rust/tantivy
1
0
Fork 0
mirror of https://github.com/quickwit-oss/tantivy.git synced 2025-12-28 13:02:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

Compare commits

base: rust:0.22.0
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:tokenizer_docs
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

1 Commits
0.22.0 ... tokenizer_

Author SHA1 Message Date
Pascal Seitz
806a1e1b1e clarify tokenizer docs 2023-04-03 22:59:38 +08:00
340 changed files with 9348 additions and 26898 deletions
Expand all files Collapse all files

13
.github/workflows/coverage.yml vendored
View File

@@ -3,23 +3,20 @@ name: Coverage
on:
push:
branches: [main]
# Ensures that we cancel running jobs for the same PR / same workflow.
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
pull_request:
branches: [main]
jobs:
coverage:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Install Rust
run: rustup toolchain install nightly-2023-09-10 --profile minimal --component llvm-tools-preview
run: rustup toolchain install nightly --profile minimal --component llvm-tools-preview
- uses: Swatinem/rust-cache@v2
- uses: taiki-e/install-action@cargo-llvm-cov
- name: Generate code coverage
run: cargo +nightly-2023-09-10 llvm-cov --all-features --workspace --doctests --lcov --output-path lcov.info
run: cargo +nightly llvm-cov --all-features --workspace --doctests --lcov --output-path lcov.info
- name: Upload coverage to Codecov
uses: codecov/codecov-action@v3
continue-on-error: true

7
.github/workflows/long_running.yml vendored
View File

@@ -8,18 +8,13 @@ env:
CARGO_TERM_COLOR: always
NUM_FUNCTIONAL_TEST_ITERATIONS: 20000
# Ensures that we cancel running jobs for the same PR / same workflow.
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
jobs:
test:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Install stable
uses: actions-rs/toolchain@v1
with:

18
.github/workflows/test.yml vendored
View File

@@ -9,18 +9,13 @@ on:
env:
CARGO_TERM_COLOR: always
# Ensures that we cancel running jobs for the same PR / same workflow.
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
jobs:
check:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Install nightly
uses: actions-rs/toolchain@v1
@@ -39,13 +34,6 @@ jobs:
- name: Check Formatting
run: cargo +nightly fmt --all -- --check
- name: Check Stable Compilation
run: cargo build --all-features
- name: Check Bench Compilation
run: cargo +nightly bench --no-run --profile=dev --all-features
- uses: actions-rs/clippy-check@v1
with:
@@ -60,14 +48,14 @@ jobs:
strategy:
matrix:
features: [
{ label: "all", flags: "mmap,stopwords,lz4-compression,zstd-compression,failpoints" },
{ label: "all", flags: "mmap,stopwords,brotli-compression,lz4-compression,snappy-compression,zstd-compression,failpoints" },
{ label: "quickwit", flags: "mmap,quickwit,failpoints" }
]
name: test-${{ matrix.features.label}}
steps:
- uses: actions/checkout@v4
- uses: actions/checkout@v3
- name: Install stable
uses: actions-rs/toolchain@v1

2
ARCHITECTURE.md
View File

@@ -254,7 +254,7 @@ The token positions of all of the terms are then stored in a separate file with
The [TermInfo](src/postings/term_info.rs) gives an offset (expressed in position this time) in this file. As we iterate through the docset,
we advance the position reader by the number of term frequencies of the current document.
## [fieldnorm/](src/fieldnorm): Here is my doc, how many tokens in this field?
## [fieldnorms/](src/fieldnorms): Here is my doc, how many tokens in this field?
The [BM25](https://en.wikipedia.org/wiki/Okapi_BM25) formula also requires to know the number of tokens stored in a specific field for a given document. We store this information on one byte per document in the fieldnorm.
The fieldnorm is therefore compressed. Values up to 40 are encoded unchanged.

184
CHANGELOG.md
View File

@@ -1,187 +1,3 @@
Tantivy 0.22
================================
Tantivy 0.22 will be able to read indices created with Tantivy 0.21.
#### Bugfixes
- Fix null byte handling in JSON paths (null bytes in json keys caused panic during indexing) [#2345](https://github.com/quickwit-oss/tantivy/pull/2345)(@PSeitz)
- Fix bug that can cause `get_docids_for_value_range` to panic. [#2295](https://github.com/quickwit-oss/tantivy/pull/2295)(@fulmicoton)
- Avoid 1 document indices by increase min memory to 15MB for indexing [#2176](https://github.com/quickwit-oss/tantivy/pull/2176)(@PSeitz)
- Fix merge panic for JSON fields [#2284](https://github.com/quickwit-oss/tantivy/pull/2284)(@PSeitz)
- Fix bug occuring when merging JSON object indexed with positions. [#2253](https://github.com/quickwit-oss/tantivy/pull/2253)(@fulmicoton)
- Fix empty DateHistogram gap bug [#2183](https://github.com/quickwit-oss/tantivy/pull/2183)(@PSeitz)
- Fix range query end check (fields with less than 1 value per doc are affected) [#2226](https://github.com/quickwit-oss/tantivy/pull/2226)(@PSeitz)
- Handle exclusive out of bounds ranges on fastfield range queries [#2174](https://github.com/quickwit-oss/tantivy/pull/2174)(@PSeitz)
#### Breaking API Changes
- rename ReloadPolicy onCommit to onCommitWithDelay [#2235](https://github.com/quickwit-oss/tantivy/pull/2235)(@giovannicuccu)
- Move exports from the root into modules [#2220](https://github.com/quickwit-oss/tantivy/pull/2220)(@PSeitz)
- Accept field name instead of `Field` in FilterCollector [#2196](https://github.com/quickwit-oss/tantivy/pull/2196)(@PSeitz)
- remove deprecated IntOptions and DateTime [#2353](https://github.com/quickwit-oss/tantivy/pull/2353)(@PSeitz)
#### Features/Improvements
- Tantivy documents as a trait: Index data directly without converting to tantivy types first [#2071](https://github.com/quickwit-oss/tantivy/pull/2071)(@ChillFish8)
- encode some part of posting list as -1 instead of direct values (smaller inverted indices) [#2185](https://github.com/quickwit-oss/tantivy/pull/2185)(@trinity-1686a)
- **Aggregation**
- Support to deserialize f64 from string [#2311](https://github.com/quickwit-oss/tantivy/pull/2311)(@PSeitz)
- Add a top_hits aggregator [#2198](https://github.com/quickwit-oss/tantivy/pull/2198)(@ditsuke)
- Support bool type in term aggregation [#2318](https://github.com/quickwit-oss/tantivy/pull/2318)(@PSeitz)
- Support ip adresses in term aggregation [#2319](https://github.com/quickwit-oss/tantivy/pull/2319)(@PSeitz)
- Support date type in term aggregation [#2172](https://github.com/quickwit-oss/tantivy/pull/2172)(@PSeitz)
- Support escaped dot when addressing field [#2250](https://github.com/quickwit-oss/tantivy/pull/2250)(@PSeitz)
- Add ExistsQuery to check documents that have a value [#2160](https://github.com/quickwit-oss/tantivy/pull/2160)(@imotov)
- Expose TopDocs::order_by_u64_field again [#2282](https://github.com/quickwit-oss/tantivy/pull/2282)(@ditsuke)
- **Memory/Performance**
- Faster TopN: replace BinaryHeap with TopNComputer [#2186](https://github.com/quickwit-oss/tantivy/pull/2186)(@PSeitz)
- reduce number of allocations during indexing [#2257](https://github.com/quickwit-oss/tantivy/pull/2257)(@PSeitz)
- Less Memory while indexing: docid deltas while indexing [#2249](https://github.com/quickwit-oss/tantivy/pull/2249)(@PSeitz)
- Faster indexing: use term hashmap in fastfield [#2243](https://github.com/quickwit-oss/tantivy/pull/2243)(@PSeitz)
- term hashmap remove copy in is_empty, unused unordered_id [#2229](https://github.com/quickwit-oss/tantivy/pull/2229)(@PSeitz)
- add method to fetch block of first values in columnar [#2330](https://github.com/quickwit-oss/tantivy/pull/2330)(@PSeitz)
- Faster aggregations: add fast path for full columns in fetch_block [#2328](https://github.com/quickwit-oss/tantivy/pull/2328)(@PSeitz)
- Faster sstable loading: use fst for sstable index [#2268](https://github.com/quickwit-oss/tantivy/pull/2268)(@trinity-1686a)
- **QueryParser**
- allow newline where we allow space in query parser [#2302](https://github.com/quickwit-oss/tantivy/pull/2302)(@trinity-1686a)
- allow some mixing of occur and bool in strict query parser [#2323](https://github.com/quickwit-oss/tantivy/pull/2323)(@trinity-1686a)
- handle * inside term in lenient query parser [#2228](https://github.com/quickwit-oss/tantivy/pull/2228)(@trinity-1686a)
- add support for exists query syntax in query parser [#2170](https://github.com/quickwit-oss/tantivy/pull/2170)(@trinity-1686a)
- Add shared search executor [#2312](https://github.com/quickwit-oss/tantivy/pull/2312)(@MochiXu)
- Truncate keys to u16::MAX in term hashmap [#2299](https://github.com/quickwit-oss/tantivy/pull/2299)(@PSeitz)
- report if a term matched when warming up posting list [#2309](https://github.com/quickwit-oss/tantivy/pull/2309)(@trinity-1686a)
- Support json fields in FuzzyTermQuery [#2173](https://github.com/quickwit-oss/tantivy/pull/2173)(@PingXia-at)
- Read list of fields encoded in term dictionary for JSON fields [#2184](https://github.com/quickwit-oss/tantivy/pull/2184)(@PSeitz)
- add collect_block to BoxableSegmentCollector [#2331](https://github.com/quickwit-oss/tantivy/pull/2331)(@PSeitz)
- expose collect_block buffer size [#2326](https://github.com/quickwit-oss/tantivy/pull/2326)(@PSeitz)
- Forward regex parser errors [#2288](https://github.com/quickwit-oss/tantivy/pull/2288)(@adamreichold)
- Make FacetCounts defaultable and cloneable. [#2322](https://github.com/quickwit-oss/tantivy/pull/2322)(@adamreichold)
- Derive Debug for SchemaBuilder [#2254](https://github.com/quickwit-oss/tantivy/pull/2254)(@GodTamIt)
- add missing inlines to tantivy options [#2245](https://github.com/quickwit-oss/tantivy/pull/2245)(@PSeitz)
Tantivy 0.21.1
================================
#### Bugfixes
- Range queries on fast fields with less values on that field than documents had an invalid end condition, leading to missing results. [#2226](https://github.com/quickwit-oss/tantivy/issues/2226)(@appaquet @PSeitz)
- Increase the minimum memory budget from 3MB to 15MB to avoid single doc segments (API fix). [#2176](https://github.com/quickwit-oss/tantivy/issues/2176)(@PSeitz)
Tantivy 0.21
================================
#### Bugfixes
- Fix track fast field memory consumption, which led to higher memory consumption than the budget allowed during indexing [#2148](https://github.com/quickwit-oss/tantivy/issues/2148)[#2147](https://github.com/quickwit-oss/tantivy/issues/2147)(@PSeitz)
- Fix a regression from 0.20 where sort index by date wasn't working anymore [#2124](https://github.com/quickwit-oss/tantivy/issues/2124)(@PSeitz)
- Fix getting the root facet on the `FacetCollector`. [#2086](https://github.com/quickwit-oss/tantivy/issues/2086)(@adamreichold)
- Align numerical type priority order of columnar and query. [#2088](https://github.com/quickwit-oss/tantivy/issues/2088)(@fmassot)
#### Breaking Changes
- Remove support for Brotli and Snappy compression [#2123](https://github.com/quickwit-oss/tantivy/issues/2123)(@adamreichold)
#### Features/Improvements
- Implement lenient query parser [#2129](https://github.com/quickwit-oss/tantivy/pull/2129)(@trinity-1686a)
- order_by_u64_field and order_by_fast_field allow sorting in ascending and descending order [#2111](https://github.com/quickwit-oss/tantivy/issues/2111)(@naveenann)
- Allow dynamic filters in text analyzer builder [#2110](https://github.com/quickwit-oss/tantivy/issues/2110)(@fulmicoton @fmassot)
- **Aggregation**
- Add missing parameter for term aggregation [#2149](https://github.com/quickwit-oss/tantivy/issues/2149)[#2103](https://github.com/quickwit-oss/tantivy/issues/2103)(@PSeitz)
- Add missing parameter for percentiles [#2157](https://github.com/quickwit-oss/tantivy/issues/2157)(@PSeitz)
- Add missing parameter for stats,min,max,count,sum,avg [#2151](https://github.com/quickwit-oss/tantivy/issues/2151)(@PSeitz)
- Improve aggregation deserialization error message [#2150](https://github.com/quickwit-oss/tantivy/issues/2150)(@PSeitz)
- Add validation for type Bytes to term_agg [#2077](https://github.com/quickwit-oss/tantivy/issues/2077)(@PSeitz)
- Alternative mixed field collection [#2135](https://github.com/quickwit-oss/tantivy/issues/2135)(@PSeitz)
- Add missing query_terms impl for TermSetQuery. [#2120](https://github.com/quickwit-oss/tantivy/issues/2120)(@adamreichold)
- Minor improvements to OwnedBytes [#2134](https://github.com/quickwit-oss/tantivy/issues/2134)(@adamreichold)
- Remove allocations in split compound words [#2080](https://github.com/quickwit-oss/tantivy/issues/2080)(@PSeitz)
- Ngram tokenizer now returns an error with invalid arguments [#2102](https://github.com/quickwit-oss/tantivy/issues/2102)(@fmassot)
- Make TextAnalyzerBuilder public [#2097](https://github.com/quickwit-oss/tantivy/issues/2097)(@adamreichold)
- Return an error when tokenizer is not found while indexing [#2093](https://github.com/quickwit-oss/tantivy/issues/2093)(@naveenann)
- Delayed column opening during merge [#2132](https://github.com/quickwit-oss/tantivy/issues/2132)(@PSeitz)
Tantivy 0.20.2
================================
- Align numerical type priority order on the search side. [#2088](https://github.com/quickwit-oss/tantivy/issues/2088) (@fmassot)
- Fix is_child_of function not considering the root facet. [#2086](https://github.com/quickwit-oss/tantivy/issues/2086) (@adamreichhold)
Tantivy 0.20.1
================================
- Fix building on windows with mmap [#2070](https://github.com/quickwit-oss/tantivy/issues/2070) (@ChillFish8)
Tantivy 0.20
================================
#### Bugfixes
- Fix phrase queries with slop (slop supports now transpositions, algorithm that carries slop so far for num terms > 2) [#2031](https://github.com/quickwit-oss/tantivy/issues/2031)[#2020](https://github.com/quickwit-oss/tantivy/issues/2020)(@PSeitz)
- Handle error for exists on MMapDirectory [#1988](https://github.com/quickwit-oss/tantivy/issues/1988) (@PSeitz)
- Aggregation
- Fix min doc_count empty merge bug [#2057](https://github.com/quickwit-oss/tantivy/issues/2057) (@PSeitz)
- Fix: Sort order for term aggregations (sort order on key was inverted) [#1858](https://github.com/quickwit-oss/tantivy/issues/1858) (@PSeitz)
#### Features/Improvements
- Add PhrasePrefixQuery [#1842](https://github.com/quickwit-oss/tantivy/issues/1842) (@trinity-1686a)
- Add `coerce` option for text and numbers types (convert the value instead of returning an error during indexing) [#1904](https://github.com/quickwit-oss/tantivy/issues/1904) (@PSeitz)
- Add regex tokenizer [#1759](https://github.com/quickwit-oss/tantivy/issues/1759)(@mkleen)
- Move tokenizer API to seperate crate. Having a seperate crate with a stable API will allow us to use tokenizers with different tantivy versions. [#1767](https://github.com/quickwit-oss/tantivy/issues/1767) (@PSeitz)
- **Columnar crate**: New fast field handling (@fulmicoton @PSeitz) [#1806](https://github.com/quickwit-oss/tantivy/issues/1806)[#1809](https://github.com/quickwit-oss/tantivy/issues/1809)
- Support for fast fields with optional values. Previously tantivy supported only single-valued and multi-value fast fields. The encoding of optional fast fields is now very compact.
- Fast field Support for JSON (schemaless fast fields). Support multiple types on the same column. [#1876](https://github.com/quickwit-oss/tantivy/issues/1876) (@fulmicoton)
- Unified access for fast fields over different cardinalities.
- Unified storage for typed and untyped fields.
- Move fastfield codecs into columnar. [#1782](https://github.com/quickwit-oss/tantivy/issues/1782) (@fulmicoton)
- Sparse dense index for optional values [#1716](https://github.com/quickwit-oss/tantivy/issues/1716) (@PSeitz)
- Switch to nanosecond precision in DateTime fastfield [#2016](https://github.com/quickwit-oss/tantivy/issues/2016) (@PSeitz)
- **Aggregation**
- Add `date_histogram` aggregation (only `fixed_interval` for now) [#1900](https://github.com/quickwit-oss/tantivy/issues/1900) (@PSeitz)
- Add `percentiles` aggregations [#1984](https://github.com/quickwit-oss/tantivy/issues/1984) (@PSeitz)
- [**breaking**] Drop JSON support on intermediate agg result (we use postcard as format in `quickwit` to send intermediate results) [#1992](https://github.com/quickwit-oss/tantivy/issues/1992) (@PSeitz)
- Set memory limit in bytes for aggregations after which they abort (Previously there was only the bucket limit) [#1942](https://github.com/quickwit-oss/tantivy/issues/1942)[#1957](https://github.com/quickwit-oss/tantivy/issues/1957)(@PSeitz)
- Add support for u64,i64,f64 fields in term aggregation [#1883](https://github.com/quickwit-oss/tantivy/issues/1883) (@PSeitz)
- Allow histogram bounds to be passed as Rfc3339 [#2076](https://github.com/quickwit-oss/tantivy/issues/2076) (@PSeitz)
- Add count, min, max, and sum aggregations [#1794](https://github.com/quickwit-oss/tantivy/issues/1794) (@guilload)
- Switch to Aggregation without serde_untagged => better deserialization errors. [#2003](https://github.com/quickwit-oss/tantivy/issues/2003) (@PSeitz)
- Switch to ms in histogram for date type (ES compatibility) [#2045](https://github.com/quickwit-oss/tantivy/issues/2045) (@PSeitz)
- Reduce term aggregation memory consumption [#2013](https://github.com/quickwit-oss/tantivy/issues/2013) (@PSeitz)
- Reduce agg memory consumption: Replace generic aggregation collector (which has a high memory requirement per instance) in aggregation tree with optimized versions behind a trait.
- Split term collection count and sub_agg (Faster term agg with less memory consumption for cases without sub-aggs) [#1921](https://github.com/quickwit-oss/tantivy/issues/1921) (@PSeitz)
- Schemaless aggregations: In combination with stacker tantivy supports now schemaless aggregations via the JSON type.
- Add aggregation support for JSON type [#1888](https://github.com/quickwit-oss/tantivy/issues/1888) (@PSeitz)
- Mixed types support on JSON fields in aggs [#1971](https://github.com/quickwit-oss/tantivy/issues/1971) (@PSeitz)
- Perf: Fetch blocks of vals in aggregation for all cardinality [#1950](https://github.com/quickwit-oss/tantivy/issues/1950) (@PSeitz)
- Allow histogram bounds to be passed as Rfc3339 [#2076](https://github.com/quickwit-oss/tantivy/issues/2076) (@PSeitz)
- `Searcher` with disabled scoring via `EnableScoring::Disabled` [#1780](https://github.com/quickwit-oss/tantivy/issues/1780) (@shikhar)
- Enable tokenizer on json fields [#2053](https://github.com/quickwit-oss/tantivy/issues/2053) (@PSeitz)
- Enforcing "NOT" and "-" queries consistency in UserInputAst [#1609](https://github.com/quickwit-oss/tantivy/issues/1609) (@bazhenov)
- Faster indexing
- Refactor tokenization pipeline to use GATs [#1924](https://github.com/quickwit-oss/tantivy/issues/1924) (@trinity-1686a)
- Faster term hash map [#2058](https://github.com/quickwit-oss/tantivy/issues/2058)[#1940](https://github.com/quickwit-oss/tantivy/issues/1940) (@PSeitz)
- tokenizer-api: reduce Tokenizer allocation overhead [#2062](https://github.com/quickwit-oss/tantivy/issues/2062) (@PSeitz)
- Refactor vint [#2010](https://github.com/quickwit-oss/tantivy/issues/2010) (@PSeitz)
- Faster search
- Work in batches of docs on the SegmentCollector (Only for cases without score for now) [#1937](https://github.com/quickwit-oss/tantivy/issues/1937) (@PSeitz)
- Faster fast field range queries using SIMD [#1954](https://github.com/quickwit-oss/tantivy/issues/1954) (@fulmicoton)
- Improve fast field range query performance [#1864](https://github.com/quickwit-oss/tantivy/issues/1864) (@PSeitz)
- Make BM25 scoring more flexible [#1855](https://github.com/quickwit-oss/tantivy/issues/1855) (@alexcole)
- Switch fs2 to fs4 as it is now unmaintained and does not support illumos [#1944](https://github.com/quickwit-oss/tantivy/issues/1944) (@Toasterson)
- Made BooleanWeight and BoostWeight public [#1991](https://github.com/quickwit-oss/tantivy/issues/1991) (@fulmicoton)
- Make index compatible with virtual drives on Windows [#1843](https://github.com/quickwit-oss/tantivy/issues/1843) (@gyk)
- Add stop words for Hungarian language [#2069](https://github.com/quickwit-oss/tantivy/issues/2069) (@tnxbutno)
- Auto downgrade index record option, instead of vint error [#1857](https://github.com/quickwit-oss/tantivy/issues/1857) (@PSeitz)
- Enable range query on fast field for u64 compatible types [#1762](https://github.com/quickwit-oss/tantivy/issues/1762) (@PSeitz) [#1876]
- sstable
- Isolating sstable and stacker in independant crates. [#1718](https://github.com/quickwit-oss/tantivy/issues/1718) (@fulmicoton)
- New sstable format [#1943](https://github.com/quickwit-oss/tantivy/issues/1943)[#1953](https://github.com/quickwit-oss/tantivy/issues/1953) (@trinity-1686a)
- Use DeltaReader directly to implement Dictionnary::ord_to_term [#1928](https://github.com/quickwit-oss/tantivy/issues/1928) (@trinity-1686a)
- Use DeltaReader directly to implement Dictionnary::term_ord [#1925](https://github.com/quickwit-oss/tantivy/issues/1925) (@trinity-1686a)
- Add seperate tokenizer manager for fast fields [#2019](https://github.com/quickwit-oss/tantivy/issues/2019) (@PSeitz)
- Make construction of LevenshteinAutomatonBuilder for FuzzyTermQuery instances lazy. [#1756](https://github.com/quickwit-oss/tantivy/issues/1756) (@adamreichold)
- Added support for madvise when opening an mmaped Index [#2036](https://github.com/quickwit-oss/tantivy/issues/2036) (@fulmicoton)
- Rename `DatePrecision` to `DateTimePrecision` [#2051](https://github.com/quickwit-oss/tantivy/issues/2051) (@guilload)
- Query Parser
- Quotation mark can now be used for phrase queries. [#2050](https://github.com/quickwit-oss/tantivy/issues/2050) (@fulmicoton)
- PhrasePrefixQuery is supported in the query parser via: `field:"phrase ter"*` [#2044](https://github.com/quickwit-oss/tantivy/issues/2044) (@adamreichold)
- Docs
- Update examples for literate docs [#1880](https://github.com/quickwit-oss/tantivy/issues/1880) (@PSeitz)
- Add ip field example [#1775](https://github.com/quickwit-oss/tantivy/issues/1775) (@PSeitz)
- Fix doc store cache documentation [#1821](https://github.com/quickwit-oss/tantivy/issues/1821) (@PSeitz)
- Fix BooleanQuery document [#1999](https://github.com/quickwit-oss/tantivy/issues/1999) (@RT_Enzyme)
- Update comments in the faceted search example [#1737](https://github.com/quickwit-oss/tantivy/issues/1737) (@DawChihLiou)
Tantivy 0.19
================================
#### Bugfixes

81
Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy"
version = "0.22.0"
version = "0.19.0"
authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT"
categories = ["database-implementations", "data-structures"]
@@ -11,57 +11,57 @@ repository = "https://github.com/quickwit-oss/tantivy"
readme = "README.md"
keywords = ["search", "information", "retrieval"]
edition = "2021"
rust-version = "1.63"
exclude = ["benches/*.json", "benches/*.txt"]
rust-version = "1.62"
[dependencies]
oneshot = "0.1.5"
base64 = "0.22.0"
base64 = "0.21.0"
byteorder = "1.4.3"
crc32fast = "1.3.2"
once_cell = "1.10.0"
regex = { version = "1.5.5", default-features = false, features = ["std", "unicode"] }
aho-corasick = "1.0"
tantivy-fst = "0.5"
memmap2 = { version = "0.9.0", optional = true }
lz4_flex = { version = "0.11", default-features = false, optional = true }
zstd = { version = "0.13", optional = true, default-features = false }
aho-corasick = "0.7"
tantivy-fst = "0.4.0"
memmap2 = { version = "0.5.3", optional = true }
lz4_flex = { version = "0.10", default-features = false, features = ["checked-decode"], optional = true }
brotli = { version = "3.3.4", optional = true }
zstd = { version = "0.12", optional = true, default-features = false }
snap = { version = "1.0.5", optional = true }
tempfile = { version = "3.3.0", optional = true }
log = "0.4.16"
serde = { version = "1.0.136", features = ["derive"] }
serde_json = "1.0.79"
num_cpus = "1.13.1"
fs4 = { version = "0.8.0", optional = true }
fs4 = { version = "0.6.3", optional = true }
levenshtein_automata = "0.2.1"
uuid = { version = "1.0.0", features = ["v4", "serde"] }
crossbeam-channel = "0.5.4"
rust-stemmers = "1.2.0"
downcast-rs = "1.2.0"
bitpacking = { version = "0.9.2", default-features = false, features = ["bitpacker4x"] }
census = "0.4.2"
bitpacking = { version = "0.8.4", default-features = false, features = ["bitpacker4x"] }
census = "0.4.0"
rustc-hash = "1.1.0"
thiserror = "1.0.30"
htmlescape = "0.3.1"
fail = { version = "0.5.0", optional = true }
fail = "0.5.0"
murmurhash32 = "0.3.0"
time = { version = "0.3.10", features = ["serde-well-known"] }
smallvec = "1.8.0"
rayon = "1.5.2"
lru = "0.12.0"
lru = "0.10.0"
fastdivide = "0.4.0"
itertools = "0.12.0"
itertools = "0.10.3"
measure_time = "0.8.2"
async-trait = "0.1.53"
arc-swap = "1.5.0"
columnar = { version= "0.3", path="./columnar", package ="tantivy-columnar" }
sstable = { version= "0.3", path="./sstable", package ="tantivy-sstable", optional = true }
stacker = { version= "0.3", path="./stacker", package ="tantivy-stacker" }
query-grammar = { version= "0.22.0", path="./query-grammar", package = "tantivy-query-grammar" }
tantivy-bitpacker = { version= "0.6", path="./bitpacker" }
common = { version= "0.7", path = "./common/", package = "tantivy-common" }
tokenizer-api = { version= "0.3", path="./tokenizer-api", package="tantivy-tokenizer-api" }
sketches-ddsketch = { version = "0.2.1", features = ["use_serde"] }
futures-util = { version = "0.3.28", optional = true }
fnv = "1.0.7"
columnar = { version="0.1", path="./columnar", package ="tantivy-columnar" }
sstable = { version="0.1", path="./sstable", package ="tantivy-sstable", optional = true }
stacker = { version="0.1", path="./stacker", package ="tantivy-stacker" }
query-grammar = { version= "0.19.0", path="./query-grammar", package = "tantivy-query-grammar" }
tantivy-bitpacker = { version= "0.3", path="./bitpacker" }
common = { version= "0.5", path = "./common/", package = "tantivy-common" }
tokenizer-api = { version="0.1", path="./tokenizer-api", package="tantivy-tokenizer-api" }
[target.'cfg(windows)'.dependencies]
winapi = "0.3.9"
@@ -72,18 +72,12 @@ maplit = "1.0.2"
matches = "0.1.9"
pretty_assertions = "1.2.1"
proptest = "1.0.0"
criterion = "0.4"
test-log = "0.2.10"
env_logger = "0.10.0"
pprof = { version = "0.11.0", features = ["flamegraph", "criterion"] }
futures = "0.3.21"
paste = "1.0.11"
more-asserts = "0.3.1"
rand_distr = "0.4.3"
time = { version = "0.3.10", features = ["serde-well-known", "macros"] }
postcard = { version = "1.0.4", features = [
"use-std",
], default-features = false }
[target.'cfg(not(windows))'.dev-dependencies]
criterion = { version = "0.5", default-features = false }
[dev-dependencies.fail]
version = "0.5.0"
@@ -94,11 +88,6 @@ opt-level = 3
debug = false
debug-assertions = false
[profile.bench]
opt-level = 3
debug = true
debug-assertions = false
[profile.test]
debug-assertions = true
overflow-checks = true
@@ -108,18 +97,15 @@ default = ["mmap", "stopwords", "lz4-compression"]
mmap = ["fs4", "tempfile", "memmap2"]
stopwords = []
brotli-compression = ["brotli"]
lz4-compression = ["lz4_flex"]
snappy-compression = ["snap"]
zstd-compression = ["zstd"]
failpoints = ["fail", "fail/failpoints"]
failpoints = ["fail/failpoints"]
unstable = [] # useful for benches.
quickwit = ["sstable", "futures-util"]
# Compares only the hash of a string when indexing data.
# Increases indexing speed, but may lead to extremely rare missing terms, when there's a hash collision.
# Uses 64bit ahash.
compare_hash_only = ["stacker/compare_hash_only"]
quickwit = ["sstable"]
[workspace]
members = ["query-grammar", "bitpacker", "common", "ownedbytes", "stacker", "sstable", "tokenizer-api", "columnar"]
@@ -134,7 +120,7 @@ members = ["query-grammar", "bitpacker", "common", "ownedbytes", "stacker", "sst
[[test]]
name = "failpoints"
path = "tests/failpoints/mod.rs"
required-features = ["failpoints"]
required-features = ["fail/failpoints"]
[[bench]]
name = "analyzer"
@@ -143,3 +129,4 @@ harness = false
[[bench]]
name = "index-bench"
harness = false

2
Makefile
View File

@@ -1,5 +1,5 @@
test:
@echo "Run test only... No examples."
echo "Run test only... No examples."
cargo test --tests --lib
fmt:

37
README.md
View File

@@ -5,18 +5,19 @@
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
[![Crates.io](https://img.shields.io/crates/v/tantivy.svg)](https://crates.io/crates/tantivy)
<img src="https://tantivy-search.github.io/logo/tantivy-logo.png" alt="Tantivy, the fastest full-text search engine library written in Rust" height="250">
![Tantivy](https://tantivy-search.github.io/logo/tantivy-logo.png)
## Fast full-text search engine library written in Rust
**Tantivy** is a **full-text search engine library** written in Rust.
**If you are looking for an alternative to Elasticsearch or Apache Solr, check out [Quickwit](https://github.com/quickwit-oss/quickwit), our distributed search engine built on top of Tantivy.**
Tantivy is closer to [Apache Lucene](https://lucene.apache.org/) than to [Elasticsearch](https://www.elastic.co/products/elasticsearch) or [Apache Solr](https://lucene.apache.org/solr/) in the sense it is not
an off-the-shelf search engine server, but rather a crate that can be used to build such a search engine.
It is closer to [Apache Lucene](https://lucene.apache.org/) than to [Elasticsearch](https://www.elastic.co/products/elasticsearch) or [Apache Solr](https://lucene.apache.org/solr/) in the sense it is not
an off-the-shelf search engine server, but rather a crate that can be used
to build such a search engine.
Tantivy is, in fact, strongly inspired by Lucene's design.
## Benchmark
If you are looking for an alternative to Elasticsearch or Apache Solr, check out [Quickwit](https://github.com/quickwit-oss/quickwit), our search engine built on top of Tantivy.
# Benchmark
The following [benchmark](https://tantivy-search.github.io/bench/) breakdowns
performance for different types of queries/collections.
@@ -25,9 +26,7 @@ Your mileage WILL vary depending on the nature of queries and their load.
<img src="doc/assets/images/searchbenchmark.png">
Details about the benchmark can be found at this [repository](https://github.com/quickwit-oss/search-benchmark-game).
## Features
# 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), [Vaporetto](https://crates.io/crates/vaporetto_tantivy), and [tantivy-tokenizer-tiny-segmenter](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))
@@ -43,7 +42,7 @@ Details about the benchmark can be found at this [repository](https://github.com
- Single valued and multivalued u64, i64, and f64 fast fields (equivalent of doc values in Lucene)
- `&[u8]` fast fields
- Text, i64, u64, f64, dates, ip, bool, and hierarchical facet fields
- Compressed document store (LZ4, Zstd, None)
- Compressed document store (LZ4, Zstd, None, Brotli, Snap)
- Range queries
- Faceted search
- Configurable indexing (optional term frequency and position indexing)
@@ -53,11 +52,11 @@ Details about the benchmark can be found at this [repository](https://github.com
- Searcher Warmer API
- Cheesy logo with a horse
### Non-features
## Non-features
Distributed search is out of the scope of Tantivy, but if you are looking for this feature, check out [Quickwit](https://github.com/quickwit-oss/quickwit/).
## Getting started
# Getting started
Tantivy works on stable Rust and supports Linux, macOS, and Windows.
@@ -67,7 +66,7 @@ index documents, and search via the CLI or a small server with a REST API.
It walks you through getting a Wikipedia search engine up and running in a few minutes.
- [Reference doc for the last released version](https://docs.rs/tantivy/)
## How can I support this project?
# How can I support this project?
There are many ways to support this project.
@@ -78,16 +77,16 @@ There are many ways to support this project.
- Contribute code (you can join [our Discord server](https://discord.gg/MT27AG5EVE))
- Talk about Tantivy around you
## Contributing code
# Contributing code
We use the GitHub Pull Request workflow: reference a GitHub ticket and/or include a comprehensive commit message when opening a PR.
Feel free to update CHANGELOG.md with your contribution.
### Tokenizer
## Tokenizer
When implementing a tokenizer for tantivy depend on the `tantivy-tokenizer-api` crate.
### Clone and build locally
## Clone and build locally
Tantivy compiles on stable Rust.
To check out and run tests, you can simply run:
@@ -98,7 +97,7 @@ cd tantivy
cargo test
```
## Companies Using Tantivy
# Companies Using Tantivy
<p align="left">
<img align="center" src="doc/assets/images/etsy.png" alt="Etsy" height="25" width="auto" />&nbsp;
@@ -110,7 +109,7 @@ cargo test
<img align="center" src="doc/assets/images/element-dark-theme.png#gh-dark-mode-only" alt="Element.io" height="25" width="auto" />
</p>
## FAQ
# FAQ
### Can I use Tantivy in other languages?

21
RELEASE.md
View File

@@ -1,21 +0,0 @@
# Release a new Tantivy Version
## Steps
1. Identify new packages in workspace since last release
2. Identify changed packages in workspace since last release
3. Bump version in `Cargo.toml` and their dependents for all changed packages
4. Update version of root `Cargo.toml`
5. Publish version starting with leaf nodes
6. Set git tag with new version
In conjucation with `cargo-release` Steps 1-4 (I'm not sure if the change detection works):
Set new packages to version 0.0.0
Replace prev-tag-name
```bash
cargo release --workspace --no-publish -v --prev-tag-name 0.19 --push-remote origin minor --no-tag --execute
```
no-tag or it will create tags for all the subpackages

23
appveyor.yml Normal file
View File

@@ -0,0 +1,23 @@
# Appveyor configuration template for Rust using rustup for Rust installation
# https://github.com/starkat99/appveyor-rust
os: Visual Studio 2015
environment:
matrix:
- channel: stable
target: x86_64-pc-windows-msvc
install:
- appveyor DownloadFile https://win.rustup.rs/ -FileName rustup-init.exe
- rustup-init -yv --default-toolchain %channel% --default-host %target%
- set PATH=%PATH%;%USERPROFILE%\.cargo\bin
- if defined msys_bits set PATH=%PATH%;C:\msys64\mingw%msys_bits%\bin
- rustc -vV
- cargo -vV
build: false
test_script:
- REM SET RUST_LOG=tantivy,test & cargo test --all --verbose --no-default-features --features lz4-compression --features mmap
- REM SET RUST_LOG=tantivy,test & cargo test test_store --verbose --no-default-features --features lz4-compression --features snappy-compression --features brotli-compression --features mmap
- REM SET RUST_BACKTRACE=1 & cargo build --examples

27
benches/analyzer.rs
View File

@@ -1,13 +1,11 @@
use criterion::{criterion_group, criterion_main, Criterion};
use tantivy::tokenizer::{
LowerCaser, RemoveLongFilter, SimpleTokenizer, TextAnalyzer, TokenizerManager,
};
use tantivy::tokenizer::TokenizerManager;
const ALICE_TXT: &str = include_str!("alice.txt");
pub fn criterion_benchmark(c: &mut Criterion) {
let tokenizer_manager = TokenizerManager::default();
let mut tokenizer = tokenizer_manager.get("default").unwrap();
let tokenizer = tokenizer_manager.get("default").unwrap();
c.bench_function("default-tokenize-alice", |b| {
b.iter(|| {
let mut word_count = 0;
@@ -18,26 +16,7 @@ pub fn criterion_benchmark(c: &mut Criterion) {
assert_eq!(word_count, 30_731);
})
});
let mut dynamic_analyzer = TextAnalyzer::builder(SimpleTokenizer::default())
.dynamic()
.filter_dynamic(RemoveLongFilter::limit(40))
.filter_dynamic(LowerCaser)
.build();
c.bench_function("dynamic-tokenize-alice", |b| {
b.iter(|| {
let mut word_count = 0;
let mut token_stream = dynamic_analyzer.token_stream(ALICE_TXT);
while token_stream.advance() {
word_count += 1;
}
assert_eq!(word_count, 30_731);
})
});
}
criterion_group! {
name = benches;
config = Criterion::default().sample_size(200);
targets = criterion_benchmark
}
criterion_group!(benches, criterion_benchmark);
criterion_main!(benches);

1000
benches/gh.json
View File

File diff suppressed because one or more lines are too long

267
benches/index-bench.rs
View File

@@ -1,100 +1,10 @@
use criterion::{criterion_group, criterion_main, BatchSize, Bencher, Criterion, Throughput};
use tantivy::schema::{TantivyDocument, FAST, INDEXED, STORED, STRING, TEXT};
use tantivy::{tokenizer, Index, IndexWriter};
use criterion::{criterion_group, criterion_main, Criterion};
use pprof::criterion::{Output, PProfProfiler};
use tantivy::schema::{INDEXED, STORED, STRING, TEXT};
use tantivy::Index;
const HDFS_LOGS: &str = include_str!("hdfs.json");
const GH_LOGS: &str = include_str!("gh.json");
const WIKI: &str = include_str!("wiki.json");
fn benchmark(
b: &mut Bencher,
input: &str,
schema: tantivy::schema::Schema,
commit: bool,
parse_json: bool,
is_dynamic: bool,
) {
if is_dynamic {
benchmark_dynamic_json(b, input, schema, commit, parse_json)
} else {
_benchmark(b, input, schema, commit, parse_json, |schema, doc_json| {
TantivyDocument::parse_json(&schema, doc_json).unwrap()
})
}
}
fn get_index(schema: tantivy::schema::Schema) -> Index {
let mut index = Index::create_in_ram(schema.clone());
let ff_tokenizer_manager = tokenizer::TokenizerManager::default();
ff_tokenizer_manager.register(
"raw",
tokenizer::TextAnalyzer::builder(tokenizer::RawTokenizer::default())
.filter(tokenizer::RemoveLongFilter::limit(255))
.build(),
);
index.set_fast_field_tokenizers(ff_tokenizer_manager.clone());
index
}
fn _benchmark(
b: &mut Bencher,
input: &str,
schema: tantivy::schema::Schema,
commit: bool,
include_json_parsing: bool,
create_doc: impl Fn(&tantivy::schema::Schema, &str) -> TantivyDocument,
) {
if include_json_parsing {
let lines: Vec<&str> = input.trim().split('\n').collect();
b.iter(|| {
let index = get_index(schema.clone());
let mut index_writer: IndexWriter =
index.writer_with_num_threads(1, 100_000_000).unwrap();
for doc_json in &lines {
let doc = create_doc(&schema, doc_json);
index_writer.add_document(doc).unwrap();
}
if commit {
index_writer.commit().unwrap();
}
})
} else {
let docs: Vec<_> = input
.trim()
.split('\n')
.map(|doc_json| create_doc(&schema, doc_json))
.collect();
b.iter_batched(
|| docs.clone(),
|docs| {
let index = get_index(schema.clone());
let mut index_writer: IndexWriter =
index.writer_with_num_threads(1, 100_000_000).unwrap();
for doc in docs {
index_writer.add_document(doc).unwrap();
}
if commit {
index_writer.commit().unwrap();
}
},
BatchSize::SmallInput,
)
}
}
fn benchmark_dynamic_json(
b: &mut Bencher,
input: &str,
schema: tantivy::schema::Schema,
commit: bool,
parse_json: bool,
) {
let json_field = schema.get_field("json").unwrap();
_benchmark(b, input, schema, commit, parse_json, |_schema, doc_json| {
let json_val: serde_json::Map<String, serde_json::Value> =
serde_json::from_str(doc_json).unwrap();
tantivy::doc!(json_field=>json_val)
})
}
const NUM_REPEATS: usize = 2;
pub fn hdfs_index_benchmark(c: &mut Criterion) {
let schema = {
@@ -104,14 +14,7 @@ pub fn hdfs_index_benchmark(c: &mut Criterion) {
schema_builder.add_text_field("severity", STRING);
schema_builder.build()
};
let schema_only_fast = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_u64_field("timestamp", FAST);
schema_builder.add_text_field("body", FAST);
schema_builder.add_text_field("severity", FAST);
schema_builder.build()
};
let _schema_with_store = {
let schema_with_store = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_u64_field("timestamp", INDEXED | STORED);
schema_builder.add_text_field("body", TEXT | STORED);
@@ -120,99 +23,99 @@ pub fn hdfs_index_benchmark(c: &mut Criterion) {
};
let dynamic_schema = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_json_field("json", TEXT | FAST);
schema_builder.add_json_field("json", TEXT);
schema_builder.build()
};
let mut group = c.benchmark_group("index-hdfs");
group.throughput(Throughput::Bytes(HDFS_LOGS.len() as u64));
group.sample_size(20);
let benches = [
("only-indexed-".to_string(), schema, false),
//("stored-".to_string(), _schema_with_store, false),
("only-fast-".to_string(), schema_only_fast, false),
("dynamic-".to_string(), dynamic_schema, true),
];
for (prefix, schema, is_dynamic) in benches {
for commit in [false, true] {
let suffix = if commit { "with-commit" } else { "no-commit" };
for parse_json in [false] {
// for parse_json in [false, true] {
let suffix = if parse_json {
format!("{}-with-json-parsing", suffix)
} else {
format!("{}", suffix)
};
let bench_name = format!("{}{}", prefix, suffix);
group.bench_function(bench_name, |b| {
benchmark(b, HDFS_LOGS, schema.clone(), commit, parse_json, is_dynamic)
});
group.bench_function("index-hdfs-no-commit", |b| {
b.iter(|| {
let index = Index::create_in_ram(schema.clone());
let index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
for _ in 0..NUM_REPEATS {
for doc_json in HDFS_LOGS.trim().split('\n') {
let doc = schema.parse_document(doc_json).unwrap();
index_writer.add_document(doc).unwrap();
}
}
}
}
}
pub fn gh_index_benchmark(c: &mut Criterion) {
let dynamic_schema = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_json_field("json", TEXT | FAST);
schema_builder.build()
};
let dynamic_schema_fast = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_json_field("json", FAST);
schema_builder.build()
};
let mut group = c.benchmark_group("index-gh");
group.throughput(Throughput::Bytes(GH_LOGS.len() as u64));
group.bench_function("index-gh-no-commit", |b| {
benchmark_dynamic_json(b, GH_LOGS, dynamic_schema.clone(), false, false)
})
});
group.bench_function("index-gh-fast", |b| {
benchmark_dynamic_json(b, GH_LOGS, dynamic_schema_fast.clone(), false, false)
group.bench_function("index-hdfs-with-commit", |b| {
b.iter(|| {
let index = Index::create_in_ram(schema.clone());
let mut index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
for _ in 0..NUM_REPEATS {
for doc_json in HDFS_LOGS.trim().split('\n') {
let doc = schema.parse_document(doc_json).unwrap();
index_writer.add_document(doc).unwrap();
}
}
index_writer.commit().unwrap();
})
});
group.bench_function("index-gh-fast-with-commit", |b| {
benchmark_dynamic_json(b, GH_LOGS, dynamic_schema_fast.clone(), true, false)
group.bench_function("index-hdfs-no-commit-with-docstore", |b| {
b.iter(|| {
let index = Index::create_in_ram(schema_with_store.clone());
let index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
for _ in 0..NUM_REPEATS {
for doc_json in HDFS_LOGS.trim().split('\n') {
let doc = schema.parse_document(doc_json).unwrap();
index_writer.add_document(doc).unwrap();
}
}
})
});
}
pub fn wiki_index_benchmark(c: &mut Criterion) {
let dynamic_schema = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_json_field("json", TEXT | FAST);
schema_builder.build()
};
let mut group = c.benchmark_group("index-wiki");
group.throughput(Throughput::Bytes(WIKI.len() as u64));
group.bench_function("index-wiki-no-commit", |b| {
benchmark_dynamic_json(b, WIKI, dynamic_schema.clone(), false, false)
group.bench_function("index-hdfs-with-commit-with-docstore", |b| {
b.iter(|| {
let index = Index::create_in_ram(schema_with_store.clone());
let mut index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
for _ in 0..NUM_REPEATS {
for doc_json in HDFS_LOGS.trim().split('\n') {
let doc = schema.parse_document(doc_json).unwrap();
index_writer.add_document(doc).unwrap();
}
}
index_writer.commit().unwrap();
})
});
group.bench_function("index-wiki-with-commit", |b| {
benchmark_dynamic_json(b, WIKI, dynamic_schema.clone(), true, false)
group.bench_function("index-hdfs-no-commit-json-without-docstore", |b| {
b.iter(|| {
let index = Index::create_in_ram(dynamic_schema.clone());
let json_field = dynamic_schema.get_field("json").unwrap();
let mut index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
for _ in 0..NUM_REPEATS {
for doc_json in HDFS_LOGS.trim().split('\n') {
let json_val: serde_json::Map<String, serde_json::Value> =
serde_json::from_str(doc_json).unwrap();
let doc = tantivy::doc!(json_field=>json_val);
index_writer.add_document(doc).unwrap();
}
}
index_writer.commit().unwrap();
})
});
group.bench_function("index-hdfs-with-commit-json-without-docstore", |b| {
b.iter(|| {
let index = Index::create_in_ram(dynamic_schema.clone());
let json_field = dynamic_schema.get_field("json").unwrap();
let mut index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
for _ in 0..NUM_REPEATS {
for doc_json in HDFS_LOGS.trim().split('\n') {
let json_val: serde_json::Map<String, serde_json::Value> =
serde_json::from_str(doc_json).unwrap();
let doc = tantivy::doc!(json_field=>json_val);
index_writer.add_document(doc).unwrap();
}
}
index_writer.commit().unwrap();
})
});
}
criterion_group! {
name = benches;
config = Criterion::default();
config = Criterion::default().with_profiler(PProfProfiler::new(100, Output::Flamegraph(None)));
targets = hdfs_index_benchmark
}
criterion_group! {
name = gh_benches;
config = Criterion::default();
targets = gh_index_benchmark
}
criterion_group! {
name = wiki_benches;
config = Criterion::default();
targets = wiki_index_benchmark
}
criterion_main!(benches, gh_benches, wiki_benches);
criterion_main!(benches);

1000
benches/wiki.json
View File

File diff suppressed because one or more lines are too long

4
bitpacker/Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy-bitpacker"
version = "0.6.0"
version = "0.3.0"
edition = "2021"
authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT"
@@ -15,7 +15,7 @@ homepage = "https://github.com/quickwit-oss/tantivy"
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
[dependencies]
bitpacking = { version = "0.9.2", default-features = false, features = ["bitpacker1x"] }
bitpacking = {version="0.8", default-features=false, features = ["bitpacker1x"]}
[dev-dependencies]
rand = "0.8"

3
bitpacker/src/bitpacker.rs
View File

@@ -1,3 +1,4 @@
use std::convert::TryInto;
use std::io;
use std::ops::{Range, RangeInclusive};
@@ -366,7 +367,7 @@ mod test {
let mut output: Vec<u32> = Vec::new();
for len in [0, 1, 2, 32, 33, 34, 64] {
for start_idx in 0u32..32u32 {
output.resize(len, 0);
output.resize(len as usize, 0);
bitunpacker.get_batch_u32s(start_idx, &buffer, &mut output);
for i in 0..len {
let expected = (start_idx + i as u32) & mask;

4
bitpacker/src/blocked_bitpacker.rs
View File

@@ -64,8 +64,10 @@ fn mem_usage<T>(items: &Vec<T>) -> usize {
impl BlockedBitpacker {
pub fn new() -> Self {
let mut compressed_blocks = vec![];
compressed_blocks.resize(8, 0);
Self {
compressed_blocks: vec![0; 8],
compressed_blocks,
buffer: vec![],
offset_and_bits: vec![],
}

2
bitpacker/src/filter_vec/avx2.rs
View File

@@ -1,5 +1,5 @@
//! SIMD filtering of a vector as described in the following blog post.
//! <https://quickwit.io/blog/filtering%20a%20vector%20with%20simd%20instructions%20avx-2%20and%20avx-512>
//! https://quickwit.io/blog/filtering%20a%20vector%20with%20simd%20instructions%20avx-2%20and%20avx-512
use std::arch::x86_64::{
__m256i as DataType, _mm256_add_epi32 as op_add, _mm256_cmpgt_epi32 as op_greater,
_mm256_lddqu_si256 as load_unaligned, _mm256_or_si256 as op_or, _mm256_set1_epi32 as set1,

2
bitpacker/src/filter_vec/mod.rs
View File

@@ -1,6 +1,6 @@
use std::ops::RangeInclusive;
#[cfg(target_arch = "x86_64")]
#[cfg(any(target_arch = "x86_64"))]
mod avx2;
mod scalar;

93
cliff.toml
View File

@@ -1,93 +0,0 @@
# configuration file for git-cliff{ pattern = "foo", replace = "bar"}
# see https://github.com/orhun/git-cliff#configuration-file
[remote.github]
owner = "quickwit-oss"
repo = "tantivy"
[changelog]
# changelog header
header = """
"""
# template for the changelog body
# https://tera.netlify.app/docs/#introduction
body = """
## What's Changed
{%- if version %} in {{ version }}{%- endif -%}
{% for commit in commits %}
{% if commit.github.pr_title -%}
{%- set commit_message = commit.github.pr_title -%}
{%- else -%}
{%- set commit_message = commit.message -%}
{%- endif -%}
- {{ commit_message | split(pat="\n") | first | trim }}\
{% if commit.github.pr_number %} \
[#{{ commit.github.pr_number }}]({{ self::remote_url() }}/pull/{{ commit.github.pr_number }}){% if commit.github.username %}(@{{ commit.github.username }}){%- endif -%} \
{%- endif %}
{%- endfor -%}
{% if github.contributors | filter(attribute="is_first_time", value=true) | length != 0 %}
{% raw %}\n{% endraw -%}
## New Contributors
{%- endif %}\
{% for contributor in github.contributors | filter(attribute="is_first_time", value=true) %}
* @{{ contributor.username }} made their first contribution
{%- if contributor.pr_number %} in \
[#{{ contributor.pr_number }}]({{ self::remote_url() }}/pull/{{ contributor.pr_number }}) \
{%- endif %}
{%- endfor -%}
{% if version %}
{% if previous.version %}
**Full Changelog**: {{ self::remote_url() }}/compare/{{ previous.version }}...{{ version }}
{% endif %}
{% else -%}
{% raw %}\n{% endraw %}
{% endif %}
{%- macro remote_url() -%}
https://github.com/{{ remote.github.owner }}/{{ remote.github.repo }}
{%- endmacro -%}
"""
# remove the leading and trailing whitespace from the template
trim = true
# changelog footer
footer = """
"""
postprocessors = [
]
[git]
# parse the commits based on https://www.conventionalcommits.org
# This is required or commit.message contains the whole commit message and not just the title
conventional_commits = false
# filter out the commits that are not conventional
filter_unconventional = true
# process each line of a commit as an individual commit
split_commits = false
# regex for preprocessing the commit messages
commit_preprocessors = [
{ pattern = '\((\w+\s)?#([0-9]+)\)', replace = ""},
]
#link_parsers = [
#{ pattern = "#(\\d+)", href = "https://github.com/quickwit-oss/tantivy/pulls/$1"},
#]
# regex for parsing and grouping commits
# protect breaking changes from being skipped due to matching a skipping commit_parser
protect_breaking_commits = false
# filter out the commits that are not matched by commit parsers
filter_commits = false
# glob pattern for matching git tags
tag_pattern = "v[0-9]*"
# regex for skipping tags
skip_tags = "v0.1.0-beta.1"
# regex for ignoring tags
ignore_tags = ""
# sort the tags topologically
topo_order = false
# sort the commits inside sections by oldest/newest order
sort_commits = "newest"
# limit the number of commits included in the changelog.
# limit_commits = 42

24
columnar/Cargo.toml
View File

@@ -1,28 +1,28 @@
[package]
name = "tantivy-columnar"
version = "0.3.0"
version = "0.1.0"
edition = "2021"
license = "MIT"
homepage = "https://github.com/quickwit-oss/tantivy"
repository = "https://github.com/quickwit-oss/tantivy"
description = "column oriented storage for tantivy"
categories = ["database-implementations", "data-structures", "compression"]
[dependencies]
itertools = "0.12.0"
itertools = "0.10.5"
log = "0.4.17"
fnv = "1.0.7"
fastdivide = "0.4.0"
rand = { version = "0.8.5", optional = true }
measure_time = { version = "0.8.2", optional = true }
prettytable-rs = { version = "0.10.0", optional = true }
stacker = { version= "0.3", path = "../stacker", package="tantivy-stacker"}
sstable = { version= "0.3", path = "../sstable", package = "tantivy-sstable" }
common = { version= "0.7", path = "../common", package = "tantivy-common" }
tantivy-bitpacker = { version= "0.6", path = "../bitpacker/" }
stacker = { path = "../stacker", package="tantivy-stacker"}
sstable = { path = "../sstable", package = "tantivy-sstable" }
common = { path = "../common", package = "tantivy-common" }
tantivy-bitpacker = { version= "0.3", path = "../bitpacker/" }
serde = "1.0.152"
downcast-rs = "1.2.0"
[dev-dependencies]
proptest = "1"
more-asserts = "0.3.1"
rand = "0.8"
rand = "0.8.5"
[features]
unstable = []

155
columnar/benches/bench_first_vals.rs
View File

@@ -1,155 +0,0 @@
#![feature(test)]
extern crate test;
use std::sync::Arc;
use rand::prelude::*;
use tantivy_columnar::column_values::{serialize_and_load_u64_based_column_values, CodecType};
use tantivy_columnar::*;
use test::{black_box, Bencher};
struct Columns {
pub optional: Column,
pub full: Column,
pub multi: Column,
}
fn get_test_columns() -> Columns {
let data = generate_permutation();
let mut dataframe_writer = ColumnarWriter::default();
for (idx, val) in data.iter().enumerate() {
dataframe_writer.record_numerical(idx as u32, "full_values", NumericalValue::U64(*val));
if idx % 2 == 0 {
dataframe_writer.record_numerical(
idx as u32,
"optional_values",
NumericalValue::U64(*val),
);
}
dataframe_writer.record_numerical(idx as u32, "multi_values", NumericalValue::U64(*val));
dataframe_writer.record_numerical(idx as u32, "multi_values", NumericalValue::U64(*val));
}
let mut buffer: Vec<u8> = Vec::new();
dataframe_writer
.serialize(data.len() as u32, None, &mut buffer)
.unwrap();
let columnar = ColumnarReader::open(buffer).unwrap();
let cols: Vec<DynamicColumnHandle> = columnar.read_columns("optional_values").unwrap();
assert_eq!(cols.len(), 1);
let optional = cols[0].open_u64_lenient().unwrap().unwrap();
assert_eq!(optional.index.get_cardinality(), Cardinality::Optional);
let cols: Vec<DynamicColumnHandle> = columnar.read_columns("full_values").unwrap();
assert_eq!(cols.len(), 1);
let column_full = cols[0].open_u64_lenient().unwrap().unwrap();
assert_eq!(column_full.index.get_cardinality(), Cardinality::Full);
let cols: Vec<DynamicColumnHandle> = columnar.read_columns("multi_values").unwrap();
assert_eq!(cols.len(), 1);
let multi = cols[0].open_u64_lenient().unwrap().unwrap();
assert_eq!(multi.index.get_cardinality(), Cardinality::Multivalued);
Columns {
optional,
full: column_full,
multi,
}
}
const NUM_VALUES: u64 = 100_000;
fn generate_permutation() -> Vec<u64> {
let mut permutation: Vec<u64> = (0u64..NUM_VALUES).collect();
permutation.shuffle(&mut StdRng::from_seed([1u8; 32]));
permutation
}
pub fn serialize_and_load(column: &[u64], codec_type: CodecType) -> Arc<dyn ColumnValues<u64>> {
serialize_and_load_u64_based_column_values(&column, &[codec_type])
}
fn run_bench_on_column_full_scan(b: &mut Bencher, column: Column) {
let num_iter = black_box(NUM_VALUES);
b.iter(|| {
let mut sum = 0u64;
for i in 0..num_iter as u32 {
let val = column.first(i);
sum += val.unwrap_or(0);
}
sum
});
}
fn run_bench_on_column_block_fetch(b: &mut Bencher, column: Column) {
let mut block: Vec<Option<u64>> = vec![None; 64];
let fetch_docids = (0..64).collect::<Vec<_>>();
b.iter(move || {
column.first_vals(&fetch_docids, &mut block);
block[0]
});
}
fn run_bench_on_column_block_single_calls(b: &mut Bencher, column: Column) {
let mut block: Vec<Option<u64>> = vec![None; 64];
let fetch_docids = (0..64).collect::<Vec<_>>();
b.iter(move || {
for i in 0..fetch_docids.len() {
block[i] = column.first(fetch_docids[i]);
}
block[0]
});
}
/// Column first method
#[bench]
fn bench_get_first_on_full_column_full_scan(b: &mut Bencher) {
let column = get_test_columns().full;
run_bench_on_column_full_scan(b, column);
}
#[bench]
fn bench_get_first_on_optional_column_full_scan(b: &mut Bencher) {
let column = get_test_columns().optional;
run_bench_on_column_full_scan(b, column);
}
#[bench]
fn bench_get_first_on_multi_column_full_scan(b: &mut Bencher) {
let column = get_test_columns().multi;
run_bench_on_column_full_scan(b, column);
}
/// Block fetch column accessor
#[bench]
fn bench_get_block_first_on_optional_column(b: &mut Bencher) {
let column = get_test_columns().optional;
run_bench_on_column_block_fetch(b, column);
}
#[bench]
fn bench_get_block_first_on_multi_column(b: &mut Bencher) {
let column = get_test_columns().multi;
run_bench_on_column_block_fetch(b, column);
}
#[bench]
fn bench_get_block_first_on_full_column(b: &mut Bencher) {
let column = get_test_columns().full;
run_bench_on_column_block_fetch(b, column);
}
#[bench]
fn bench_get_block_first_on_optional_column_single_calls(b: &mut Bencher) {
let column = get_test_columns().optional;
run_bench_on_column_block_single_calls(b, column);
}
#[bench]
fn bench_get_block_first_on_multi_column_single_calls(b: &mut Bencher) {
let column = get_test_columns().multi;
run_bench_on_column_block_single_calls(b, column);
}
#[bench]
fn bench_get_block_first_on_full_column_single_calls(b: &mut Bencher) {
let column = get_test_columns().full;
run_bench_on_column_block_single_calls(b, column);
}

0
columnar/benches/bench_values_u128.rs → columnar/benches/bench_u128.rs
View File

8
columnar/benches/bench_values_u64.rs → columnar/benches/bench_u64.rs
View File

@@ -16,6 +16,14 @@ fn generate_permutation() -> Vec<u64> {
permutation
}
fn generate_random() -> Vec<u64> {
let mut permutation: Vec<u64> = (0u64..100_000u64)
.map(|el| el + random::<u16>() as u64)
.collect();
permutation.shuffle(&mut StdRng::from_seed([1u8; 32]));
permutation
}
// Warning: this generates the same permutation at each call
fn generate_permutation_gcd() -> Vec<u64> {
let mut permutation: Vec<u64> = (1u64..100_000u64).map(|el| el * 1000).collect();

1
columnar/columnar-cli/Cargo.toml
View File

@@ -8,6 +8,7 @@ license = "MIT"
columnar = {path="../", package="tantivy-columnar"}
serde_json = "1"
serde_json_borrow = {git="https://github.com/PSeitz/serde_json_borrow/"}
serde = "1"
[workspace]
members = []

148
columnar/src/block_accessor.rs
View File

@@ -1,12 +1,9 @@
use std::cmp::Ordering;
use crate::{Column, DocId, RowId};
#[derive(Debug, Default, Clone)]
pub struct ColumnBlockAccessor<T> {
val_cache: Vec<T>,
docid_cache: Vec<DocId>,
missing_docids_cache: Vec<DocId>,
row_id_cache: Vec<RowId>,
}
@@ -14,40 +11,14 @@ impl<T: PartialOrd + Copy + std::fmt::Debug + Send + Sync + 'static + Default>
ColumnBlockAccessor<T>
{
#[inline]
pub fn fetch_block<'a>(&'a mut self, docs: &'a [u32], accessor: &Column<T>) {
if accessor.index.get_cardinality().is_full() {
self.val_cache.resize(docs.len(), T::default());
accessor.values.get_vals(docs, &mut self.val_cache);
} else {
self.docid_cache.clear();
self.row_id_cache.clear();
accessor.row_ids_for_docs(docs, &mut self.docid_cache, &mut self.row_id_cache);
self.val_cache.resize(self.row_id_cache.len(), T::default());
accessor
.values
.get_vals(&self.row_id_cache, &mut self.val_cache);
}
}
#[inline]
pub fn fetch_block_with_missing(&mut self, docs: &[u32], accessor: &Column<T>, missing: T) {
self.fetch_block(docs, accessor);
// no missing values
if accessor.index.get_cardinality().is_full() {
return;
}
// We can compare docid_cache length with docs to find missing docs
// For multi value columns we can't rely on the length and always need to scan
if accessor.index.get_cardinality().is_multivalue() || docs.len() != self.docid_cache.len()
{
self.missing_docids_cache.clear();
find_missing_docs(docs, &self.docid_cache, |doc| {
self.missing_docids_cache.push(doc);
self.val_cache.push(missing);
});
self.docid_cache
.extend_from_slice(&self.missing_docids_cache);
}
pub fn fetch_block(&mut self, docs: &[u32], accessor: &Column<T>) {
self.docid_cache.clear();
self.row_id_cache.clear();
accessor.row_ids_for_docs(docs, &mut self.docid_cache, &mut self.row_id_cache);
self.val_cache.resize(self.row_id_cache.len(), T::default());
accessor
.values
.get_vals(&self.row_id_cache, &mut self.val_cache);
}
#[inline]
@@ -56,103 +27,10 @@ impl<T: PartialOrd + Copy + std::fmt::Debug + Send + Sync + 'static + Default>
}
#[inline]
/// Returns an iterator over the docids and values
/// The passed in `docs` slice needs to be the same slice that was passed to `fetch_block` or
/// `fetch_block_with_missing`.
///
/// The docs is used if the column is full (each docs has exactly one value), otherwise the
/// internal docid vec is used for the iterator, which e.g. may contain duplicate docs.
pub fn iter_docid_vals<'a>(
&'a self,
docs: &'a [u32],
accessor: &Column<T>,
) -> impl Iterator<Item = (DocId, T)> + '_ {
if accessor.index.get_cardinality().is_full() {
docs.iter().cloned().zip(self.val_cache.iter().cloned())
} else {
self.docid_cache
.iter()
.cloned()
.zip(self.val_cache.iter().cloned())
}
}
}
/// Given two sorted lists of docids `docs` and `hits`, hits is a subset of `docs`.
/// Return all docs that are not in `hits`.
fn find_missing_docs<F>(docs: &[u32], hits: &[u32], mut callback: F)
where F: FnMut(u32) {
let mut docs_iter = docs.iter();
let mut hits_iter = hits.iter();
let mut doc = docs_iter.next();
let mut hit = hits_iter.next();
while let (Some(&current_doc), Some(&current_hit)) = (doc, hit) {
match current_doc.cmp(&current_hit) {
Ordering::Less => {
callback(current_doc);
doc = docs_iter.next();
}
Ordering::Equal => {
doc = docs_iter.next();
hit = hits_iter.next();
}
Ordering::Greater => {
hit = hits_iter.next();
}
}
}
while let Some(&current_doc) = doc {
callback(current_doc);
doc = docs_iter.next();
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn test_find_missing_docs() {
let docs: Vec<u32> = vec![1, 2, 3, 4, 5, 6, 7, 8, 9, 10];
let hits: Vec<u32> = vec![2, 4, 6, 8, 10];
let mut missing_docs: Vec<u32> = Vec::new();
find_missing_docs(&docs, &hits, |missing_doc| {
missing_docs.push(missing_doc);
});
assert_eq!(missing_docs, vec![1, 3, 5, 7, 9]);
}
#[test]
fn test_find_missing_docs_empty() {
let docs: Vec<u32> = Vec::new();
let hits: Vec<u32> = vec![2, 4, 6, 8, 10];
let mut missing_docs: Vec<u32> = Vec::new();
find_missing_docs(&docs, &hits, |missing_doc| {
missing_docs.push(missing_doc);
});
assert_eq!(missing_docs, vec![]);
}
#[test]
fn test_find_missing_docs_all_missing() {
let docs: Vec<u32> = vec![1, 2, 3, 4, 5];
let hits: Vec<u32> = Vec::new();
let mut missing_docs: Vec<u32> = Vec::new();
find_missing_docs(&docs, &hits, |missing_doc| {
missing_docs.push(missing_doc);
});
assert_eq!(missing_docs, vec![1, 2, 3, 4, 5]);
pub fn iter_docid_vals(&self) -> impl Iterator<Item = (DocId, T)> + '_ {
self.docid_cache
.iter()
.cloned()
.zip(self.val_cache.iter().cloned())
}
}

9
columnar/src/column/dictionary_encoded.rs
View File

@@ -30,13 +30,6 @@ impl fmt::Debug for BytesColumn {
}
impl BytesColumn {
pub fn empty(num_docs: u32) -> BytesColumn {
BytesColumn {
dictionary: Arc::new(Dictionary::empty()),
term_ord_column: Column::build_empty_column(num_docs),
}
}
/// Fills the given `output` buffer with the term associated to the ordinal `ord`.
///
/// Returns `false` if the term does not exist (e.g. `term_ord` is greater or equal to the
@@ -84,7 +77,7 @@ impl From<StrColumn> for BytesColumn {
}
impl StrColumn {
pub fn wrap(bytes_column: BytesColumn) -> StrColumn {
pub(crate) fn wrap(bytes_column: BytesColumn) -> StrColumn {
StrColumn(bytes_column)
}

48
columnar/src/column/mod.rs
View File

@@ -3,17 +3,17 @@ mod serialize;
use std::fmt::{self, Debug};
use std::io::Write;
use std::ops::{Range, RangeInclusive};
use std::ops::{Deref, Range, RangeInclusive};
use std::sync::Arc;
use common::BinarySerializable;
pub use dictionary_encoded::{BytesColumn, StrColumn};
pub use serialize::{
open_column_bytes, open_column_str, open_column_u128, open_column_u128_as_compact_u64,
open_column_u64, serialize_column_mappable_to_u128, serialize_column_mappable_to_u64,
open_column_bytes, open_column_str, open_column_u128, open_column_u64,
serialize_column_mappable_to_u128, serialize_column_mappable_to_u64,
};
use crate::column_index::{ColumnIndex, Set};
use crate::column_index::ColumnIndex;
use crate::column_values::monotonic_mapping::StrictlyMonotonicMappingToInternal;
use crate::column_values::{monotonic_map_column, ColumnValues};
use crate::{Cardinality, DocId, EmptyColumnValues, MonotonicallyMappableToU64, RowId};
@@ -83,36 +83,10 @@ impl<T: PartialOrd + Copy + Debug + Send + Sync + 'static> Column<T> {
self.values.max_value()
}
#[inline]
pub fn first(&self, row_id: RowId) -> Option<T> {
self.values_for_doc(row_id).next()
}
/// Load the first value for each docid in the provided slice.
#[inline]
pub fn first_vals(&self, docids: &[DocId], output: &mut [Option<T>]) {
match &self.index {
ColumnIndex::Empty { .. } => {}
ColumnIndex::Full => self.values.get_vals_opt(docids, output),
ColumnIndex::Optional(optional_index) => {
for (i, docid) in docids.iter().enumerate() {
output[i] = optional_index
.rank_if_exists(*docid)
.map(|rowid| self.values.get_val(rowid));
}
}
ColumnIndex::Multivalued(multivalued_index) => {
for (i, docid) in docids.iter().enumerate() {
let range = multivalued_index.range(*docid);
let is_empty = range.start == range.end;
if !is_empty {
output[i] = Some(self.values.get_val(range.start));
}
}
}
}
}
/// Translates a block of docis to row_ids.
///
/// returns the row_ids and the matching docids on the same index
@@ -131,8 +105,7 @@ impl<T: PartialOrd + Copy + Debug + Send + Sync + 'static> Column<T> {
}
pub fn values_for_doc(&self, doc_id: DocId) -> impl Iterator<Item = T> + '_ {
self.index
.value_row_ids(doc_id)
self.value_row_ids(doc_id)
.map(|value_row_id: RowId| self.values.get_val(value_row_id))
}
@@ -157,7 +130,7 @@ impl<T: PartialOrd + Copy + Debug + Send + Sync + 'static> Column<T> {
.select_batch_in_place(selected_docid_range.start, doc_ids);
}
/// Fills the output vector with the (possibly multiple values that are associated_with
/// Fils the output vector with the (possibly multiple values that are associated_with
/// `row_id`.
///
/// This method clears the `output` vector.
@@ -174,6 +147,14 @@ impl<T: PartialOrd + Copy + Debug + Send + Sync + 'static> Column<T> {
}
}
impl<T> Deref for Column<T> {
type Target = ColumnIndex;
fn deref(&self) -> &Self::Target {
&self.index
}
}
impl BinarySerializable for Cardinality {
fn serialize<W: Write + ?Sized>(&self, writer: &mut W) -> std::io::Result<()> {
self.to_code().serialize(writer)
@@ -195,7 +176,6 @@ struct FirstValueWithDefault<T: Copy> {
impl<T: PartialOrd + Debug + Send + Sync + Copy + 'static> ColumnValues<T>
for FirstValueWithDefault<T>
{
#[inline(always)]
fn get_val(&self, idx: u32) -> T {
self.column.first(idx).unwrap_or(self.default_value)
}

20
columnar/src/column/serialize.rs
View File

@@ -76,26 +76,6 @@ pub fn open_column_u128<T: MonotonicallyMappableToU128>(
})
}
/// Open the column as u64.
///
/// See [`open_u128_as_compact_u64`] for more details.
pub fn open_column_u128_as_compact_u64(bytes: OwnedBytes) -> io::Result<Column<u64>> {
let (body, column_index_num_bytes_payload) = bytes.rsplit(4);
let column_index_num_bytes = u32::from_le_bytes(
column_index_num_bytes_payload
.as_slice()
.try_into()
.unwrap(),
);
let (column_index_data, column_values_data) = body.split(column_index_num_bytes as usize);
let column_index = crate::column_index::open_column_index(column_index_data)?;
let column_values = crate::column_values::open_u128_as_compact_u64(column_values_data)?;
Ok(Column {
index: column_index,
values: column_values,
})
}
pub fn open_column_bytes(data: OwnedBytes) -> io::Result<BytesColumn> {
let (body, dictionary_len_bytes) = data.rsplit(4);
let dictionary_len = u32::from_le_bytes(dictionary_len_bytes.as_slice().try_into().unwrap());

120
columnar/src/column_index/merge/mod.rs
View File

@@ -1,73 +1,19 @@
mod shuffled;
mod stacked;
use common::ReadOnlyBitSet;
use shuffled::merge_column_index_shuffled;
use stacked::merge_column_index_stacked;
use crate::column_index::SerializableColumnIndex;
use crate::{Cardinality, ColumnIndex, MergeRowOrder};
fn detect_cardinality_single_column_index(
column_index: &ColumnIndex,
alive_bitset_opt: &Option<ReadOnlyBitSet>,
) -> Cardinality {
let Some(alive_bitset) = alive_bitset_opt else {
return column_index.get_cardinality();
};
let cardinality_before_deletes = column_index.get_cardinality();
if cardinality_before_deletes == Cardinality::Full {
// The columnar cardinality can only become more restrictive in the presence of deletes
// (where cardinality sorted from the more restrictive to the least restrictive are Full,
// Optional, Multivalued)
//
// If we are already "Full", we are guaranteed to stay "Full" after deletes.
return Cardinality::Full;
}
let mut cardinality_so_far = Cardinality::Full;
for doc_id in alive_bitset.iter() {
let num_values = column_index.value_row_ids(doc_id).len();
let row_cardinality = match num_values {
0 => Cardinality::Optional,
1 => Cardinality::Full,
_ => Cardinality::Multivalued,
};
cardinality_so_far = cardinality_so_far.max(row_cardinality);
if cardinality_so_far >= cardinality_before_deletes {
// There won't be any improvement in the cardinality.
// We can early exit.
return cardinality_before_deletes;
}
}
cardinality_so_far
}
fn detect_cardinality(
column_indexes: &[ColumnIndex],
merge_row_order: &MergeRowOrder,
) -> Cardinality {
match merge_row_order {
MergeRowOrder::Stack(_) => column_indexes
.iter()
.map(ColumnIndex::get_cardinality)
.max()
.unwrap_or(Cardinality::Full),
MergeRowOrder::Shuffled(shuffle_merge_order) => {
let mut merged_cardinality = Cardinality::Full;
for (column_index, alive_bitset_opt) in column_indexes
.iter()
.zip(shuffle_merge_order.alive_bitsets.iter())
{
let cardinality: Cardinality =
detect_cardinality_single_column_index(column_index, alive_bitset_opt);
if cardinality == Cardinality::Multivalued {
return cardinality;
}
merged_cardinality = merged_cardinality.max(cardinality);
}
merged_cardinality
}
}
// For simplification, we never have cardinality go down due to deletes.
fn detect_cardinality(columns: &[ColumnIndex]) -> Cardinality {
columns
.iter()
.map(ColumnIndex::get_cardinality)
.max()
.unwrap_or(Cardinality::Full)
}
pub fn merge_column_index<'a>(
@@ -76,7 +22,7 @@ pub fn merge_column_index<'a>(
) -> SerializableColumnIndex<'a> {
// For simplification, we do not try to detect whether the cardinality could be
// downgraded thanks to deletes.
let cardinality_after_merge = detect_cardinality(columns, merge_row_order);
let cardinality_after_merge = detect_cardinality(columns);
match merge_row_order {
MergeRowOrder::Stack(stack_merge_order) => {
merge_column_index_stacked(columns, cardinality_after_merge, stack_merge_order)
@@ -98,54 +44,34 @@ mod tests {
use crate::column_index::merge::detect_cardinality;
use crate::column_index::multivalued_index::MultiValueIndex;
use crate::column_index::{merge_column_index, OptionalIndex, SerializableColumnIndex};
use crate::{
Cardinality, ColumnIndex, MergeRowOrder, RowAddr, RowId, ShuffleMergeOrder, StackMergeOrder,
};
use crate::{Cardinality, ColumnIndex, MergeRowOrder, RowAddr, RowId, ShuffleMergeOrder};
#[test]
fn test_detect_cardinality() {
assert_eq!(
detect_cardinality(&[], &StackMergeOrder::stack_for_test(&[]).into()),
Cardinality::Full
);
assert_eq!(detect_cardinality(&[]), Cardinality::Full);
let optional_index: ColumnIndex = OptionalIndex::for_test(1, &[]).into();
let multivalued_index: ColumnIndex = MultiValueIndex::for_test(&[0, 1]).into();
assert_eq!(
detect_cardinality(
&[optional_index.clone(), ColumnIndex::Empty { num_docs: 0 }],
&StackMergeOrder::stack_for_test(&[1, 0]).into()
),
detect_cardinality(&[optional_index.clone(), ColumnIndex::Empty { num_docs: 0 }]),
Cardinality::Optional
);
assert_eq!(
detect_cardinality(
&[optional_index.clone(), ColumnIndex::Full],
&StackMergeOrder::stack_for_test(&[1, 1]).into()
),
detect_cardinality(&[optional_index.clone(), ColumnIndex::Full]),
Cardinality::Optional
);
assert_eq!(
detect_cardinality(
&[
multivalued_index.clone(),
ColumnIndex::Empty { num_docs: 0 }
],
&StackMergeOrder::stack_for_test(&[1, 0]).into()
),
detect_cardinality(&[
multivalued_index.clone(),
ColumnIndex::Empty { num_docs: 0 }
]),
Cardinality::Multivalued
);
assert_eq!(
detect_cardinality(
&[multivalued_index.clone(), optional_index.clone()],
&StackMergeOrder::stack_for_test(&[1, 1]).into()
),
detect_cardinality(&[multivalued_index.clone(), optional_index.clone()]),
Cardinality::Multivalued
);
assert_eq!(
detect_cardinality(
&[optional_index, multivalued_index],
&StackMergeOrder::stack_for_test(&[1, 1]).into()
),
detect_cardinality(&[optional_index, multivalued_index]),
Cardinality::Multivalued
);
}
@@ -168,9 +94,8 @@ mod tests {
)
.into();
let merged_column_index = merge_column_index(&column_indexes[..], &merge_row_order);
let SerializableColumnIndex::Multivalued(start_index_iterable) = merged_column_index else {
panic!("Excpected a multivalued index")
};
let SerializableColumnIndex::Multivalued(start_index_iterable) = merged_column_index
else { panic!("Excpected a multivalued index") };
let start_indexes: Vec<RowId> = start_index_iterable.boxed_iter().collect();
assert_eq!(&start_indexes, &[0, 3, 5]);
}
@@ -201,9 +126,8 @@ mod tests {
)
.into();
let merged_column_index = merge_column_index(&column_indexes[..], &merge_row_order);
let SerializableColumnIndex::Multivalued(start_index_iterable) = merged_column_index else {
panic!("Excpected a multivalued index")
};
let SerializableColumnIndex::Multivalued(start_index_iterable) = merged_column_index
else { panic!("Excpected a multivalued index") };
let start_indexes: Vec<RowId> = start_index_iterable.boxed_iter().collect();
assert_eq!(&start_indexes, &[0, 3, 5, 6]);
}

10
columnar/src/column_index/merge/shuffled.rs
View File

@@ -140,7 +140,7 @@ mod tests {
#[test]
fn test_merge_column_index_optional_shuffle() {
let optional_index: ColumnIndex = OptionalIndex::for_test(2, &[0]).into();
let column_indexes = [optional_index, ColumnIndex::Full];
let column_indexes = vec![optional_index, ColumnIndex::Full];
let row_addrs = vec![
RowAddr {
segment_ord: 0u32,
@@ -157,13 +157,7 @@ mod tests {
Cardinality::Optional,
&shuffle_merge_order,
);
let SerializableColumnIndex::Optional {
non_null_row_ids,
num_rows,
} = serializable_index
else {
panic!()
};
let SerializableColumnIndex::Optional { non_null_row_ids, num_rows } = serializable_index else { panic!() };
assert_eq!(num_rows, 2);
let non_null_rows: Vec<RowId> = non_null_row_ids.boxed_iter().collect();
assert_eq!(&non_null_rows, &[1]);

5
columnar/src/column_index/merge/stacked.rs
View File

@@ -111,7 +111,10 @@ fn stack_multivalued_indexes<'a>(
let mut last_row_id = 0;
let mut current_it = multivalued_indexes.next();
Box::new(std::iter::from_fn(move || loop {
if let Some(row_id) = current_it.as_mut()?.next() {
let Some(multivalued_index) = current_it.as_mut() else {
return None;
};
if let Some(row_id) = multivalued_index.next() {
last_row_id = offset + row_id;
return Some(last_row_id);
}

25
columnar/src/column_index/mod.rs
View File

@@ -1,8 +1,3 @@
//! # `column_index`
//!
//! `column_index` provides rank and select operations to associate positions when not all
//! documents have exactly one element.
mod merge;
mod multivalued_index;
mod optional_index;
@@ -42,10 +37,10 @@ impl From<MultiValueIndex> for ColumnIndex {
}
impl ColumnIndex {
/// Returns the cardinality of the column index.
///
/// By convention, if the column contains no docs, we consider that it is
/// full.
// Returns the cardinality of the column index.
//
// By convention, if the column contains no docs, we consider that it is
// full.
#[inline]
pub fn get_cardinality(&self) -> Cardinality {
match self {
@@ -122,18 +117,18 @@ impl ColumnIndex {
}
}
pub fn docid_range_to_rowids(&self, doc_id_range: Range<DocId>) -> Range<RowId> {
pub fn docid_range_to_rowids(&self, doc_id: Range<DocId>) -> Range<RowId> {
match self {
ColumnIndex::Empty { .. } => 0..0,
ColumnIndex::Full => doc_id_range,
ColumnIndex::Full => doc_id,
ColumnIndex::Optional(optional_index) => {
let row_start = optional_index.rank(doc_id_range.start);
let row_end = optional_index.rank(doc_id_range.end);
let row_start = optional_index.rank(doc_id.start);
let row_end = optional_index.rank(doc_id.end);
row_start..row_end
}
ColumnIndex::Multivalued(multivalued_index) => {
let end_docid = doc_id_range.end.min(multivalued_index.num_docs() - 1) + 1;
let start_docid = doc_id_range.start.min(end_docid);
let end_docid = doc_id.end.min(multivalued_index.num_docs() - 1) + 1;
let start_docid = doc_id.start.min(end_docid);
let row_start = multivalued_index.start_index_column.get_val(start_docid);
let row_end = multivalued_index.start_index_column.get_val(end_docid);

17
columnar/src/column_index/optional_index/mod.rs
View File

@@ -21,6 +21,8 @@ const DENSE_BLOCK_THRESHOLD: u32 =
const ELEMENTS_PER_BLOCK: u32 = u16::MAX as u32 + 1;
const BLOCK_SIZE: RowId = 1 << 16;
#[derive(Copy, Clone, Debug)]
struct BlockMeta {
non_null_rows_before_block: u32,
@@ -107,8 +109,8 @@ struct RowAddr {
#[inline(always)]
fn row_addr_from_row_id(row_id: RowId) -> RowAddr {
RowAddr {
block_id: (row_id / ELEMENTS_PER_BLOCK) as u16,
in_block_row_id: (row_id % ELEMENTS_PER_BLOCK) as u16,
block_id: (row_id / BLOCK_SIZE) as u16,
in_block_row_id: (row_id % BLOCK_SIZE) as u16,
}
}
@@ -183,13 +185,8 @@ impl Set<RowId> for OptionalIndex {
}
}
/// Any value doc_id is allowed.
/// In particular, doc_id = num_rows.
#[inline]
fn rank(&self, doc_id: DocId) -> RowId {
if doc_id >= self.num_docs() {
return self.num_non_nulls();
}
let RowAddr {
block_id,
in_block_row_id,
@@ -203,15 +200,13 @@ impl Set<RowId> for OptionalIndex {
block_meta.non_null_rows_before_block + block_offset_row_id
}
/// Any value doc_id is allowed.
/// In particular, doc_id = num_rows.
#[inline]
fn rank_if_exists(&self, doc_id: DocId) -> Option<RowId> {
let RowAddr {
block_id,
in_block_row_id,
} = row_addr_from_row_id(doc_id);
let block_meta = *self.block_metas.get(block_id as usize)?;
let block_meta = self.block_metas[block_id as usize];
let block = self.block(block_meta);
let block_offset_row_id = match block {
Block::Dense(dense_block) => dense_block.rank_if_exists(in_block_row_id),
@@ -496,7 +491,7 @@ fn deserialize_optional_index_block_metadatas(
non_null_rows_before_block += num_non_null_rows;
}
block_metas.resize(
((num_rows + ELEMENTS_PER_BLOCK - 1) / ELEMENTS_PER_BLOCK) as usize,
((num_rows + BLOCK_SIZE - 1) / BLOCK_SIZE) as usize,
BlockMeta {
non_null_rows_before_block,
start_byte_offset,

3
columnar/src/column_index/optional_index/set.rs
View File

@@ -39,8 +39,7 @@ pub trait Set<T> {
///
/// # Panics
///
/// May panic if rank is greater or equal to the number of
/// elements in the Set.
/// May panic if rank is greater than the number of elements in the Set.
fn select(&self, rank: T) -> T;
/// Creates a brand new select cursor.

1
columnar/src/column_index/optional_index/set_block/dense.rs
View File

@@ -1,3 +1,4 @@
use std::convert::TryInto;
use std::io::{self, Write};
use common::BinarySerializable;

43
columnar/src/column_index/optional_index/tests.rs
View File

@@ -1,31 +1,8 @@
use proptest::prelude::*;
use proptest::prelude::{any, prop, *};
use proptest::strategy::Strategy;
use proptest::{prop_oneof, proptest};
use super::*;
use crate::{ColumnarReader, ColumnarWriter, DynamicColumnHandle};
#[test]
fn test_optional_index_bug_2293() {
// tests for panic in docid_range_to_rowids for docid == num_docs
test_optional_index_with_num_docs(ELEMENTS_PER_BLOCK - 1);
test_optional_index_with_num_docs(ELEMENTS_PER_BLOCK);
test_optional_index_with_num_docs(ELEMENTS_PER_BLOCK + 1);
}
fn test_optional_index_with_num_docs(num_docs: u32) {
let mut dataframe_writer = ColumnarWriter::default();
dataframe_writer.record_numerical(100, "score", 80i64);
let mut buffer: Vec<u8> = Vec::new();
dataframe_writer
.serialize(num_docs, None, &mut buffer)
.unwrap();
let columnar = ColumnarReader::open(buffer).unwrap();
assert_eq!(columnar.num_columns(), 1);
let cols: Vec<DynamicColumnHandle> = columnar.read_columns("score").unwrap();
assert_eq!(cols.len(), 1);
let col = cols[0].open().unwrap();
col.column_index().docid_range_to_rowids(0..num_docs);
}
#[test]
fn test_dense_block_threshold() {
@@ -58,7 +35,7 @@ proptest! {
#[test]
fn test_with_random_sets_simple() {
let vals = 10..ELEMENTS_PER_BLOCK * 2;
let vals = 10..BLOCK_SIZE * 2;
let mut out: Vec<u8> = Vec::new();
serialize_optional_index(&vals, 100, &mut out).unwrap();
let null_index = open_optional_index(OwnedBytes::new(out)).unwrap();
@@ -194,7 +171,7 @@ fn test_optional_index_rank() {
test_optional_index_rank_aux(&[0u32, 1u32]);
let mut block = Vec::new();
block.push(3u32);
block.extend((0..ELEMENTS_PER_BLOCK).map(|i| i + ELEMENTS_PER_BLOCK + 1));
block.extend((0..BLOCK_SIZE).map(|i| i + BLOCK_SIZE + 1));
test_optional_index_rank_aux(&block);
}
@@ -208,8 +185,8 @@ fn test_optional_index_iter_empty_one() {
fn test_optional_index_iter_dense_block() {
let mut block = Vec::new();
block.push(3u32);
block.extend((0..ELEMENTS_PER_BLOCK).map(|i| i + ELEMENTS_PER_BLOCK + 1));
test_optional_index_iter_aux(&block, 3 * ELEMENTS_PER_BLOCK);
block.extend((0..BLOCK_SIZE).map(|i| i + BLOCK_SIZE + 1));
test_optional_index_iter_aux(&block, 3 * BLOCK_SIZE);
}
#[test]
@@ -238,12 +215,12 @@ mod bench {
let vals: Vec<RowId> = (0..TOTAL_NUM_VALUES)
.map(|_| rng.gen_bool(fill_ratio))
.enumerate()
.filter(|(_pos, val)| *val)
.filter(|(pos, val)| *val)
.map(|(pos, _)| pos as RowId)
.collect();
serialize_optional_index(&&vals[..], TOTAL_NUM_VALUES, &mut out).unwrap();
open_optional_index(OwnedBytes::new(out)).unwrap()
let codec = open_optional_index(OwnedBytes::new(out)).unwrap();
codec
}
fn random_range_iterator(
@@ -265,7 +242,7 @@ mod bench {
}
fn n_percent_step_iterator(percent: f32, num_values: u32) -> impl Iterator<Item = u32> {
let ratio = percent / 100.0;
let ratio = percent as f32 / 100.0;
let step_size = (1f32 / ratio) as u32;
let deviation = step_size - 1;
random_range_iterator(0, num_values, step_size, deviation)

2
columnar/src/column_index/serialize.rs
View File

@@ -30,7 +30,6 @@ impl<'a> SerializableColumnIndex<'a> {
}
}
/// Serialize a column index.
pub fn serialize_column_index(
column_index: SerializableColumnIndex,
output: &mut impl Write,
@@ -52,7 +51,6 @@ pub fn serialize_column_index(
Ok(column_index_num_bytes)
}
/// Open a serialized column index.
pub fn open_column_index(mut bytes: OwnedBytes) -> io::Result<ColumnIndex> {
if bytes.is_empty() {
return Err(io::Error::new(

2
columnar/src/column_values/merge.rs
View File

@@ -10,7 +10,7 @@ pub(crate) struct MergedColumnValues<'a, T> {
pub(crate) merge_row_order: &'a MergeRowOrder,
}
impl<'a, T: Copy + PartialOrd + Debug + 'static> Iterable<T> for MergedColumnValues<'a, T> {
impl<'a, T: Copy + PartialOrd + Debug> Iterable<T> for MergedColumnValues<'a, T> {
fn boxed_iter(&self) -> Box<dyn Iterator<Item = T> + '_> {
match self.merge_row_order {
MergeRowOrder::Stack(_) => Box::new(

57
columnar/src/column_values/mod.rs
View File

@@ -2,7 +2,7 @@
//! # `fastfield_codecs`
//!
//! - Columnar storage of data for tantivy [`crate::Column`].
//! - Columnar storage of data for tantivy [`Column`].
//! - Encode data in different codecs.
//! - Monotonically map values to u64/u128
@@ -10,7 +10,6 @@ use std::fmt::Debug;
use std::ops::{Range, RangeInclusive};
use std::sync::Arc;
use downcast_rs::DowncastSync;
pub use monotonic_mapping::{MonotonicallyMappableToU64, StrictlyMonotonicFn};
pub use monotonic_mapping_u128::MonotonicallyMappableToU128;
@@ -26,10 +25,7 @@ mod monotonic_column;
pub(crate) use merge::MergedColumnValues;
pub use stats::ColumnStats;
pub use u128_based::{
open_u128_as_compact_u64, open_u128_mapped, serialize_column_values_u128,
CompactSpaceU64Accessor,
};
pub use u128_based::{open_u128_mapped, serialize_column_values_u128};
pub use u64_based::{
load_u64_based_column_values, serialize_and_load_u64_based_column_values,
serialize_u64_based_column_values, CodecType, ALL_U64_CODEC_TYPES,
@@ -45,7 +41,7 @@ use crate::RowId;
///
/// Any methods with a default and specialized implementation need to be called in the
/// wrappers that implement the trait: Arc and MonotonicMappingColumn
pub trait ColumnValues<T: PartialOrd = u64>: Send + Sync + DowncastSync {
pub trait ColumnValues<T: PartialOrd = u64>: Send + Sync {
/// Return the value associated with the given idx.
///
/// This accessor should return as fast as possible.
@@ -72,40 +68,11 @@ pub trait ColumnValues<T: PartialOrd = u64>: Send + Sync + DowncastSync {
out_x4[3] = self.get_val(idx_x4[3]);
}
let out_and_idx_chunks = output
.chunks_exact_mut(4)
.into_remainder()
.iter_mut()
.zip(indexes.chunks_exact(4).remainder());
for (out, idx) in out_and_idx_chunks {
*out = self.get_val(*idx);
}
}
let step_size = 4;
let cutoff = indexes.len() - indexes.len() % step_size;
/// Allows to push down multiple fetch calls, to avoid dynamic dispatch overhead.
/// The slightly weird `Option<T>` in output allows pushdown to full columns.
///
/// idx and output should have the same length
///
/// # Panics
///
/// May panic if `idx` is greater than the column length.
fn get_vals_opt(&self, indexes: &[u32], output: &mut [Option<T>]) {
assert!(indexes.len() == output.len());
let out_and_idx_chunks = output.chunks_exact_mut(4).zip(indexes.chunks_exact(4));
for (out_x4, idx_x4) in out_and_idx_chunks {
out_x4[0] = Some(self.get_val(idx_x4[0]));
out_x4[1] = Some(self.get_val(idx_x4[1]));
out_x4[2] = Some(self.get_val(idx_x4[2]));
out_x4[3] = Some(self.get_val(idx_x4[3]));
}
let out_and_idx_chunks = output
.chunks_exact_mut(4)
.into_remainder()
.iter_mut()
.zip(indexes.chunks_exact(4).remainder());
for (out, idx) in out_and_idx_chunks {
*out = Some(self.get_val(*idx));
for idx in cutoff..indexes.len() {
output[idx] = self.get_val(indexes[idx]);
}
}
@@ -134,7 +101,7 @@ pub trait ColumnValues<T: PartialOrd = u64>: Send + Sync + DowncastSync {
row_id_hits: &mut Vec<RowId>,
) {
let row_id_range = row_id_range.start..row_id_range.end.min(self.num_vals());
for idx in row_id_range {
for idx in row_id_range.start..row_id_range.end {
let val = self.get_val(idx);
if value_range.contains(&val) {
row_id_hits.push(idx);
@@ -172,7 +139,6 @@ pub trait ColumnValues<T: PartialOrd = u64>: Send + Sync + DowncastSync {
Box::new((0..self.num_vals()).map(|idx| self.get_val(idx)))
}
}
downcast_rs::impl_downcast!(sync ColumnValues<T> where T: PartialOrd);
/// Empty column of values.
pub struct EmptyColumnValues;
@@ -195,17 +161,12 @@ impl<T: PartialOrd + Default> ColumnValues<T> for EmptyColumnValues {
}
}
impl<T: Copy + PartialOrd + Debug + 'static> ColumnValues<T> for Arc<dyn ColumnValues<T>> {
impl<T: Copy + PartialOrd + Debug> ColumnValues<T> for Arc<dyn ColumnValues<T>> {
#[inline(always)]
fn get_val(&self, idx: u32) -> T {
self.as_ref().get_val(idx)
}
#[inline(always)]
fn get_vals_opt(&self, indexes: &[u32], output: &mut [Option<T>]) {
self.as_ref().get_vals_opt(indexes, output)
}
#[inline(always)]
fn min_value(&self) -> T {
self.as_ref().min_value()

18
columnar/src/column_values/monotonic_column.rs
View File

@@ -31,10 +31,10 @@ pub fn monotonic_map_column<C, T, Input, Output>(
monotonic_mapping: T,
) -> impl ColumnValues<Output>
where
C: ColumnValues<Input> + 'static,
T: StrictlyMonotonicFn<Input, Output> + Send + Sync + 'static,
Input: PartialOrd + Debug + Send + Sync + Clone + 'static,
Output: PartialOrd + Debug + Send + Sync + Clone + 'static,
C: ColumnValues<Input>,
T: StrictlyMonotonicFn<Input, Output> + Send + Sync,
Input: PartialOrd + Debug + Send + Sync + Clone,
Output: PartialOrd + Debug + Send + Sync + Clone,
{
MonotonicMappingColumn {
from_column,
@@ -45,10 +45,10 @@ where
impl<C, T, Input, Output> ColumnValues<Output> for MonotonicMappingColumn<C, T, Input>
where
C: ColumnValues<Input> + 'static,
T: StrictlyMonotonicFn<Input, Output> + Send + Sync + 'static,
Input: PartialOrd + Send + Debug + Sync + Clone + 'static,
Output: PartialOrd + Send + Debug + Sync + Clone + 'static,
C: ColumnValues<Input>,
T: StrictlyMonotonicFn<Input, Output> + Send + Sync,
Input: PartialOrd + Send + Debug + Sync + Clone,
Output: PartialOrd + Send + Debug + Sync + Clone,
{
#[inline(always)]
fn get_val(&self, idx: u32) -> Output {
@@ -107,7 +107,7 @@ mod tests {
#[test]
fn test_monotonic_mapping_iter() {
let vals: Vec<u64> = (0..100u64).map(|el| el * 10).collect();
let col = VecColumn::from(vals);
let col = VecColumn::from(&vals);
let mapped = monotonic_map_column(
col,
StrictlyMonotonicMappingInverter::from(StrictlyMonotonicMappingToInternal::<i64>::new()),

4
columnar/src/column_values/monotonic_mapping.rs
View File

@@ -139,12 +139,12 @@ impl MonotonicallyMappableToU64 for i64 {
impl MonotonicallyMappableToU64 for DateTime {
#[inline(always)]
fn to_u64(self) -> u64 {
common::i64_to_u64(self.into_timestamp_nanos())
common::i64_to_u64(self.into_timestamp_micros())
}
#[inline(always)]
fn from_u64(val: u64) -> Self {
DateTime::from_timestamp_nanos(common::u64_to_i64(val))
DateTime::from_timestamp_micros(common::u64_to_i64(val))
}
}

2
columnar/src/column_values/u128_based/compact_space/blank_range.rs
View File

@@ -38,6 +38,6 @@ impl Ord for BlankRange {
}
impl PartialOrd for BlankRange {
fn partial_cmp(&self, other: &Self) -> Option<std::cmp::Ordering> {
Some(self.cmp(other))
Some(self.blank_size().cmp(&other.blank_size()))
}
}

68
columnar/src/column_values/u128_based/compact_space/mod.rs
View File

@@ -22,7 +22,7 @@ mod build_compact_space;
use build_compact_space::get_compact_space;
use common::{BinarySerializable, CountingWriter, OwnedBytes, VInt, VIntU128};
use tantivy_bitpacker::{BitPacker, BitUnpacker};
use tantivy_bitpacker::{self, BitPacker, BitUnpacker};
use crate::column_values::ColumnValues;
use crate::RowId;
@@ -148,7 +148,7 @@ impl CompactSpace {
.binary_search_by_key(&compact, |range_mapping| range_mapping.compact_start)
// Correctness: Overflow. The first range starts at compact space 0, the error from
// binary search can never be 0
.unwrap_or_else(|e| e - 1);
.map_or_else(|e| e - 1, |v| v);
let range_mapping = &self.ranges_mapping[pos];
let diff = compact - range_mapping.compact_start;
@@ -292,63 +292,6 @@ impl BinarySerializable for IPCodecParams {
}
}
/// Exposes the compact space compressed values as u64.
///
/// This allows faster access to the values, as u64 is faster to work with than u128.
/// It also allows to handle u128 values like u64, via the `open_u64_lenient` as a uniform
/// access interface.
///
/// When converting from the internal u64 to u128 `compact_to_u128` can be used.
pub struct CompactSpaceU64Accessor(CompactSpaceDecompressor);
impl CompactSpaceU64Accessor {
pub(crate) fn open(data: OwnedBytes) -> io::Result<CompactSpaceU64Accessor> {
let decompressor = CompactSpaceU64Accessor(CompactSpaceDecompressor::open(data)?);
Ok(decompressor)
}
/// Convert a compact space value to u128
pub fn compact_to_u128(&self, compact: u32) -> u128 {
self.0.compact_to_u128(compact)
}
}
impl ColumnValues<u64> for CompactSpaceU64Accessor {
#[inline]
fn get_val(&self, doc: u32) -> u64 {
let compact = self.0.get_compact(doc);
compact as u64
}
fn min_value(&self) -> u64 {
self.0.u128_to_compact(self.0.min_value()).unwrap() as u64
}
fn max_value(&self) -> u64 {
self.0.u128_to_compact(self.0.max_value()).unwrap() as u64
}
fn num_vals(&self) -> u32 {
self.0.params.num_vals
}
#[inline]
fn iter(&self) -> Box<dyn Iterator<Item = u64> + '_> {
Box::new(self.0.iter_compact().map(|el| el as u64))
}
#[inline]
fn get_row_ids_for_value_range(
&self,
value_range: RangeInclusive<u64>,
position_range: Range<u32>,
positions: &mut Vec<u32>,
) {
let value_range = self.0.compact_to_u128(*value_range.start() as u32)
..=self.0.compact_to_u128(*value_range.end() as u32);
self.0
.get_row_ids_for_value_range(value_range, position_range, positions)
}
}
impl ColumnValues<u128> for CompactSpaceDecompressor {
#[inline]
fn get_val(&self, doc: u32) -> u128 {
@@ -459,14 +402,9 @@ impl CompactSpaceDecompressor {
.map(|compact| self.compact_to_u128(compact))
}
#[inline]
pub fn get_compact(&self, idx: u32) -> u32 {
self.params.bit_unpacker.get(idx, &self.data) as u32
}
#[inline]
pub fn get(&self, idx: u32) -> u128 {
let compact = self.get_compact(idx);
let compact = self.params.bit_unpacker.get(idx, &self.data) as u32;
self.compact_to_u128(compact)
}

21
columnar/src/column_values/u128_based/mod.rs
View File

@@ -6,9 +6,7 @@ use std::sync::Arc;
mod compact_space;
use common::{BinarySerializable, OwnedBytes, VInt};
pub use compact_space::{
CompactSpaceCompressor, CompactSpaceDecompressor, CompactSpaceU64Accessor,
};
use compact_space::{CompactSpaceCompressor, CompactSpaceDecompressor};
use crate::column_values::monotonic_map_column;
use crate::column_values::monotonic_mapping::{
@@ -110,23 +108,6 @@ pub fn open_u128_mapped<T: MonotonicallyMappableToU128 + Debug>(
StrictlyMonotonicMappingToInternal::<T>::new().into();
Ok(Arc::new(monotonic_map_column(reader, inverted)))
}
/// Returns the u64 representation of the u128 data.
/// The internal representation of the data as u64 is useful for faster processing.
///
/// In order to convert to u128 back cast to `CompactSpaceU64Accessor` and call
/// `compact_to_u128`.
///
/// # Notice
/// In case there are new codecs added, check for usages of `CompactSpaceDecompressorU64` and
/// also handle the new codecs.
pub fn open_u128_as_compact_u64(mut bytes: OwnedBytes) -> io::Result<Arc<dyn ColumnValues<u64>>> {
let header = U128Header::deserialize(&mut bytes)?;
assert_eq!(header.codec_type, U128FastFieldCodecType::CompactSpace);
let reader = CompactSpaceU64Accessor::open(bytes)?;
Ok(Arc::new(reader))
}
#[cfg(test)]
pub mod tests {
use super::*;

4
columnar/src/column_values/u64_based/bitpacked.rs
View File

@@ -63,6 +63,7 @@ impl ColumnValues for BitpackedReader {
fn get_val(&self, doc: u32) -> u64 {
self.stats.min_value + self.stats.gcd.get() * self.bit_unpacker.get(doc, &self.data)
}
#[inline]
fn min_value(&self) -> u64 {
self.stats.min_value
@@ -82,8 +83,7 @@ impl ColumnValues for BitpackedReader {
doc_id_range: Range<u32>,
positions: &mut Vec<u32>,
) {
let Some(transformed_range) =
transform_range_before_linear_transformation(&self.stats, range)
let Some(transformed_range) = transform_range_before_linear_transformation(&self.stats, range)
else {
positions.clear();
return;

7
columnar/src/column_values/u64_based/blockwise_linear.rs
View File

@@ -63,10 +63,7 @@ impl BlockwiseLinearEstimator {
if self.block.is_empty() {
return;
}
let column = VecColumn::from(std::mem::take(&mut self.block));
let line = Line::train(&column);
self.block = column.into();
let line = Line::train(&VecColumn::from(&self.block));
let mut max_value = 0u64;
for (i, buffer_val) in self.block.iter().enumerate() {
let interpolated_val = line.eval(i as u32);
@@ -128,7 +125,7 @@ impl ColumnCodecEstimator for BlockwiseLinearEstimator {
*buffer_val = gcd_divider.divide(*buffer_val - stats.min_value);
}
let line = Line::train(&VecColumn::from(buffer.to_vec()));
let line = Line::train(&VecColumn::from(&buffer));
assert!(!buffer.is_empty());

2
columnar/src/column_values/u64_based/line.rs
View File

@@ -184,7 +184,7 @@ mod tests {
}
fn test_eval_max_err(ys: &[u64]) -> Option<u64> {
let line = Line::train(&VecColumn::from(ys.to_vec()));
let line = Line::train(&VecColumn::from(&ys));
ys.iter()
.enumerate()
.map(|(x, y)| y.wrapping_sub(line.eval(x as u32)))

4
columnar/src/column_values/u64_based/linear.rs
View File

@@ -173,9 +173,7 @@ impl LinearCodecEstimator {
fn collect_before_line_estimation(&mut self, value: u64) {
self.block.push(value);
if self.block.len() == LINE_ESTIMATION_BLOCK_LEN {
let column = VecColumn::from(std::mem::take(&mut self.block));
let line = Line::train(&column);
self.block = column.into();
let line = Line::train(&VecColumn::from(&self.block));
let block = std::mem::take(&mut self.block);
for val in block {
self.collect_after_line_estimation(&line, val);

2
columnar/src/column_values/u64_based/stats_collector.rs
View File

@@ -27,7 +27,7 @@ pub struct StatsCollector {
// This is the same as computing the difference between the values and the first value.
//
// This way, we can compress i64-converted-to-u64 (e.g. timestamp that were supplied in
// seconds, only to be converted in nanoseconds).
// seconds, only to be converted in microseconds).
increment_gcd_opt: Option<(NonZeroU64, DividerU64)>,
first_value_opt: Option<u64>,
}

1
columnar/src/column_values/u64_based/tests.rs
View File

@@ -1,4 +1,5 @@
use proptest::prelude::*;
use proptest::strategy::Strategy;
use proptest::{prop_oneof, proptest};
#[test]

20
columnar/src/column_values/vec_column.rs
View File

@@ -4,14 +4,14 @@ use tantivy_bitpacker::minmax;
use crate::ColumnValues;
/// VecColumn provides `Column` over a `Vec<T>`.
pub struct VecColumn<T = u64> {
pub(crate) values: Vec<T>,
/// VecColumn provides `Column` over a slice.
pub struct VecColumn<'a, T = u64> {
pub(crate) values: &'a [T],
pub(crate) min_value: T,
pub(crate) max_value: T,
}
impl<T: Copy + PartialOrd + Send + Sync + Debug + 'static> ColumnValues<T> for VecColumn<T> {
impl<'a, T: Copy + PartialOrd + Send + Sync + Debug> ColumnValues<T> for VecColumn<'a, T> {
fn get_val(&self, position: u32) -> T {
self.values[position as usize]
}
@@ -37,8 +37,11 @@ impl<T: Copy + PartialOrd + Send + Sync + Debug + 'static> ColumnValues<T> for V
}
}
impl<T: Copy + PartialOrd + Default> From<Vec<T>> for VecColumn<T> {
fn from(values: Vec<T>) -> Self {
impl<'a, T: Copy + PartialOrd + Default, V> From<&'a V> for VecColumn<'a, T>
where V: AsRef<[T]> + ?Sized
{
fn from(values: &'a V) -> Self {
let values = values.as_ref();
let (min_value, max_value) = minmax(values.iter().copied()).unwrap_or_default();
Self {
values,
@@ -47,8 +50,3 @@ impl<T: Copy + PartialOrd + Default> From<Vec<T>> for VecColumn<T> {
}
}
}
impl From<VecColumn> for Vec<u64> {
fn from(column: VecColumn) -> Self {
column.values
}
}

5
columnar/src/columnar/column_type.rs
View File

@@ -34,7 +34,7 @@ impl fmt::Display for ColumnType {
ColumnType::IpAddr => "ip",
ColumnType::DateTime => "datetime",
};
write!(f, "{short_str}")
write!(f, "{}", short_str)
}
}
@@ -54,9 +54,6 @@ impl ColumnType {
pub fn to_code(self) -> u8 {
self as u8
}
pub fn is_date_time(&self) -> bool {
self == &ColumnType::DateTime
}
pub(crate) fn try_from_code(code: u8) -> Result<ColumnType, InvalidData> {
COLUMN_TYPES.get(code as usize).copied().ok_or(InvalidData)

34
columnar/src/columnar/merge/merge_dict_column.rs
View File

@@ -1,7 +1,7 @@
use std::io::{self, Write};
use common::{BitSet, CountingWriter, ReadOnlyBitSet};
use sstable::{SSTable, Streamer, TermOrdinal, VoidSSTable};
use sstable::{SSTable, TermOrdinal};
use super::term_merger::TermMerger;
use crate::column::serialize_column_mappable_to_u64;
@@ -52,23 +52,18 @@ impl<'a> Iterable for RemappedTermOrdinalsValues<'a> {
impl<'a> RemappedTermOrdinalsValues<'a> {
fn boxed_iter_stacked(&self) -> Box<dyn Iterator<Item = u64> + '_> {
let iter = self
.bytes_columns
.iter()
.enumerate()
.flat_map(|(seg_ord, bytes_column_opt)| {
let bytes_column = bytes_column_opt.as_ref()?;
Some((seg_ord, bytes_column))
})
.flat_map(move |(seg_ord, bytes_column)| {
let term_ord_after_merge_mapping =
self.term_ord_mapping.get_segment(seg_ord as u32);
let iter = self.bytes_columns.iter().flatten().enumerate().flat_map(
move |(seg_ord_with_column, bytes_column)| {
let term_ord_after_merge_mapping = self
.term_ord_mapping
.get_segment(seg_ord_with_column as u32);
bytes_column
.ords()
.values
.iter()
.map(move |term_ord| term_ord_after_merge_mapping[term_ord as usize])
});
},
);
Box::new(iter)
}
@@ -126,15 +121,10 @@ fn serialize_merged_dict(
let mut term_ord_mapping = TermOrdinalMapping::default();
let mut field_term_streams = Vec::new();
for column_opt in bytes_columns.iter() {
if let Some(column) = column_opt {
term_ord_mapping.add_segment(column.dictionary.num_terms());
let terms: Streamer<VoidSSTable> = column.dictionary.stream()?;
field_term_streams.push(terms);
} else {
term_ord_mapping.add_segment(0);
field_term_streams.push(Streamer::empty());
}
for column in bytes_columns.iter().flatten() {
term_ord_mapping.add_segment(column.dictionary.num_terms());
let terms = column.dictionary.stream()?;
field_term_streams.push(terms);
}
let mut merged_terms = TermMerger::new(field_term_streams);

15
columnar/src/columnar/merge/merge_mapping.rs
View File

@@ -11,17 +11,6 @@ pub struct StackMergeOrder {
}
impl StackMergeOrder {
#[cfg(test)]
pub fn stack_for_test(num_rows_per_columnar: &[u32]) -> StackMergeOrder {
let mut cumulated_row_ids: Vec<RowId> = Vec::with_capacity(num_rows_per_columnar.len());
let mut cumulated_row_id = 0;
for &num_rows in num_rows_per_columnar {
cumulated_row_id += num_rows;
cumulated_row_ids.push(cumulated_row_id);
}
StackMergeOrder { cumulated_row_ids }
}
pub fn stack(columnars: &[&ColumnarReader]) -> StackMergeOrder {
let mut cumulated_row_ids: Vec<RowId> = Vec::with_capacity(columnars.len());
let mut cumulated_row_id = 0;
@@ -52,8 +41,8 @@ pub enum MergeRowOrder {
/// Columnar tables are simply stacked one above the other.
/// If the i-th columnar_readers has n_rows_i rows, then
/// in the resulting columnar,
/// rows [r0..n_row_0) contains the row of `columnar_readers[0]`, in ordder
/// rows [n_row_0..n_row_0 + n_row_1 contains the row of `columnar_readers[1]`, in order.
/// rows [r0..n_row_0) contains the row of columnar_readers[0], in ordder
/// rows [n_row_0..n_row_0 + n_row_1 contains the row of columnar_readers[1], in order.
/// ..
/// No documents is deleted.
Stack(StackMergeOrder),

243
columnar/src/columnar/merge/mod.rs
View File

@@ -2,12 +2,11 @@ mod merge_dict_column;
mod merge_mapping;
mod term_merger;
use std::collections::{BTreeMap, HashSet};
use std::collections::{BTreeMap, HashMap, HashSet};
use std::io;
use std::net::Ipv6Addr;
use std::sync::Arc;
use itertools::Itertools;
pub use merge_mapping::{MergeRowOrder, ShuffleMergeOrder, StackMergeOrder};
use super::writer::ColumnarSerializer;
@@ -18,8 +17,7 @@ use crate::columnar::writer::CompatibleNumericalTypes;
use crate::columnar::ColumnarReader;
use crate::dynamic_column::DynamicColumn;
use crate::{
BytesColumn, Column, ColumnIndex, ColumnType, ColumnValues, DynamicColumnHandle, NumericalType,
NumericalValue,
BytesColumn, Column, ColumnIndex, ColumnType, ColumnValues, NumericalType, NumericalValue,
};
/// Column types are grouped into different categories.
@@ -29,16 +27,14 @@ use crate::{
/// In practise, today, only Numerical colummns are coerced into one type today.
///
/// See also [README.md].
///
/// The ordering has to match the ordering of the variants in [ColumnType].
#[derive(Copy, Clone, Eq, PartialOrd, Ord, PartialEq, Hash, Debug)]
#[derive(Copy, Clone, Eq, PartialEq, Hash, Debug)]
pub(crate) enum ColumnTypeCategory {
Numerical,
Bytes,
Str,
Bool,
IpAddr,
Str,
Numerical,
DateTime,
Bytes,
IpAddr,
}
impl From<ColumnType> for ColumnTypeCategory {
@@ -86,22 +82,10 @@ pub fn merge_columnar(
.iter()
.map(|reader| reader.num_rows())
.collect::<Vec<u32>>();
let columns_to_merge =
group_columns_for_merge(columnar_readers, required_columns, &merge_row_order)?;
for res in columns_to_merge {
let ((column_name, _column_type_category), grouped_columns) = res;
let grouped_columns = grouped_columns.open(&merge_row_order)?;
if grouped_columns.is_empty() {
continue;
}
let column_type = grouped_columns.column_type_after_merge();
let mut columns = grouped_columns.columns;
coerce_columns(column_type, &mut columns)?;
let columns_to_merge = group_columns_for_merge(columnar_readers, required_columns)?;
for ((column_name, column_type), columns) in columns_to_merge {
let mut column_serializer =
serializer.start_serialize_column(column_name.as_bytes(), column_type);
serializer.serialize_column(column_name.as_bytes(), column_type);
merge_column(
column_type,
&num_rows_per_columnar,
@@ -109,9 +93,7 @@ pub fn merge_columnar(
&merge_row_order,
&mut column_serializer,
)?;
column_serializer.finalize()?;
}
serializer.finalize(merge_row_order.num_rows())?;
Ok(())
}
@@ -225,12 +207,40 @@ fn merge_column(
struct GroupedColumns {
required_column_type: Option<ColumnType>,
columns: Vec<Option<DynamicColumn>>,
column_category: ColumnTypeCategory,
}
impl GroupedColumns {
/// Check is column group can be skipped during serialization.
fn is_empty(&self) -> bool {
self.required_column_type.is_none() && self.columns.iter().all(Option::is_none)
fn for_category(column_category: ColumnTypeCategory, num_columnars: usize) -> Self {
GroupedColumns {
required_column_type: None,
columns: vec![None; num_columnars],
column_category,
}
}
/// Set the dynamic column for a given columnar.
fn set_column(&mut self, columnar_id: usize, column: DynamicColumn) {
self.columns[columnar_id] = Some(column);
}
/// Force the existence of a column, as well as its type.
fn require_type(&mut self, required_type: ColumnType) -> io::Result<()> {
if let Some(existing_required_type) = self.required_column_type {
if existing_required_type == required_type {
// This was just a duplicate in the `required_columns`.
// Nothing to do.
return Ok(());
} else {
return Err(io::Error::new(
io::ErrorKind::InvalidInput,
"Required column conflicts with another required column of the same type \
category.",
));
}
}
self.required_column_type = Some(required_type);
Ok(())
}
/// Returns the column type after merge.
@@ -252,76 +262,11 @@ impl GroupedColumns {
}
// At the moment, only the numerical categorical column type has more than one possible
// column type.
assert!(self
.columns
.iter()
.flatten()
.all(|el| ColumnTypeCategory::from(el.column_type()) == ColumnTypeCategory::Numerical));
assert_eq!(self.column_category, ColumnTypeCategory::Numerical);
merged_numerical_columns_type(self.columns.iter().flatten()).into()
}
}
struct GroupedColumnsHandle {
required_column_type: Option<ColumnType>,
columns: Vec<Option<DynamicColumnHandle>>,
}
impl GroupedColumnsHandle {
fn new(num_columnars: usize) -> Self {
GroupedColumnsHandle {
required_column_type: None,
columns: vec![None; num_columnars],
}
}
fn open(self, merge_row_order: &MergeRowOrder) -> io::Result<GroupedColumns> {
let mut columns: Vec<Option<DynamicColumn>> = Vec::new();
for (columnar_id, column) in self.columns.iter().enumerate() {
if let Some(column) = column {
let column = column.open()?;
// We skip columns that end up with 0 documents.
// That way, we make sure they don't end up influencing the merge type or
// creating empty columns.
if is_empty_after_merge(merge_row_order, &column, columnar_id) {
columns.push(None);
} else {
columns.push(Some(column));
}
} else {
columns.push(None);
}
}
Ok(GroupedColumns {
required_column_type: self.required_column_type,
columns,
})
}
/// Set the dynamic column for a given columnar.
fn set_column(&mut self, columnar_id: usize, column: DynamicColumnHandle) {
self.columns[columnar_id] = Some(column);
}
/// Force the existence of a column, as well as its type.
fn require_type(&mut self, required_type: ColumnType) -> io::Result<()> {
if let Some(existing_required_type) = self.required_column_type {
if existing_required_type == required_type {
// This was just a duplicate in the `required_columns`.
// Nothing to do.
return Ok(());
} else {
return Err(io::Error::new(
io::ErrorKind::InvalidInput,
"Required column conflicts with another required column of the same type \
category.",
));
}
}
self.required_column_type = Some(required_type);
Ok(())
}
}
/// Returns the type of the merged numerical column.
///
/// This function picks the first numerical type out of i64, u64, f64 (order matters
@@ -342,92 +287,48 @@ fn merged_numerical_columns_type<'a>(
compatible_numerical_types.to_numerical_type()
}
fn is_empty_after_merge(
merge_row_order: &MergeRowOrder,
column: &DynamicColumn,
columnar_ord: usize,
) -> bool {
if column.num_values() == 0u32 {
// It was empty before the merge.
return true;
}
match merge_row_order {
MergeRowOrder::Stack(_) => {
// If we are stacking the columnar, no rows are being deleted.
false
}
MergeRowOrder::Shuffled(shuffled) => {
if let Some(alive_bitset) = &shuffled.alive_bitsets[columnar_ord] {
let column_index = column.column_index();
match column_index {
ColumnIndex::Empty { .. } => true,
ColumnIndex::Full => alive_bitset.len() == 0,
ColumnIndex::Optional(optional_index) => {
for doc in optional_index.iter_rows() {
if alive_bitset.contains(doc) {
return false;
}
}
true
}
ColumnIndex::Multivalued(multivalued_index) => {
for (doc_id, (start_index, end_index)) in multivalued_index
.start_index_column
.iter()
.tuple_windows()
.enumerate()
{
let doc_id = doc_id as u32;
if start_index == end_index {
// There are no values in this document
continue;
}
// The document contains values and is present in the alive bitset.
// The column is therefore not empty.
if alive_bitset.contains(doc_id) {
return false;
}
}
true
}
}
} else {
// No document is being deleted.
// The shuffle is applying a permutation.
false
}
}
}
}
/// Iterates over the columns of the columnar readers, grouped by column name.
/// Key functionality is that `open` of the Columns is done lazy per group.
fn group_columns_for_merge<'a>(
columnar_readers: &'a [&'a ColumnarReader],
required_columns: &'a [(String, ColumnType)],
_merge_row_order: &'a MergeRowOrder,
) -> io::Result<BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle>> {
let mut columns: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> = BTreeMap::new();
#[allow(clippy::type_complexity)]
fn group_columns_for_merge(
columnar_readers: &[&ColumnarReader],
required_columns: &[(String, ColumnType)],
) -> io::Result<BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>>> {
// Each column name may have multiple types of column associated.
// For merging we are interested in the same column type category since they can be merged.
let mut columns_grouped: HashMap<(String, ColumnTypeCategory), GroupedColumns> = HashMap::new();
for &(ref column_name, column_type) in required_columns {
columns
columns_grouped
.entry((column_name.clone(), column_type.into()))
.or_insert_with(|| GroupedColumnsHandle::new(columnar_readers.len()))
.or_insert_with(|| {
GroupedColumns::for_category(column_type.into(), columnar_readers.len())
})
.require_type(column_type)?;
}
for (columnar_id, columnar_reader) in columnar_readers.iter().enumerate() {
let column_name_and_handle = columnar_reader.iter_columns()?;
let column_name_and_handle = columnar_reader.list_columns()?;
for (column_name, handle) in column_name_and_handle {
let column_category: ColumnTypeCategory = handle.column_type().into();
columns
let column = handle.open()?;
columns_grouped
.entry((column_name, column_category))
.or_insert_with(|| GroupedColumnsHandle::new(columnar_readers.len()))
.set_column(columnar_id, handle);
.or_insert_with(|| {
GroupedColumns::for_category(column_category, columnar_readers.len())
})
.set_column(columnar_id, column);
}
}
Ok(columns)
let mut merge_columns: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
Default::default();
for ((column_name, _), mut grouped_columns) in columns_grouped {
let column_type = grouped_columns.column_type_after_merge();
coerce_columns(column_type, &mut grouped_columns.columns)?;
merge_columns.insert((column_name, column_type), grouped_columns.columns);
}
Ok(merge_columns)
}
fn coerce_columns(

132
columnar/src/columnar/merge/tests.rs
View File

@@ -1,3 +1,5 @@
use itertools::Itertools;
use super::*;
use crate::{Cardinality, ColumnarWriter, HasAssociatedColumnType, RowId};
@@ -23,73 +25,70 @@ fn test_column_coercion_to_u64() {
let columnar1 = make_columnar("numbers", &[1i64]);
// u64 type
let columnar2 = make_columnar("numbers", &[u64::MAX]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(columnars, &[], &merge_order).unwrap();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(&[&columnar1, &columnar2], &[]).unwrap();
assert_eq!(column_map.len(), 1);
assert!(column_map.contains_key(&("numbers".to_string(), ColumnTypeCategory::Numerical)));
assert!(column_map.contains_key(&("numbers".to_string(), ColumnType::U64)));
}
#[test]
fn test_column_no_coercion_if_all_the_same() {
let columnar1 = make_columnar("numbers", &[1u64]);
let columnar2 = make_columnar("numbers", &[2u64]);
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(&[&columnar1, &columnar2], &[]).unwrap();
assert_eq!(column_map.len(), 1);
assert!(column_map.contains_key(&("numbers".to_string(), ColumnType::U64)));
}
#[test]
fn test_column_coercion_to_i64() {
let columnar1 = make_columnar("numbers", &[-1i64]);
let columnar2 = make_columnar("numbers", &[2u64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(columnars, &[], &merge_order).unwrap();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(&[&columnar1, &columnar2], &[]).unwrap();
assert_eq!(column_map.len(), 1);
assert!(column_map.contains_key(&("numbers".to_string(), ColumnTypeCategory::Numerical)));
assert!(column_map.contains_key(&("numbers".to_string(), ColumnType::I64)));
}
//#[test]
// fn test_impossible_coercion_returns_an_error() {
// let columnar1 = make_columnar("numbers", &[u64::MAX]);
// let merge_order = StackMergeOrder::stack(&[&columnar1]).into();
// let group_error = group_columns_for_merge_iter(
//&[&columnar1],
//&[("numbers".to_string(), ColumnType::I64)],
//&merge_order,
//)
//.unwrap_err();
// assert_eq!(group_error.kind(), io::ErrorKind::InvalidInput);
//}
#[test]
fn test_impossible_coercion_returns_an_error() {
let columnar1 = make_columnar("numbers", &[u64::MAX]);
let group_error =
group_columns_for_merge(&[&columnar1], &[("numbers".to_string(), ColumnType::I64)])
.map(|_| ())
.unwrap_err();
assert_eq!(group_error.kind(), io::ErrorKind::InvalidInput);
}
#[test]
fn test_group_columns_with_required_column() {
let columnar1 = make_columnar("numbers", &[1i64]);
let columnar2 = make_columnar("numbers", &[2u64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(
&[&columnar1, &columnar2],
&[("numbers".to_string(), ColumnType::U64)],
&merge_order,
)
.unwrap();
assert_eq!(column_map.len(), 1);
assert!(column_map.contains_key(&("numbers".to_string(), ColumnTypeCategory::Numerical)));
assert!(column_map.contains_key(&("numbers".to_string(), ColumnType::U64)));
}
#[test]
fn test_group_columns_required_column_with_no_existing_columns() {
let columnar1 = make_columnar("numbers", &[2u64]);
let columnar2 = make_columnar("numbers", &[2u64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<_, _> = group_columns_for_merge(
columnars,
&[("required_col".to_string(), ColumnType::Str)],
&merge_order,
)
.unwrap();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(
&[&columnar1, &columnar2],
&[("required_col".to_string(), ColumnType::Str)],
)
.unwrap();
assert_eq!(column_map.len(), 2);
let columns = &column_map
.get(&("required_col".to_string(), ColumnTypeCategory::Str))
.unwrap()
.columns;
let columns = column_map
.get(&("required_col".to_string(), ColumnType::Str))
.unwrap();
assert_eq!(columns.len(), 2);
assert!(columns[0].is_none());
assert!(columns[1].is_none());
@@ -99,42 +98,35 @@ fn test_group_columns_required_column_with_no_existing_columns() {
fn test_group_columns_required_column_is_above_all_columns_have_the_same_type_rule() {
let columnar1 = make_columnar("numbers", &[2i64]);
let columnar2 = make_columnar("numbers", &[2i64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(
columnars,
&[&columnar1, &columnar2],
&[("numbers".to_string(), ColumnType::U64)],
&merge_order,
)
.unwrap();
assert_eq!(column_map.len(), 1);
assert!(column_map.contains_key(&("numbers".to_string(), ColumnTypeCategory::Numerical)));
assert!(column_map.contains_key(&("numbers".to_string(), ColumnType::U64)));
}
#[test]
fn test_missing_column() {
let columnar1 = make_columnar("numbers", &[-1i64]);
let columnar2 = make_columnar("numbers2", &[2u64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(columnars, &[], &merge_order).unwrap();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(&[&columnar1, &columnar2], &[]).unwrap();
assert_eq!(column_map.len(), 2);
assert!(column_map.contains_key(&("numbers".to_string(), ColumnTypeCategory::Numerical)));
assert!(column_map.contains_key(&("numbers".to_string(), ColumnType::I64)));
{
let columns = &column_map
.get(&("numbers".to_string(), ColumnTypeCategory::Numerical))
.unwrap()
.columns;
let columns = column_map
.get(&("numbers".to_string(), ColumnType::I64))
.unwrap();
assert!(columns[0].is_some());
assert!(columns[1].is_none());
}
{
let columns = &column_map
.get(&("numbers2".to_string(), ColumnTypeCategory::Numerical))
.unwrap()
.columns;
let columns = column_map
.get(&("numbers2".to_string(), ColumnType::U64))
.unwrap();
assert!(columns[0].is_none());
assert!(columns[1].is_some());
}
@@ -232,9 +224,7 @@ fn test_merge_columnar_numbers() {
assert_eq!(columnar_reader.num_columns(), 1);
let cols = columnar_reader.read_columns("numbers").unwrap();
let dynamic_column = cols[0].open().unwrap();
let DynamicColumn::F64(vals) = dynamic_column else {
panic!()
};
let DynamicColumn::F64(vals) = dynamic_column else { panic!() };
assert_eq!(vals.get_cardinality(), Cardinality::Optional);
assert_eq!(vals.first(0u32), Some(-1f64));
assert_eq!(vals.first(1u32), None);
@@ -260,9 +250,7 @@ fn test_merge_columnar_texts() {
assert_eq!(columnar_reader.num_columns(), 1);
let cols = columnar_reader.read_columns("texts").unwrap();
let dynamic_column = cols[0].open().unwrap();
let DynamicColumn::Str(vals) = dynamic_column else {
panic!()
};
let DynamicColumn::Str(vals) = dynamic_column else { panic!() };
assert_eq!(vals.ords().get_cardinality(), Cardinality::Optional);
let get_str_for_ord = |ord| {
@@ -309,9 +297,7 @@ fn test_merge_columnar_byte() {
assert_eq!(columnar_reader.num_columns(), 1);
let cols = columnar_reader.read_columns("bytes").unwrap();
let dynamic_column = cols[0].open().unwrap();
let DynamicColumn::Bytes(vals) = dynamic_column else {
panic!()
};
let DynamicColumn::Bytes(vals) = dynamic_column else { panic!() };
let get_bytes_for_ord = |ord| {
let mut out = Vec::new();
vals.ord_to_bytes(ord, &mut out).unwrap();
@@ -365,9 +351,7 @@ fn test_merge_columnar_byte_with_missing() {
assert_eq!(columnar_reader.num_columns(), 2);
let cols = columnar_reader.read_columns("col").unwrap();
let dynamic_column = cols[0].open().unwrap();
let DynamicColumn::Bytes(vals) = dynamic_column else {
panic!()
};
let DynamicColumn::Bytes(vals) = dynamic_column else { panic!() };
let get_bytes_for_ord = |ord| {
let mut out = Vec::new();
vals.ord_to_bytes(ord, &mut out).unwrap();
@@ -419,9 +403,7 @@ fn test_merge_columnar_different_types() {
// numeric column
let dynamic_column = cols[0].open().unwrap();
let DynamicColumn::I64(vals) = dynamic_column else {
panic!()
};
let DynamicColumn::I64(vals) = dynamic_column else { panic!() };
assert_eq!(vals.get_cardinality(), Cardinality::Optional);
assert_eq!(vals.values_for_doc(0).collect_vec(), vec![]);
assert_eq!(vals.values_for_doc(1).collect_vec(), vec![]);
@@ -431,9 +413,7 @@ fn test_merge_columnar_different_types() {
// text column
let dynamic_column = cols[1].open().unwrap();
let DynamicColumn::Str(vals) = dynamic_column else {
panic!()
};
let DynamicColumn::Str(vals) = dynamic_column else { panic!() };
assert_eq!(vals.ords().get_cardinality(), Cardinality::Optional);
let get_str_for_ord = |ord| {
let mut out = String::new();

53
columnar/src/columnar/reader/mod.rs
View File

@@ -102,41 +102,30 @@ impl ColumnarReader {
pub fn num_rows(&self) -> RowId {
self.num_rows
}
// Iterate over the columns in a sorted way
pub fn iter_columns(
&self,
) -> io::Result<impl Iterator<Item = (String, DynamicColumnHandle)> + '_> {
let mut stream = self.column_dictionary.stream()?;
Ok(std::iter::from_fn(move || {
if stream.advance() {
let key_bytes: &[u8] = stream.key();
let column_code: u8 = key_bytes.last().cloned().unwrap();
// TODO Error Handling. The API gets quite ugly when returning the error here, so
// instead we could just check the first N columns upfront.
let column_type: ColumnType = ColumnType::try_from_code(column_code)
.map_err(|_| io_invalid_data(format!("Unknown column code `{column_code}`")))
.unwrap();
let range = stream.value().clone();
let column_name =
// The last two bytes are respectively the 0u8 separator and the column_type.
String::from_utf8_lossy(&key_bytes[..key_bytes.len() - 2]).to_string();
let file_slice = self
.column_data
.slice(range.start as usize..range.end as usize);
let column_handle = DynamicColumnHandle {
file_slice,
column_type,
};
Some((column_name, column_handle))
} else {
None
}
}))
}
// TODO Add unit tests
pub fn list_columns(&self) -> io::Result<Vec<(String, DynamicColumnHandle)>> {
Ok(self.iter_columns()?.collect())
let mut stream = self.column_dictionary.stream()?;
let mut results = Vec::new();
while stream.advance() {
let key_bytes: &[u8] = stream.key();
let column_code: u8 = key_bytes.last().cloned().unwrap();
let column_type: ColumnType = ColumnType::try_from_code(column_code)
.map_err(|_| io_invalid_data(format!("Unknown column code `{column_code}`")))?;
let range = stream.value().clone();
let column_name =
// The last two bytes are respectively the 0u8 separator and the column_type.
String::from_utf8_lossy(&key_bytes[..key_bytes.len() - 2]).to_string();
let file_slice = self
.column_data
.slice(range.start as usize..range.end as usize);
let column_handle = DynamicColumnHandle {
file_slice,
column_type,
};
results.push((column_name, column_handle));
}
Ok(results)
}
fn stream_for_column_range(&self, column_name: &str) -> sstable::StreamerBuilder<RangeSSTable> {

3
columnar/src/columnar/writer/column_writers.rs
View File

@@ -269,8 +269,7 @@ impl StrOrBytesColumnWriter {
dictionaries: &mut [DictionaryBuilder],
arena: &mut MemoryArena,
) {
let unordered_id =
dictionaries[self.dictionary_id as usize].get_or_allocate_id(bytes, arena);
let unordered_id = dictionaries[self.dictionary_id as usize].get_or_allocate_id(bytes);
self.column_writer.record(doc, unordered_id, arena);
}

67
columnar/src/columnar/writer/mod.rs
View File

@@ -13,7 +13,9 @@ pub(crate) use serializer::ColumnarSerializer;
use stacker::{Addr, ArenaHashMap, MemoryArena};
use crate::column_index::SerializableColumnIndex;
use crate::column_values::{MonotonicallyMappableToU128, MonotonicallyMappableToU64};
use crate::column_values::{
ColumnValues, MonotonicallyMappableToU128, MonotonicallyMappableToU64, VecColumn,
};
use crate::columnar::column_type::ColumnType;
use crate::columnar::writer::column_writers::{
ColumnWriter, NumericalColumnWriter, StrOrBytesColumnWriter,
@@ -77,6 +79,7 @@ fn mutate_or_create_column<V, TMutator>(
impl ColumnarWriter {
pub fn mem_usage(&self) -> usize {
// TODO add dictionary builders.
self.arena.mem_usage()
+ self.numerical_field_hash_map.mem_usage()
+ self.bool_field_hash_map.mem_usage()
@@ -84,11 +87,6 @@ impl ColumnarWriter {
+ self.str_field_hash_map.mem_usage()
+ self.ip_addr_field_hash_map.mem_usage()
+ self.datetime_field_hash_map.mem_usage()
+ self
.dictionaries
.iter()
.map(|dict| dict.mem_usage())
.sum::<usize>()
}
/// Returns the list of doc ids from 0..num_docs sorted by the `sort_field`
@@ -100,15 +98,9 @@ impl ColumnarWriter {
///
/// The sort applied is stable.
pub fn sort_order(&self, sort_field: &str, num_docs: RowId, reversed: bool) -> Vec<u32> {
let Some(numerical_col_writer) = self
.numerical_field_hash_map
.get::<NumericalColumnWriter>(sort_field.as_bytes())
.or_else(|| {
self.datetime_field_hash_map
.get::<NumericalColumnWriter>(sort_field.as_bytes())
})
else {
return Vec::new();
let Some(numerical_col_writer) =
self.numerical_field_hash_map.get::<NumericalColumnWriter>(sort_field.as_bytes()) else {
return Vec::new();
};
let mut symbols_buffer = Vec::new();
let mut values = Vec::new();
@@ -274,7 +266,7 @@ impl ColumnarWriter {
let mut column: ColumnWriter = column_opt.unwrap_or_default();
column.record(
doc,
NumericalValue::I64(datetime.into_timestamp_nanos()),
NumericalValue::I64(datetime.into_timestamp_micros()),
arena,
);
column
@@ -336,7 +328,7 @@ impl ColumnarWriter {
let mut columns: Vec<(&[u8], ColumnType, Addr)> = self
.numerical_field_hash_map
.iter()
.map(|(column_name, addr)| {
.map(|(column_name, addr, _)| {
let numerical_column_writer: NumericalColumnWriter =
self.numerical_field_hash_map.read(addr);
let column_type = numerical_column_writer.numerical_type().into();
@@ -346,27 +338,27 @@ impl ColumnarWriter {
columns.extend(
self.bytes_field_hash_map
.iter()
.map(|(term, addr)| (term, ColumnType::Bytes, addr)),
.map(|(term, addr, _)| (term, ColumnType::Bytes, addr)),
);
columns.extend(
self.str_field_hash_map
.iter()
.map(|(column_name, addr)| (column_name, ColumnType::Str, addr)),
.map(|(column_name, addr, _)| (column_name, ColumnType::Str, addr)),
);
columns.extend(
self.bool_field_hash_map
.iter()
.map(|(column_name, addr)| (column_name, ColumnType::Bool, addr)),
.map(|(column_name, addr, _)| (column_name, ColumnType::Bool, addr)),
);
columns.extend(
self.ip_addr_field_hash_map
.iter()
.map(|(column_name, addr)| (column_name, ColumnType::IpAddr, addr)),
.map(|(column_name, addr, _)| (column_name, ColumnType::IpAddr, addr)),
);
columns.extend(
self.datetime_field_hash_map
.iter()
.map(|(column_name, addr)| (column_name, ColumnType::DateTime, addr)),
.map(|(column_name, addr, _)| (column_name, ColumnType::DateTime, addr)),
);
columns.sort_unstable_by_key(|(column_name, col_type, _)| (*column_name, *col_type));
@@ -378,7 +370,7 @@ impl ColumnarWriter {
let column_writer: ColumnWriter = self.bool_field_hash_map.read(addr);
let cardinality = column_writer.get_cardinality(num_docs);
let mut column_serializer =
serializer.start_serialize_column(column_name, column_type);
serializer.serialize_column(column_name, column_type);
serialize_bool_column(
cardinality,
num_docs,
@@ -390,13 +382,12 @@ impl ColumnarWriter {
buffers,
&mut column_serializer,
)?;
column_serializer.finalize()?;
}
ColumnType::IpAddr => {
let column_writer: ColumnWriter = self.ip_addr_field_hash_map.read(addr);
let cardinality = column_writer.get_cardinality(num_docs);
let mut column_serializer =
serializer.start_serialize_column(column_name, ColumnType::IpAddr);
serializer.serialize_column(column_name, ColumnType::IpAddr);
serialize_ip_addr_column(
cardinality,
num_docs,
@@ -408,7 +399,6 @@ impl ColumnarWriter {
buffers,
&mut column_serializer,
)?;
column_serializer.finalize()?;
}
ColumnType::Bytes | ColumnType::Str => {
let str_or_bytes_column_writer: StrOrBytesColumnWriter =
@@ -423,7 +413,7 @@ impl ColumnarWriter {
.column_writer
.get_cardinality(num_docs);
let mut column_serializer =
serializer.start_serialize_column(column_name, column_type);
serializer.serialize_column(column_name, column_type);
serialize_bytes_or_str_column(
cardinality,
num_docs,
@@ -435,17 +425,15 @@ impl ColumnarWriter {
&mut symbol_byte_buffer,
),
buffers,
&self.arena,
&mut column_serializer,
)?;
column_serializer.finalize()?;
}
ColumnType::F64 | ColumnType::I64 | ColumnType::U64 => {
let numerical_column_writer: NumericalColumnWriter =
self.numerical_field_hash_map.read(addr);
let cardinality = numerical_column_writer.cardinality(num_docs);
let mut column_serializer =
serializer.start_serialize_column(column_name, column_type);
serializer.serialize_column(column_name, column_type);
let numerical_type = column_type.numerical_type().unwrap();
serialize_numerical_column(
cardinality,
@@ -459,13 +447,12 @@ impl ColumnarWriter {
buffers,
&mut column_serializer,
)?;
column_serializer.finalize()?;
}
ColumnType::DateTime => {
let column_writer: ColumnWriter = self.datetime_field_hash_map.read(addr);
let cardinality = column_writer.get_cardinality(num_docs);
let mut column_serializer =
serializer.start_serialize_column(column_name, ColumnType::DateTime);
serializer.serialize_column(column_name, ColumnType::DateTime);
serialize_numerical_column(
cardinality,
num_docs,
@@ -478,7 +465,6 @@ impl ColumnarWriter {
buffers,
&mut column_serializer,
)?;
column_serializer.finalize()?;
}
};
}
@@ -489,7 +475,6 @@ impl ColumnarWriter {
// Serialize [Dictionary, Column, dictionary num bytes U32::LE]
// Column: [Column Index, Column Values, column index num bytes U32::LE]
#[allow(clippy::too_many_arguments)]
fn serialize_bytes_or_str_column(
cardinality: Cardinality,
num_docs: RowId,
@@ -497,7 +482,6 @@ fn serialize_bytes_or_str_column(
dictionary_builder: &DictionaryBuilder,
operation_it: impl Iterator<Item = ColumnOperation<UnorderedId>>,
buffers: &mut SpareBuffers,
arena: &MemoryArena,
wrt: impl io::Write,
) -> io::Result<()> {
let SpareBuffers {
@@ -506,8 +490,7 @@ fn serialize_bytes_or_str_column(
..
} = buffers;
let mut counting_writer = CountingWriter::wrap(wrt);
let term_id_mapping: TermIdMapping =
dictionary_builder.serialize(arena, &mut counting_writer)?;
let term_id_mapping: TermIdMapping = dictionary_builder.serialize(&mut counting_writer)?;
let dictionary_num_bytes: u32 = counting_writer.written_bytes() as u32;
let mut wrt = counting_writer.finish();
let operation_iterator = operation_it.map(|symbol: ColumnOperation<UnorderedId>| {
@@ -643,7 +626,10 @@ fn send_to_serialize_column_mappable_to_u128<
value_index_builders: &mut PreallocatedIndexBuilders,
values: &mut Vec<T>,
mut wrt: impl io::Write,
) -> io::Result<()> {
) -> io::Result<()>
where
for<'a> VecColumn<'a, T>: ColumnValues<T>,
{
values.clear();
// TODO: split index and values
let serializable_column_index = match cardinality {
@@ -696,7 +682,10 @@ fn send_to_serialize_column_mappable_to_u64(
value_index_builders: &mut PreallocatedIndexBuilders,
values: &mut Vec<u64>,
mut wrt: impl io::Write,
) -> io::Result<()> {
) -> io::Result<()>
where
for<'a> VecColumn<'a, u64>: ColumnValues<u64>,
{
values.clear();
let serializable_column_index = match cardinality {
Cardinality::Full => {

26
columnar/src/columnar/writer/serializer.rs
View File

@@ -18,12 +18,7 @@ pub struct ColumnarSerializer<W: io::Write> {
/// code.
fn prepare_key(key: &[u8], column_type: ColumnType, buffer: &mut Vec<u8>) {
buffer.clear();
// Convert 0 bytes to '0' string, as 0 bytes are reserved for the end of the path.
if key.contains(&0u8) {
buffer.extend(key.iter().map(|&b| if b == 0 { b'0' } else { b }));
} else {
buffer.extend_from_slice(key);
}
buffer.extend_from_slice(key);
buffer.push(0u8);
buffer.push(column_type.to_code());
}
@@ -39,12 +34,11 @@ impl<W: io::Write> ColumnarSerializer<W> {
}
}
/// Creates a ColumnSerializer.
pub fn start_serialize_column<'a>(
pub fn serialize_column<'a>(
&'a mut self,
column_name: &[u8],
column_type: ColumnType,
) -> ColumnSerializer<'a, W> {
) -> impl io::Write + 'a {
let start_offset = self.wrt.written_bytes();
prepare_key(column_name, column_type, &mut self.prepare_key_buffer);
ColumnSerializer {
@@ -66,21 +60,20 @@ impl<W: io::Write> ColumnarSerializer<W> {
}
}
pub struct ColumnSerializer<'a, W: io::Write> {
struct ColumnSerializer<'a, W: io::Write> {
columnar_serializer: &'a mut ColumnarSerializer<W>,
start_offset: u64,
}
impl<'a, W: io::Write> ColumnSerializer<'a, W> {
pub fn finalize(self) -> io::Result<()> {
impl<'a, W: io::Write> Drop for ColumnSerializer<'a, W> {
fn drop(&mut self) {
let end_offset: u64 = self.columnar_serializer.wrt.written_bytes();
let byte_range = self.start_offset..end_offset;
self.columnar_serializer.sstable_range.insert(
self.columnar_serializer.sstable_range.insert_cannot_fail(
&self.columnar_serializer.prepare_key_buffer[..],
&byte_range,
)?;
);
self.columnar_serializer.prepare_key_buffer.clear();
Ok(())
}
}
@@ -101,13 +94,14 @@ impl<'a, W: io::Write> io::Write for ColumnSerializer<'a, W> {
#[cfg(test)]
mod tests {
use super::*;
use crate::columnar::column_type::ColumnType;
#[test]
fn test_prepare_key_bytes() {
let mut buffer: Vec<u8> = b"somegarbage".to_vec();
prepare_key(b"root\0child", ColumnType::Str, &mut buffer);
assert_eq!(buffer.len(), 12);
assert_eq!(&buffer[..10], b"root0child");
assert_eq!(&buffer[..10], b"root\0child");
assert_eq!(buffer[10], 0u8);
assert_eq!(buffer[11], ColumnType::Str.to_code());
}

49
columnar/src/dictionary.rs
View File

@@ -1,7 +1,7 @@
use std::io;
use fnv::FnvHashMap;
use sstable::SSTable;
use stacker::{MemoryArena, SharedArenaHashMap};
pub(crate) struct TermIdMapping {
unordered_to_ord: Vec<OrderedId>,
@@ -31,38 +31,26 @@ pub struct OrderedId(pub u32);
/// mapping.
#[derive(Default)]
pub(crate) struct DictionaryBuilder {
dict: SharedArenaHashMap,
dict: FnvHashMap<Vec<u8>, UnorderedId>,
}
impl DictionaryBuilder {
/// Get or allocate an unordered id.
/// (This ID is simply an auto-incremented id.)
pub fn get_or_allocate_id(&mut self, term: &[u8], arena: &mut MemoryArena) -> UnorderedId {
let next_id = self.dict.len() as u32;
let unordered_id = self
.dict
.mutate_or_create(term, arena, |unordered_id: Option<u32>| {
if let Some(unordered_id) = unordered_id {
unordered_id
} else {
next_id
}
});
UnorderedId(unordered_id)
pub fn get_or_allocate_id(&mut self, term: &[u8]) -> UnorderedId {
if let Some(term_id) = self.dict.get(term) {
return *term_id;
}
let new_id = UnorderedId(self.dict.len() as u32);
self.dict.insert(term.to_vec(), new_id);
new_id
}
/// Serialize the dictionary into an fst, and returns the
/// `UnorderedId -> TermOrdinal` map.
pub fn serialize<'a, W: io::Write + 'a>(
&self,
arena: &MemoryArena,
wrt: &mut W,
) -> io::Result<TermIdMapping> {
let mut terms: Vec<(&[u8], UnorderedId)> = self
.dict
.iter(arena)
.map(|(k, v)| (k, arena.read(v)))
.collect();
pub fn serialize<'a, W: io::Write + 'a>(&self, wrt: &mut W) -> io::Result<TermIdMapping> {
let mut terms: Vec<(&[u8], UnorderedId)> =
self.dict.iter().map(|(k, v)| (k.as_slice(), *v)).collect();
terms.sort_unstable_by_key(|(key, _)| *key);
// TODO Remove the allocation.
let mut unordered_to_ord: Vec<OrderedId> = vec![OrderedId(0u32); terms.len()];
@@ -75,10 +63,6 @@ impl DictionaryBuilder {
sstable_builder.finish()?;
Ok(TermIdMapping { unordered_to_ord })
}
pub(crate) fn mem_usage(&self) -> usize {
self.dict.mem_usage()
}
}
#[cfg(test)]
@@ -87,13 +71,12 @@ mod tests {
#[test]
fn test_dictionary_builder() {
let mut arena = MemoryArena::default();
let mut dictionary_builder = DictionaryBuilder::default();
let hello_uid = dictionary_builder.get_or_allocate_id(b"hello", &mut arena);
let happy_uid = dictionary_builder.get_or_allocate_id(b"happy", &mut arena);
let tax_uid = dictionary_builder.get_or_allocate_id(b"tax", &mut arena);
let hello_uid = dictionary_builder.get_or_allocate_id(b"hello");
let happy_uid = dictionary_builder.get_or_allocate_id(b"happy");
let tax_uid = dictionary_builder.get_or_allocate_id(b"tax");
let mut buffer = Vec::new();
let id_mapping = dictionary_builder.serialize(&arena, &mut buffer).unwrap();
let id_mapping = dictionary_builder.serialize(&mut buffer).unwrap();
assert_eq!(id_mapping.to_ord(hello_uid), OrderedId(1));
assert_eq!(id_mapping.to_ord(happy_uid), OrderedId(0));
assert_eq!(id_mapping.to_ord(tax_uid), OrderedId(2));

39
columnar/src/dynamic_column.rs
View File

@@ -8,7 +8,7 @@ use common::{ByteCount, DateTime, HasLen, OwnedBytes};
use crate::column::{BytesColumn, Column, StrColumn};
use crate::column_values::{monotonic_map_column, StrictlyMonotonicFn};
use crate::columnar::ColumnType;
use crate::{Cardinality, ColumnIndex, ColumnValues, NumericalType};
use crate::{Cardinality, ColumnIndex, NumericalType};
#[derive(Clone)]
pub enum DynamicColumn {
@@ -26,14 +26,14 @@ impl fmt::Debug for DynamicColumn {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "[{} {} |", self.get_cardinality(), self.column_type())?;
match self {
DynamicColumn::Bool(col) => write!(f, " {col:?}")?,
DynamicColumn::I64(col) => write!(f, " {col:?}")?,
DynamicColumn::U64(col) => write!(f, " {col:?}")?,
DynamicColumn::F64(col) => write!(f, "{col:?}")?,
DynamicColumn::IpAddr(col) => write!(f, "{col:?}")?,
DynamicColumn::DateTime(col) => write!(f, "{col:?}")?,
DynamicColumn::Bytes(col) => write!(f, "{col:?}")?,
DynamicColumn::Str(col) => write!(f, "{col:?}")?,
DynamicColumn::Bool(col) => write!(f, " {:?}", col)?,
DynamicColumn::I64(col) => write!(f, " {:?}", col)?,
DynamicColumn::U64(col) => write!(f, " {:?}", col)?,
DynamicColumn::F64(col) => write!(f, "{:?}", col)?,
DynamicColumn::IpAddr(col) => write!(f, "{:?}", col)?,
DynamicColumn::DateTime(col) => write!(f, "{:?}", col)?,
DynamicColumn::Bytes(col) => write!(f, "{:?}", col)?,
DynamicColumn::Str(col) => write!(f, "{:?}", col)?,
}
write!(f, "]")
}
@@ -228,7 +228,7 @@ static_dynamic_conversions!(StrColumn, Str);
static_dynamic_conversions!(BytesColumn, Bytes);
static_dynamic_conversions!(Column<Ipv6Addr>, IpAddr);
#[derive(Clone, Debug)]
#[derive(Clone)]
pub struct DynamicColumnHandle {
pub(crate) file_slice: FileSlice,
pub(crate) column_type: ColumnType,
@@ -247,12 +247,7 @@ impl DynamicColumnHandle {
}
/// Returns the `u64` fast field reader reader associated with `fields` of types
/// Str, u64, i64, f64, bool, ip, or datetime.
///
/// Notice that for IpAddr, the fastfield reader will return the u64 representation of the
/// IpAddr.
/// In order to convert to u128 back cast to `CompactSpaceU64Accessor` and call
/// `compact_to_u128`.
/// Str, u64, i64, f64, or datetime.
///
/// If not, the fastfield reader will returns the u64-value associated with the original
/// FastValue.
@@ -263,15 +258,9 @@ impl DynamicColumnHandle {
let column: BytesColumn = crate::column::open_column_bytes(column_bytes)?;
Ok(Some(column.term_ord_column))
}
ColumnType::IpAddr => {
let column = crate::column::open_column_u128_as_compact_u64(column_bytes)?;
Ok(Some(column))
}
ColumnType::Bool
| ColumnType::I64
| ColumnType::U64
| ColumnType::F64
| ColumnType::DateTime => {
ColumnType::Bool => Ok(None),
ColumnType::IpAddr => Ok(None),
ColumnType::I64 | ColumnType::U64 | ColumnType::F64 | ColumnType::DateTime => {
let column = crate::column::open_column_u64::<u64>(column_bytes)?;
Ok(Some(column))
}

26
columnar/src/lib.rs
View File

@@ -1,22 +1,3 @@
//! # Tantivy-Columnar
//!
//! `tantivy-columnar`provides a columnar storage for tantivy.
//! The crate allows for efficient read operations on specific columns rather than entire records.
//!
//! ## Overview
//!
//! - **columnar**: Reading, writing, and merging multiple columns:
//! - **[ColumnarWriter]**: Makes it possible to create a new columnar.
//! - **[ColumnarReader]**: The ColumnarReader makes it possible to access a set of columns
//! associated to field names.
//! - **[merge_columnar]**: Contains the functionalities to merge multiple ColumnarReader or
//! segments into a single one.
//!
//! - **column**: A single column, which contains
//! - [column_index]: Resolves the rows for a document id. Manages the cardinality of the
//! column.
//! - [column_values]: Stores the values of a column in a dense format.
#![cfg_attr(all(feature = "unstable", test), feature(test))]
#[cfg(test)]
@@ -31,7 +12,7 @@ use std::io;
mod block_accessor;
mod column;
pub mod column_index;
mod column_index;
pub mod column_values;
mod columnar;
mod dictionary;
@@ -58,7 +39,7 @@ pub use self::dynamic_column::{DynamicColumn, DynamicColumnHandle};
pub type RowId = u32;
pub type DocId = u32;
#[derive(Clone, Copy, Debug)]
#[derive(Clone, Copy)]
pub struct RowAddr {
pub segment_ord: u32,
pub row_id: RowId,
@@ -113,9 +94,6 @@ impl Cardinality {
pub fn is_multivalue(&self) -> bool {
matches!(self, Cardinality::Multivalued)
}
pub fn is_full(&self) -> bool {
matches!(self, Cardinality::Full)
}
pub(crate) fn to_code(self) -> u8 {
self as u8
}

276
columnar/src/tests.rs
View File

@@ -4,15 +4,13 @@ use std::net::Ipv6Addr;
use common::DateTime;
use proptest::prelude::*;
use proptest::sample::subsequence;
use crate::column_values::MonotonicallyMappableToU128;
use crate::columnar::{ColumnType, ColumnTypeCategory};
use crate::dynamic_column::{DynamicColumn, DynamicColumnHandle};
use crate::value::{Coerce, NumericalValue};
use crate::{
BytesColumn, Cardinality, Column, ColumnarReader, ColumnarWriter, RowAddr, RowId,
ShuffleMergeOrder, StackMergeOrder,
BytesColumn, Cardinality, Column, ColumnarReader, ColumnarWriter, RowId, StackMergeOrder,
};
#[test]
@@ -26,7 +24,7 @@ fn test_dataframe_writer_str() {
assert_eq!(columnar.num_columns(), 1);
let cols: Vec<DynamicColumnHandle> = columnar.read_columns("my_string").unwrap();
assert_eq!(cols.len(), 1);
assert_eq!(cols[0].num_bytes(), 73);
assert_eq!(cols[0].num_bytes(), 89);
}
#[test]
@@ -40,7 +38,7 @@ fn test_dataframe_writer_bytes() {
assert_eq!(columnar.num_columns(), 1);
let cols: Vec<DynamicColumnHandle> = columnar.read_columns("my_string").unwrap();
assert_eq!(cols.len(), 1);
assert_eq!(cols[0].num_bytes(), 73);
assert_eq!(cols[0].num_bytes(), 89);
}
#[test]
@@ -57,9 +55,7 @@ fn test_dataframe_writer_bool() {
assert_eq!(cols[0].num_bytes(), 22);
assert_eq!(cols[0].column_type(), ColumnType::Bool);
let dyn_bool_col = cols[0].open().unwrap();
let DynamicColumn::Bool(bool_col) = dyn_bool_col else {
panic!();
};
let DynamicColumn::Bool(bool_col) = dyn_bool_col else { panic!(); };
let vals: Vec<Option<bool>> = (0..5).map(|row_id| bool_col.first(row_id)).collect();
assert_eq!(&vals, &[None, Some(false), None, Some(true), None,]);
}
@@ -81,9 +77,7 @@ fn test_dataframe_writer_u64_multivalued() {
assert_eq!(cols.len(), 1);
assert_eq!(cols[0].num_bytes(), 29);
let dyn_i64_col = cols[0].open().unwrap();
let DynamicColumn::I64(divisor_col) = dyn_i64_col else {
panic!();
};
let DynamicColumn::I64(divisor_col) = dyn_i64_col else { panic!(); };
assert_eq!(
divisor_col.get_cardinality(),
crate::Cardinality::Multivalued
@@ -105,9 +99,7 @@ fn test_dataframe_writer_ip_addr() {
assert_eq!(cols[0].num_bytes(), 42);
assert_eq!(cols[0].column_type(), ColumnType::IpAddr);
let dyn_bool_col = cols[0].open().unwrap();
let DynamicColumn::IpAddr(ip_col) = dyn_bool_col else {
panic!();
};
let DynamicColumn::IpAddr(ip_col) = dyn_bool_col else { panic!(); };
let vals: Vec<Option<Ipv6Addr>> = (0..5).map(|row_id| ip_col.first(row_id)).collect();
assert_eq!(
&vals,
@@ -140,9 +132,7 @@ fn test_dataframe_writer_numerical() {
// - null footer 6 bytes
assert_eq!(cols[0].num_bytes(), 33);
let column = cols[0].open().unwrap();
let DynamicColumn::I64(column_i64) = column else {
panic!();
};
let DynamicColumn::I64(column_i64) = column else { panic!(); };
assert_eq!(column_i64.index.get_cardinality(), Cardinality::Optional);
assert_eq!(column_i64.first(0), None);
assert_eq!(column_i64.first(1), Some(12i64));
@@ -206,9 +196,7 @@ fn test_dictionary_encoded_str() {
assert_eq!(columnar_reader.num_columns(), 2);
let col_handles = columnar_reader.read_columns("my.column").unwrap();
assert_eq!(col_handles.len(), 1);
let DynamicColumn::Str(str_col) = col_handles[0].open().unwrap() else {
panic!();
};
let DynamicColumn::Str(str_col) = col_handles[0].open().unwrap() else { panic!(); };
let index: Vec<Option<u64>> = (0..5).map(|row_id| str_col.ords().first(row_id)).collect();
assert_eq!(index, &[None, Some(0), None, Some(2), Some(1)]);
assert_eq!(str_col.num_rows(), 5);
@@ -240,9 +228,7 @@ fn test_dictionary_encoded_bytes() {
assert_eq!(columnar_reader.num_columns(), 2);
let col_handles = columnar_reader.read_columns("my.column").unwrap();
assert_eq!(col_handles.len(), 1);
let DynamicColumn::Bytes(bytes_col) = col_handles[0].open().unwrap() else {
panic!();
};
let DynamicColumn::Bytes(bytes_col) = col_handles[0].open().unwrap() else { panic!(); };
let index: Vec<Option<u64>> = (0..5)
.map(|row_id| bytes_col.ords().first(row_id))
.collect();
@@ -274,15 +260,12 @@ fn test_dictionary_encoded_bytes() {
fn num_strategy() -> impl Strategy<Value = NumericalValue> {
prop_oneof![
3 => Just(NumericalValue::U64(0u64)),
3 => Just(NumericalValue::U64(u64::MAX)),
3 => Just(NumericalValue::I64(0i64)),
3 => Just(NumericalValue::I64(i64::MIN)),
3 => Just(NumericalValue::I64(i64::MAX)),
3 => Just(NumericalValue::F64(1.2f64)),
1 => any::<f64>().prop_map(NumericalValue::from),
1 => any::<u64>().prop_map(NumericalValue::from),
1 => any::<i64>().prop_map(NumericalValue::from),
Just(NumericalValue::U64(0u64)),
Just(NumericalValue::U64(u64::MAX)),
Just(NumericalValue::I64(0i64)),
Just(NumericalValue::I64(i64::MIN)),
Just(NumericalValue::I64(i64::MAX)),
Just(NumericalValue::F64(1.2f64)),
]
}
@@ -296,12 +279,6 @@ enum ColumnValue {
DateTime(DateTime),
}
impl<T: Into<NumericalValue>> From<T> for ColumnValue {
fn from(val: T) -> ColumnValue {
ColumnValue::Numerical(val.into())
}
}
impl ColumnValue {
pub(crate) fn column_type_category(&self) -> ColumnTypeCategory {
match self {
@@ -330,9 +307,9 @@ fn bytes_strategy() -> impl Strategy<Value = &'static [u8]> {
// A random column value
fn column_value_strategy() -> impl Strategy<Value = ColumnValue> {
prop_oneof![
10 => string_strategy().prop_map(ColumnValue::Str),
1 => bytes_strategy().prop_map(ColumnValue::Bytes),
40 => num_strategy().prop_map(ColumnValue::Numerical),
10 => string_strategy().prop_map(|s| ColumnValue::Str(s)),
1 => bytes_strategy().prop_map(|b| ColumnValue::Bytes(b)),
40 => num_strategy().prop_map(|n| ColumnValue::Numerical(n)),
1 => (1u16..3u16).prop_map(|ip_addr_byte| ColumnValue::IpAddr(Ipv6Addr::new(
127,
0,
@@ -343,7 +320,7 @@ fn column_value_strategy() -> impl Strategy<Value = ColumnValue> {
0,
ip_addr_byte
))),
1 => any::<bool>().prop_map(ColumnValue::Bool),
1 => any::<bool>().prop_map(|b| ColumnValue::Bool(b)),
1 => (0_679_723_993i64..1_679_723_995i64)
.prop_map(|val| { ColumnValue::DateTime(DateTime::from_timestamp_secs(val)) })
]
@@ -351,22 +328,12 @@ fn column_value_strategy() -> impl Strategy<Value = ColumnValue> {
// A document contains up to 4 values.
fn doc_strategy() -> impl Strategy<Value = Vec<(&'static str, ColumnValue)>> {
proptest::collection::vec((column_name_strategy(), column_value_strategy()), 0..=4)
}
fn num_docs_strategy() -> impl Strategy<Value = usize> {
prop_oneof!(
// We focus heavily on the 0..2 case as we assume it is sufficient to cover all edge cases.
0usize..=3usize,
// We leave 50% of the effort exploring more defensively.
3usize..=12usize
)
proptest::collection::vec((column_name_strategy(), column_value_strategy()), 0..4)
}
// A columnar contains up to 2 docs.
fn columnar_docs_strategy() -> impl Strategy<Value = Vec<Vec<(&'static str, ColumnValue)>>> {
num_docs_strategy()
.prop_flat_map(|num_docs| proptest::collection::vec(doc_strategy(), num_docs))
proptest::collection::vec(doc_strategy(), 0..=2)
}
fn columnar_docs_and_mapping_strategy(
@@ -380,11 +347,6 @@ fn permutation_strategy(n: usize) -> impl Strategy<Value = Vec<RowId>> {
Just((0u32..n as RowId).collect()).prop_shuffle()
}
fn permutation_and_subset_strategy(n: usize) -> impl Strategy<Value = Vec<usize>> {
let vals: Vec<usize> = (0..n).collect();
subsequence(vals, 0..=n).prop_shuffle()
}
fn build_columnar_with_mapping(
docs: &[Vec<(&'static str, ColumnValue)>],
old_to_new_row_ids_opt: Option<&[RowId]>,
@@ -419,23 +381,15 @@ fn build_columnar_with_mapping(
columnar_writer
.serialize(num_docs, old_to_new_row_ids_opt, &mut buffer)
.unwrap();
ColumnarReader::open(buffer).unwrap()
let columnar_reader = ColumnarReader::open(buffer).unwrap();
columnar_reader
}
fn build_columnar(docs: &[Vec<(&'static str, ColumnValue)>]) -> ColumnarReader {
build_columnar_with_mapping(docs, None)
}
fn assert_columnar_eq_strict(left: &ColumnarReader, right: &ColumnarReader) {
assert_columnar_eq(left, right, false);
}
fn assert_columnar_eq(
left: &ColumnarReader,
right: &ColumnarReader,
lenient_on_numerical_value: bool,
) {
fn assert_columnar_eq(left: &ColumnarReader, right: &ColumnarReader) {
assert_eq!(left.num_rows(), right.num_rows());
let left_columns = left.list_columns().unwrap();
let right_columns = right.list_columns().unwrap();
@@ -444,7 +398,7 @@ fn assert_columnar_eq(
assert_eq!(left_columns[i].0, right_columns[i].0);
let left_column = left_columns[i].1.open().unwrap();
let right_column = right_columns[i].1.open().unwrap();
assert_dyn_column_eq(&left_column, &right_column, lenient_on_numerical_value);
assert_dyn_column_eq(&left_column, &right_column);
}
}
@@ -488,11 +442,11 @@ fn assert_bytes_column_eq(left: &BytesColumn, right: &BytesColumn) {
assert!(!right_terms.advance());
}
fn assert_dyn_column_eq(
left_dyn_column: &DynamicColumn,
right_dyn_column: &DynamicColumn,
lenient_on_numerical_value: bool,
) {
fn assert_dyn_column_eq(left_dyn_column: &DynamicColumn, right_dyn_column: &DynamicColumn) {
assert_eq!(
&left_dyn_column.column_type(),
&right_dyn_column.column_type()
);
assert_eq!(
&left_dyn_column.get_cardinality(),
&right_dyn_column.get_cardinality()
@@ -522,19 +476,8 @@ fn assert_dyn_column_eq(
(DynamicColumn::Str(left_col), DynamicColumn::Str(right_col)) => {
assert_bytes_column_eq(left_col, right_col);
}
(left, right) => {
if lenient_on_numerical_value {
assert_eq!(
ColumnTypeCategory::from(left.column_type()),
ColumnTypeCategory::from(right.column_type())
);
} else {
panic!(
"Column type are not the same: {:?} vs {:?}",
left.column_type(),
right.column_type()
);
}
_ => {
unreachable!()
}
}
}
@@ -545,36 +488,28 @@ trait AssertEqualToColumnValue {
impl AssertEqualToColumnValue for bool {
fn assert_equal_to_column_value(&self, column_value: &ColumnValue) {
let ColumnValue::Bool(val) = column_value else {
panic!()
};
let ColumnValue::Bool(val) = column_value else { panic!() };
assert_eq!(self, val);
}
}
impl AssertEqualToColumnValue for Ipv6Addr {
fn assert_equal_to_column_value(&self, column_value: &ColumnValue) {
let ColumnValue::IpAddr(val) = column_value else {
panic!()
};
let ColumnValue::IpAddr(val) = column_value else { panic!() };
assert_eq!(self, val);
}
}
impl<T: Coerce + PartialEq + Debug + Into<NumericalValue>> AssertEqualToColumnValue for T {
fn assert_equal_to_column_value(&self, column_value: &ColumnValue) {
let ColumnValue::Numerical(num) = column_value else {
panic!()
};
let ColumnValue::Numerical(num) = column_value else { panic!() };
assert_eq!(self, &T::coerce(*num));
}
}
impl AssertEqualToColumnValue for DateTime {
fn assert_equal_to_column_value(&self, column_value: &ColumnValue) {
let ColumnValue::DateTime(dt) = column_value else {
panic!()
};
let ColumnValue::DateTime(dt) = column_value else { panic!() };
assert_eq!(self, dt);
}
}
@@ -746,9 +681,9 @@ proptest! {
let stack_merge_order = StackMergeOrder::stack(&columnar_readers_arr[..]).into();
crate::merge_columnar(&columnar_readers_arr[..], &[], stack_merge_order, &mut output).unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
let concat_rows: Vec<Vec<(&'static str, ColumnValue)>> = columnar_docs.iter().flatten().cloned().collect();
let concat_rows: Vec<Vec<(&'static str, ColumnValue)>> = columnar_docs.iter().cloned().flatten().collect();
let expected_merged_columnar = build_columnar(&concat_rows[..]);
assert_columnar_eq_strict(&merged_columnar, &expected_merged_columnar);
assert_columnar_eq(&merged_columnar, &expected_merged_columnar);
}
}
@@ -772,9 +707,9 @@ fn test_columnar_merging_empty_columnar() {
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
let concat_rows: Vec<Vec<(&'static str, ColumnValue)>> =
columnar_docs.iter().flatten().cloned().collect();
columnar_docs.iter().cloned().flatten().collect();
let expected_merged_columnar = build_columnar(&concat_rows[..]);
assert_columnar_eq_strict(&merged_columnar, &expected_merged_columnar);
assert_columnar_eq(&merged_columnar, &expected_merged_columnar);
}
#[test]
@@ -809,137 +744,10 @@ fn test_columnar_merging_number_columns() {
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
let concat_rows: Vec<Vec<(&'static str, ColumnValue)>> =
columnar_docs.iter().flatten().cloned().collect();
columnar_docs.iter().cloned().flatten().collect();
let expected_merged_columnar = build_columnar(&concat_rows[..]);
assert_columnar_eq_strict(&merged_columnar, &expected_merged_columnar);
assert_columnar_eq(&merged_columnar, &expected_merged_columnar);
}
// TODO add non trivial remap and merge
// TODO test required_columns
// TODO document edge case: required_columns incompatible with values.
fn columnar_docs_and_remap(
) -> impl Strategy<Value = (Vec<Vec<Vec<(&'static str, ColumnValue)>>>, Vec<RowAddr>)> {
proptest::collection::vec(columnar_docs_strategy(), 2..=3).prop_flat_map(
|columnars_docs: Vec<Vec<Vec<(&str, ColumnValue)>>>| {
let row_addrs: Vec<RowAddr> = columnars_docs
.iter()
.enumerate()
.flat_map(|(segment_ord, columnar_docs)| {
(0u32..columnar_docs.len() as u32).map(move |row_id| RowAddr {
segment_ord: segment_ord as u32,
row_id,
})
})
.collect();
permutation_and_subset_strategy(row_addrs.len()).prop_map(move |shuffled_subset| {
let shuffled_row_addr_subset: Vec<RowAddr> =
shuffled_subset.iter().map(|ord| row_addrs[*ord]).collect();
(columnars_docs.clone(), shuffled_row_addr_subset)
})
},
)
}
proptest! {
#![proptest_config(ProptestConfig::with_cases(1000))]
#[test]
fn test_columnar_merge_and_remap_proptest((columnar_docs, shuffle_merge_order) in columnar_docs_and_remap()) {
let shuffled_rows: Vec<Vec<(&'static str, ColumnValue)>> = shuffle_merge_order.iter()
.map(|row_addr| columnar_docs[row_addr.segment_ord as usize][row_addr.row_id as usize].clone())
.collect();
let expected_merged_columnar = build_columnar(&shuffled_rows[..]);
let columnar_readers: Vec<ColumnarReader> = columnar_docs.iter()
.map(|docs| build_columnar(&docs[..]))
.collect::<Vec<_>>();
let columnar_readers_arr: Vec<&ColumnarReader> = columnar_readers.iter().collect();
let mut output: Vec<u8> = Vec::new();
let segment_num_rows: Vec<RowId> = columnar_docs.iter().map(|docs| docs.len() as RowId).collect();
let shuffle_merge_order = ShuffleMergeOrder::for_test(&segment_num_rows, shuffle_merge_order);
crate::merge_columnar(&columnar_readers_arr[..], &[], shuffle_merge_order.into(), &mut output).unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
assert_columnar_eq(&merged_columnar, &expected_merged_columnar, true);
}
}
#[test]
fn test_columnar_merge_empty() {
let columnar_reader_1 = build_columnar(&[]);
let rows: &[Vec<_>] = &[vec![("c1", ColumnValue::Str("a"))]][..];
let columnar_reader_2 = build_columnar(rows);
let mut output: Vec<u8> = Vec::new();
let segment_num_rows: Vec<RowId> = vec![0, 0];
let shuffle_merge_order = ShuffleMergeOrder::for_test(&segment_num_rows, vec![]);
crate::merge_columnar(
&[&columnar_reader_1, &columnar_reader_2],
&[],
shuffle_merge_order.into(),
&mut output,
)
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
assert_eq!(merged_columnar.num_rows(), 0);
assert_eq!(merged_columnar.num_columns(), 0);
}
#[test]
fn test_columnar_merge_single_str_column() {
let columnar_reader_1 = build_columnar(&[]);
let rows: &[Vec<_>] = &[vec![("c1", ColumnValue::Str("a"))]][..];
let columnar_reader_2 = build_columnar(rows);
let mut output: Vec<u8> = Vec::new();
let segment_num_rows: Vec<RowId> = vec![0, 1];
let shuffle_merge_order = ShuffleMergeOrder::for_test(
&segment_num_rows,
vec![RowAddr {
segment_ord: 1u32,
row_id: 0u32,
}],
);
crate::merge_columnar(
&[&columnar_reader_1, &columnar_reader_2],
&[],
shuffle_merge_order.into(),
&mut output,
)
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
assert_eq!(merged_columnar.num_rows(), 1);
assert_eq!(merged_columnar.num_columns(), 1);
}
#[test]
fn test_delete_decrease_cardinality() {
let columnar_reader_1 = build_columnar(&[]);
let rows: &[Vec<_>] = &[
vec![
("c", ColumnValue::from(0i64)),
("c", ColumnValue::from(0i64)),
],
vec![("c", ColumnValue::from(0i64))],
][..];
// c is multivalued here
let columnar_reader_2 = build_columnar(rows);
let mut output: Vec<u8> = Vec::new();
let shuffle_merge_order = ShuffleMergeOrder::for_test(
&[0, 2],
vec![RowAddr {
segment_ord: 1u32,
row_id: 1u32,
}],
);
crate::merge_columnar(
&[&columnar_reader_1, &columnar_reader_2],
&[],
shuffle_merge_order.into(),
&mut output,
)
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
assert_eq!(merged_columnar.num_rows(), 1);
assert_eq!(merged_columnar.num_columns(), 1);
let cols = merged_columnar.read_columns("c").unwrap();
assert_eq!(cols.len(), 1);
assert_eq!(cols[0].column_type(), ColumnType::I64);
assert_eq!(cols[0].open().unwrap().get_cardinality(), Cardinality::Full);
}

2
columnar/src/value.rs
View File

@@ -109,7 +109,7 @@ impl Coerce for f64 {
impl Coerce for DateTime {
fn coerce(value: NumericalValue) -> Self {
let timestamp_micros = i64::coerce(value);
DateTime::from_timestamp_nanos(timestamp_micros)
DateTime::from_timestamp_micros(timestamp_micros)
}
}

4
common/Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy-common"
version = "0.7.0"
version = "0.5.0"
authors = ["Paul Masurel <paul@quickwit.io>", "Pascal Seitz <pascal@quickwit.io>"]
license = "MIT"
edition = "2021"
@@ -14,7 +14,7 @@ repository = "https://github.com/quickwit-oss/tantivy"
[dependencies]
byteorder = "1.4.3"
ownedbytes = { version= "0.7", path="../ownedbytes" }
ownedbytes = { version= "0.5", path="../ownedbytes" }
async-trait = "0.1"
time = { version = "0.3.10", features = ["serde-well-known"] }
serde = { version = "1.0.136", features = ["derive"] }

39
common/benches/bench.rs
View File

@@ -1,39 +0,0 @@
#![feature(test)]
extern crate test;
#[cfg(test)]
mod tests {
use rand::seq::IteratorRandom;
use rand::thread_rng;
use tantivy_common::serialize_vint_u32;
use test::Bencher;
#[bench]
fn bench_vint(b: &mut Bencher) {
let vals: Vec<u32> = (0..20_000).collect();
b.iter(|| {
let mut out = 0u64;
for val in vals.iter().cloned() {
let mut buf = [0u8; 8];
serialize_vint_u32(val, &mut buf);
out += u64::from(buf[0]);
}
out
});
}
#[bench]
fn bench_vint_rand(b: &mut Bencher) {
let vals: Vec<u32> = (0..20_000).choose_multiple(&mut thread_rng(), 100_000);
b.iter(|| {
let mut out = 0u64;
for val in vals.iter().cloned() {
let mut buf = [0u8; 8];
serialize_vint_u32(val, &mut buf);
out += u64::from(buf[0]);
}
out
});
}
}

1
common/src/bitset.rs
View File

@@ -1,3 +1,4 @@
use std::convert::TryInto;
use std::io::Write;
use std::{fmt, io, u64};

2
common/src/byte_count.rs
View File

@@ -37,7 +37,7 @@ impl ByteCount {
for (suffix, threshold) in SUFFIX_AND_THRESHOLD.iter().rev() {
if self.get_bytes() >= *threshold {
let unit_num = self.get_bytes() as f64 / *threshold as f64;
return format!("{unit_num:.2} {suffix}");
return format!("{:.2} {}", unit_num, suffix);
}
}
format!("{:.2} B", self.get_bytes())

86
common/src/datetime.rs
View File

@@ -1,31 +1,25 @@
use std::fmt;
use std::io::{Read, Write};
use serde::{Deserialize, Serialize};
use time::format_description::well_known::Rfc3339;
use time::{OffsetDateTime, PrimitiveDateTime, UtcOffset};
use crate::BinarySerializable;
/// Precision with which datetimes are truncated when stored in fast fields. This setting is only
/// relevant for fast fields. In the docstore, datetimes are always saved with nanosecond precision.
/// DateTime Precision
#[derive(
Clone, Copy, Debug, Hash, PartialEq, Eq, PartialOrd, Ord, Serialize, Deserialize, Default,
)]
#[serde(rename_all = "lowercase")]
pub enum DateTimePrecision {
/// Second precision.
pub enum DatePrecision {
/// Seconds precision
#[default]
Seconds,
/// Millisecond precision.
/// Milli-seconds precision.
Milliseconds,
/// Microsecond precision.
/// Micro-seconds precision.
Microseconds,
/// Nanosecond precision.
Nanoseconds,
}
/// A date/time value with nanoseconds precision.
/// A date/time value with microsecond precision.
///
/// This timestamp does not carry any explicit time zone information.
/// Users are responsible for applying the provided conversion
@@ -35,48 +29,41 @@ pub enum DateTimePrecision {
/// All constructors and conversions are provided as explicit
/// functions and not by implementing any `From`/`Into` traits
/// to prevent unintended usage.
#[derive(Clone, Default, Copy, PartialEq, Eq, PartialOrd, Ord, Hash, Serialize, Deserialize)]
#[derive(Clone, Default, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
pub struct DateTime {
// Timestamp in nanoseconds.
pub(crate) timestamp_nanos: i64,
// Timestamp in microseconds.
pub(crate) timestamp_micros: i64,
}
impl DateTime {
/// Minimum possible `DateTime` value.
pub const MIN: DateTime = DateTime {
timestamp_nanos: i64::MIN,
timestamp_micros: i64::MIN,
};
/// Maximum possible `DateTime` value.
pub const MAX: DateTime = DateTime {
timestamp_nanos: i64::MAX,
timestamp_micros: i64::MAX,
};
/// Create new from UNIX timestamp in seconds
pub const fn from_timestamp_secs(seconds: i64) -> Self {
Self {
timestamp_nanos: seconds * 1_000_000_000,
timestamp_micros: seconds * 1_000_000,
}
}
/// Create new from UNIX timestamp in milliseconds
pub const fn from_timestamp_millis(milliseconds: i64) -> Self {
Self {
timestamp_nanos: milliseconds * 1_000_000,
timestamp_micros: milliseconds * 1_000,
}
}
/// Create new from UNIX timestamp in microseconds.
pub const fn from_timestamp_micros(microseconds: i64) -> Self {
Self {
timestamp_nanos: microseconds * 1_000,
}
}
/// Create new from UNIX timestamp in nanoseconds.
pub const fn from_timestamp_nanos(nanoseconds: i64) -> Self {
Self {
timestamp_nanos: nanoseconds,
timestamp_micros: microseconds,
}
}
@@ -84,9 +71,9 @@ impl DateTime {
///
/// The given date/time is converted to UTC and the actual
/// time zone is discarded.
pub fn from_utc(dt: OffsetDateTime) -> Self {
let timestamp_nanos = dt.unix_timestamp_nanos() as i64;
Self { timestamp_nanos }
pub const fn from_utc(dt: OffsetDateTime) -> Self {
let timestamp_micros = dt.unix_timestamp() * 1_000_000 + dt.microsecond() as i64;
Self { timestamp_micros }
}
/// Create new from `PrimitiveDateTime`
@@ -100,27 +87,23 @@ impl DateTime {
/// Convert to UNIX timestamp in seconds.
pub const fn into_timestamp_secs(self) -> i64 {
self.timestamp_nanos / 1_000_000_000
self.timestamp_micros / 1_000_000
}
/// Convert to UNIX timestamp in milliseconds.
pub const fn into_timestamp_millis(self) -> i64 {
self.timestamp_nanos / 1_000_000
self.timestamp_micros / 1_000
}
/// Convert to UNIX timestamp in microseconds.
pub const fn into_timestamp_micros(self) -> i64 {
self.timestamp_nanos / 1_000
}
/// Convert to UNIX timestamp in nanoseconds.
pub const fn into_timestamp_nanos(self) -> i64 {
self.timestamp_nanos
self.timestamp_micros
}
/// Convert to UTC `OffsetDateTime`
pub fn into_utc(self) -> OffsetDateTime {
let utc_datetime = OffsetDateTime::from_unix_timestamp_nanos(self.timestamp_nanos as i128)
let timestamp_nanos = self.timestamp_micros as i128 * 1000;
let utc_datetime = OffsetDateTime::from_unix_timestamp_nanos(timestamp_nanos)
.expect("valid UNIX timestamp");
debug_assert_eq!(UtcOffset::UTC, utc_datetime.offset());
utc_datetime
@@ -143,34 +126,21 @@ impl DateTime {
}
/// Truncates the microseconds value to the corresponding precision.
pub fn truncate(self, precision: DateTimePrecision) -> Self {
pub fn truncate(self, precision: DatePrecision) -> Self {
let truncated_timestamp_micros = match precision {
DateTimePrecision::Seconds => (self.timestamp_nanos / 1_000_000_000) * 1_000_000_000,
DateTimePrecision::Milliseconds => (self.timestamp_nanos / 1_000_000) * 1_000_000,
DateTimePrecision::Microseconds => (self.timestamp_nanos / 1_000) * 1_000,
DateTimePrecision::Nanoseconds => self.timestamp_nanos,
DatePrecision::Seconds => (self.timestamp_micros / 1_000_000) * 1_000_000,
DatePrecision::Milliseconds => (self.timestamp_micros / 1_000) * 1_000,
DatePrecision::Microseconds => self.timestamp_micros,
};
Self {
timestamp_nanos: truncated_timestamp_micros,
timestamp_micros: truncated_timestamp_micros,
}
}
}
impl fmt::Debug for DateTime {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
let utc_rfc3339 = self.into_utc().format(&Rfc3339).map_err(|_| fmt::Error)?;
f.write_str(&utc_rfc3339)
}
}
impl BinarySerializable for DateTime {
fn serialize<W: Write + ?Sized>(&self, writer: &mut W) -> std::io::Result<()> {
let timestamp_micros = self.into_timestamp_micros();
<i64 as BinarySerializable>::serialize(&timestamp_micros, writer)
}
fn deserialize<R: Read>(reader: &mut R) -> std::io::Result<Self> {
let timestamp_micros = <i64 as BinarySerializable>::deserialize(reader)?;
Ok(Self::from_timestamp_micros(timestamp_micros))
}
}

63
common/src/dictionary_footer.rs Normal file
View File

@@ -0,0 +1,63 @@
use std::io::{self, Read, Write};
use crate::BinarySerializable;
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
#[repr(u32)]
pub enum DictionaryKind {
Fst = 1,
SSTable = 2,
}
#[derive(Debug, Clone, PartialEq)]
pub struct DictionaryFooter {
pub kind: DictionaryKind,
pub version: u32,
}
impl DictionaryFooter {
pub fn verify_equal(&self, other: &DictionaryFooter) -> io::Result<()> {
if self.kind != other.kind {
return Err(io::Error::new(
io::ErrorKind::Other,
format!(
"Invalid dictionary type, expected {:?}, found {:?}",
self.kind, other.kind
),
));
}
if self.version != other.version {
return Err(io::Error::new(
io::ErrorKind::Other,
format!(
"Unsuported dictionary version, expected {}, found {}",
self.version, other.version
),
));
}
Ok(())
}
}
impl BinarySerializable for DictionaryFooter {
fn serialize<W: Write + ?Sized>(&self, writer: &mut W) -> io::Result<()> {
self.version.serialize(writer)?;
(self.kind as u32).serialize(writer)
}
fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
let version = u32::deserialize(reader)?;
let kind = u32::deserialize(reader)?;
let kind = match kind {
1 => DictionaryKind::Fst,
2 => DictionaryKind::SSTable,
_ => {
return Err(io::Error::new(
io::ErrorKind::Other,
format!("invalid dictionary kind: {kind}"),
))
}
};
Ok(DictionaryFooter { kind, version })
}
}

81
common/src/file_slice.rs
View File

@@ -1,4 +1,3 @@
use std::fs::File;
use std::ops::{Deref, Range, RangeBounds};
use std::sync::Arc;
use std::{fmt, io};
@@ -33,62 +32,6 @@ pub trait FileHandle: 'static + Send + Sync + HasLen + fmt::Debug {
}
}
#[derive(Debug)]
/// A File with it's length included.
pub struct WrapFile {
file: File,
len: usize,
}
impl WrapFile {
/// Creates a new WrapFile and stores its length.
pub fn new(file: File) -> io::Result<Self> {
let len = file.metadata()?.len() as usize;
Ok(WrapFile { file, len })
}
}
#[async_trait]
impl FileHandle for WrapFile {
fn read_bytes(&self, range: Range<usize>) -> io::Result<OwnedBytes> {
let file_len = self.len();
// Calculate the actual range to read, ensuring it stays within file boundaries
let start = range.start;
let end = range.end.min(file_len);
// Ensure the start is before the end of the range
if start >= end {
return Err(io::Error::new(io::ErrorKind::InvalidInput, "Invalid range"));
}
let mut buffer = vec![0; end - start];
#[cfg(unix)]
{
use std::os::unix::prelude::FileExt;
self.file.read_exact_at(&mut buffer, start as u64)?;
}
#[cfg(not(unix))]
{
use std::io::{Read, Seek};
let mut file = self.file.try_clone()?; // Clone the file to read from it separately
// Seek to the start position in the file
file.seek(io::SeekFrom::Start(start as u64))?;
// Read the data into the buffer
file.read_exact(&mut buffer)?;
}
Ok(OwnedBytes::new(buffer))
}
// todo implement async
}
impl HasLen for WrapFile {
fn len(&self) -> usize {
self.len
}
}
#[async_trait]
impl FileHandle for &'static [u8] {
fn read_bytes(&self, range: Range<usize>) -> io::Result<OwnedBytes> {
@@ -124,30 +67,6 @@ impl fmt::Debug for FileSlice {
}
}
impl FileSlice {
pub fn stream_file_chunks(&self) -> impl Iterator<Item = io::Result<OwnedBytes>> + '_ {
let len = self.range.end;
let mut start = self.range.start;
std::iter::from_fn(move || {
/// Returns chunks of 1MB of data from the FileHandle.
const CHUNK_SIZE: usize = 1024 * 1024; // 1MB
if start < len {
let end = (start + CHUNK_SIZE).min(len);
let range = start..end;
let chunk = self.data.read_bytes(range);
start += CHUNK_SIZE;
match chunk {
Ok(chunk) => Some(Ok(chunk)),
Err(e) => Some(Err(e)),
}
} else {
None
}
})
}
}
/// Takes a range, a `RangeBounds` object, and returns
/// a `Range` that corresponds to the relative application of the
/// `RangeBounds` object to the original `Range`.

26
common/src/group_by.rs
View File

@@ -27,15 +27,15 @@ pub trait GroupByIteratorExtended: Iterator {
where
Self: Sized,
F: FnMut(&Self::Item) -> K,
K: PartialEq + Clone,
Self::Item: Clone,
K: PartialEq + Copy,
Self::Item: Copy,
{
GroupByIterator::new(self, key)
}
}
impl<I: Iterator> GroupByIteratorExtended for I {}
pub struct GroupByIterator<I, F, K: Clone>
pub struct GroupByIterator<I, F, K: Copy>
where
I: Iterator,
F: FnMut(&I::Item) -> K,
@@ -50,7 +50,7 @@ where
inner: Rc<RefCell<GroupByShared<I, F, K>>>,
}
struct GroupByShared<I, F, K: Clone>
struct GroupByShared<I, F, K: Copy>
where
I: Iterator,
F: FnMut(&I::Item) -> K,
@@ -63,7 +63,7 @@ impl<I, F, K> GroupByIterator<I, F, K>
where
I: Iterator,
F: FnMut(&I::Item) -> K,
K: Clone,
K: Copy,
{
fn new(inner: I, group_by_fn: F) -> Self {
let inner = GroupByShared {
@@ -80,28 +80,28 @@ where
impl<I, F, K> Iterator for GroupByIterator<I, F, K>
where
I: Iterator,
I::Item: Clone,
I::Item: Copy,
F: FnMut(&I::Item) -> K,
K: Clone,
K: Copy,
{
type Item = (K, GroupIterator<I, F, K>);
fn next(&mut self) -> Option<Self::Item> {
let mut inner = self.inner.borrow_mut();
let value = inner.iter.peek()?.clone();
let value = *inner.iter.peek()?;
let key = (inner.group_by_fn)(&value);
let inner = self.inner.clone();
let group_iter = GroupIterator {
inner,
group_key: key.clone(),
group_key: key,
};
Some((key, group_iter))
}
}
pub struct GroupIterator<I, F, K: Clone>
pub struct GroupIterator<I, F, K: Copy>
where
I: Iterator,
F: FnMut(&I::Item) -> K,
@@ -110,10 +110,10 @@ where
group_key: K,
}
impl<I, F, K: PartialEq + Clone> Iterator for GroupIterator<I, F, K>
impl<I, F, K: PartialEq + Copy> Iterator for GroupIterator<I, F, K>
where
I: Iterator,
I::Item: Clone,
I::Item: Copy,
F: FnMut(&I::Item) -> K,
{
type Item = I::Item;
@@ -121,7 +121,7 @@ where
fn next(&mut self) -> Option<Self::Item> {
let mut inner = self.inner.borrow_mut();
// peek if next value is in group
let peek_val = inner.iter.peek()?.clone();
let peek_val = *inner.iter.peek()?;
if (inner.group_by_fn)(&peek_val) == self.group_key {
inner.iter.next()
} else {

112
common/src/json_path_writer.rs
View File

@@ -1,112 +0,0 @@
use crate::replace_in_place;
/// Separates the different segments of a json path.
pub const JSON_PATH_SEGMENT_SEP: u8 = 1u8;
pub const JSON_PATH_SEGMENT_SEP_STR: &str =
unsafe { std::str::from_utf8_unchecked(&[JSON_PATH_SEGMENT_SEP]) };
/// Create a new JsonPathWriter, that creates flattened json paths for tantivy.
#[derive(Clone, Debug, Default)]
pub struct JsonPathWriter {
path: String,
indices: Vec<usize>,
expand_dots: bool,
}
impl JsonPathWriter {
pub fn new() -> Self {
JsonPathWriter {
path: String::new(),
indices: Vec::new(),
expand_dots: false,
}
}
/// When expand_dots is enabled, json object like
/// `{"k8s.node.id": 5}` is processed as if it was
/// `{"k8s": {"node": {"id": 5}}}`.
/// This option has the merit of allowing users to
/// write queries like `k8s.node.id:5`.
/// On the other, enabling that feature can lead to
/// ambiguity.
#[inline]
pub fn set_expand_dots(&mut self, expand_dots: bool) {
self.expand_dots = expand_dots;
}
/// Push a new segment to the path.
#[inline]
pub fn push(&mut self, segment: &str) {
let len_path = self.path.len();
self.indices.push(len_path);
if !self.path.is_empty() {
self.path.push_str(JSON_PATH_SEGMENT_SEP_STR);
}
self.path.push_str(segment);
if self.expand_dots {
// This might include the separation byte, which is ok because it is not a dot.
let appended_segment = &mut self.path[len_path..];
// The unsafe below is safe as long as b'.' and JSON_PATH_SEGMENT_SEP are
// valid single byte ut8 strings.
// By utf-8 design, they cannot be part of another codepoint.
unsafe {
replace_in_place(b'.', JSON_PATH_SEGMENT_SEP, appended_segment.as_bytes_mut())
};
}
}
/// Remove the last segment. Does nothing if the path is empty.
#[inline]
pub fn pop(&mut self) {
if let Some(last_idx) = self.indices.pop() {
self.path.truncate(last_idx);
}
}
/// Clear the path.
#[inline]
pub fn clear(&mut self) {
self.path.clear();
self.indices.clear();
}
/// Get the current path.
#[inline]
pub fn as_str(&self) -> &str {
&self.path
}
}
impl From<JsonPathWriter> for String {
#[inline]
fn from(value: JsonPathWriter) -> Self {
value.path
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn json_path_writer_test() {
let mut writer = JsonPathWriter::new();
writer.push("root");
assert_eq!(writer.as_str(), "root");
writer.push("child");
assert_eq!(writer.as_str(), "root\u{1}child");
writer.pop();
assert_eq!(writer.as_str(), "root");
writer.push("k8s.node.id");
assert_eq!(writer.as_str(), "root\u{1}k8s.node.id");
writer.set_expand_dots(true);
writer.pop();
writer.push("k8s.node.id");
assert_eq!(writer.as_str(), "root\u{1}k8s\u{1}node\u{1}id");
}
}

10
common/src/lib.rs
View File

@@ -7,21 +7,22 @@ pub use byteorder::LittleEndian as Endianness;
mod bitset;
mod byte_count;
mod datetime;
mod dictionary_footer;
pub mod file_slice;
mod group_by;
mod json_path_writer;
mod serialize;
mod vint;
mod writer;
pub use bitset::*;
pub use byte_count::ByteCount;
pub use datetime::{DateTime, DateTimePrecision};
pub use datetime::{DatePrecision, DateTime};
pub use dictionary_footer::*;
pub use group_by::GroupByIteratorExtended;
pub use json_path_writer::JsonPathWriter;
pub use ownedbytes::{OwnedBytes, StableDeref};
pub use serialize::{BinarySerializable, DeserializeFrom, FixedSize};
pub use vint::{
read_u32_vint, read_u32_vint_no_advance, serialize_vint_u32, write_u32_vint, VInt, VIntU128,
deserialize_vint_u128, read_u32_vint, read_u32_vint_no_advance, serialize_vint_u128,
serialize_vint_u32, write_u32_vint, VInt, VIntU128,
};
pub use writer::{AntiCallToken, CountingWriter, TerminatingWrite};
@@ -116,7 +117,6 @@ pub fn u64_to_f64(val: u64) -> f64 {
///
/// This function assumes that the needle is rarely contained in the bytes string
/// and offers a fast path if the needle is not present.
#[inline]
pub fn replace_in_place(needle: u8, replacement: u8, bytes: &mut [u8]) {
if !bytes.contains(&needle) {
return;

41
common/src/serialize.rs
View File

@@ -1,4 +1,3 @@
use std::borrow::Cow;
use std::io::{Read, Write};
use std::{fmt, io};
@@ -250,47 +249,11 @@ impl BinarySerializable for String {
}
}
impl<'a> BinarySerializable for Cow<'a, str> {
fn serialize<W: Write + ?Sized>(&self, writer: &mut W) -> io::Result<()> {
let data: &[u8] = self.as_bytes();
VInt(data.len() as u64).serialize(writer)?;
writer.write_all(data)
}
fn deserialize<R: Read>(reader: &mut R) -> io::Result<Cow<'a, str>> {
let string_length = VInt::deserialize(reader)?.val() as usize;
let mut result = String::with_capacity(string_length);
reader
.take(string_length as u64)
.read_to_string(&mut result)?;
Ok(Cow::Owned(result))
}
}
impl<'a> BinarySerializable for Cow<'a, [u8]> {
fn serialize<W: Write + ?Sized>(&self, writer: &mut W) -> io::Result<()> {
VInt(self.len() as u64).serialize(writer)?;
for it in self.iter() {
it.serialize(writer)?;
}
Ok(())
}
fn deserialize<R: Read>(reader: &mut R) -> io::Result<Cow<'a, [u8]>> {
let num_items = VInt::deserialize(reader)?.val();
let mut items: Vec<u8> = Vec::with_capacity(num_items as usize);
for _ in 0..num_items {
let item = u8::deserialize(reader)?;
items.push(item);
}
Ok(Cow::Owned(items))
}
}
#[cfg(test)]
pub mod test {
use super::*;
use super::{VInt, *};
use crate::serialize::BinarySerializable;
pub fn fixed_size_test<O: BinarySerializable + FixedSize + Default>() {
let mut buffer = Vec::new();
O::default().serialize(&mut buffer).unwrap();

77
common/src/vint.rs
View File

@@ -1,6 +1,8 @@
use std::io;
use std::io::{Read, Write};
use byteorder::{ByteOrder, LittleEndian};
use super::BinarySerializable;
/// Variable int serializes a u128 number
@@ -17,6 +19,26 @@ pub fn serialize_vint_u128(mut val: u128, output: &mut Vec<u8>) {
}
}
/// Deserializes a u128 number
///
/// Returns the number and the slice after the vint
pub fn deserialize_vint_u128(data: &[u8]) -> io::Result<(u128, &[u8])> {
let mut result = 0u128;
let mut shift = 0u64;
for i in 0..19 {
let b = data[i];
result |= u128::from(b % 128u8) << shift;
if b >= STOP_BIT {
return Ok((result, &data[i + 1..]));
}
shift += 7;
}
Err(io::Error::new(
io::ErrorKind::InvalidData,
"Failed to deserialize u128 vint",
))
}
/// Wrapper over a `u128` that serializes as a variable int.
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
pub struct VIntU128(pub u128);
@@ -58,13 +80,17 @@ pub struct VInt(pub u64);
const STOP_BIT: u8 = 128;
#[inline]
pub fn serialize_vint_u32(val: u32, buf: &mut [u8; 8]) -> &[u8] {
const START_2: u64 = 1 << 7;
const START_3: u64 = 1 << 14;
const START_4: u64 = 1 << 21;
const START_5: u64 = 1 << 28;
const STOP_1: u64 = START_2 - 1;
const STOP_2: u64 = START_3 - 1;
const STOP_3: u64 = START_4 - 1;
const STOP_4: u64 = START_5 - 1;
const MASK_1: u64 = 127;
const MASK_2: u64 = MASK_1 << 7;
const MASK_3: u64 = MASK_2 << 7;
@@ -73,29 +99,25 @@ 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) = if val < START_2 {
(val | STOP_BIT, 1)
} else if val < START_3 {
(
let (res, num_bytes) = match val {
0..=STOP_1 => (val | STOP_BIT, 1),
START_2..=STOP_2 => (
(val & MASK_1) | ((val & MASK_2) << 1) | (STOP_BIT << (8)),
2,
)
} else if val < START_4 {
(
),
START_3..=STOP_3 => (
(val & MASK_1) | ((val & MASK_2) << 1) | ((val & MASK_3) << 2) | (STOP_BIT << (8 * 2)),
3,
)
} else if val < START_5 {
(
),
START_4..=STOP_4 => (
(val & MASK_1)
| ((val & MASK_2) << 1)
| ((val & MASK_3) << 2)
| ((val & MASK_4) << 3)
| (STOP_BIT << (8 * 3)),
4,
)
} else {
(
),
_ => (
(val & MASK_1)
| ((val & MASK_2) << 1)
| ((val & MASK_3) << 2)
@@ -103,9 +125,9 @@ pub fn serialize_vint_u32(val: u32, buf: &mut [u8; 8]) -> &[u8] {
| ((val & MASK_5) << 4)
| (STOP_BIT << (8 * 4)),
5,
)
),
};
*buf = res.to_le_bytes();
LittleEndian::write_u64(&mut buf[..], res);
&buf[0..num_bytes]
}
@@ -223,6 +245,7 @@ impl BinarySerializable for VInt {
mod tests {
use super::{serialize_vint_u32, BinarySerializable, VInt};
use crate::vint::{deserialize_vint_u128, serialize_vint_u128, VIntU128};
fn aux_test_vint(val: u64) {
let mut v = [14u8; 10];
@@ -261,7 +284,27 @@ mod tests {
let mut buffer2 = [0u8; 8];
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}");
assert_eq!(&buffer[..len_vint], res2, "array wrong for {}", val);
}
fn aux_test_vint_u128(val: u128) {
let mut data = vec![];
serialize_vint_u128(val, &mut data);
let (deser_val, _data) = deserialize_vint_u128(&data).unwrap();
assert_eq!(val, deser_val);
let mut out = vec![];
VIntU128(val).serialize(&mut out).unwrap();
let deser_val = VIntU128::deserialize(&mut &out[..]).unwrap();
assert_eq!(val, deser_val.0);
}
#[test]
fn test_vint_u128() {
aux_test_vint_u128(0);
aux_test_vint_u128(1);
aux_test_vint_u128(u128::MAX / 3);
aux_test_vint_u128(u128::MAX);
}
#[test]

65
examples/aggregation.rs
View File

@@ -7,12 +7,17 @@
// ---
use serde_json::{Deserializer, Value};
use tantivy::aggregation::agg_req::Aggregations;
use tantivy::aggregation::agg_req::{
Aggregation, Aggregations, BucketAggregation, BucketAggregationType, MetricAggregation,
RangeAggregation,
};
use tantivy::aggregation::agg_result::AggregationResults;
use tantivy::aggregation::bucket::RangeAggregationRange;
use tantivy::aggregation::metric::AverageAggregation;
use tantivy::aggregation::AggregationCollector;
use tantivy::query::AllQuery;
use tantivy::schema::{self, IndexRecordOption, Schema, TextFieldIndexing, FAST};
use tantivy::{Index, IndexWriter, TantivyDocument};
use tantivy::Index;
fn main() -> tantivy::Result<()> {
// # Create Schema
@@ -132,10 +137,10 @@ fn main() -> tantivy::Result<()> {
let stream = Deserializer::from_str(data).into_iter::<Value>();
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let mut index_writer = index.writer(50_000_000)?;
let mut num_indexed = 0;
for value in stream {
let doc = TantivyDocument::parse_json(&schema, &serde_json::to_string(&value.unwrap())?)?;
let doc = schema.parse_document(&serde_json::to_string(&value.unwrap())?)?;
index_writer.add_document(doc)?;
num_indexed += 1;
if num_indexed > 4 {
@@ -189,9 +194,56 @@ fn main() -> tantivy::Result<()> {
let agg_req: Aggregations = serde_json::from_str(agg_req_str)?;
let collector = AggregationCollector::from_aggs(agg_req, Default::default());
let agg_res: AggregationResults = searcher.search(&AllQuery, &collector).unwrap();
let res2: Value = serde_json::to_value(agg_res)?;
// ### Request Rust API
//
// This is exactly the same request as above, but via the rust structures.
//
let agg_req: Aggregations = vec![(
"group_by_stock".to_string(),
Aggregation::Bucket(Box::new(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "stock".to_string(),
ranges: vec![
RangeAggregationRange {
key: Some("few".into()),
from: None,
to: Some(1f64),
},
RangeAggregationRange {
key: Some("some".into()),
from: Some(1f64),
to: Some(10f64),
},
RangeAggregationRange {
key: Some("many".into()),
from: Some(10f64),
to: None,
},
],
..Default::default()
}),
sub_aggregation: vec![(
"average_price".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("price".to_string()),
)),
)]
.into_iter()
.collect(),
})),
)]
.into_iter()
.collect();
let collector = AggregationCollector::from_aggs(agg_req, Default::default());
// We use the `AllQuery` which will pass all documents to the AggregationCollector.
let agg_res: AggregationResults = searcher.search(&AllQuery, &collector).unwrap();
let res: Value = serde_json::to_value(agg_res)?;
let res1: Value = serde_json::to_value(agg_res)?;
// ### Aggregation Result
//
@@ -209,7 +261,8 @@ fn main() -> tantivy::Result<()> {
}
"#;
let expected_json: Value = serde_json::from_str(expected_res)?;
assert_eq!(expected_json, res);
assert_eq!(expected_json, res1);
assert_eq!(expected_json, res2);
// ### Request 2
//

26
examples/basic_search.rs
View File

@@ -15,7 +15,7 @@
use tantivy::collector::TopDocs;
use tantivy::query::QueryParser;
use tantivy::schema::*;
use tantivy::{doc, Index, IndexWriter, ReloadPolicy};
use tantivy::{doc, Index, ReloadPolicy};
use tempfile::TempDir;
fn main() -> tantivy::Result<()> {
@@ -75,7 +75,7 @@ fn main() -> tantivy::Result<()> {
// Here we give tantivy a budget of `50MB`.
// Using a bigger memory_arena for the indexer may increase
// throughput, but 50 MB is already plenty.
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let mut index_writer = index.writer(50_000_000)?;
// Let's index our documents!
// We first need a handle on the title and the body field.
@@ -87,7 +87,7 @@ fn main() -> tantivy::Result<()> {
let title = schema.get_field("title").unwrap();
let body = schema.get_field("body").unwrap();
let mut old_man_doc = TantivyDocument::default();
let mut old_man_doc = Document::default();
old_man_doc.add_text(title, "The Old Man and the Sea");
old_man_doc.add_text(
body,
@@ -164,7 +164,7 @@ fn main() -> tantivy::Result<()> {
// will reload the index automatically after each commit.
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::OnCommitWithDelay)
.reload_policy(ReloadPolicy::OnCommit)
.try_into()?;
// We now need to acquire a searcher.
@@ -217,23 +217,9 @@ fn main() -> tantivy::Result<()> {
// the document returned will only contain
// a title.
for (_score, doc_address) in top_docs {
let retrieved_doc: TantivyDocument = searcher.doc(doc_address)?;
println!("{}", retrieved_doc.to_json(&schema));
let retrieved_doc = searcher.doc(doc_address)?;
println!("{}", schema.to_json(&retrieved_doc));
}
// We can also get an explanation to understand
// how a found document got its score.
let query = query_parser.parse_query("title:sea^20 body:whale^70")?;
let (_score, doc_address) = searcher
.search(&query, &TopDocs::with_limit(1))?
.into_iter()
.next()
.unwrap();
let explanation = query.explain(&searcher, doc_address)?;
println!("{}", explanation.to_pretty_json());
Ok(())
}

4
examples/custom_collector.rs
View File

@@ -13,7 +13,7 @@ use columnar::Column;
use tantivy::collector::{Collector, SegmentCollector};
use tantivy::query::QueryParser;
use tantivy::schema::{Schema, FAST, INDEXED, TEXT};
use tantivy::{doc, Index, IndexWriter, Score, SegmentReader};
use tantivy::{doc, Index, Score, SegmentReader};
#[derive(Default)]
struct Stats {
@@ -142,7 +142,7 @@ fn main() -> tantivy::Result<()> {
// this example.
let index = Index::create_in_ram(schema);
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let mut index_writer = index.writer(50_000_000)?;
index_writer.add_document(doc!(
product_name => "Super Broom 2000",
product_description => "While it is ok for short distance travel, this broom \

10
examples/custom_tokenizer.rs
View File

@@ -6,7 +6,7 @@ use tantivy::collector::TopDocs;
use tantivy::query::QueryParser;
use tantivy::schema::*;
use tantivy::tokenizer::NgramTokenizer;
use tantivy::{doc, Index, IndexWriter};
use tantivy::{doc, Index};
fn main() -> tantivy::Result<()> {
// # Defining the schema
@@ -53,7 +53,7 @@ fn main() -> tantivy::Result<()> {
// this will store tokens of 3 characters each
index
.tokenizers()
.register("ngram3", NgramTokenizer::new(3, 3, false).unwrap());
.register("ngram3", NgramTokenizer::new(3, 3, false));
// To insert document we need an index writer.
// There must be only one writer at a time.
@@ -62,7 +62,7 @@ fn main() -> tantivy::Result<()> {
//
// Here we use a buffer of 50MB per thread. Using a bigger
// memory arena for the indexer can increase its throughput.
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let mut index_writer = index.writer(50_000_000)?;
index_writer.add_document(doc!(
title => "The Old Man and the Sea",
body => "He was an old man who fished alone in a skiff in the Gulf Stream and \
@@ -103,8 +103,8 @@ fn main() -> tantivy::Result<()> {
let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
for (_, doc_address) in top_docs {
let retrieved_doc: TantivyDocument = searcher.doc(doc_address)?;
println!("{}", retrieved_doc.to_json(&schema));
let retrieved_doc = searcher.doc(doc_address)?;
println!("{}", schema.to_json(&retrieved_doc));
}
Ok(())

20
examples/date_time_field.rs
View File

@@ -4,8 +4,8 @@
use tantivy::collector::TopDocs;
use tantivy::query::QueryParser;
use tantivy::schema::{DateOptions, Document, OwnedValue, Schema, INDEXED, STORED, STRING};
use tantivy::{Index, IndexWriter, TantivyDocument};
use tantivy::schema::{DateOptions, Schema, Value, INDEXED, STORED, STRING};
use tantivy::Index;
fn main() -> tantivy::Result<()> {
// # Defining the schema
@@ -13,7 +13,7 @@ fn main() -> tantivy::Result<()> {
let opts = DateOptions::from(INDEXED)
.set_stored()
.set_fast()
.set_precision(tantivy::DateTimePrecision::Seconds);
.set_precision(tantivy::DatePrecision::Seconds);
// Add `occurred_at` date field type
let occurred_at = schema_builder.add_date_field("occurred_at", opts);
let event_type = schema_builder.add_text_field("event", STRING | STORED);
@@ -22,18 +22,16 @@ fn main() -> tantivy::Result<()> {
// # Indexing documents
let index = Index::create_in_ram(schema.clone());
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let mut index_writer = index.writer(50_000_000)?;
// The dates are passed as string in the RFC3339 format
let doc = TantivyDocument::parse_json(
&schema,
let doc = schema.parse_document(
r#"{
"occurred_at": "2022-06-22T12:53:50.53Z",
"event": "pull-request"
}"#,
)?;
index_writer.add_document(doc)?;
let doc = TantivyDocument::parse_json(
&schema,
let doc = schema.parse_document(
r#"{
"occurred_at": "2022-06-22T13:00:00.22Z",
"event": "comment"
@@ -60,13 +58,13 @@ fn main() -> tantivy::Result<()> {
let count_docs = searcher.search(&*query, &TopDocs::with_limit(4))?;
assert_eq!(count_docs.len(), 1);
for (_score, doc_address) in count_docs {
let retrieved_doc = searcher.doc::<TantivyDocument>(doc_address)?;
let retrieved_doc = searcher.doc(doc_address)?;
assert!(matches!(
retrieved_doc.get_first(occurred_at),
Some(OwnedValue::Date(_))
Some(Value::Date(_))
));
assert_eq!(
retrieved_doc.to_json(&schema),
schema.to_json(&retrieved_doc),
r#"{"event":["comment"],"occurred_at":["2022-06-22T13:00:00.22Z"]}"#
);
}

12
examples/deleting_updating_documents.rs
View File

@@ -11,7 +11,7 @@
use tantivy::collector::TopDocs;
use tantivy::query::TermQuery;
use tantivy::schema::*;
use tantivy::{doc, Index, IndexReader, IndexWriter};
use tantivy::{doc, Index, IndexReader};
// A simple helper function to fetch a single document
// given its id from our index.
@@ -19,7 +19,7 @@ use tantivy::{doc, Index, IndexReader, IndexWriter};
fn extract_doc_given_isbn(
reader: &IndexReader,
isbn_term: &Term,
) -> tantivy::Result<Option<TantivyDocument>> {
) -> tantivy::Result<Option<Document>> {
let searcher = reader.searcher();
// This is the simplest query you can think of.
@@ -69,10 +69,10 @@ fn main() -> tantivy::Result<()> {
let index = Index::create_in_ram(schema.clone());
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let mut index_writer = index.writer(50_000_000)?;
// Let's add a couple of documents, for the sake of the example.
let mut old_man_doc = TantivyDocument::default();
let mut old_man_doc = Document::default();
old_man_doc.add_text(title, "The Old Man and the Sea");
index_writer.add_document(doc!(
isbn => "978-0099908401",
@@ -94,7 +94,7 @@ fn main() -> tantivy::Result<()> {
// Oops our frankenstein doc seems misspelled
let frankenstein_doc_misspelled = extract_doc_given_isbn(&reader, &frankenstein_isbn)?.unwrap();
assert_eq!(
frankenstein_doc_misspelled.to_json(&schema),
schema.to_json(&frankenstein_doc_misspelled),
r#"{"isbn":["978-9176370711"],"title":["Frankentein"]}"#,
);
@@ -136,7 +136,7 @@ fn main() -> tantivy::Result<()> {
// No more typo!
let frankenstein_new_doc = extract_doc_given_isbn(&reader, &frankenstein_isbn)?.unwrap();
assert_eq!(
frankenstein_new_doc.to_json(&schema),
schema.to_json(&frankenstein_new_doc),
r#"{"isbn":["978-9176370711"],"title":["Frankenstein"]}"#,
);

4
examples/faceted_search.rs
View File

@@ -17,7 +17,7 @@
use tantivy::collector::FacetCollector;
use tantivy::query::{AllQuery, TermQuery};
use tantivy::schema::*;
use tantivy::{doc, Index, IndexWriter};
use tantivy::{doc, Index};
fn main() -> tantivy::Result<()> {
// Let's create a temporary directory for the sake of this example
@@ -30,7 +30,7 @@ fn main() -> tantivy::Result<()> {
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer: IndexWriter = index.writer(30_000_000)?;
let mut index_writer = index.writer(30_000_000)?;
// For convenience, tantivy also comes with a macro to
// reduce the boilerplate above.

9
examples/faceted_search_with_tweaked_score.rs
View File

@@ -12,7 +12,7 @@ use std::collections::HashSet;
use tantivy::collector::TopDocs;
use tantivy::query::BooleanQuery;
use tantivy::schema::*;
use tantivy::{doc, DocId, Index, IndexWriter, Score, SegmentReader};
use tantivy::{doc, DocId, Index, Score, SegmentReader};
fn main() -> tantivy::Result<()> {
let mut schema_builder = Schema::builder();
@@ -23,7 +23,7 @@ fn main() -> tantivy::Result<()> {
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer: IndexWriter = index.writer(30_000_000)?;
let mut index_writer = index.writer(30_000_000)?;
index_writer.add_document(doc!(
title => "Fried egg",
@@ -91,10 +91,11 @@ fn main() -> tantivy::Result<()> {
.iter()
.map(|(_, doc_id)| {
searcher
.doc::<TantivyDocument>(*doc_id)
.doc(*doc_id)
.unwrap()
.get_first(title)
.and_then(|v| v.as_str())
.unwrap()
.as_text()
.unwrap()
.to_owned()
})

10
examples/fuzzy_search.rs
View File

@@ -14,7 +14,7 @@
use tantivy::collector::{Count, TopDocs};
use tantivy::query::FuzzyTermQuery;
use tantivy::schema::*;
use tantivy::{doc, Index, IndexWriter, ReloadPolicy};
use tantivy::{doc, Index, ReloadPolicy};
use tempfile::TempDir;
fn main() -> tantivy::Result<()> {
@@ -66,7 +66,7 @@ fn main() -> tantivy::Result<()> {
// Here we give tantivy a budget of `50MB`.
// Using a bigger memory_arena for the indexer may increase
// throughput, but 50 MB is already plenty.
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let mut index_writer = index.writer(50_000_000)?;
// Let's index our documents!
// We first need a handle on the title and the body field.
@@ -123,7 +123,7 @@ fn main() -> tantivy::Result<()> {
// will reload the index automatically after each commit.
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::OnCommitWithDelay)
.reload_policy(ReloadPolicy::OnCommit)
.try_into()?;
// We now need to acquire a searcher.
@@ -151,10 +151,10 @@ fn main() -> tantivy::Result<()> {
assert_eq!(count, 3);
assert_eq!(top_docs.len(), 3);
for (score, doc_address) in top_docs {
let retrieved_doc = searcher.doc(doc_address)?;
// Note that the score is not lower for the fuzzy hit.
// There's an issue open for that: https://github.com/quickwit-oss/tantivy/issues/563
let retrieved_doc: TantivyDocument = searcher.doc(doc_address)?;
println!("score {score:?} doc {}", retrieved_doc.to_json(&schema));
println!("score {score:?} doc {}", schema.to_json(&retrieved_doc));
// score 1.0 doc {"title":["The Diary of Muadib"]}
//
// score 1.0 doc {"title":["The Diary of a Young Girl"]}

2
examples/index_from_multiple_threads.rs
View File

@@ -96,7 +96,7 @@ fn main() -> tantivy::Result<()> {
let mut index_writer_wlock = index_writer.write().unwrap();
index_writer_wlock.commit()?
};
println!("committed with opstamp {opstamp}");
println!("committed with opstamp {}", opstamp);
thread::sleep(Duration::from_millis(500));
}

4
examples/index_with_json.rs
View File

@@ -21,7 +21,7 @@ fn main() -> tantivy::Result<()> {
}"#;
// We can parse our document
let _mice_and_men_doc = TantivyDocument::parse_json(&schema, mice_and_men_doc_json)?;
let _mice_and_men_doc = schema.parse_document(mice_and_men_doc_json)?;
// Multi-valued field are allowed, they are
// expressed in JSON by an array.
@@ -30,7 +30,7 @@ fn main() -> tantivy::Result<()> {
"title": ["Frankenstein", "The Modern Prometheus"],
"year": 1818
}"#;
let _frankenstein_doc = TantivyDocument::parse_json(&schema, frankenstein_json)?;
let _frankenstein_doc = schema.parse_document(frankenstein_json)?;
// Note that the schema is saved in your index directory.
//

4
examples/integer_range_search.rs
View File

@@ -5,7 +5,7 @@
use tantivy::collector::Count;
use tantivy::query::RangeQuery;
use tantivy::schema::{Schema, INDEXED};
use tantivy::{doc, Index, IndexWriter, Result};
use tantivy::{doc, Index, Result};
fn main() -> Result<()> {
// For the sake of simplicity, this schema will only have 1 field
@@ -17,7 +17,7 @@ fn main() -> Result<()> {
let index = Index::create_in_ram(schema);
let reader = index.reader()?;
{
let mut index_writer: IndexWriter = index.writer_with_num_threads(1, 6_000_000)?;
let mut index_writer = index.writer_with_num_threads(1, 6_000_000)?;
for year in 1950u64..2019u64 {
index_writer.add_document(doc!(year_field => year))?;
}

13
examples/ip_field.rs
View File

@@ -6,7 +6,7 @@
use tantivy::collector::{Count, TopDocs};
use tantivy::query::QueryParser;
use tantivy::schema::{Schema, FAST, INDEXED, STORED, STRING};
use tantivy::{Index, IndexWriter, TantivyDocument};
use tantivy::Index;
fn main() -> tantivy::Result<()> {
// # Defining the schema
@@ -22,22 +22,20 @@ fn main() -> tantivy::Result<()> {
// # Indexing documents
let index = Index::create_in_ram(schema.clone());
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let mut index_writer = index.writer(50_000_000)?;
// ### IPv4
// Adding documents that contain an IPv4 address. Notice that the IP addresses are passed as
// `String`. Since the field is of type ip, we parse the IP address from the string and store it
// internally as IPv6.
let doc = TantivyDocument::parse_json(
&schema,
let doc = schema.parse_document(
r#"{
"ip": "192.168.0.33",
"event_type": "login"
}"#,
)?;
index_writer.add_document(doc)?;
let doc = TantivyDocument::parse_json(
&schema,
let doc = schema.parse_document(
r#"{
"ip": "192.168.0.80",
"event_type": "checkout"
@@ -46,8 +44,7 @@ fn main() -> tantivy::Result<()> {
index_writer.add_document(doc)?;
// ### IPv6
// Adding a document that contains an IPv6 address.
let doc = TantivyDocument::parse_json(
&schema,
let doc = schema.parse_document(
r#"{
"ip": "2001:0db8:85a3:0000:0000:8a2e:0370:7334",
"event_type": "checkout"

8
examples/iterating_docs_and_positions.rs
View File

@@ -10,7 +10,7 @@
// ---
// Importing tantivy...
use tantivy::schema::*;
use tantivy::{doc, DocSet, Index, IndexWriter, Postings, TERMINATED};
use tantivy::{doc, DocSet, Index, Postings, TERMINATED};
fn main() -> tantivy::Result<()> {
// We first create a schema for the sake of the
@@ -24,7 +24,7 @@ fn main() -> tantivy::Result<()> {
let index = Index::create_in_ram(schema);
let mut index_writer: IndexWriter = index.writer_with_num_threads(1, 50_000_000)?;
let mut index_writer = index.writer_with_num_threads(1, 50_000_000)?;
index_writer.add_document(doc!(title => "The Old Man and the Sea"))?;
index_writer.add_document(doc!(title => "Of Mice and Men"))?;
index_writer.add_document(doc!(title => "The modern Promotheus"))?;
@@ -84,7 +84,7 @@ fn main() -> tantivy::Result<()> {
// Doc 0: TermFreq 2: [0, 4]
// Doc 2: TermFreq 1: [0]
// ```
println!("Doc {doc_id}: TermFreq {term_freq}: {positions:?}");
println!("Doc {}: TermFreq {}: {:?}", doc_id, term_freq, positions);
doc_id = segment_postings.advance();
}
}
@@ -125,7 +125,7 @@ fn main() -> tantivy::Result<()> {
// Once again these docs MAY contains deleted documents as well.
let docs = block_segment_postings.docs();
// Prints `Docs [0, 2].`
println!("Docs {docs:?}");
println!("Docs {:?}", docs);
block_segment_postings.advance();
}
}

10
examples/json_field.rs
View File

@@ -7,7 +7,7 @@
use tantivy::collector::{Count, TopDocs};
use tantivy::query::QueryParser;
use tantivy::schema::{Schema, FAST, STORED, STRING, TEXT};
use tantivy::{Index, IndexWriter, TantivyDocument};
use tantivy::Index;
fn main() -> tantivy::Result<()> {
// # Defining the schema
@@ -20,9 +20,8 @@ fn main() -> tantivy::Result<()> {
// # Indexing documents
let index = Index::create_in_ram(schema.clone());
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let doc = TantivyDocument::parse_json(
&schema,
let mut index_writer = index.writer(50_000_000)?;
let doc = schema.parse_document(
r#"{
"timestamp": "2022-02-22T23:20:50.53Z",
"event_type": "click",
@@ -34,8 +33,7 @@ fn main() -> tantivy::Result<()> {
}"#,
)?;
index_writer.add_document(doc)?;
let doc = TantivyDocument::parse_json(
&schema,
let doc = schema.parse_document(
r#"{
"timestamp": "2022-02-22T23:20:51.53Z",
"event_type": "click",

83
examples/phrase_prefix_search.rs
View File

@@ -1,83 +0,0 @@
use tantivy::collector::TopDocs;
use tantivy::query::QueryParser;
use tantivy::schema::*;
use tantivy::{doc, Index, IndexWriter, ReloadPolicy, Result};
use tempfile::TempDir;
fn main() -> Result<()> {
let index_path = TempDir::new()?;
let mut schema_builder = Schema::builder();
schema_builder.add_text_field("title", TEXT | STORED);
schema_builder.add_text_field("body", TEXT);
let schema = schema_builder.build();
let title = schema.get_field("title").unwrap();
let body = schema.get_field("body").unwrap();
let index = Index::create_in_dir(&index_path, schema)?;
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
index_writer.add_document(doc!(
title => "The Old Man and the Sea",
body => "He was an old man who fished alone in a skiff in the Gulf Stream and he had gone \
eighty-four days now without taking a fish.",
))?;
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",
title => "The Modern Prometheus",
body => "You will rejoice to hear that no disaster has accompanied the commencement of an \
enterprise which you have regarded with such evil forebodings. I arrived here \
yesterday, and my first task is to assure my dear sister of my welfare and \
increasing confidence in the success of my undertaking."
))?;
index_writer.commit()?;
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::OnCommitWithDelay)
.try_into()?;
let searcher = reader.searcher();
let query_parser = QueryParser::for_index(&index, vec![title, body]);
// This will match documents containing the phrase "in the"
// followed by some word starting with "su",
// i.e. it will match "in the sunlight" and "in the success",
// but not "in the Gulf Stream".
let query = query_parser.parse_query("\"in the su\"*")?;
let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
let mut titles = top_docs
.into_iter()
.map(|(_score, doc_address)| {
let doc = searcher.doc::<TantivyDocument>(doc_address)?;
let title = doc
.get_first(title)
.and_then(|v| v.as_str())
.unwrap()
.to_owned();
Ok(title)
})
.collect::<Result<Vec<_>>>()?;
titles.sort_unstable();
assert_eq!(titles, ["Frankenstein", "Of Mice and Men"]);
Ok(())
}

15
examples/pre_tokenized_text.rs
View File

@@ -13,12 +13,11 @@ use tantivy::collector::{Count, TopDocs};
use tantivy::query::TermQuery;
use tantivy::schema::*;
use tantivy::tokenizer::{PreTokenizedString, SimpleTokenizer, Token, TokenStream, Tokenizer};
use tantivy::{doc, Index, IndexWriter, ReloadPolicy};
use tantivy::{doc, Index, ReloadPolicy};
use tempfile::TempDir;
fn pre_tokenize_text(text: &str) -> Vec<Token> {
let mut tokenizer = SimpleTokenizer::default();
let mut token_stream = tokenizer.token_stream(text);
let mut token_stream = SimpleTokenizer.token_stream(text);
let mut tokens = vec![];
while token_stream.advance() {
tokens.push(token_stream.token().clone());
@@ -38,7 +37,7 @@ fn main() -> tantivy::Result<()> {
let index = Index::create_in_dir(&index_path, schema.clone())?;
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let mut index_writer = index.writer(50_000_000)?;
// We can create a document manually, by setting the fields
// one by one in a Document object.
@@ -83,7 +82,7 @@ fn main() -> tantivy::Result<()> {
}]
}"#;
let short_man_doc = TantivyDocument::parse_json(&schema, short_man_json)?;
let short_man_doc = schema.parse_document(short_man_json)?;
index_writer.add_document(short_man_doc)?;
@@ -94,7 +93,7 @@ fn main() -> tantivy::Result<()> {
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::OnCommitWithDelay)
.reload_policy(ReloadPolicy::OnCommit)
.try_into()?;
let searcher = reader.searcher();
@@ -115,8 +114,8 @@ fn main() -> tantivy::Result<()> {
// Note that the tokens are not stored along with the original text
// in the document store
for (_score, doc_address) in top_docs {
let retrieved_doc: TantivyDocument = searcher.doc(doc_address)?;
println!("{}", retrieved_doc.to_json(&schema));
let retrieved_doc = searcher.doc(doc_address)?;
println!("Document: {}", schema.to_json(&retrieved_doc));
}
// In contrary to the previous query, when we search for the "man" term we

14
examples/snippet.rs
View File

@@ -10,8 +10,7 @@
use tantivy::collector::TopDocs;
use tantivy::query::QueryParser;
use tantivy::schema::*;
use tantivy::snippet::{Snippet, SnippetGenerator};
use tantivy::{doc, Index, IndexWriter};
use tantivy::{doc, Index, Snippet, SnippetGenerator};
use tempfile::TempDir;
fn main() -> tantivy::Result<()> {
@@ -28,7 +27,7 @@ fn main() -> tantivy::Result<()> {
// # Indexing documents
let index = Index::create_in_dir(&index_path, schema)?;
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let mut index_writer = index.writer(50_000_000)?;
// we'll only need one doc for this example.
index_writer.add_document(doc!(
@@ -55,10 +54,13 @@ fn main() -> tantivy::Result<()> {
let snippet_generator = SnippetGenerator::create(&searcher, &*query, body)?;
for (score, doc_address) in top_docs {
let doc = searcher.doc::<TantivyDocument>(doc_address)?;
let doc = searcher.doc(doc_address)?;
let snippet = snippet_generator.snippet_from_doc(&doc);
println!("Document score {score}:");
println!("title: {}", doc.get_first(title).unwrap().as_str().unwrap());
println!("Document score {}:", score);
println!(
"title: {}",
doc.get_first(title).unwrap().as_text().unwrap()
);
println!("snippet: {}", snippet.to_html());
println!("custom highlighting: {}", highlight(snippet));
}

12
examples/stop_words.rs
View File

@@ -15,7 +15,7 @@ use tantivy::collector::TopDocs;
use tantivy::query::QueryParser;
use tantivy::schema::*;
use tantivy::tokenizer::*;
use tantivy::{doc, Index, IndexWriter};
use tantivy::{doc, Index};
fn main() -> tantivy::Result<()> {
// this example assumes you understand the content in `basic_search`
@@ -50,7 +50,7 @@ fn main() -> tantivy::Result<()> {
// This tokenizer lowers all of the text (to help with stop word matching)
// then removes all instances of `the` and `and` from the corpus
let tokenizer = TextAnalyzer::builder(SimpleTokenizer::default())
let tokenizer = TextAnalyzer::builder(SimpleTokenizer)
.filter(LowerCaser)
.filter(StopWordFilter::remove(vec![
"the".to_string(),
@@ -60,7 +60,7 @@ fn main() -> tantivy::Result<()> {
index.tokenizers().register("stoppy", tokenizer);
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let mut index_writer = index.writer(50_000_000)?;
let title = schema.get_field("title").unwrap();
let body = schema.get_field("body").unwrap();
@@ -105,9 +105,9 @@ fn main() -> tantivy::Result<()> {
let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
for (score, doc_address) in top_docs {
let retrieved_doc: TantivyDocument = searcher.doc(doc_address)?;
println!("\n==\nDocument score {score}:");
println!("{}", retrieved_doc.to_json(&schema));
let retrieved_doc = searcher.doc(doc_address)?;
println!("\n==\nDocument score {}:", score);
println!("{}", schema.to_json(&retrieved_doc));
}
Ok(())

75
examples/warmer.rs
View File

@@ -6,14 +6,12 @@ use tantivy::collector::TopDocs;
use tantivy::query::QueryParser;
use tantivy::schema::{Schema, FAST, TEXT};
use tantivy::{
doc, DocAddress, DocId, Index, IndexWriter, Opstamp, Searcher, SearcherGeneration, SegmentId,
doc, DocAddress, DocId, Index, IndexReader, Opstamp, Searcher, SearcherGeneration, SegmentId,
SegmentReader, Warmer,
};
// This example shows how warmers can be used to
// load values from an external sources and
// tie their lifecycle to that of the index segments
// using the Warmer API.
// load a values from an external sources using the Warmer API.
//
// In this example, we assume an e-commerce search engine.
@@ -25,11 +23,9 @@ pub trait PriceFetcher: Send + Sync + 'static {
fn fetch_prices(&self, product_ids: &[ProductId]) -> Vec<Price>;
}
type SegmentKey = (SegmentId, Option<Opstamp>);
struct DynamicPriceColumn {
field: String,
price_cache: RwLock<HashMap<SegmentKey, Arc<Vec<Price>>>>,
price_cache: RwLock<HashMap<(SegmentId, Option<Opstamp>), Arc<Vec<Price>>>>,
price_fetcher: Box<dyn PriceFetcher>,
}
@@ -50,6 +46,7 @@ impl DynamicPriceColumn {
impl Warmer for DynamicPriceColumn {
fn warm(&self, searcher: &Searcher) -> tantivy::Result<()> {
for segment in searcher.segment_readers() {
let key = (segment.segment_id(), segment.delete_opstamp());
let product_id_reader = segment
.fast_fields()
.u64(&self.field)?
@@ -58,40 +55,37 @@ impl Warmer for DynamicPriceColumn {
.doc_ids_alive()
.map(|doc| product_id_reader.get_val(doc))
.collect();
let mut prices = self.price_fetcher.fetch_prices(&product_ids).into_iter();
let prices: Vec<Price> = (0..segment.max_doc())
.map(|doc| {
if !segment.is_deleted(doc) {
prices.next().unwrap()
} else {
0
}
})
.collect();
let key = (segment.segment_id(), segment.delete_opstamp());
let mut prices_it = self.price_fetcher.fetch_prices(&product_ids).into_iter();
let mut price_vals: Vec<Price> = Vec::new();
for doc in 0..segment.max_doc() {
if segment.is_deleted(doc) {
price_vals.push(0);
} else {
price_vals.push(prices_it.next().unwrap())
}
}
self.price_cache
.write()
.unwrap()
.insert(key, Arc::new(prices));
.insert(key, Arc::new(price_vals));
}
Ok(())
}
fn garbage_collect(&self, live_generations: &[&SearcherGeneration]) {
let live_keys: HashSet<SegmentKey> = live_generations
.iter()
.flat_map(|gen| gen.segments())
.map(|(&segment_id, &opstamp)| (segment_id, opstamp))
let live_segment_id_and_delete_ops: HashSet<(SegmentId, Option<Opstamp>)> =
live_generations
.iter()
.flat_map(|gen| gen.segments())
.map(|(&segment_id, &opstamp)| (segment_id, opstamp))
.collect();
let mut price_cache_wrt = self.price_cache.write().unwrap();
// let price_cache = std::mem::take(&mut *price_cache_wrt);
// Drain would be nicer here.
*price_cache_wrt = std::mem::take(&mut *price_cache_wrt)
.into_iter()
.filter(|(seg_id_and_op, _)| !live_segment_id_and_delete_ops.contains(seg_id_and_op))
.collect();
self.price_cache
.write()
.unwrap()
.retain(|key, _| live_keys.contains(key));
}
}
@@ -106,17 +100,17 @@ pub struct ExternalPriceTable {
impl ExternalPriceTable {
pub fn update_price(&self, product_id: ProductId, price: Price) {
self.prices.write().unwrap().insert(product_id, price);
let mut prices_wrt = self.prices.write().unwrap();
prices_wrt.insert(product_id, price);
}
}
impl PriceFetcher for ExternalPriceTable {
fn fetch_prices(&self, product_ids: &[ProductId]) -> Vec<Price> {
let prices = self.prices.read().unwrap();
let prices_read = self.prices.read().unwrap();
product_ids
.iter()
.map(|product_id| prices.get(product_id).cloned().unwrap_or(0))
.map(|product_id| prices_read.get(product_id).cloned().unwrap_or(0))
.collect()
}
}
@@ -143,14 +137,17 @@ fn main() -> tantivy::Result<()> {
const SNEAKERS: ProductId = 23222;
let index = Index::create_in_ram(schema);
let mut writer: IndexWriter = index.writer_with_num_threads(1, 15_000_000)?;
let mut writer = index.writer_with_num_threads(1, 10_000_000)?;
writer.add_document(doc!(product_id=>OLIVE_OIL, text=>"cooking olive oil from greece"))?;
writer.add_document(doc!(product_id=>GLOVES, text=>"kitchen gloves, perfect for cooking"))?;
writer.add_document(doc!(product_id=>SNEAKERS, text=>"uber sweet sneakers"))?;
writer.commit()?;
let warmers = vec![Arc::downgrade(&price_dynamic_column) as Weak<dyn Warmer>];
let reader = index.reader_builder().warmers(warmers).try_into()?;
let warmers: Vec<Weak<dyn Warmer>> = vec![Arc::downgrade(
&(price_dynamic_column.clone() as Arc<dyn Warmer>),
)];
let reader: IndexReader = index.reader_builder().warmers(warmers).try_into()?;
reader.reload()?;
let query_parser = QueryParser::for_index(&index, vec![text]);
let query = query_parser.parse_query("cooking")?;

2
ownedbytes/Cargo.toml
View File

@@ -1,7 +1,7 @@
[package]
authors = ["Paul Masurel <paul@quickwit.io>", "Pascal Seitz <pascal@quickwit.io>"]
name = "ownedbytes"
version = "0.7.0"
version = "0.5.0"
edition = "2021"
description = "Expose data as static slice"
license = "MIT"

110
ownedbytes/src/lib.rs
View File

@@ -1,6 +1,7 @@
use std::convert::TryInto;
use std::ops::{Deref, Range};
use std::sync::Arc;
use std::{fmt, io};
use std::{fmt, io, mem};
pub use stable_deref_trait::StableDeref;
@@ -25,8 +26,8 @@ impl OwnedBytes {
data_holder: T,
) -> OwnedBytes {
let box_stable_deref = Arc::new(data_holder);
let bytes: &[u8] = box_stable_deref.deref();
let data = unsafe { &*(bytes as *const [u8]) };
let bytes: &[u8] = box_stable_deref.as_ref();
let data = unsafe { mem::transmute::<_, &'static [u8]>(bytes.deref()) };
OwnedBytes {
data,
box_stable_deref,
@@ -56,12 +57,6 @@ impl OwnedBytes {
self.data.len()
}
/// Returns true iff this `OwnedBytes` is empty.
#[inline]
pub fn is_empty(&self) -> bool {
self.data.is_empty()
}
/// Splits the OwnedBytes into two OwnedBytes `(left, right)`.
///
/// Left will hold `split_len` bytes.
@@ -73,14 +68,13 @@ impl OwnedBytes {
#[inline]
#[must_use]
pub fn split(self, split_len: usize) -> (OwnedBytes, OwnedBytes) {
let (left_data, right_data) = self.data.split_at(split_len);
let right_box_stable_deref = self.box_stable_deref.clone();
let left = OwnedBytes {
data: left_data,
data: &self.data[..split_len],
box_stable_deref: self.box_stable_deref,
};
let right = OwnedBytes {
data: right_data,
data: &self.data[split_len..],
box_stable_deref: right_box_stable_deref,
};
(left, right)
@@ -105,45 +99,45 @@ impl OwnedBytes {
///
/// `self` is truncated to `split_len`, left with the remaining bytes.
pub fn split_off(&mut self, split_len: usize) -> OwnedBytes {
let (left, right) = self.data.split_at(split_len);
let right_box_stable_deref = self.box_stable_deref.clone();
let right_piece = OwnedBytes {
data: right,
data: &self.data[split_len..],
box_stable_deref: right_box_stable_deref,
};
self.data = left;
self.data = &self.data[..split_len];
right_piece
}
/// Returns true iff this `OwnedBytes` is empty.
#[inline]
pub fn is_empty(&self) -> bool {
self.as_slice().is_empty()
}
/// Drops the left most `advance_len` bytes.
#[inline]
pub fn advance(&mut self, advance_len: usize) -> &[u8] {
let (data, rest) = self.data.split_at(advance_len);
self.data = rest;
data
pub fn advance(&mut self, advance_len: usize) {
self.data = &self.data[advance_len..]
}
/// Reads an `u8` from the `OwnedBytes` and advance by one byte.
#[inline]
pub fn read_u8(&mut self) -> u8 {
self.advance(1)[0]
}
assert!(!self.is_empty());
#[inline]
fn read_n<const N: usize>(&mut self) -> [u8; N] {
self.advance(N).try_into().unwrap()
}
/// Reads an `u32` encoded as little-endian from the `OwnedBytes` and advance by 4 bytes.
#[inline]
pub fn read_u32(&mut self) -> u32 {
u32::from_le_bytes(self.read_n())
let byte = self.as_slice()[0];
self.advance(1);
byte
}
/// Reads an `u64` encoded as little-endian from the `OwnedBytes` and advance by 8 bytes.
#[inline]
pub fn read_u64(&mut self) -> u64 {
u64::from_le_bytes(self.read_n())
assert!(self.len() > 7);
let octlet: [u8; 8] = self.as_slice()[..8].try_into().unwrap();
self.advance(8);
u64::from_le_bytes(octlet)
}
}
@@ -156,7 +150,7 @@ impl fmt::Debug for OwnedBytes {
} else {
self.as_slice()
};
write!(f, "OwnedBytes({bytes_truncated:?}, len={})", self.len())
write!(f, "OwnedBytes({:?}, len={})", bytes_truncated, self.len())
}
}
@@ -197,33 +191,32 @@ impl Deref for OwnedBytes {
}
}
impl AsRef<[u8]> for OwnedBytes {
#[inline]
fn as_ref(&self) -> &[u8] {
self.as_slice()
}
}
impl io::Read for OwnedBytes {
#[inline]
fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
let data_len = self.data.len();
let buf_len = buf.len();
if data_len >= buf_len {
let data = self.advance(buf_len);
buf.copy_from_slice(data);
Ok(buf_len)
} else {
buf[..data_len].copy_from_slice(self.data);
self.data = &[];
Ok(data_len)
}
let read_len = {
let data = self.as_slice();
if data.len() >= buf.len() {
let buf_len = buf.len();
buf.copy_from_slice(&data[..buf_len]);
buf.len()
} else {
let data_len = data.len();
buf[..data_len].copy_from_slice(data);
data_len
}
};
self.advance(read_len);
Ok(read_len)
}
#[inline]
fn read_to_end(&mut self, buf: &mut Vec<u8>) -> io::Result<usize> {
buf.extend(self.data);
let read_len = self.data.len();
self.data = &[];
let read_len = {
let data = self.as_slice();
buf.extend(data);
data.len()
};
self.advance(read_len);
Ok(read_len)
}
#[inline]
@@ -239,6 +232,13 @@ impl io::Read for OwnedBytes {
}
}
impl AsRef<[u8]> for OwnedBytes {
#[inline]
fn as_ref(&self) -> &[u8] {
self.as_slice()
}
}
#[cfg(test)]
mod tests {
use std::io::{self, Read};
@@ -249,12 +249,12 @@ mod tests {
fn test_owned_bytes_debug() {
let short_bytes = OwnedBytes::new(b"abcd".as_ref());
assert_eq!(
format!("{short_bytes:?}"),
format!("{:?}", short_bytes),
"OwnedBytes([97, 98, 99, 100], len=4)"
);
let long_bytes = OwnedBytes::new(b"abcdefghijklmnopq".as_ref());
assert_eq!(
format!("{long_bytes:?}"),
format!("{:?}", long_bytes),
"OwnedBytes([97, 98, 99, 100, 101, 102, 103, 104, 105, 106], len=17)"
);
}

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