: Score, right: Score) -> bool {
(left - right).abs() < 0.000001 * (left + right).abs()
}
fn compute_checkpoints_for_each_pruning(
term_scorers: Vec<TermScorer>,
n: usize,
) -> Vec<(DocId, Score)> {
let mut heap: BinaryHeap<Float> = BinaryHeap::with_capacity(n);
let mut checkpoints: Vec<(DocId, Score)> = Vec::new();
let mut limit: Score = 0.0;
super::block_wand(term_scorers, Score::MIN, &mut |doc, score| {
heap.push(Float(score));
if heap.len() > n {
heap.pop().unwrap();
}
if heap.len() == n {
limit = heap.peek().unwrap().0;
}
if !nearly_equals(score, limit) {
checkpoints.push((doc, score));
}
return limit;
});
checkpoints
}
fn compute_checkpoints_manual(term_scorers: Vec<TermScorer>, n: usize) -> Vec<(DocId, Score)> {
let mut heap: BinaryHeap<Float> = BinaryHeap::with_capacity(n);
let mut checkpoints: Vec<(DocId, Score)> = Vec::new();
let mut scorer: Union<TermScorer, SumCombiner> = Union::from(term_scorers);
let mut limit = Score::MIN;
loop {
if scorer.doc() == TERMINATED {
break;
}
let doc = scorer.doc();
let score = scorer.score();
if score > limit {
heap.push(Float(score));
if heap.len() > n {
heap.pop().unwrap();
}
if heap.len() == n {
limit = heap.peek().unwrap().0;
}
if !nearly_equals(score, limit) {
checkpoints.push((doc, score));
}
}
scorer.advance();
}
checkpoints
}
const MAX_TERM_FREQ: u32 = 100u32;
fn posting_list(max_doc: u32) -> BoxedStrategy<Vec<(DocId, u32)>> {
(1..max_doc + 1)
.prop_flat_map(move |doc_freq| {
(
proptest::bits::bitset::sampled(doc_freq as usize, 0..max_doc as usize),
proptest::collection::vec(1u32..MAX_TERM_FREQ, doc_freq as usize),
)
})
.prop_map(|(docset, term_freqs)| {
docset
.iter()
.map(|doc| doc as u32)
.zip(term_freqs.iter().cloned())
.collect::<Vec<_>>()
})
.boxed()
}
fn gen_term_scorers(num_scorers: usize) -> BoxedStrategy<(Vec<Vec<(DocId, u32)>>, Vec<u32>)> {
(1u32..100u32)
.prop_flat_map(move |max_doc: u32| {
(
proptest::collection::vec(posting_list(max_doc), num_scorers),
proptest::collection::vec(2u32..10u32 * MAX_TERM_FREQ, max_doc as usize),
)
})
.boxed()
}
fn test_block_wand_aux(posting_lists: &[Vec<(DocId, u32)>], fieldnorms: &[u32]) {
// We virtually repeat all docs 64 times in order to emulate blocks of 2 documents
// and surface blogs more easily.
const REPEAT: usize = 64;
let fieldnorms_expanded = fieldnorms
.iter()
.cloned()
.flat_map(|fieldnorm| iter::repeat(fieldnorm).take(REPEAT))
.collect::<Vec<u32>>();
let postings_lists_expanded: Vec<Vec<(DocId, u32)>> = posting_lists
.iter()
.map(|posting_list| {
posting_list
.into_iter()
.cloned()
.flat_map(|(doc, term_freq)| {
(0 as u32..REPEAT as u32).map(move |offset| {
(
doc * (REPEAT as u32) + offset,
if offset == 0 { term_freq } else { 1 },
)
})
})
.collect::<Vec<(DocId, u32)>>()
})
.collect::<Vec<_>>();
let total_fieldnorms: u64 = fieldnorms_expanded
.iter()
.cloned()
.map(|fieldnorm| fieldnorm as u64)
.sum();
let average_fieldnorm = (total_fieldnorms as Score) / (fieldnorms_expanded.len() as Score);
let max_doc = fieldnorms_expanded.len();
let term_scorers: Vec<TermScorer> = postings_lists_expanded
.iter()
.map(|postings| {
let bm25_weight = BM25Weight::for_one_term(
postings.len() as u64,
max_doc as u64,
average_fieldnorm,
);
TermScorer::create_for_test(postings, &fieldnorms_expanded[..], bm25_weight)
})
.collect();
for top_k in 1..4 {
let checkpoints_for_each_pruning =
compute_checkpoints_for_each_pruning(term_scorers.clone(), top_k);
let checkpoints_manual = compute_checkpoints_manual(term_scorers.clone(), top_k);
assert_eq!(checkpoints_for_each_pruning.len(), checkpoints_manual.len());
for (&(left_doc, left_score), &(right_doc, right_score)) in checkpoints_for_each_pruning
.iter()
.zip(checkpoints_manual.iter())
{
assert_eq!(left_doc, right_doc);
assert!(nearly_equals(left_score, right_score));
}
}
}
proptest! {
#![proptest_config(ProptestConfig::with_cases(500))]
#[test]
fn test_block_wand_two_term_scorers((posting_lists, fieldnorms) in gen_term_scorers(2)) {
test_block_wand_aux(&posting_lists[..], &fieldnorms[..]);
}
}
proptest! {
#![proptest_config(ProptestConfig::with_cases(500))]
#[test]
fn test_block_wand_three_term_scorers((posting_lists, fieldnorms) in gen_term_scorers(3)) {
test_block_wand_aux(&posting_lists[..], &fieldnorms[..]);
}
}
}

142
src/query/boolean_query/boolean_weight.rs
View File

@@ -1,9 +1,7 @@
use crate::core::SegmentReader;
use crate::postings::FreqReadingOption;
use crate::query::explanation::does_not_match;
use crate::query::score_combiner::{DoNothingCombiner, ScoreCombiner, SumWithCoordsCombiner};
use crate::query::term_query::TermScorer;
use crate::query::weight::{for_each_pruning_scorer, for_each_scorer};
use crate::query::EmptyScorer;
use crate::query::Exclude;
use crate::query::Occur;
@@ -12,21 +10,16 @@ use crate::query::Scorer;
use crate::query::Union;
use crate::query::Weight;
use crate::query::{intersect_scorers, Explanation};
use crate::{DocId, Score};
use crate::{DocId, SkipResult};
use std::collections::HashMap;
enum SpecializedScorer {
TermUnion(Vec<TermScorer>),
Other(Box<dyn Scorer>),
}
fn scorer_union<TScoreCombiner>(scorers: Vec<Box<dyn Scorer>>) -> SpecializedScorer
fn scorer_union<TScoreCombiner>(scorers: Vec<Box<dyn Scorer>>) -> Box<dyn Scorer>
where
TScoreCombiner: ScoreCombiner,
{
assert!(!scorers.is_empty());
if scorers.len() == 1 {
return SpecializedScorer::Other(scorers.into_iter().next().unwrap()); //< we checked the size beforehands
return scorers.into_iter().next().unwrap(); //< we checked the size beforehands
}
{
@@ -36,30 +29,14 @@ where
.into_iter()
.map(|scorer| *(scorer.downcast::<TermScorer>().map_err(|_| ()).unwrap()))
.collect();
if scorers
.iter()
.all(|scorer| scorer.freq_reading_option() == FreqReadingOption::ReadFreq)
{
// Block wand is only available iff we read frequencies.
return SpecializedScorer::TermUnion(scorers);
} else {
return SpecializedScorer::Other(Box::new(Union::<_, TScoreCombiner>::from(
scorers,
)));
}
let scorer: Box<dyn Scorer> =
Box::new(Union::<TermScorer, TScoreCombiner>::from(scorers));
return scorer;
}
}
SpecializedScorer::Other(Box::new(Union::<_, TScoreCombiner>::from(scorers)))
}
fn into_box_scorer<TScoreCombiner: ScoreCombiner>(scorer: SpecializedScorer) -> Box<dyn Scorer> {
match scorer {
SpecializedScorer::TermUnion(term_scorers) => {
let union_scorer = Union::<TermScorer, TScoreCombiner>::from(term_scorers);
Box::new(union_scorer)
}
SpecializedScorer::Other(scorer) => scorer,
}
let scorer: Box<dyn Scorer> = Box::new(Union::<_, TScoreCombiner>::from(scorers));
scorer
}
pub struct BooleanWeight {
@@ -78,11 +55,10 @@ impl BooleanWeight {
fn per_occur_scorers(
&self,
reader: &SegmentReader,
boost: Score,
) -> crate::Result<HashMap<Occur, Vec<Box<dyn Scorer>>>> {
let mut per_occur_scorers: HashMap<Occur, Vec<Box<dyn Scorer>>> = HashMap::new();
for &(ref occur, ref subweight) in &self.weights {
let sub_scorer: Box<dyn Scorer> = subweight.scorer(reader, boost)?;
let sub_scorer: Box<dyn Scorer> = subweight.scorer(reader)?;
per_occur_scorers
.entry(*occur)
.or_insert_with(Vec::new)
@@ -94,52 +70,41 @@ impl BooleanWeight {
fn complex_scorer<TScoreCombiner: ScoreCombiner>(
&self,
reader: &SegmentReader,
boost: Score,
) -> crate::Result<SpecializedScorer> {
let mut per_occur_scorers = self.per_occur_scorers(reader, boost)?;
) -> crate::Result<Box<dyn Scorer>> {
let mut per_occur_scorers = self.per_occur_scorers(reader)?;
let should_scorer_opt: Option<SpecializedScorer> = per_occur_scorers
let should_scorer_opt: Option<Box<dyn Scorer>> = per_occur_scorers
.remove(&Occur::Should)
.map(scorer_union::<TScoreCombiner>);
let exclude_scorer_opt: Option<Box<dyn Scorer>> = per_occur_scorers
.remove(&Occur::MustNot)
.map(scorer_union::<DoNothingCombiner>)
.map(into_box_scorer::<DoNothingCombiner>);
.map(scorer_union::<TScoreCombiner>);
let must_scorer_opt: Option<Box<dyn Scorer>> = per_occur_scorers
.remove(&Occur::Must)
.map(intersect_scorers);
let positive_scorer: SpecializedScorer = match (should_scorer_opt, must_scorer_opt) {
let positive_scorer: Box<dyn Scorer> = match (should_scorer_opt, must_scorer_opt) {
(Some(should_scorer), Some(must_scorer)) => {
if self.scoring_enabled {
SpecializedScorer::Other(Box::new(RequiredOptionalScorer::<
Box<dyn Scorer>,
Box<dyn Scorer>,
TScoreCombiner,
>::new(
Box::new(RequiredOptionalScorer::<_, _, TScoreCombiner>::new(
must_scorer,
into_box_scorer::<TScoreCombiner>(should_scorer),
)))
should_scorer,
))
} else {
SpecializedScorer::Other(must_scorer)
must_scorer
}
}
(None, Some(must_scorer)) => SpecializedScorer::Other(must_scorer),
(None, Some(must_scorer)) => must_scorer,
(Some(should_scorer), None) => should_scorer,
(None, None) => {
return Ok(SpecializedScorer::Other(Box::new(EmptyScorer)));
return Ok(Box::new(EmptyScorer));
}
};
if let Some(exclude_scorer) = exclude_scorer_opt {
let positive_scorer_boxed: Box<dyn Scorer> =
into_box_scorer::<TScoreCombiner>(positive_scorer);
Ok(SpecializedScorer::Other(Box::new(Exclude::new(
positive_scorer_boxed,
exclude_scorer,
))))
Ok(Box::new(Exclude::new(positive_scorer, exclude_scorer)))
} else {
Ok(positive_scorer)
}
@@ -147,7 +112,7 @@ impl BooleanWeight {
}
impl Weight for BooleanWeight {
fn scorer(&self, reader: &SegmentReader, boost: Score) -> crate::Result<Box<dyn Scorer>> {
fn scorer(&self, reader: &SegmentReader) -> crate::Result<Box<dyn Scorer>> {
if self.weights.is_empty() {
Ok(Box::new(EmptyScorer))
} else if self.weights.len() == 1 {
@@ -155,26 +120,22 @@ impl Weight for BooleanWeight {
if occur == Occur::MustNot {
Ok(Box::new(EmptyScorer))
} else {
weight.scorer(reader, boost)
weight.scorer(reader)
}
} else if self.scoring_enabled {
self.complex_scorer::<SumWithCoordsCombiner>(reader, boost)
.map(|specialized_scorer| {
into_box_scorer::<SumWithCoordsCombiner>(specialized_scorer)
})
self.complex_scorer::<SumWithCoordsCombiner>(reader)
} else {
self.complex_scorer::<DoNothingCombiner>(reader, boost)
.map(into_box_scorer::<DoNothingCombiner>)
self.complex_scorer::<DoNothingCombiner>(reader)
}
}
fn explain(&self, reader: &SegmentReader, doc: DocId) -> crate::Result<Explanation> {
let mut scorer = self.scorer(reader, 1.0)?;
if scorer.seek(doc) != doc {
let mut scorer = self.scorer(reader)?;
if scorer.skip_next(doc) != SkipResult::Reached {
return Err(does_not_match(doc));
}
if !self.scoring_enabled {
return Ok(Explanation::new("BooleanQuery with no scoring", 1.0));
return Ok(Explanation::new("BooleanQuery with no scoring", 1f32));
}
let mut explanation = Explanation::new("BooleanClause. Sum of ...", scorer.score());
@@ -187,53 +148,6 @@ impl Weight for BooleanWeight {
}
Ok(explanation)
}
fn for_each(
&self,
reader: &SegmentReader,
callback: &mut dyn FnMut(DocId, Score),
) -> crate::Result<()> {
let scorer = self.complex_scorer::<SumWithCoordsCombiner>(reader, 1.0)?;
match scorer {
SpecializedScorer::TermUnion(term_scorers) => {
let mut union_scorer =
Union::<TermScorer, SumWithCoordsCombiner>::from(term_scorers);
for_each_scorer(&mut union_scorer, callback);
}
SpecializedScorer::Other(mut scorer) => {
for_each_scorer(scorer.as_mut(), callback);
}
}
Ok(())
}
/// Calls `callback` with all of the `(doc, score)` for which score
/// is exceeding a given threshold.
///
/// This method is useful for the TopDocs collector.
/// For all docsets, the blanket implementation has the benefit
/// of prefiltering (doc, score) pairs, avoiding the
/// virtual dispatch cost.
///
/// More importantly, it makes it possible for scorers to implement
/// important optimization (e.g. BlockWAND for union).
fn for_each_pruning(
&self,
threshold: Score,
reader: &SegmentReader,
callback: &mut dyn FnMut(DocId, Score) -> Score,
) -> crate::Result<()> {
let scorer = self.complex_scorer::<SumWithCoordsCombiner>(reader, 1.0)?;
match scorer {
SpecializedScorer::TermUnion(term_scorers) => {
super::block_wand(term_scorers, threshold, callback);
}
SpecializedScorer::Other(mut scorer) => {
for_each_pruning_scorer(scorer.as_mut(), threshold, callback);
}
}
Ok(())
}
}
fn is_positive_occur(occur: Occur) -> bool {

297
src/query/boolean_query/mod.rs
View File

@@ -1,17 +1,13 @@
mod block_wand;
mod boolean_query;
mod boolean_weight;
pub(crate) use self::block_wand::block_wand;
pub use self::boolean_query::BooleanQuery;
#[cfg(test)]
mod tests {
use super::*;
use crate::assert_nearly_equals;
use crate::collector::tests::TEST_COLLECTOR_WITH_SCORE;
use crate::collector::TopDocs;
use crate::query::score_combiner::SumWithCoordsCombiner;
use crate::query::term_query::TermScorer;
use crate::query::Intersection;
@@ -23,7 +19,7 @@ mod tests {
use crate::query::TermQuery;
use crate::schema::*;
use crate::Index;
use crate::{DocAddress, DocId, Score};
use crate::{DocAddress, DocId};
fn aux_test_helper() -> (Index, Field) {
let mut schema_builder = Schema::builder();
@@ -32,13 +28,26 @@ mod tests {
let index = Index::create_in_ram(schema);
{
// writing the segment
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
{
index_writer.add_document(doc!(text_field => "a b c"));
index_writer.add_document(doc!(text_field => "a c"));
index_writer.add_document(doc!(text_field => "b c"));
index_writer.add_document(doc!(text_field => "a b c d"));
index_writer.add_document(doc!(text_field => "d"));
let doc = doc!(text_field => "a b c");
index_writer.add_document(doc);
}
{
let doc = doc!(text_field => "a c");
index_writer.add_document(doc);
}
{
let doc = doc!(text_field => "b c");
index_writer.add_document(doc);
}
{
let doc = doc!(text_field => "a b c d");
index_writer.add_document(doc);
}
{
let doc = doc!(text_field => "d");
index_writer.add_document(doc);
}
assert!(index_writer.commit().is_ok());
}
@@ -61,7 +70,7 @@ mod tests {
let query = query_parser.parse_query("+a").unwrap();
let searcher = index.reader().unwrap().searcher();
let weight = query.weight(&searcher, true).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32), 1.0).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap();
assert!(scorer.is::<TermScorer>());
}
@@ -73,13 +82,13 @@ mod tests {
{
let query = query_parser.parse_query("+a +b +c").unwrap();
let weight = query.weight(&searcher, true).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32), 1.0).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap();
assert!(scorer.is::<Intersection<TermScorer>>());
}
{
let query = query_parser.parse_query("+a +(b c)").unwrap();
let weight = query.weight(&searcher, true).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32), 1.0).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap();
assert!(scorer.is::<Intersection<Box<dyn Scorer>>>());
}
}
@@ -92,7 +101,7 @@ mod tests {
{
let query = query_parser.parse_query("+a b").unwrap();
let weight = query.weight(&searcher, true).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32), 1.0).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap();
assert!(scorer.is::<RequiredOptionalScorer<
Box<dyn Scorer>,
Box<dyn Scorer>,
@@ -102,7 +111,7 @@ mod tests {
{
let query = query_parser.parse_query("+a b").unwrap();
let weight = query.weight(&searcher, false).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32), 1.0).unwrap();
let scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap();
assert!(scorer.is::<TermScorer>());
}
}
@@ -133,6 +142,7 @@ mod tests {
.map(|doc| doc.1)
.collect::<Vec<DocId>>()
};
{
let boolean_query = BooleanQuery::from(vec![(Occur::Must, make_term_query("a"))]);
assert_eq!(matching_docs(&boolean_query), vec![0, 1, 3]);
@@ -169,96 +179,6 @@ mod tests {
}
}
#[test]
pub fn test_boolean_query_two_excluded() {
let (index, text_field) = aux_test_helper();
let make_term_query = |text: &str| {
let term_query = TermQuery::new(
Term::from_field_text(text_field, text),
IndexRecordOption::Basic,
);
let query: Box<dyn Query> = Box::new(term_query);
query
};
let reader = index.reader().unwrap();
let matching_topdocs = |query: &dyn Query| {
reader
.searcher()
.search(query, &TopDocs::with_limit(3))
.unwrap()
};
let score_doc_4: Score; // score of doc 4 should not be influenced by exclusion
{
let boolean_query_no_excluded =
BooleanQuery::from(vec![(Occur::Must, make_term_query("d"))]);
let topdocs_no_excluded = matching_topdocs(&boolean_query_no_excluded);
assert_eq!(topdocs_no_excluded.len(), 2);
let (top_score, top_doc) = topdocs_no_excluded[0];
assert_eq!(top_doc, DocAddress(0, 4));
assert_eq!(topdocs_no_excluded[1].1, DocAddress(0, 3)); // ignore score of doc 3.
score_doc_4 = top_score;
}
{
let boolean_query_two_excluded = BooleanQuery::from(vec![
(Occur::Must, make_term_query("d")),
(Occur::MustNot, make_term_query("a")),
(Occur::MustNot, make_term_query("b")),
]);
let topdocs_excluded = matching_topdocs(&boolean_query_two_excluded);
assert_eq!(topdocs_excluded.len(), 1);
let (top_score, top_doc) = topdocs_excluded[0];
assert_eq!(top_doc, DocAddress(0, 4));
assert_eq!(top_score, score_doc_4);
}
}
#[test]
pub fn test_boolean_query_with_weight() {
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);
{
let mut index_writer = index.writer_for_tests().unwrap();
index_writer.add_document(doc!(text_field => "a b c"));
index_writer.add_document(doc!(text_field => "a c"));
index_writer.add_document(doc!(text_field => "b c"));
assert!(index_writer.commit().is_ok());
}
let term_a: Box<dyn Query> = Box::new(TermQuery::new(
Term::from_field_text(text_field, "a"),
IndexRecordOption::WithFreqs,
));
let term_b: Box<dyn Query> = Box::new(TermQuery::new(
Term::from_field_text(text_field, "b"),
IndexRecordOption::WithFreqs,
));
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let boolean_query =
BooleanQuery::from(vec![(Occur::Should, term_a), (Occur::Should, term_b)]);
let boolean_weight = boolean_query.weight(&searcher, true).unwrap();
{
let mut boolean_scorer = boolean_weight
.scorer(searcher.segment_reader(0u32), 1.0)
.unwrap();
assert_eq!(boolean_scorer.doc(), 0u32);
assert_nearly_equals!(boolean_scorer.score(), 0.84163445);
}
{
let mut boolean_scorer = boolean_weight
.scorer(searcher.segment_reader(0u32), 2.0)
.unwrap();
assert_eq!(boolean_scorer.doc(), 0u32);
assert_nearly_equals!(boolean_scorer.score(), 1.6832689);
}
}
#[test]
pub fn test_intersection_score() {
let (index, text_field) = aux_test_helper();
@@ -285,9 +205,168 @@ mod tests {
(Occur::Must, make_term_query("a")),
(Occur::Must, make_term_query("b")),
]);
let scores = score_docs(&boolean_query);
assert_nearly_equals!(scores[0], 0.977973);
assert_nearly_equals!(scores[1], 0.84699446);
assert_eq!(score_docs(&boolean_query), vec![0.977973, 0.84699446]);
}
}
// motivated by #554
#[test]
fn test_bm25_several_fields() {
let mut schema_builder = Schema::builder();
let title = schema_builder.add_text_field("title", TEXT);
let text = schema_builder.add_text_field("text", TEXT);
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();
index_writer.add_document(doc!(
// tf = 1 0
title => "Законы притяжения Оксана Кулакова",
// tf = 1 0
text => "Законы притяжения Оксана Кулакова] \n\nТема: Сексуальное искусство, Женственность\nТип товара: Запись вебинара (аудио)\nПродолжительность: 1,5 часа\n\nСсылка на вебинар:\n ",
));
index_writer.add_document(doc!(
// tf = 1 0
title => "Любимые русские пироги (Оксана Путан)",
// tf = 2 0
text => "http://i95.fastpic.ru/big/2017/0628/9a/615b9c8504d94a3893d7f496ac53539a.jpg \n\nОт издателя\nОксана Путан профессиональный повар, автор кулинарных книг и известный кулинарный блогер. Ее рецепты отличаются практичностью, доступностью и пользуются огромной популярностью в русскоязычном интернете. Это третья книга автора о самом вкусном и ароматном настоящих русских пирогах и выпечке!\nДаже новички на кухне легко готовят по ее рецептам. Оксана описывает процесс приготовления настолько подробно и понятно, что вам остается только наслаждаться готовкой и не тратить время на лишние усилия. Готовьте легко и просто!\n\nhttps://www.ozon.ru/context/detail/id/139872462/"
));
index_writer.add_document(doc!(
// tf = 1 1
title => "PDF Мастер Класс \"Морячок\" (Оксана Лифенко)",
// tf = 0 0
text => "https://i.ibb.co/pzvHrDN/I3d U T6 Gg TM.jpg\nhttps://i.ibb.co/NFrb6v6/N0ls Z9nwjb U.jpg\nВ описание входит штаны, кофта, берет, матросский воротник. Описание продается в формате PDF, состоит из 12 страниц формата А4 и может быть напечатано на любом принтере.\nОписание предназначено для кукол BJD RealPuki от FairyLand, но может подойти и другим подобным куклам. Также вы можете вязать этот наряд из обычной пряжи, и он подойдет для куколок побольше.\nhttps://vk.com/market 95724412?w=product 95724412_2212"
));
for _ in 0..1_000 {
index_writer.add_document(doc!(
title => "a b d e f g",
text => "maitre corbeau sur un arbre perche tenait dans son bec un fromage Maitre rnard par lodeur alleche lui tint a peu pres ce langage."
));
}
index_writer.commit().unwrap();
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let query_parser = QueryParser::for_index(&index, vec![title, text]);
let query = query_parser.parse_query("Оксана Лифенко").unwrap();
let weight = query.weight(&searcher, true).unwrap();
let mut scorer = weight.scorer(searcher.segment_reader(0u32)).unwrap();
scorer.advance();
let explanation = query.explain(&searcher, DocAddress(0u32, 0u32)).unwrap();
assert_eq!(
explanation.to_pretty_json(),
r#"{
"value": 12.997711,
"description": "BooleanClause. Sum of ...",
"details": [
{
"value": 12.997711,
"description": "BooleanClause. Sum of ...",
"details": [
{
"value": 6.551476,
"description": "TermQuery, product of...",
"details": [
{
"value": 2.2,
"description": "(K1+1)"
},
{
"value": 5.658984,
"description": "idf, computed as log(1 + (N - n + 0.5) / (n + 0.5))",
"details": [
{
"value": 3.0,
"description": "n, number of docs containing this term"
},
{
"value": 1003.0,
"description": "N, total number of docs"
}
]
},
{
"value": 0.5262329,
"description": "freq / (freq + k1 * (1 - b + b * dl / avgdl))",
"details": [
{
"value": 1.0,
"description": "freq, occurrences of term within document"
},
{
"value": 1.2,
"description": "k1, term saturation parameter"
},
{
"value": 0.75,
"description": "b, length normalization parameter"
},
{
"value": 4.0,
"description": "dl, length of field"
},
{
"value": 5.997009,
"description": "avgdl, average length of field"
}
]
}
]
},
{
"value": 6.446235,
"description": "TermQuery, product of...",
"details": [
{
"value": 2.2,
"description": "(K1+1)"
},
{
"value": 5.9954567,
"description": "idf, computed as log(1 + (N - n + 0.5) / (n + 0.5))",
"details": [
{
"value": 2.0,
"description": "n, number of docs containing this term"
},
{
"value": 1003.0,
"description": "N, total number of docs"
}
]
},
{
"value": 0.4887212,
"description": "freq / (freq + k1 * (1 - b + b * dl / avgdl))",
"details": [
{
"value": 1.0,
"description": "freq, occurrences of term within document"
},
{
"value": 1.2,
"description": "k1, term saturation parameter"
},
{
"value": 0.75,
"description": "b, length normalization parameter"
},
{
"value": 20.0,
"description": "dl, length of field"
},
{
"value": 24.123629,
"description": "avgdl, average length of field"
}
]
}
]
}
]
}
]
}"#
);
}
}

159
src/query/boost_query.rs
View File

@@ -1,159 +0,0 @@
use crate::fastfield::DeleteBitSet;
use crate::query::explanation::does_not_match;
use crate::query::{Explanation, Query, Scorer, Weight};
use crate::{DocId, DocSet, Score, Searcher, SegmentReader, Term};
use std::collections::BTreeSet;
use std::fmt;
/// `BoostQuery` is a wrapper over a query used to boost its score.
///
/// The document set matched by the `BoostQuery` is strictly the same as the underlying query.
/// The score of each document, is the score of the underlying query multiplied by the `boost`
/// factor.
pub struct BoostQuery {
query: Box<dyn Query>,
boost: Score,
}
impl BoostQuery {
/// Builds a boost query.
pub fn new(query: Box<dyn Query>, boost: Score) -> BoostQuery {
BoostQuery { query, boost }
}
}
impl Clone for BoostQuery {
fn clone(&self) -> Self {
BoostQuery {
query: self.query.box_clone(),
boost: self.boost,
}
}
}
impl fmt::Debug for BoostQuery {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "Boost(query={:?}, boost={})", self.query, self.boost)
}
}
impl Query for BoostQuery {
fn weight(&self, searcher: &Searcher, scoring_enabled: bool) -> crate::Result<Box<dyn Weight>> {
let weight_without_boost = self.query.weight(searcher, scoring_enabled)?;
let boosted_weight = if scoring_enabled {
Box::new(BoostWeight::new(weight_without_boost, self.boost))
} else {
weight_without_boost
};
Ok(boosted_weight)
}
fn query_terms(&self, term_set: &mut BTreeSet<Term>) {
self.query.query_terms(term_set)
}
}
pub(crate) struct BoostWeight {
weight: Box<dyn Weight>,
boost: Score,
}
impl BoostWeight {
pub fn new(weight: Box<dyn Weight>, boost: Score) -> Self {
BoostWeight { weight, boost }
}
}
impl Weight for BoostWeight {
fn scorer(&self, reader: &SegmentReader, boost: Score) -> crate::Result<Box<dyn Scorer>> {
self.weight.scorer(reader, boost * self.boost)
}
fn explain(&self, reader: &SegmentReader, doc: u32) -> crate::Result<Explanation> {
let mut scorer = self.scorer(reader, 1.0)?;
if scorer.seek(doc) != doc {
return Err(does_not_match(doc));
}
let mut explanation =
Explanation::new(format!("Boost x{} of ...", self.boost), scorer.score());
let underlying_explanation = self.weight.explain(reader, doc)?;
explanation.add_detail(underlying_explanation);
Ok(explanation)
}
fn count(&self, reader: &SegmentReader) -> crate::Result<u32> {
self.weight.count(reader)
}
}
pub(crate) struct BoostScorer<S: Scorer> {
underlying: S,
boost: Score,
}
impl<S: Scorer> BoostScorer<S> {
pub fn new(underlying: S, boost: Score) -> BoostScorer<S> {
BoostScorer { underlying, boost }
}
}
impl<S: Scorer> DocSet for BoostScorer<S> {
fn advance(&mut self) -> DocId {
self.underlying.advance()
}
fn seek(&mut self, target: DocId) -> DocId {
self.underlying.seek(target)
}
fn fill_buffer(&mut self, buffer: &mut [DocId]) -> usize {
self.underlying.fill_buffer(buffer)
}
fn doc(&self) -> u32 {
self.underlying.doc()
}
fn size_hint(&self) -> u32 {
self.underlying.size_hint()
}
fn count(&mut self, delete_bitset: &DeleteBitSet) -> u32 {
self.underlying.count(delete_bitset)
}
fn count_including_deleted(&mut self) -> u32 {
self.underlying.count_including_deleted()
}
}
impl<S: Scorer> Scorer for BoostScorer<S> {
fn score(&mut self) -> Score {
self.underlying.score() * self.boost
}
}
#[cfg(test)]
mod tests {
use super::BoostQuery;
use crate::query::{AllQuery, Query};
use crate::schema::Schema;
use crate::{DocAddress, Document, Index};
#[test]
fn test_boost_query_explain() {
let schema = Schema::builder().build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
index_writer.add_document(Document::new());
assert!(index_writer.commit().is_ok());
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let query = BoostQuery::new(Box::new(AllQuery), 0.2);
let explanation = query.explain(&searcher, DocAddress(0, 0u32)).unwrap();
assert_eq!(
explanation.to_pretty_json(),
"{\n \"value\": 0.2,\n \"description\": \"Boost x0.2 of ...\",\n \"details\": [\n {\n \"value\": 1.0,\n \"description\": \"AllQuery\"\n }\n ]\n}"
)
}
}

25
src/query/empty_query.rs
View File

@@ -1,5 +1,4 @@
use super::Scorer;
use crate::docset::TERMINATED;
use crate::query::explanation::does_not_match;
use crate::query::Weight;
use crate::query::{Explanation, Query};
@@ -34,7 +33,7 @@ impl Query for EmptyQuery {
/// It is useful for tests and handling edge cases.
pub struct EmptyWeight;
impl Weight for EmptyWeight {
fn scorer(&self, _reader: &SegmentReader, _boost: Score) -> crate::Result<Box<dyn Scorer>> {
fn scorer(&self, _reader: &SegmentReader) -> crate::Result<Box<dyn Scorer>> {
Ok(Box::new(EmptyScorer))
}
@@ -49,12 +48,15 @@ impl Weight for EmptyWeight {
pub struct EmptyScorer;
impl DocSet for EmptyScorer {
fn advance(&mut self) -> DocId {
TERMINATED
fn advance(&mut self) -> bool {
false
}
fn doc(&self) -> DocId {
TERMINATED
panic!(
"You may not call .doc() on a scorer \
where the last call to advance() did not return true."
);
}
fn size_hint(&self) -> u32 {
@@ -64,21 +66,24 @@ impl DocSet for EmptyScorer {
impl Scorer for EmptyScorer {
fn score(&mut self) -> Score {
0.0
0f32
}
}
#[cfg(test)]
mod tests {
use crate::docset::TERMINATED;
use crate::query::EmptyScorer;
use crate::DocSet;
#[test]
fn test_empty_scorer() {
let mut empty_scorer = EmptyScorer;
assert_eq!(empty_scorer.doc(), TERMINATED);
assert_eq!(empty_scorer.advance(), TERMINATED);
assert_eq!(empty_scorer.doc(), TERMINATED);
assert!(!empty_scorer.advance());
}
#[test]
#[should_panic]
fn test_empty_scorer_panic_on_doc_call() {
EmptyScorer.doc();
}
}

107
src/query/exclude.rs
View File

@@ -1,11 +1,12 @@
use crate::docset::{DocSet, TERMINATED};
use crate::docset::{DocSet, SkipResult};
use crate::query::Scorer;
use crate::DocId;
use crate::Score;
#[inline(always)]
fn is_within<TDocSetExclude: DocSet>(docset: &mut TDocSetExclude, doc: DocId) -> bool {
docset.doc() <= doc && docset.seek(doc) == doc
#[derive(Clone, Copy, Debug)]
enum State {
ExcludeOne(DocId),
Finished,
}
/// Filters a given `DocSet` by removing the docs from a given `DocSet`.
@@ -14,6 +15,29 @@ fn is_within<TDocSetExclude: DocSet>(docset: &mut TDocSetExclude, doc: DocId) ->
pub struct Exclude<TDocSet, TDocSetExclude> {
underlying_docset: TDocSet,
excluding_docset: TDocSetExclude,
excluding_state: State,
}
impl<TDocSet, TDocSetExclude> Exclude<TDocSet, TDocSetExclude>
where
TDocSetExclude: DocSet,
{
/// Creates a new `ExcludeScorer`
pub fn new(
underlying_docset: TDocSet,
mut excluding_docset: TDocSetExclude,
) -> Exclude<TDocSet, TDocSetExclude> {
let state = if excluding_docset.advance() {
State::ExcludeOne(excluding_docset.doc())
} else {
State::Finished
};
Exclude {
underlying_docset,
excluding_docset,
excluding_state: state,
}
}
}
impl<TDocSet, TDocSetExclude> Exclude<TDocSet, TDocSetExclude>
@@ -21,21 +45,33 @@ where
TDocSet: DocSet,
TDocSetExclude: DocSet,
{
/// Creates a new `ExcludeScorer`
pub fn new(
mut underlying_docset: TDocSet,
mut excluding_docset: TDocSetExclude,
) -> Exclude<TDocSet, TDocSetExclude> {
while underlying_docset.doc() != TERMINATED {
let target = underlying_docset.doc();
if !is_within(&mut excluding_docset, target) {
break;
/// Returns true iff the doc is not removed.
///
/// The method has to be called with non strictly
/// increasing `doc`.
fn accept(&mut self) -> bool {
let doc = self.underlying_docset.doc();
match self.excluding_state {
State::ExcludeOne(excluded_doc) => {
if doc == excluded_doc {
return false;
}
if excluded_doc > doc {
return true;
}
match self.excluding_docset.skip_next(doc) {
SkipResult::OverStep => {
self.excluding_state = State::ExcludeOne(self.excluding_docset.doc());
true
}
SkipResult::End => {
self.excluding_state = State::Finished;
true
}
SkipResult::Reached => false,
}
}
underlying_docset.advance();
}
Exclude {
underlying_docset,
excluding_docset,
State::Finished => true,
}
}
}
@@ -45,27 +81,27 @@ where
TDocSet: DocSet,
TDocSetExclude: DocSet,
{
fn advance(&mut self) -> DocId {
loop {
let candidate = self.underlying_docset.advance();
if candidate == TERMINATED {
return TERMINATED;
}
if !is_within(&mut self.excluding_docset, candidate) {
return candidate;
fn advance(&mut self) -> bool {
while self.underlying_docset.advance() {
if self.accept() {
return true;
}
}
false
}
fn seek(&mut self, target: DocId) -> DocId {
let candidate = self.underlying_docset.seek(target);
if candidate == TERMINATED {
return TERMINATED;
fn skip_next(&mut self, target: DocId) -> SkipResult {
let underlying_skip_result = self.underlying_docset.skip_next(target);
if underlying_skip_result == SkipResult::End {
return SkipResult::End;
}
if !is_within(&mut self.excluding_docset, candidate) {
return candidate;
if self.accept() {
underlying_skip_result
} else if self.advance() {
SkipResult::OverStep
} else {
SkipResult::End
}
self.advance()
}
fn doc(&self) -> DocId {
@@ -105,9 +141,8 @@ mod tests {
VecDocSet::from(vec![1, 2, 3, 10, 16, 24]),
);
let mut els = vec![];
while exclude_scorer.doc() != TERMINATED {
while exclude_scorer.advance() {
els.push(exclude_scorer.doc());
exclude_scorer.advance();
}
assert_eq!(els, vec![5, 8, 15]);
}
@@ -121,7 +156,7 @@ mod tests {
VecDocSet::from(vec![1, 2, 3, 10, 16, 24]),
))
},
vec![5, 8, 10, 15, 24],
vec![1, 2, 5, 8, 10, 15, 24],
);
}

18
src/query/explanation.rs
View File

@@ -1,6 +1,4 @@
use crate::{DocId, Score, TantivyError};
use serde::Serialize;
use std::fmt;
use crate::{DocId, TantivyError};
pub(crate) fn does_not_match(doc: DocId) -> TantivyError {
TantivyError::InvalidArgument(format!("Document #({}) does not match", doc))
@@ -13,21 +11,15 @@ pub(crate) fn does_not_match(doc: DocId) -> TantivyError {
/// representation of this tree when debugging a given score.
#[derive(Clone, Serialize)]
pub struct Explanation {
value: Score,
value: f32,
description: String,
#[serde(skip_serializing_if = "Vec::is_empty")]
details: Vec<Explanation>,
}
impl fmt::Debug for Explanation {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "Explanation({})", self.to_pretty_json())
}
}
impl Explanation {
/// Creates a new explanation object.
pub fn new<T: ToString>(description: T, value: Score) -> Explanation {
pub fn new<T: ToString>(description: T, value: f32) -> Explanation {
Explanation {
value,
description: description.to_string(),
@@ -36,7 +28,7 @@ impl Explanation {
}
/// Returns the value associated to the current node.
pub fn value(&self) -> Score {
pub fn value(&self) -> f32 {
self.value
}
@@ -48,7 +40,7 @@ impl Explanation {
}
/// Shortcut for `self.details.push(Explanation::new(name, value));`
pub fn add_const<T: ToString>(&mut self, name: T, value: Score) {
pub fn add_const<T: ToString>(&mut self, name: T, value: f32) {
self.details.push(Explanation::new(name, value));
}

77
src/query/fuzzy_query.rs
View File

@@ -2,40 +2,14 @@ use crate::query::{AutomatonWeight, Query, Weight};
use crate::schema::Term;
use crate::Searcher;
use crate::TantivyError::InvalidArgument;
use levenshtein_automata::{Distance, LevenshteinAutomatonBuilder, DFA};
use levenshtein_automata::{LevenshteinAutomatonBuilder, DFA};
use once_cell::sync::Lazy;
use std::collections::HashMap;
use std::ops::Range;
use tantivy_fst::Automaton;
pub(crate) struct DFAWrapper(pub DFA);
impl Automaton for DFAWrapper {
type State = u32;
fn start(&self) -> Self::State {
self.0.initial_state()
}
fn is_match(&self, state: &Self::State) -> bool {
match self.0.distance(*state) {
Distance::Exact(_) => true,
Distance::AtLeast(_) => false,
}
}
fn can_match(&self, state: &u32) -> bool {
*state != levenshtein_automata::SINK_STATE
}
fn accept(&self, state: &Self::State, byte: u8) -> Self::State {
self.0.transition(*state, byte)
}
}
/// A range of Levenshtein distances that we will build DFAs for our terms
/// The computation is exponential, so best keep it to low single digits
const VALID_LEVENSHTEIN_DISTANCE_RANGE: Range<u8> = 0..3;
const VALID_LEVENSHTEIN_DISTANCE_RANGE: Range<u8> = (0..3);
static LEV_BUILDER: Lazy<HashMap<(u8, bool), LevenshteinAutomatonBuilder>> = Lazy::new(|| {
let mut lev_builder_cache = HashMap::new();
@@ -117,7 +91,7 @@ impl FuzzyTermQuery {
}
}
/// Creates a new Fuzzy Query of the Term prefix
/// Creates a new Fuzzy Query that treats transpositions as cost one rather than two
pub fn new_prefix(term: Term, distance: u8, transposition_cost_one: bool) -> FuzzyTermQuery {
FuzzyTermQuery {
term,
@@ -127,20 +101,13 @@ impl FuzzyTermQuery {
}
}
fn specialized_weight(&self) -> crate::Result<AutomatonWeight<DFAWrapper>> {
fn specialized_weight(&self) -> crate::Result<AutomatonWeight<DFA>> {
// LEV_BUILDER is a HashMap, whose `get` method returns an Option
match LEV_BUILDER.get(&(self.distance, false)) {
// Unwrap the option and build the Ok(AutomatonWeight)
Some(automaton_builder) => {
let automaton = if self.prefix {
automaton_builder.build_prefix_dfa(self.term.text())
} else {
automaton_builder.build_dfa(self.term.text())
};
Ok(AutomatonWeight::new(
self.term.field(),
DFAWrapper(automaton),
))
let automaton = automaton_builder.build_dfa(self.term.text());
Ok(AutomatonWeight::new(self.term.field(), automaton))
}
None => Err(InvalidArgument(format!(
"Levenshtein distance of {} is not allowed. Choose a value in the {:?} range",
@@ -163,10 +130,10 @@ impl Query for FuzzyTermQuery {
#[cfg(test)]
mod test {
use super::FuzzyTermQuery;
use crate::assert_nearly_equals;
use crate::collector::TopDocs;
use crate::schema::Schema;
use crate::schema::TEXT;
use crate::tests::assert_nearly_equals;
use crate::Index;
use crate::Term;
@@ -177,7 +144,7 @@ mod test {
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 10_000_000).unwrap();
index_writer.add_document(doc!(
country_field => "japan",
));
@@ -188,8 +155,6 @@ mod test {
}
let reader = index.reader().unwrap();
let searcher = reader.searcher();
// passes because Levenshtein distance is 1 (substitute 'o' with 'a')
{
let term = Term::from_field_text(country_field, "japon");
@@ -199,31 +164,7 @@ mod test {
.unwrap();
assert_eq!(top_docs.len(), 1, "Expected only 1 document");
let (score, _) = top_docs[0];
assert_nearly_equals!(1.0, score);
}
// fails because non-prefix Levenshtein distance is more than 1 (add 'a' and 'n')
{
let term = Term::from_field_text(country_field, "jap");
let fuzzy_query = FuzzyTermQuery::new(term, 1, true);
let top_docs = searcher
.search(&fuzzy_query, &TopDocs::with_limit(2))
.unwrap();
assert_eq!(top_docs.len(), 0, "Expected no document");
}
// passes because prefix Levenshtein distance is 0
{
let term = Term::from_field_text(country_field, "jap");
let fuzzy_query = FuzzyTermQuery::new_prefix(term, 1, true);
let top_docs = searcher
.search(&fuzzy_query, &TopDocs::with_limit(2))
.unwrap();
assert_eq!(top_docs.len(), 1, "Expected only 1 document");
let (score, _) = top_docs[0];
assert_nearly_equals!(1.0, score);
assert_nearly_equals(1f32, score);
}
}
}

192
src/query/intersection.rs
View File

@@ -1,4 +1,4 @@
use crate::docset::{DocSet, TERMINATED};
use crate::docset::{DocSet, SkipResult};
use crate::query::term_query::TermScorer;
use crate::query::EmptyScorer;
use crate::query::Scorer;
@@ -20,14 +20,12 @@ pub fn intersect_scorers(mut scorers: Vec<Box<dyn Scorer>>) -> Box<dyn Scorer> {
if scorers.len() == 1 {
return scorers.pop().unwrap();
}
scorers.sort_by_key(|scorer| scorer.size_hint());
let doc = go_to_first_doc(&mut scorers[..]);
if doc == TERMINATED {
return Box::new(EmptyScorer);
}
// We know that we have at least 2 elements.
let left = scorers.remove(0);
let right = scorers.remove(0);
let num_docsets = scorers.len();
scorers.sort_by(|left, right| right.size_hint().cmp(&left.size_hint()));
let left = scorers.pop().unwrap();
let right = scorers.pop().unwrap();
scorers.reverse();
let all_term_scorers = [&left, &right]
.iter()
.all(|&scorer| scorer.is::<TermScorer>());
@@ -36,12 +34,14 @@ pub fn intersect_scorers(mut scorers: Vec<Box<dyn Scorer>>) -> Box<dyn Scorer> {
left: *(left.downcast::<TermScorer>().map_err(|_| ()).unwrap()),
right: *(right.downcast::<TermScorer>().map_err(|_| ()).unwrap()),
others: scorers,
num_docsets,
});
}
Box::new(Intersection {
left,
right,
others: scorers,
num_docsets,
})
}
@@ -50,35 +50,22 @@ pub struct Intersection<TDocSet: DocSet, TOtherDocSet: DocSet = Box<dyn Scorer>>
left: TDocSet,
right: TDocSet,
others: Vec<TOtherDocSet>,
}
fn go_to_first_doc<TDocSet: DocSet>(docsets: &mut [TDocSet]) -> DocId {
assert!(!docsets.is_empty());
let mut candidate = docsets.iter().map(TDocSet::doc).max().unwrap();
'outer: loop {
for docset in docsets.iter_mut() {
let seek_doc = docset.seek(candidate);
if seek_doc > candidate {
candidate = docset.doc();
continue 'outer;
}
}
return candidate;
}
num_docsets: usize,
}
impl<TDocSet: DocSet> Intersection<TDocSet, TDocSet> {
pub(crate) fn new(mut docsets: Vec<TDocSet>) -> Intersection<TDocSet, TDocSet> {
let num_docsets = docsets.len();
assert!(num_docsets >= 2);
docsets.sort_by_key(|docset| docset.size_hint());
go_to_first_doc(&mut docsets);
let left = docsets.remove(0);
let right = docsets.remove(0);
docsets.sort_by(|left, right| right.size_hint().cmp(&left.size_hint()));
let left = docsets.pop().unwrap();
let right = docsets.pop().unwrap();
docsets.reverse();
Intersection {
left,
right,
others: docsets,
num_docsets,
}
}
}
@@ -93,49 +80,128 @@ impl<TDocSet: DocSet> Intersection<TDocSet, TDocSet> {
}
}
impl<TDocSet: DocSet, TOtherDocSet: DocSet> Intersection<TDocSet, TOtherDocSet> {
pub(crate) fn docset_mut(&mut self, ord: usize) -> &mut dyn DocSet {
match ord {
0 => &mut self.left,
1 => &mut self.right,
n => &mut self.others[n - 2],
}
}
}
impl<TDocSet: DocSet, TOtherDocSet: DocSet> DocSet for Intersection<TDocSet, TOtherDocSet> {
fn advance(&mut self) -> DocId {
fn advance(&mut self) -> bool {
let (left, right) = (&mut self.left, &mut self.right);
let mut candidate = left.advance();
if !left.advance() {
return false;
}
let mut candidate = left.doc();
let mut other_candidate_ord: usize = usize::max_value();
'outer: loop {
// In the first part we look for a document in the intersection
// of the two rarest `DocSet` in the intersection.
loop {
let right_doc = right.seek(candidate);
candidate = left.seek(right_doc);
if candidate == right_doc {
break;
match right.skip_next(candidate) {
SkipResult::Reached => {
break;
}
SkipResult::OverStep => {
candidate = right.doc();
other_candidate_ord = usize::max_value();
}
SkipResult::End => {
return false;
}
}
match left.skip_next(candidate) {
SkipResult::Reached => {
break;
}
SkipResult::OverStep => {
candidate = left.doc();
other_candidate_ord = usize::max_value();
}
SkipResult::End => {
return false;
}
}
}
debug_assert_eq!(left.doc(), right.doc());
// test the remaining scorers;
for docset in self.others.iter_mut() {
let seek_doc = docset.seek(candidate);
if seek_doc > candidate {
candidate = left.seek(seek_doc);
continue 'outer;
for (ord, docset) in self.others.iter_mut().enumerate() {
if ord == other_candidate_ord {
continue;
}
// `candidate_ord` is already at the
// right position.
//
// Calling `skip_next` would advance this docset
// and miss it.
match docset.skip_next(candidate) {
SkipResult::Reached => {}
SkipResult::OverStep => {
// this is not in the intersection,
// let's update our candidate.
candidate = docset.doc();
match left.skip_next(candidate) {
SkipResult::Reached => {
other_candidate_ord = ord;
}
SkipResult::OverStep => {
candidate = left.doc();
other_candidate_ord = usize::max_value();
}
SkipResult::End => {
return false;
}
}
continue 'outer;
}
SkipResult::End => {
return false;
}
}
}
debug_assert_eq!(candidate, self.left.doc());
debug_assert_eq!(candidate, self.right.doc());
debug_assert!(self.others.iter().all(|docset| docset.doc() == candidate));
return candidate;
return true;
}
}
fn seek(&mut self, target: DocId) -> DocId {
self.left.seek(target);
let mut docsets: Vec<&mut dyn DocSet> = vec![&mut self.left, &mut self.right];
for docset in &mut self.others {
docsets.push(docset);
fn skip_next(&mut self, target: DocId) -> SkipResult {
// We optimize skipping by skipping every single member
// of the intersection to target.
let mut current_target: DocId = target;
let mut current_ord = self.num_docsets;
'outer: loop {
for ord in 0..self.num_docsets {
let docset = self.docset_mut(ord);
if ord == current_ord {
continue;
}
match docset.skip_next(current_target) {
SkipResult::End => {
return SkipResult::End;
}
SkipResult::OverStep => {
// update the target
// for the remaining members of the intersection.
current_target = docset.doc();
current_ord = ord;
continue 'outer;
}
SkipResult::Reached => {}
}
}
if target == current_target {
return SkipResult::Reached;
} else {
assert!(current_target > target);
return SkipResult::OverStep;
}
}
let doc = go_to_first_doc(&mut docsets[..]);
debug_assert!(docsets.iter().all(|docset| docset.doc() == doc));
debug_assert!(doc >= target);
doc
}
fn doc(&self) -> DocId {
@@ -162,7 +228,7 @@ where
#[cfg(test)]
mod tests {
use super::Intersection;
use crate::docset::{DocSet, TERMINATED};
use crate::docset::{DocSet, SkipResult};
use crate::postings::tests::test_skip_against_unoptimized;
use crate::query::VecDocSet;
@@ -172,18 +238,20 @@ mod tests {
let left = VecDocSet::from(vec![1, 3, 9]);
let right = VecDocSet::from(vec![3, 4, 9, 18]);
let mut intersection = Intersection::new(vec![left, right]);
assert!(intersection.advance());
assert_eq!(intersection.doc(), 3);
assert_eq!(intersection.advance(), 9);
assert!(intersection.advance());
assert_eq!(intersection.doc(), 9);
assert_eq!(intersection.advance(), TERMINATED);
assert!(!intersection.advance());
}
{
let a = VecDocSet::from(vec![1, 3, 9]);
let b = VecDocSet::from(vec![3, 4, 9, 18]);
let c = VecDocSet::from(vec![1, 5, 9, 111]);
let mut intersection = Intersection::new(vec![a, b, c]);
assert!(intersection.advance());
assert_eq!(intersection.doc(), 9);
assert_eq!(intersection.advance(), TERMINATED);
assert!(!intersection.advance());
}
}
@@ -192,8 +260,8 @@ mod tests {
let left = VecDocSet::from(vec![0]);
let right = VecDocSet::from(vec![0]);
let mut intersection = Intersection::new(vec![left, right]);
assert!(intersection.advance());
assert_eq!(intersection.doc(), 0);
assert_eq!(intersection.advance(), TERMINATED);
}
#[test]
@@ -201,7 +269,7 @@ mod tests {
let left = VecDocSet::from(vec![0, 1, 2, 4]);
let right = VecDocSet::from(vec![2, 5]);
let mut intersection = Intersection::new(vec![left, right]);
assert_eq!(intersection.seek(2), 2);
assert_eq!(intersection.skip_next(2), SkipResult::Reached);
assert_eq!(intersection.doc(), 2);
}
@@ -244,7 +312,7 @@ mod tests {
let a = VecDocSet::from(vec![1, 3]);
let b = VecDocSet::from(vec![1, 4]);
let c = VecDocSet::from(vec![3, 9]);
let intersection = Intersection::new(vec![a, b, c]);
assert_eq!(intersection.doc(), TERMINATED);
let mut intersection = Intersection::new(vec![a, b, c]);
assert!(!intersection.advance());
}
}

9
src/query/mod.rs
View File

@@ -1,11 +1,12 @@
/*! Query Module */
/*!
Query
*/
mod all_query;
mod automaton_weight;
mod bitset;
mod bm25;
mod boolean_query;
mod boost_query;
mod empty_query;
mod exclude;
mod explanation;
@@ -26,7 +27,6 @@ mod weight;
mod vec_docset;
pub(crate) mod score_combiner;
pub(crate) use self::bm25::BM25Weight;
pub use self::intersection::Intersection;
pub use self::union::Union;
@@ -37,12 +37,9 @@ pub use self::all_query::{AllQuery, AllScorer, AllWeight};
pub use self::automaton_weight::AutomatonWeight;
pub use self::bitset::BitSetDocSet;
pub use self::boolean_query::BooleanQuery;
pub use self::boost_query::BoostQuery;
pub use self::empty_query::{EmptyQuery, EmptyScorer, EmptyWeight};
pub use self::exclude::Exclude;
pub use self::explanation::Explanation;
#[cfg(test)]
pub(crate) use self::fuzzy_query::DFAWrapper;
pub use self::fuzzy_query::FuzzyTermQuery;
pub use self::intersection::intersect_scorers;
pub use self::phrase_query::PhraseQuery;

88
src/query/phrase_query/mod.rs
View File

@@ -7,24 +7,24 @@ pub use self::phrase_scorer::PhraseScorer;
pub use self::phrase_weight::PhraseWeight;
#[cfg(test)]
pub mod tests {
mod tests {
use super::*;
use crate::assert_nearly_equals;
use crate::collector::tests::{TEST_COLLECTOR_WITHOUT_SCORE, TEST_COLLECTOR_WITH_SCORE};
use crate::core::Index;
use crate::query::Weight;
use crate::error::TantivyError;
use crate::schema::{Schema, Term, TEXT};
use crate::tests::assert_nearly_equals;
use crate::DocId;
use crate::{DocAddress, TERMINATED};
use crate::{DocAddress, DocSet};
pub fn create_index(texts: &[&'static str]) -> Index {
fn create_index(texts: &[&'static str]) -> 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);
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
for &text in texts {
let doc = doc!(text_field=>text);
index_writer.add_document(doc);
@@ -61,30 +61,13 @@ pub mod tests {
.map(|docaddr| docaddr.1)
.collect::<Vec<_>>()
};
assert_eq!(test_query(vec!["a", "b"]), vec![1, 2, 3, 4]);
assert_eq!(test_query(vec!["a", "b", "c"]), vec![2, 4]);
assert_eq!(test_query(vec!["a", "b"]), vec![1, 2, 3, 4]);
assert_eq!(test_query(vec!["b", "b"]), vec![0, 1]);
assert!(test_query(vec!["g", "ewrwer"]).is_empty());
assert!(test_query(vec!["g", "a"]).is_empty());
}
#[test]
pub fn test_phrase_query_simple() -> crate::Result<()> {
let index = create_index(&["a b b d c g c", "a b a b c"]);
let text_field = index.schema().get_field("text").unwrap();
let searcher = index.reader()?.searcher();
let terms: Vec<Term> = vec!["a", "b", "c"]
.iter()
.map(|text| Term::from_field_text(text_field, text))
.collect();
let phrase_query = PhraseQuery::new(terms);
let phrase_weight = phrase_query.phrase_weight(&searcher, false)?;
let mut phrase_scorer = phrase_weight.scorer(searcher.segment_reader(0), 1.0)?;
assert_eq!(phrase_scorer.doc(), 1);
assert_eq!(phrase_scorer.advance(), TERMINATED);
Ok(())
}
#[test]
pub fn test_phrase_query_no_score() {
let index = create_index(&[
@@ -119,6 +102,30 @@ pub mod tests {
assert!(test_query(vec!["g", "a"]).is_empty());
}
#[test]
pub fn test_phrase_count() {
let index = create_index(&["a c", "a a b d a b c", " a b"]);
let schema = index.schema();
let text_field = schema.get_field("text").unwrap();
let searcher = index.reader().unwrap().searcher();
let phrase_query = PhraseQuery::new(vec![
Term::from_field_text(text_field, "a"),
Term::from_field_text(text_field, "b"),
]);
let phrase_weight = phrase_query.phrase_weight(&searcher, true).unwrap();
let mut phrase_scorer = phrase_weight
.phrase_scorer(searcher.segment_reader(0u32))
.unwrap()
.unwrap();
assert!(phrase_scorer.advance());
assert_eq!(phrase_scorer.doc(), 1);
assert_eq!(phrase_scorer.phrase_count(), 2);
assert!(phrase_scorer.advance());
assert_eq!(phrase_scorer.doc(), 2);
assert_eq!(phrase_scorer.phrase_count(), 1);
assert!(!phrase_scorer.advance());
}
#[test]
pub fn test_phrase_query_no_positions() {
let mut schema_builder = Schema::builder();
@@ -135,7 +142,7 @@ pub mod tests {
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"a b c"));
assert!(index_writer.commit().is_ok());
}
@@ -144,16 +151,21 @@ pub mod tests {
Term::from_field_text(text_field, "a"),
Term::from_field_text(text_field, "b"),
]);
let search_result = searcher
match searcher
.search(&phrase_query, &TEST_COLLECTOR_WITH_SCORE)
.map(|_| ());
assert!(matches!(
search_result,
Err(crate::TantivyError::SchemaError(msg))
if msg == "Applied phrase query on field \"text\", which does not have positions \
indexed"
));
.map(|_| ())
.unwrap_err()
{
TantivyError::SchemaError(ref msg) => {
assert_eq!(
"Applied phrase query on field \"text\", which does not have positions indexed",
msg.as_str()
);
}
_ => {
panic!("Should have returned an error");
}
}
}
#[test]
@@ -175,8 +187,8 @@ pub mod tests {
.to_vec()
};
let scores = test_query(vec!["a", "b"]);
assert_nearly_equals!(scores[0], 0.40618482);
assert_nearly_equals!(scores[1], 0.46844664);
assert_nearly_equals(scores[0], 0.40618482);
assert_nearly_equals(scores[1], 0.46844664);
}
#[test] // motivated by #234
@@ -186,7 +198,7 @@ pub mod tests {
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"b"));
index_writer.add_document(doc!(text_field=>"a b"));
index_writer.add_document(doc!(text_field=>"b a"));
@@ -217,7 +229,7 @@ pub mod tests {
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{
let mut index_writer = index.writer_for_tests().unwrap();
let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
index_writer.add_document(doc!(text_field=>"a b c d e f g h"));
assert!(index_writer.commit().is_ok());
}

50
src/query/phrase_query/phrase_scorer.rs
View File

@@ -1,9 +1,9 @@
use crate::docset::{DocSet, TERMINATED};
use crate::docset::{DocSet, SkipResult};
use crate::fieldnorm::FieldNormReader;
use crate::postings::Postings;
use crate::query::bm25::BM25Weight;
use crate::query::{Intersection, Scorer};
use crate::{DocId, Score};
use crate::DocId;
use std::cmp::Ordering;
struct PostingsWithOffset<TPostings> {
@@ -25,12 +25,12 @@ impl<TPostings: Postings> PostingsWithOffset<TPostings> {
}
impl<TPostings: Postings> DocSet for PostingsWithOffset<TPostings> {
fn advance(&mut self) -> DocId {
fn advance(&mut self) -> bool {
self.postings.advance()
}
fn seek(&mut self, target: DocId) -> DocId {
self.postings.seek(target)
fn skip_next(&mut self, target: DocId) -> SkipResult {
self.postings.skip_next(target)
}
fn doc(&self) -> DocId {
@@ -149,7 +149,7 @@ impl<TPostings: Postings> PhraseScorer<TPostings> {
PostingsWithOffset::new(postings, (max_offset - offset) as u32)
})
.collect::<Vec<_>>();
let mut scorer = PhraseScorer {
PhraseScorer {
intersection_docset: Intersection::new(postings_with_offsets),
num_terms: num_docsets,
left: Vec::with_capacity(100),
@@ -158,11 +158,7 @@ impl<TPostings: Postings> PhraseScorer<TPostings> {
similarity_weight,
fieldnorm_reader,
score_needed,
};
if scorer.doc() != TERMINATED && !scorer.phrase_match() {
scorer.advance();
}
scorer
}
pub fn phrase_count(&self) -> u32 {
@@ -229,22 +225,31 @@ impl<TPostings: Postings> PhraseScorer<TPostings> {
}
impl<TPostings: Postings> DocSet for PhraseScorer<TPostings> {
fn advance(&mut self) -> DocId {
loop {
let doc = self.intersection_docset.advance();
if doc == TERMINATED || self.phrase_match() {
return doc;
fn advance(&mut self) -> bool {
while self.intersection_docset.advance() {
if self.phrase_match() {
return true;
}
}
false
}
fn seek(&mut self, target: DocId) -> DocId {
debug_assert!(target >= self.doc());
let doc = self.intersection_docset.seek(target);
if doc == TERMINATED || self.phrase_match() {
return doc;
fn skip_next(&mut self, target: DocId) -> SkipResult {
if self.intersection_docset.skip_next(target) == SkipResult::End {
return SkipResult::End;
}
if self.phrase_match() {
if self.doc() == target {
return SkipResult::Reached;
} else {
return SkipResult::OverStep;
}
}
if self.advance() {
SkipResult::OverStep
} else {
SkipResult::End
}
self.advance()
}
fn doc(&self) -> DocId {
@@ -257,7 +262,7 @@ impl<TPostings: Postings> DocSet for PhraseScorer<TPostings> {
}
impl<TPostings: Postings> Scorer for PhraseScorer<TPostings> {
fn score(&mut self) -> Score {
fn score(&mut self) -> f32 {
let doc = self.doc();
let fieldnorm_id = self.fieldnorm_reader.fieldnorm_id(doc);
self.similarity_weight
@@ -267,6 +272,7 @@ impl<TPostings: Postings> Scorer for PhraseScorer<TPostings> {
#[cfg(test)]
mod tests {
use super::{intersection, intersection_count};
fn test_intersection_sym(left: &[u32], right: &[u32], expected: &[u32]) {

56
src/query/phrase_query/phrase_weight.rs
View File

@@ -9,8 +9,8 @@ use crate::query::Weight;
use crate::query::{EmptyScorer, Explanation};
use crate::schema::IndexRecordOption;
use crate::schema::Term;
use crate::Score;
use crate::{DocId, DocSet};
use crate::{Result, SkipResult};
pub struct PhraseWeight {
phrase_terms: Vec<(usize, Term)>,
@@ -32,18 +32,17 @@ impl PhraseWeight {
}
}
fn fieldnorm_reader(&self, reader: &SegmentReader) -> crate::Result<FieldNormReader> {
fn fieldnorm_reader(&self, reader: &SegmentReader) -> FieldNormReader {
let field = self.phrase_terms[0].1.field();
reader.get_fieldnorms_reader(field)
}
fn phrase_scorer(
pub fn phrase_scorer(
&self,
reader: &SegmentReader,
boost: Score,
) -> crate::Result<Option<PhraseScorer<SegmentPostings>>> {
let similarity_weight = self.similarity_weight.boost_by(boost);
let fieldnorm_reader = self.fieldnorm_reader(reader)?;
) -> Result<Option<PhraseScorer<SegmentPostings>>> {
let similarity_weight = self.similarity_weight.clone();
let fieldnorm_reader = self.fieldnorm_reader(reader);
if reader.has_deletes() {
let mut term_postings_list = Vec::new();
for &(offset, ref term) in &self.phrase_terms {
@@ -85,24 +84,24 @@ impl PhraseWeight {
}
impl Weight for PhraseWeight {
fn scorer(&self, reader: &SegmentReader, boost: Score) -> crate::Result<Box<dyn Scorer>> {
if let Some(scorer) = self.phrase_scorer(reader, boost)? {
fn scorer(&self, reader: &SegmentReader) -> Result<Box<dyn Scorer>> {
if let Some(scorer) = self.phrase_scorer(reader)? {
Ok(Box::new(scorer))
} else {
Ok(Box::new(EmptyScorer))
}
}
fn explain(&self, reader: &SegmentReader, doc: DocId) -> crate::Result<Explanation> {
let scorer_opt = self.phrase_scorer(reader, 1.0)?;
fn explain(&self, reader: &SegmentReader, doc: DocId) -> Result<Explanation> {
let scorer_opt = self.phrase_scorer(reader)?;
if scorer_opt.is_none() {
return Err(does_not_match(doc));
}
let mut scorer = scorer_opt.unwrap();
if scorer.seek(doc) != doc {
if scorer.skip_next(doc) != SkipResult::Reached {
return Err(does_not_match(doc));
}
let fieldnorm_reader = self.fieldnorm_reader(reader)?;
let fieldnorm_reader = self.fieldnorm_reader(reader);
let fieldnorm_id = fieldnorm_reader.fieldnorm_id(doc);
let phrase_count = scorer.phrase_count();
let mut explanation = Explanation::new("Phrase Scorer", scorer.score());
@@ -110,34 +109,3 @@ impl Weight for PhraseWeight {
Ok(explanation)
}
}
#[cfg(test)]
mod tests {
use super::super::tests::create_index;
use crate::docset::TERMINATED;
use crate::query::PhraseQuery;
use crate::{DocSet, Term};
#[test]
pub fn test_phrase_count() {
let index = create_index(&["a c", "a a b d a b c", " a b"]);
let schema = index.schema();
let text_field = schema.get_field("text").unwrap();
let searcher = index.reader().unwrap().searcher();
let phrase_query = PhraseQuery::new(vec![
Term::from_field_text(text_field, "a"),
Term::from_field_text(text_field, "b"),
]);
let phrase_weight = phrase_query.phrase_weight(&searcher, true).unwrap();
let mut phrase_scorer = phrase_weight
.phrase_scorer(searcher.segment_reader(0u32), 1.0)
.unwrap()
.unwrap();
assert_eq!(phrase_scorer.doc(), 1);
assert_eq!(phrase_scorer.phrase_count(), 2);
assert_eq!(phrase_scorer.advance(), 2);
assert_eq!(phrase_scorer.doc(), 2);
assert_eq!(phrase_scorer.phrase_count(), 1);
assert_eq!(phrase_scorer.advance(), TERMINATED);
}
}

2
src/query/query.rs
View File

@@ -40,7 +40,7 @@ use std::fmt;
///
/// When implementing a new type of `Query`, it is normal to implement a
/// dedicated `Query`, `Weight` and `Scorer`.
pub trait Query: QueryClone + Send + Sync + downcast_rs::Downcast + fmt::Debug {
pub trait Query: QueryClone + downcast_rs::Downcast + fmt::Debug {
/// Create the weight associated to a query.
///
/// If scoring is not required, setting `scoring_enabled` to `false`

13
src/query/query_parser/logical_ast.rs
View File

@@ -2,7 +2,6 @@ use crate::query::Occur;
use crate::schema::Field;
use crate::schema::Term;
use crate::schema::Type;
use crate::Score;
use std::fmt;
use std::ops::Bound;
@@ -22,17 +21,6 @@ pub enum LogicalLiteral {
pub enum LogicalAST {
Clause(Vec<(Occur, LogicalAST)>),
Leaf(Box<LogicalLiteral>),
Boost(Box<LogicalAST>, Score),
}
impl LogicalAST {
pub fn boost(self, boost: Score) -> LogicalAST {
if (boost - 1.0).abs() < Score::EPSILON {
self
} else {
LogicalAST::Boost(Box::new(self), boost)
}
}
}
fn occur_letter(occur: Occur) -> &'static str {
@@ -59,7 +47,6 @@ impl fmt::Debug for LogicalAST {
}
Ok(())
}
LogicalAST::Boost(ref ast, boost) => write!(formatter, "{:?}^{}", ast, boost),
LogicalAST::Leaf(ref literal) => write!(formatter, "{:?}", literal),
}
}

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