rust/tantivy
1
0
Fork 0
mirror of https://github.com/quickwit-oss/tantivy.git synced 2025-12-31 14:32:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

Compare commits

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

91 Commits
warming ... fastfield-

Author SHA1 Message Date
François Massot
14d53851a8 Fix clippy, clean comment. 2022-03-28 01:17:07 +02:00
François Massot
2d176e66b6 Format. 2022-03-26 22:29:43 +01:00
François Massot
838a332db0 Fix fmt. 2022-03-26 21:33:08 +01:00
François Massot
defbd9139b Update fastfield codecs readme. 2022-03-26 21:33:08 +01:00
François Massot
0c87732459 Fix makefile. 2022-03-26 21:33:08 +01:00
François Massot
4d66a3f0a0 Put deprecated attributes on deprecated codecs. Clean. 2022-03-26 21:33:06 +01:00
François Massot
977f01a8a3 Deprecate linear and multilienar fast field coded, add piece wise and FOR. Update tests and clean. 2022-03-26 21:27:15 +01:00
François Massot
c14bdd26d4 Clean. 2022-03-26 21:18:13 +01:00
François Massot
3272f80171 Fix clippy. 2022-03-26 21:17:32 +01:00
François Massot
23d5ab5656 Rename new codecs. 2022-03-26 21:17:32 +01:00
François Massot
245ed5fed1 Add float dataset for comparing fast field codec. 2022-03-26 21:17:32 +01:00
François Massot
33bed01168 Clean frame of ref codec. 2022-03-26 21:17:32 +01:00
François Massot
17a5f4f0ff Seed random datasets in fast field codecs comparison. 2022-03-26 21:17:30 +01:00
François Massot
c969582308 Add frame of reference codecs. 2022-03-26 21:16:50 +01:00
François Massot
18d2ee5bb7 Add another multilinear interpolation and real world dataset. 2022-03-26 21:15:50 +01:00
Maxim Kraynyuchenko
447811c111 Update README following sections: features, benchmark illustration & FAQ. (#1318) 
* Updated features, benchmark illustration & FAQ.
* Updated README: Feat,Graph,Non-Feat,Companies,FAQ
 2022-03-23 10:02:09 +09:00
PSeitz
f29acf5d8c fix clippy (#1321) 2022-03-22 12:48:23 +09:00
Uwe Klotz
125707dbe0 Replace chrono with time (#1307) 
For date values `chrono` has been replaced with `time` 
- The `time` crate is re-exported as `tantivy::time` instead of `tantivy::chrono`.
- The type alias `tantivy::DateTime` has been removed.
- `Value::Date` wraps `time::PrimitiveDateTime` without time zone information.
- Internally date/time values are stored as seconds since UNIX epoch in UTC.
- Converting a `time::OffsetDateTime` to `Value::Date` implicitly converts the value into UTC.
If this is not desired do the time zone conversion yourself and use `time::PrimitiveDateTime`
directly instead.

Closes #1304
 2022-03-21 10:50:19 +09:00
Paul Masurel
46d5de920d Removes all usage of block_on, and use a oneshot channel instead. (#1315) 
* Removes all usage of block_on, and use a oneshot channel instead.

Calling `block_on` panics in certain context.
For instance, it panics when it is called in a the context of another
call to block.

Using it in tantivy is unnecessary. We replace it by a thin wrapper
around a oneshot channel that supports both async/sync.

* Removing needless uses of async in the API.

Co-authored-by: PSeitz <PSeitz@users.noreply.github.com>
 2022-03-18 16:54:58 +09:00
PSeitz
d2a7bcf217 fix fmt (#1317) 2022-03-18 15:53:27 +09:00
PSeitz
141b9aa245 Merge pull request #1306 from PSeitz/histogram 
add Histogram aggregation
 2022-03-18 05:03:46 +01:00
PSeitz
c5a6282fa8 Update src/aggregation/bucket/histogram/histogram.rs 
Co-authored-by: Paul Masurel <paul@quickwit.io>
 2022-03-18 04:55:31 +01:00
PSeitz
c0f524e1a3 Update src/aggregation/bucket/histogram/histogram.rs 
Co-authored-by: Paul Masurel <paul@quickwit.io>
 2022-03-18 04:55:25 +01:00
Paul Masurel
958b2bee08 Clippy comments (#1316) 2022-03-17 18:57:55 +09:00
Pascal Seitz
f619658e2c rename 2022-03-17 16:37:57 +08:00
Pascal Seitz
aa391bf843 refactor parameters 2022-03-17 16:28:37 +08:00
Pascal Seitz
47dcbdbeae handle empty results, empty indices, add tests 2022-03-17 10:24:34 +08:00
Pascal Seitz
691245bf20 make code more concise 2022-03-16 14:21:58 +08:00
Pascal Seitz
90798d4b39 address comments, add single bucket test 2022-03-16 13:58:13 +08:00
Pascal Seitz
0b6d9f90cf improve docs 2022-03-16 12:39:26 +08:00
PSeitz
8a5a12d961 add setter to json object options (#1311) 2022-03-16 10:36:30 +09:00
Pascal Seitz
e73542e2e8 Elasticsearch behaviour on hard/extended_bounds 2022-03-15 16:46:45 +08:00
Pascal Seitz
0262e44bbd merge_fruits pass by value 2022-03-15 12:59:22 +08:00
Pascal Seitz
613aad7a8a vec optional, improve performance 2022-03-14 21:29:07 +08:00
Pascal Seitz
1aa88b0c51 improve performance 2022-03-14 20:28:08 +08:00
Pascal Seitz
564fa38085 move sub_aggregations to own vec, use itertools minmax 2022-03-14 16:20:26 +08:00
dependabot[bot]
59ec21479f Update pprof requirement from 0.6 to 0.7 (#1305) 
Updates the requirements on [pprof](https://github.com/tikv/pprof-rs) to permit the latest version.
- [Release notes](https://github.com/tikv/pprof-rs/releases)
- [Changelog](https://github.com/tikv/pprof-rs/blob/master/CHANGELOG.md)
- [Commits](https://github.com/tikv/pprof-rs/commits)

---
updated-dependencies:
- dependency-name: pprof
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>

Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2022-03-14 13:57:22 +09:00
PSeitz
42283f9e91 fix error message UnknownTokenizer (#1308) 
closes #1303
 2022-03-14 13:54:47 +09:00
PSeitz
b105bf72e1 use defaults in meta.json (#1310) 
This change allows to have unset fields in meta.json and fall back to their defaults
Currently it is required to explicitly put e.g. fieldnorms: false
 2022-03-14 13:54:06 +09:00
Pascal Seitz
226f577803 Add Histogram aggregation 2022-03-11 21:52:07 +08:00
Paul Masurel
2e255c4bef Preparing for release 2022-03-09 09:59:08 +09:00
Paul Masurel
387592809f Updated CHANGELOG 2022-03-07 15:31:35 +09:00
Halvor Fladsrud Bø
cedced5bb0 Slop support for phrase queries (#1241) 
Closes #1068
 2022-03-07 15:29:18 +09:00
dependabot[bot]
d31f045872 Bump actions/checkout from 2 to 3 (#1300) 
Bumps [actions/checkout](https://github.com/actions/checkout) from 2 to 3.
- [Release notes](https://github.com/actions/checkout/releases)
- [Changelog](https://github.com/actions/checkout/blob/main/CHANGELOG.md)
- [Commits](https://github.com/actions/checkout/compare/v2...v3)

---
updated-dependencies:
- dependency-name: actions/checkout
  dependency-type: direct:production
  update-type: version-update:semver-major
...

Signed-off-by: dependabot[bot] <support@github.com>

Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2022-03-07 11:54:26 +09:00
PSeitz
6656a70d1b Merge pull request #1301 from saroh/1232-doc-fastfield 
update fastfield doc
 2022-03-04 08:18:21 +01:00
saroh
d36e0a9549 fix fastfield doc 2022-03-03 17:43:18 +01:00
Antoine G
8771b2673f Update src/fastfield/writer.rs 
Co-authored-by: PSeitz <PSeitz@users.noreply.github.com>
 2022-03-03 11:25:24 +01:00
Antoine G
a41d3d51a4 Update fastfield_codecs/src/lib.rs 2022-03-03 11:25:06 +01:00
Saroh
cae34ffe47 update fastfield doc 2022-03-02 16:04:15 +01:00
PSeitz
4b62f7907d Merge pull request #1297 from PSeitz/fix_clippy 
fix clippy issues
 2022-03-02 10:11:56 +01:00
Pascal Seitz
7fa6a0b665 cargo fmt 2022-03-02 09:24:14 +01:00
PSeitz
458ed29a31 Merge pull request #1299 from saroh/1232-doc-lint 
doc lint for errors and aggregations
 2022-03-02 09:22:07 +01:00
Antoine G
e37775fe21 iff->if or if and only if (#1298) 
* has_xxx is_xxx -> if, these function usualy define equivalence
xxx returns bool -> specify equivalence when appropriate

* fix doc
 2022-03-02 11:00:00 +09:00
Saroh
1cd2434a32 fix(aggregations) Readme 2022-03-01 20:37:48 +01:00
Saroh
de2cba6d1e error definitions 2022-03-01 20:13:59 +01:00
Paul Masurel
c0b1a58d27 Apply suggestions from code review 2022-03-01 18:41:58 +09:00
Paul Masurel
848b795b9f Apply suggestions from code review 2022-03-01 18:37:51 +09:00
Pascal Seitz
091b668624 fix clippy issues 2022-03-01 08:58:51 +01:00
Paul Masurel
5004290daa Return an error on certain type of corruption. (#1296) 2022-03-01 11:35:56 +09:00
StyMaar
5d2c2b804c Fix link to RamDirectory and MMapDirectory in Directory's documentation (#1295) 2022-03-01 09:46:53 +09:00
PSeitz
1a92b588e0 Merge pull request #1294 from PSeitz/aggregation 
fix intermediate result de/serialization
 2022-02-28 08:39:23 +01:00
Pascal Seitz
010e92c118 fix intermediate result de/serialization 
return None for empty average/stats metric
add test for de/serialization of intermediate result
add test for metric on empty result
 2022-02-25 16:39:57 +01:00
Paul Masurel
2ead010c83 Tantivy quickwit (#1293) 
* Added sstable and enabling it by default, and parallel boolean query.
* Added async API for FileSlice.
* Added async get_doc
* Reduce blocksize to 32_000
* Added debug logs

Quickwit specific feature a hidden behind the quickwit feature flag.
 2022-02-25 17:32:49 +09:00
PSeitz
c4f66eb185 improve validation in aggregation, extend invalid field test (#1292) 
* improve validation in aggregation, extend invalid field test

improve validation in aggregation
extend invalid field test
Fixes #1291

* collect fast field names on request structure

* fix visibility of AggregationSegmentCollector
 2022-02-25 15:21:19 +09:00
Paul Masurel
d7b46d2137 Added JSON Type (#1270) 
- Removed useless copy when ingesting JSON.
- Bugfix in phrase query with a missing field norms.
- Disabled range query on default fields

Closes #1251
 2022-02-24 16:25:22 +09:00
PSeitz
d042ce74c7 Merge pull request #1289 from PSeitz/numeric_options 
rename IntOptions to NumericOptions
 2022-02-23 14:04:40 +01:00
PSeitz
7ba9e662b8 Merge pull request #1290 from PSeitz/improve_docs 
improve aggregation docs
 2022-02-23 14:04:20 +01:00
Pascal Seitz
fdd5ef85e5 improve aggregation docs 2022-02-22 10:37:54 +01:00
Pascal Seitz
704498a1ac rename IntOptions to NumericOptions 
keep IntOptions with deprecation warning
Fixes #1286
 2022-02-21 22:20:07 +01:00
PSeitz
1232af7928 fix docs (#1288) 2022-02-21 23:15:58 +09:00
Paul Masurel
d37633e034 Minor changes in indexing. (#1285) 2022-02-21 17:16:52 +09:00
Paul Masurel
9815067171 Minor changes 2022-02-21 13:55:01 +09:00
PSeitz
972cb6c26d Aggregation (#1276) 
Added support for aggregation compatible with Elasticsearch's API.
 2022-02-21 09:59:11 +09:00
Paul Masurel
4dc80cfa25 Removes TokenStream chain. (#1283) 
This change is mostly motivated by the introduction of json object.

We need to be able to inject a position object to make the position
shift.
 2022-02-21 09:51:27 +09:00
PSeitz
cef145790c Fix opening bytes index with dynamic codec (#1279) 
* Fix opening bytes index with dynamic codec

Fix #1278

* extend proptest to cover bytes field codec bug
 2022-02-18 20:44:21 +09:00
Paul Masurel
e05e2a0c51 Added profiling to indexing bench (#1282) 2022-02-18 20:43:28 +09:00
Paul Masurel
e028515caf Simplified expull code. (#1281) 2022-02-18 18:57:10 +09:00
Paul Masurel
850b9eaea4 added a bench to measure the perf of indexing logs (#1275) 2022-02-18 16:48:29 +09:00
Shikhar Bhushan
505e6a440c Remove test assertion sensitive to background segment merging (#1274) 2022-02-17 10:59:46 +09:00
Koichi Akabe
fcd651f6a9 Add Vaporetto tokenizer to README (#1271) 
* Add Vaporetto tokenizer to README

* Update README.md
 2022-02-14 18:19:57 +09:00
Paul Masurel
e6653228a9 Renamed github workflows (#1269) 2022-02-04 15:10:24 +09:00
Paul Masurel
bdedefe07d Adding an IndexingContext object (#1268) 2022-02-04 15:08:01 +09:00
Paul Masurel
13a4473faa Removing obsolete clippy allow thingy. 2022-02-01 11:54:01 +09:00
Paul Masurel
2069e3e52b Fixing clippy comments 2022-02-01 10:24:05 +09:00
Paul Masurel
0d8263cba1 Using nightly to format 2022-01-31 16:10:11 +09:00
Paul Masurel
65b365b81c Fixing all-features build. 2022-01-31 14:41:14 +09:00
dependabot[bot]
4c1366da87 Update fastdivide requirement from 0.3 to 0.4 (#1265) 
Updates the requirements on fastdivide to permit the latest version.

---
updated-dependencies:
- dependency-name: fastdivide
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>

Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2022-01-31 11:26:50 +09:00
Paul Masurel
eca6628b3c Minor refactoring (#1266) 2022-01-28 15:55:55 +09:00
Paul Masurel
9679c5f306 Rename quickwit-inc -> quickwit-oss 2022-01-27 15:37:09 +09:00
Shikhar Bhushan
5a2497b6fd Avoid exposing TrackedObject from Warmer API (#1264) 2022-01-25 10:04:08 +09:00
Shikhar Bhushan
99d4b1a177 Searcher Warming API (#1261) 
Adds an API to register Warmers in the IndexReader.

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2022-01-20 23:40:25 +09:00
273 changed files with 114975 additions and 4379 deletions
Expand all files Collapse all files

2
.github/workflows/coverage.yml vendored
View File

@@ -10,7 +10,7 @@ jobs:
coverage:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Install Rust
run: rustup toolchain install nightly --component llvm-tools-preview
- name: Install cargo-llvm-cov

6
.github/workflows/long_running.yml vendored
View File

@@ -1,4 +1,4 @@
name: Rust
name: Long running tests
on:
push:
@@ -12,13 +12,13 @@ jobs:
functional_test_unsorted:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Run indexing_unsorted
run: cargo test indexing_unsorted -- --ignored
functional_test_sorted:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Run indexing_sorted
run: cargo test indexing_sorted -- --ignored

29
.github/workflows/test.yml vendored
View File

@@ -1,4 +1,4 @@
name: Rust
name: Unit tests
on:
push:
@@ -15,16 +15,35 @@ jobs:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v3
- name: Build
run: cargo build --verbose --workspace
- name: Install latest nightly to test also against unstable feature flag
uses: actions-rs/toolchain@v1
with:
toolchain: stable
toolchain: nightly
override: true
components: rustfmt
- name: Install latest nightly to test also against unstable feature flag
uses: actions-rs/toolchain@v1
with:
toolchain: stable
override: true
components: rustfmt, clippy
- name: Run tests
run: cargo test --features mmap,brotli-compression,lz4-compression,snappy-compression,failpoints --verbose --workspace
run: cargo +stable test --features mmap,brotli-compression,lz4-compression,snappy-compression,failpoints --verbose --workspace
- name: Run tests quickwit feature
run: cargo +stable test --features mmap,quickwit,failpoints --verbose --workspace
- name: Check Formatting
run: cargo fmt --all -- --check
run: cargo +nightly fmt --all -- --check
- uses: actions-rs/clippy-check@v1
with:
toolchain: stable
token: ${{ secrets.GITHUB_TOKEN }}
args: --tests

27
CHANGELOG.md
View File

@@ -1,12 +1,27 @@
Unreleased
================================
- For date values `chrono` has been replaced with `time` (@uklotzde) #1304 :
- The `time` crate is re-exported as `tantivy::time` instead of `tantivy::chrono`.
- The type alias `tantivy::DateTime` has been removed.
- `Value::Date` wraps `time::PrimitiveDateTime` without time zone information.
- Internally date/time values are stored as seconds since UNIX epoch in UTC.
- Converting a `time::OffsetDateTime` to `Value::Date` implicitly converts the value into UTC.
If this is not desired do the time zone conversion yourself and use `time::PrimitiveDateTime`
directly instead.
Tantivy 0.17
================================
- LogMergePolicy now triggers merges if the ratio of deleted documents reaches a threshold (@shikhar) [#115](https://github.com/quickwit-inc/tantivy/issues/115)
- Adds a searcher Warmer API (@shikhar)
- LogMergePolicy now triggers merges if the ratio of deleted documents reaches a threshold (@shikhar @fulmicoton) [#115](https://github.com/quickwit-oss/tantivy/issues/115)
- Adds a searcher Warmer API (@shikhar @fulmicoton)
- Change to non-strict schema. Ignore fields in data which are not defined in schema. Previously this returned an error. #1211
- Facets are necessarily indexed. Existing index with indexed facets should work out of the box. Index without facets that are marked with index: false should be broken (but they were already broken in a sense). (@fulmicoton) #1195 .
- Bugfix that could in theory impact durability in theory on some filesystems [#1224](https://github.com/quickwit-inc/tantivy/issues/1224)
- Schema now offers not indexing fieldnorms (@lpouget) [#922](https://github.com/quickwit-inc/tantivy/issues/922)
- Reduce the number of fsync calls [#1225](https://github.com/quickwit-inc/tantivy/issues/1225)
- Bugfix that could in theory impact durability in theory on some filesystems [#1224](https://github.com/quickwit-oss/tantivy/issues/1224)
- Schema now offers not indexing fieldnorms (@lpouget) [#922](https://github.com/quickwit-oss/tantivy/issues/922)
- Reduce the number of fsync calls [#1225](https://github.com/quickwit-oss/tantivy/issues/1225)
- Fix opening bytes index with dynamic codec (@PSeitz) [#1278](https://github.com/quickwit-oss/tantivy/issues/1278)
- Added an aggregation collector compatible with Elasticsearch (@PSeitz)
- Added a JSON schema type @fulmicoton [#1251](https://github.com/quickwit-oss/tantivy/issues/1251)
- Added support for slop in phrase queries @halvorboe [#1068](https://github.com/quickwit-oss/tantivy/issues/1068)
Tantivy 0.16.2
================================
@@ -128,7 +143,7 @@ Tantivy 0.12.0
## How to update?
Crates relying on custom tokenizer, or registering tokenizer in the manager will require some
minor changes. Check https://github.com/quickwit-inc/tantivy/blob/main/examples/custom_tokenizer.rs
minor changes. Check https://github.com/quickwit-oss/tantivy/blob/main/examples/custom_tokenizer.rs
to check for some code sample.
Tantivy 0.11.3

29
Cargo.toml
View File

@@ -1,18 +1,19 @@
[package]
name = "tantivy"
version = "0.17.0-dev"
version = "0.17.0"
authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT"
categories = ["database-implementations", "data-structures"]
description = """Search engine library"""
documentation = "https://docs.rs/tantivy/"
homepage = "https://github.com/quickwit-inc/tantivy"
repository = "https://github.com/quickwit-inc/tantivy"
homepage = "https://github.com/quickwit-oss/tantivy"
repository = "https://github.com/quickwit-oss/tantivy"
readme = "README.md"
keywords = ["search", "information", "retrieval"]
edition = "2018"
[dependencies]
oneshot = "0.1"
base64 = "0.13"
byteorder = "1.4.3"
crc32fast = "1.2.1"
@@ -32,10 +33,9 @@ fs2={ version = "0.4.3", optional = true }
levenshtein_automata = "0.2"
uuid = { version = "0.8.2", features = ["v4", "serde"] }
crossbeam = "0.8.1"
futures = { version = "0.3.15", features = ["thread-pool"] }
tantivy-query-grammar = { version="0.15.0", path="./query-grammar" }
tantivy-bitpacker = { version="0.1", path="./bitpacker" }
common = { version = "0.1", path = "./common/", package = "tantivy-common" }
common = { version = "0.2", path = "./common/", package = "tantivy-common" }
fastfield_codecs = { version="0.1", path="./fastfield_codecs", default-features = false }
ownedbytes = { version="0.2", path="./ownedbytes" }
stable_deref_trait = "1.2"
@@ -48,13 +48,16 @@ thiserror = "1.0.24"
htmlescape = "0.3.1"
fail = "0.5"
murmurhash32 = "0.2"
chrono = "0.4.19"
time = { version = "0.3.7", features = ["serde-well-known"] }
smallvec = "1.6.1"
rayon = "1.5"
lru = "0.7.0"
fastdivide = "0.3"
fastdivide = "0.4"
itertools = "0.10.0"
measure_time = "0.8.0"
pretty_assertions = "1.1.0"
serde_cbor = {version="0.11", optional=true}
async-trait = "0.1"
[target.'cfg(windows)'.dependencies]
winapi = "0.3.9"
@@ -67,6 +70,8 @@ proptest = "1.0"
criterion = "0.3.5"
test-log = "0.2.8"
env_logger = "0.9.0"
pprof = {version= "0.7", features=["flamegraph", "criterion"]}
futures = "0.3.15"
[dev-dependencies.fail]
version = "0.5"
@@ -92,12 +97,11 @@ snappy-compression = ["snap"]
failpoints = ["fail/failpoints"]
unstable = [] # useful for benches.
quickwit = ["serde_cbor"]
[workspace]
members = ["query-grammar", "bitpacker", "common", "fastfield_codecs", "ownedbytes"]
[badges]
travis-ci = { repository = "tantivy-search/tantivy" }
# Following the "fail" crate best practises, we isolate
# tests that define specific behavior in fail check points
# in a different binary.
@@ -113,3 +117,8 @@ required-features = ["fail/failpoints"]
[[bench]]
name = "analyzer"
harness = false
[[bench]]
name = "index-bench"
harness = false

3
Makefile
View File

@@ -1,3 +1,6 @@
test:
echo "Run test only... No examples."
cargo test --tests --lib
fmt:
cargo +nightly fmt --all

72
README.md
View File

@@ -1,23 +1,14 @@
[![Docs](https://docs.rs/tantivy/badge.svg)](https://docs.rs/crate/tantivy/)
[![Build Status](https://github.com/quickwit-inc/tantivy/actions/workflows/test.yml/badge.svg)](https://github.com/quickwit-inc/tantivy/actions/workflows/test.yml)
[![codecov](https://codecov.io/gh/quickwit-inc/tantivy/branch/main/graph/badge.svg)](https://codecov.io/gh/quickwit-inc/tantivy)
[![Build Status](https://github.com/quickwit-oss/tantivy/actions/workflows/test.yml/badge.svg)](https://github.com/quickwit-oss/tantivy/actions/workflows/test.yml)
[![codecov](https://codecov.io/gh/quickwit-oss/tantivy/branch/main/graph/badge.svg)](https://codecov.io/gh/quickwit-oss/tantivy)
[![Join the chat at https://discord.gg/MT27AG5EVE](https://shields.io/discord/908281611840282624?label=chat%20on%20discord)](https://discord.gg/MT27AG5EVE)
[![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)
![Tantivy](https://tantivy-search.github.io/logo/tantivy-logo.png)
[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/0)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/0)
[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/1)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/1)
[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/2)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/2)
[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/3)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/3)
[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/4)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/4)
[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/5)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/5)
[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/6)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/6)
[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/7)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/7)
**Tantivy** is a **full text search engine library** written in Rust.
**Tantivy** is a **full-text search engine library** written in Rust.
It is closer to [Apache Lucene](https://lucene.apache.org/) than to [Elasticsearch](https://www.elastic.co/products/elasticsearch) 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
@@ -25,19 +16,23 @@ to build such a search engine.
Tantivy is, in fact, strongly inspired by Lucene's design.
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/) break downs
performance for different type of queries / collection.
The following [benchmark](https://tantivy-search.github.io/bench/) breakdowns
performance for different types of queries/collections.
Your mileage WILL vary depending on the nature of queries and their load.
<img src="doc/assets/images/searchbenchmark.png">
# Features
- Full-text search
- Configurable tokenizer (stemming available for 17 Latin languages with third party support for Chinese ([tantivy-jieba](https://crates.io/crates/tantivy-jieba) and [cang-jie](https://crates.io/crates/cang-jie)), Japanese ([lindera](https://github.com/lindera-morphology/lindera-tantivy) and [tantivy-tokenizer-tiny-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))
- 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))
- Fast (check out the :racehorse: :sparkles: [benchmark](https://tantivy-search.github.io/bench/) :sparkles: :racehorse:)
- Tiny startup time (<10ms), perfect for command line tools
- Tiny startup time (<10ms), perfect for command-line tools
- BM25 scoring (the same as Lucene)
- Natural query language (e.g. `(michael AND jackson) OR "king of pop"`)
- Phrase queries search (e.g. `"michael jackson"`)
@@ -52,28 +47,30 @@ Your mileage WILL vary depending on the nature of queries and their load.
- Range queries
- Faceted search
- Configurable indexing (optional term frequency and position indexing)
- JSON Field
- Aggregation Collector: range buckets, average, and stats metrics
- LogMergePolicy with deletes
- Searcher Warmer API
- Cheesy logo with a horse
## Non-features
- Distributed search is out of the scope of Tantivy. That being said, Tantivy is a
library upon which one could build a distributed search. Serializable/mergeable collector state for instance,
are within the scope of Tantivy.
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
Tantivy works on stable Rust (>= 1.27) and supports Linux, MacOS, and Windows.
Tantivy works on stable Rust (>= 1.27) and supports Linux, macOS, and Windows.
- [Tantivy's simple search example](https://tantivy-search.github.io/examples/basic_search.html)
- [tantivy-cli and its tutorial](https://github.com/tantivy-search/tantivy-cli) - `tantivy-cli` is an actual command line interface that makes it easy for you to create a search engine,
- [tantivy-cli and its tutorial](https://github.com/quickwit-oss/tantivy-cli) - `tantivy-cli` is an actual command-line interface that makes it easy for you to create a search engine,
index documents, and search via the CLI or a small server with a REST API.
It walks you through getting a wikipedia search engine up and running in a few minutes.
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?
There are many ways to support this project.
There are many ways to support this project.
- Use Tantivy and tell us about your experience on [Discord](https://discord.gg/MT27AG5EVE) or by email (paul.masurel@gmail.com)
- Report bugs
@@ -92,7 +89,7 @@ Tantivy compiles on stable Rust but requires `Rust >= 1.27`.
To check out and run tests, you can simply run:
```bash
git clone https://github.com/quickwit-inc/tantivy.git
git clone https://github.com/quickwit-oss/tantivy.git
cd tantivy
cargo build
```
@@ -128,3 +125,28 @@ By default, `rustc` compiles everything in the `examples/` directory in debug mo
rust-gdb target/debug/examples/$EXAMPLE_NAME
$ gdb run
```
# Companies Using Tantivy
<p align="left">
<img align="center" src="doc/assets/images/Nuclia.png" alt="Nuclia" height="25" width="auto" /> &nbsp;
<img align="center" src="doc/assets/images/humanfirst.png" alt="Humanfirst.ai" height="30" width="auto" />&nbsp;
<img align="center" src="doc/assets/images/element.io.svg" alt="Element.io" height="25" width="auto" />
</p>
# FAQ
### Can I use Tantivy in other languages?
- Python → [tantivy-py](https://github.com/quickwit-oss/tantivy-py)
- Ruby → [tantiny](https://github.com/baygeldin/tantiny)
You can also find other bindings on [GitHub](https://github.com/search?q=tantivy) but they may be less maintained.
### What are some examples of Tantivy use?
- [seshat](https://github.com/matrix-org/seshat/): A matrix message database/indexer
- [tantiny](https://github.com/baygeldin/tantiny): Tiny full-text search for Ruby
- [lnx](https://github.com/lnx-search/lnx): adaptable, typo tolerant search engine with a REST API
- and [more](https://github.com/search?q=tantivy)!
### On average, how much faster is Tantivy compared to Lucene?
- According to our [search latency benchmark](https://tantivy-search.github.io/bench/), Tantivy is approximately 2x faster than Lucene.

100000
benches/hdfs.json Normal file
View File

File diff suppressed because it is too large Load Diff

121
benches/index-bench.rs Normal file
View File

@@ -0,0 +1,121 @@
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 NUM_REPEATS: usize = 2;
pub fn hdfs_index_benchmark(c: &mut Criterion) {
let schema = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_u64_field("timestamp", INDEXED);
schema_builder.add_text_field("body", TEXT);
schema_builder.add_text_field("severity", STRING);
schema_builder.build()
};
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);
schema_builder.add_text_field("severity", STRING | STORED);
schema_builder.build()
};
let dynamic_schema = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_json_field("json", TEXT);
schema_builder.build()
};
let mut group = c.benchmark_group("index-hdfs");
group.sample_size(20);
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();
}
}
})
});
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-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();
}
}
})
});
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-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().with_profiler(PProfProfiler::new(100, Output::Flamegraph(None)));
targets = hdfs_index_benchmark
}
criterion_main!(benches);

2
bitpacker/Cargo.toml
View File

@@ -6,7 +6,7 @@ authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT"
categories = []
description = """Tantivy-sub crate: bitpacking"""
repository = "https://github.com/quickwit-inc/tantivy"
repository = "https://github.com/quickwit-oss/tantivy"
keywords = []

2
bitpacker/benches/bench.rs
View File

@@ -6,6 +6,7 @@ extern crate test;
mod tests {
use tantivy_bitpacker::BlockedBitpacker;
use test::Bencher;
#[bench]
fn bench_blockedbitp_read(b: &mut Bencher) {
let mut blocked_bitpacker = BlockedBitpacker::new();
@@ -20,6 +21,7 @@ mod tests {
out
});
}
#[bench]
fn bench_blockedbitp_create(b: &mut Bencher) {
b.iter(|| {

3
bitpacker/src/bitpacker.rs
View File

@@ -1,4 +1,5 @@
use std::{convert::TryInto, io};
use std::convert::TryInto;
use std::io;
pub struct BitPacker {
mini_buffer: u64,

5
bitpacker/src/blocked_bitpacker.rs
View File

@@ -1,12 +1,11 @@
use super::bitpacker::BitPacker;
use super::compute_num_bits;
use crate::{minmax, BitUnpacker};
use super::{bitpacker::BitPacker, compute_num_bits};
const BLOCK_SIZE: usize = 128;
/// `BlockedBitpacker` compresses data in blocks of
/// 128 elements, while keeping an index on it
///
#[derive(Debug, Clone)]
pub struct BlockedBitpacker {
// bitpacked blocks

3
bitpacker/src/lib.rs
View File

@@ -1,8 +1,7 @@
mod bitpacker;
mod blocked_bitpacker;
pub use crate::bitpacker::BitPacker;
pub use crate::bitpacker::BitUnpacker;
pub use crate::bitpacker::{BitPacker, BitUnpacker};
pub use crate::blocked_bitpacker::BlockedBitpacker;
/// Computes the number of bits that will be used for bitpacking.

2
common/Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy-common"
version = "0.1.0"
version = "0.2.0"
authors = ["Paul Masurel <paul@quickwit.io>", "Pascal Seitz <pascal@quickwit.io>"]
license = "MIT"
edition = "2018"

23
common/src/bitset.rs
View File

@@ -1,8 +1,8 @@
use ownedbytes::OwnedBytes;
use std::convert::TryInto;
use std::io::Write;
use std::u64;
use std::{fmt, io};
use std::{fmt, io, u64};
use ownedbytes::OwnedBytes;
#[derive(Clone, Copy, Eq, PartialEq)]
pub struct TinySet(u64);
@@ -187,7 +187,6 @@ fn num_buckets(max_val: u32) -> u32 {
impl BitSet {
/// serialize a `BitSet`.
///
pub fn serialize<T: Write>(&self, writer: &mut T) -> io::Result<()> {
writer.write_all(self.max_value.to_le_bytes().as_ref())?;
for tinyset in self.tinysets.iter().cloned() {
@@ -353,7 +352,6 @@ impl ReadOnlyBitSet {
}
/// Iterate the tinyset on the fly from serialized data.
///
#[inline]
fn iter_tinysets(&self) -> impl Iterator<Item = TinySet> + '_ {
self.data.chunks_exact(8).map(move |chunk| {
@@ -363,7 +361,6 @@ impl ReadOnlyBitSet {
}
/// Iterate over the positions of the elements.
///
#[inline]
pub fn iter(&self) -> impl Iterator<Item = u32> + '_ {
self.iter_tinysets()
@@ -415,14 +412,14 @@ impl<'a> From<&'a BitSet> for ReadOnlyBitSet {
#[cfg(test)]
mod tests {
use super::BitSet;
use super::ReadOnlyBitSet;
use super::TinySet;
use std::collections::HashSet;
use ownedbytes::OwnedBytes;
use rand::distributions::Bernoulli;
use rand::rngs::StdRng;
use rand::{Rng, SeedableRng};
use std::collections::HashSet;
use super::{BitSet, ReadOnlyBitSet, TinySet};
#[test]
fn test_read_serialized_bitset_full_multi() {
@@ -443,7 +440,7 @@ mod tests {
bitset.serialize(&mut out).unwrap();
let bitset = ReadOnlyBitSet::open(OwnedBytes::new(out));
assert_eq!(bitset.len() as usize, 64 as usize);
assert_eq!(bitset.len() as usize, 64);
}
#[test]
@@ -710,10 +707,10 @@ mod tests {
#[cfg(all(test, feature = "unstable"))]
mod bench {
use super::BitSet;
use super::TinySet;
use test;
use super::{BitSet, TinySet};
#[bench]
fn bench_tinyset_pop(b: &mut test::Bencher) {
b.iter(|| {

15
common/src/lib.rs
View File

@@ -104,11 +104,12 @@ pub fn u64_to_f64(val: u64) -> f64 {
#[cfg(test)]
pub mod test {
use super::{f64_to_u64, i64_to_u64, u64_to_f64, u64_to_i64};
use super::{BinarySerializable, FixedSize};
use proptest::prelude::*;
use std::f64;
use proptest::prelude::*;
use super::{f64_to_u64, i64_to_u64, u64_to_f64, u64_to_i64, BinarySerializable, FixedSize};
fn test_i64_converter_helper(val: i64) {
assert_eq!(u64_to_i64(i64_to_u64(val)), val);
}
@@ -157,10 +158,10 @@ pub mod test {
#[test]
fn test_f64_order() {
assert!(!(f64_to_u64(f64::NEG_INFINITY)..f64_to_u64(f64::INFINITY))
.contains(&f64_to_u64(f64::NAN))); //nan is not a number
assert!(f64_to_u64(1.5) > f64_to_u64(1.0)); //same exponent, different mantissa
assert!(f64_to_u64(2.0) > f64_to_u64(1.0)); //same mantissa, different exponent
assert!(f64_to_u64(2.0) > f64_to_u64(1.5)); //different exponent and mantissa
.contains(&f64_to_u64(f64::NAN))); // nan is not a number
assert!(f64_to_u64(1.5) > f64_to_u64(1.0)); // same exponent, different mantissa
assert!(f64_to_u64(2.0) > f64_to_u64(1.0)); // same mantissa, different exponent
assert!(f64_to_u64(2.0) > f64_to_u64(1.5)); // different exponent and mantissa
assert!(f64_to_u64(1.0) > f64_to_u64(-1.0)); // pos > neg
assert!(f64_to_u64(-1.5) < f64_to_u64(-1.0));
assert!(f64_to_u64(-2.0) < f64_to_u64(1.0));

14
common/src/serialize.rs
View File

@@ -1,10 +1,9 @@
use crate::Endianness;
use crate::VInt;
use std::io::{Read, Write};
use std::{fmt, io};
use byteorder::{ReadBytesExt, WriteBytesExt};
use std::fmt;
use std::io;
use std::io::Read;
use std::io::Write;
use crate::{Endianness, VInt};
/// Trait for a simple binary serialization.
pub trait BinarySerializable: fmt::Debug + Sized {
@@ -202,8 +201,7 @@ impl BinarySerializable for String {
#[cfg(test)]
pub mod test {
use super::VInt;
use super::*;
use super::{VInt, *};
use crate::serialize::BinarySerializable;
pub fn fixed_size_test<O: BinarySerializable + FixedSize + Default>() {
let mut buffer = Vec::new();

13
common/src/vint.rs
View File

@@ -1,8 +1,9 @@
use super::BinarySerializable;
use byteorder::{ByteOrder, LittleEndian};
use std::io;
use std::io::Read;
use std::io::Write;
use std::io::{Read, Write};
use byteorder::{ByteOrder, LittleEndian};
use super::BinarySerializable;
/// Wrapper over a `u64` that serializes as a variable int.
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
@@ -174,9 +175,7 @@ impl BinarySerializable for VInt {
#[cfg(test)]
mod tests {
use super::serialize_vint_u32;
use super::BinarySerializable;
use super::VInt;
use super::{serialize_vint_u32, BinarySerializable, VInt};
fn aux_test_vint(val: u64) {
let mut v = [14u8; 10];

10
common/src/writer.rs
View File

@@ -54,7 +54,8 @@ impl<W: TerminatingWrite> TerminatingWrite for CountingWriter<W> {
}
}
/// Struct used to prevent from calling [`terminate_ref`](trait.TerminatingWrite.html#tymethod.terminate_ref) directly
/// Struct used to prevent from calling
/// [`terminate_ref`](trait.TerminatingWrite.html#tymethod.terminate_ref) directly
///
/// The point is that while the type is public, it cannot be built by anyone
/// outside of this module.
@@ -64,9 +65,7 @@ pub struct AntiCallToken(());
pub trait TerminatingWrite: Write {
/// Indicate that the writer will no longer be used. Internally call terminate_ref.
fn terminate(mut self) -> io::Result<()>
where
Self: Sized,
{
where Self: Sized {
self.terminate_ref(AntiCallToken(()))
}
@@ -97,9 +96,10 @@ impl<'a> TerminatingWrite for &'a mut Vec<u8> {
#[cfg(test)]
mod test {
use super::CountingWriter;
use std::io::Write;
use super::CountingWriter;
#[test]
fn test_counting_writer() {
let buffer: Vec<u8> = vec![];

BIN
doc/assets/images/Nuclia.png Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 3.1 KiB

8
doc/assets/images/element.io.svg Normal file
View File

@@ -0,0 +1,8 @@
<svg width="518" height="112" viewBox="0 0 518 112" fill="none" xmlns="http://www.w3.org/2000/svg">
<path fill-rule="evenodd" clip-rule="evenodd" d="M56 112C86.9279 112 112 86.9279 112 56C112 25.0721 86.9279 0 56 0C25.0721 0 0 25.0721 0 56C0 86.9279 25.0721 112 56 112Z" fill="#0DBD8B"/>
<path fill-rule="evenodd" clip-rule="evenodd" d="M45.7615 26.093C45.7615 23.8325 47.5977 22.0001 49.8629 22.0001C65.2154 22.0001 77.6611 34.4199 77.6611 49.7406C77.6611 52.001 75.8248 53.8335 73.5597 53.8335C71.2945 53.8335 69.4583 52.001 69.4583 49.7406C69.4583 38.9408 60.6851 30.1859 49.8629 30.1859C47.5977 30.1859 45.7615 28.3534 45.7615 26.093Z" fill="white"/>
<path fill-rule="evenodd" clip-rule="evenodd" d="M85.8986 45.6477C88.1637 45.6477 89.9999 47.4801 89.9999 49.7406C89.9999 65.0612 77.5543 77.4811 62.2017 77.4811C59.9366 77.4811 58.1003 75.6486 58.1003 73.3882C58.1003 71.1277 59.9366 69.2953 62.2017 69.2953C73.024 69.2953 81.7972 60.5403 81.7972 49.7406C81.7972 47.4801 83.6334 45.6477 85.8986 45.6477Z" fill="white"/>
<path fill-rule="evenodd" clip-rule="evenodd" d="M66.3031 85.907C66.3031 88.1675 64.4668 89.9999 62.2017 89.9999C46.8492 89.9999 34.4035 77.58 34.4035 62.2594C34.4035 59.9989 36.2398 58.1665 38.5049 58.1665C40.77 58.1665 42.6063 59.9989 42.6063 62.2594C42.6063 73.0592 51.3795 81.8141 62.2017 81.8141C64.4668 81.8141 66.3031 83.6466 66.3031 85.907Z" fill="white"/>
<path fill-rule="evenodd" clip-rule="evenodd" d="M26.1014 66.3523C23.8363 66.3523 22.0001 64.5199 22.0001 62.2594C22 46.9388 34.4457 34.5189 49.7983 34.5189C52.0634 34.5189 53.8997 36.3514 53.8997 38.6118C53.8997 40.8723 52.0634 42.7047 49.7983 42.7047C38.976 42.7047 30.2028 51.4597 30.2028 62.2594C30.2028 64.5199 28.3666 66.3523 26.1014 66.3523Z" fill="white"/>
<path d="M197 63.5H157.5C157.967 67.6333 159.467 70.9333 162 73.4C164.533 75.8 167.867 77 172 77C174.733 77 177.2 76.3333 179.4 75C181.6 73.6667 183.167 71.8667 184.1 69.6H196.1C194.5 74.8667 191.5 79.1333 187.1 82.4C182.767 85.6 177.633 87.2 171.7 87.2C163.967 87.2 157.7 84.6333 152.9 79.5C148.167 74.3667 145.8 67.8667 145.8 60C145.8 52.3333 148.2 45.9 153 40.7C157.8 35.5 164 32.9 171.6 32.9C179.2 32.9 185.333 35.4667 190 40.6C194.733 45.6667 197.1 52.0667 197.1 59.8L197 63.5ZM171.6 42.6C167.867 42.6 164.767 43.7 162.3 45.9C159.833 48.1 158.3 51.0333 157.7 54.7H185.3C184.767 51.0333 183.3 48.1 180.9 45.9C178.5 43.7 175.4 42.6 171.6 42.6ZM205.289 70.5V11H217.189V70.7C217.189 73.3667 218.656 74.7 221.589 74.7L223.689 74.6V85.9C222.556 86.1 221.356 86.2 220.089 86.2C214.956 86.2 211.189 84.9 208.789 82.3C206.456 79.7 205.289 75.7667 205.289 70.5ZM279.109 63.5H239.609C240.076 67.6333 241.576 70.9333 244.109 73.4C246.643 75.8 249.976 77 254.109 77C256.843 77 259.309 76.3333 261.509 75C263.709 73.6667 265.276 71.8667 266.209 69.6H278.209C276.609 74.8667 273.609 79.1333 269.209 82.4C264.876 85.6 259.743 87.2 253.809 87.2C246.076 87.2 239.809 84.6333 235.009 79.5C230.276 74.3667 227.909 67.8667 227.909 60C227.909 52.3333 230.309 45.9 235.109 40.7C239.909 35.5 246.109 32.9 253.709 32.9C261.309 32.9 267.443 35.4667 272.109 40.6C276.843 45.6667 279.209 52.0667 279.209 59.8L279.109 63.5ZM253.709 42.6C249.976 42.6 246.876 43.7 244.409 45.9C241.943 48.1 240.409 51.0333 239.809 54.7H267.409C266.876 51.0333 265.409 48.1 263.009 45.9C260.609 43.7 257.509 42.6 253.709 42.6ZM332.798 56.2V86H320.898V54.9C320.898 47.0333 317.632 43.1 311.098 43.1C307.565 43.1 304.732 44.2333 302.598 46.5C300.532 48.7667 299.498 51.8667 299.498 55.8V86H287.598V34.1H298.598V41C299.865 38.6667 301.798 36.7333 304.398 35.2C306.998 33.6667 310.232 32.9 314.098 32.9C321.298 32.9 326.498 35.6333 329.698 41.1C334.098 35.6333 339.965 32.9 347.298 32.9C353.365 32.9 358.032 34.8 361.298 38.6C364.565 42.3333 366.198 47.2667 366.198 53.4V86H354.298V54.9C354.298 47.0333 351.032 43.1 344.498 43.1C340.898 43.1 338.032 44.2667 335.898 46.6C333.832 48.8667 332.798 52.0667 332.798 56.2ZM425.379 63.5H385.879C386.346 67.6333 387.846 70.9333 390.379 73.4C392.912 75.8 396.246 77 400.379 77C403.112 77 405.579 76.3333 407.779 75C409.979 73.6667 411.546 71.8667 412.479 69.6H424.479C422.879 74.8667 419.879 79.1333 415.479 82.4C411.146 85.6 406.012 87.2 400.079 87.2C392.346 87.2 386.079 84.6333 381.279 79.5C376.546 74.3667 374.179 67.8667 374.179 60C374.179 52.3333 376.579 45.9 381.379 40.7C386.179 35.5 392.379 32.9 399.979 32.9C407.579 32.9 413.712 35.4667 418.379 40.6C423.112 45.6667 425.479 52.0667 425.479 59.8L425.379 63.5ZM399.979 42.6C396.246 42.6 393.146 43.7 390.679 45.9C388.212 48.1 386.679 51.0333 386.079 54.7H413.679C413.146 51.0333 411.679 48.1 409.279 45.9C406.879 43.7 403.779 42.6 399.979 42.6ZM444.868 34.1V41C446.068 38.7333 448.035 36.8333 450.768 35.3C453.568 33.7 456.935 32.9 460.868 32.9C467.001 32.9 471.735 34.7667 475.068 38.5C478.468 42.2333 480.168 47.2 480.168 53.4V86H468.268V54.9C468.268 51.2333 467.401 48.3667 465.668 46.3C464.001 44.1667 461.435 43.1 457.968 43.1C454.168 43.1 451.168 44.2333 448.968 46.5C446.835 48.7667 445.768 51.9 445.768 55.9V86H433.868V34.1H444.868ZM514.922 75.4V85.7C513.455 86.1 511.389 86.3 508.722 86.3C498.589 86.3 493.522 81.2 493.522 71V43.6H485.622V34.1H493.522V20.6H505.422V34.1H515.122V43.6H505.422V69.8C505.422 73.8667 507.355 75.9 511.222 75.9L514.922 75.4Z" fill="black"/>
</svg>
Side by Side

After

Width:  |  Height:  |  Size: 5.2 KiB

BIN
doc/assets/images/humanfirst.png Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 102 KiB

BIN
doc/assets/images/searchbenchmark.png Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 653 KiB

4
doc/src/index_sorting.md
View File

@@ -38,7 +38,7 @@ Note: Tantivy 0.16 does not do this optimization yet.
In principle there are many algorithms possible that exploit the monotonically increasing nature. (aggregations maybe?)
## Usage
The index sorting can be configured setting [`sort_by_field`](https://github.com/quickwit-inc/tantivy/blob/000d76b11a139a84b16b9b95060a1c93e8b9851c/src/core/index_meta.rs#L238) on `IndexSettings` and passing it to a `IndexBuilder`. As of tantvy 0.16 only fast fields are allowed to be used.
The index sorting can be configured setting [`sort_by_field`](https://github.com/quickwit-oss/tantivy/blob/000d76b11a139a84b16b9b95060a1c93e8b9851c/src/core/index_meta.rs#L238) on `IndexSettings` and passing it to a `IndexBuilder`. As of tantvy 0.16 only fast fields are allowed to be used.
```
let settings = IndexSettings {
@@ -55,7 +55,7 @@ let index = index_builder.create_in_ram().unwrap();
## Implementation details
Sorting an index is applied in the serialization step. In general there are two serialization steps: [Finishing a single segment](https://github.com/quickwit-inc/tantivy/blob/000d76b11a139a84b16b9b95060a1c93e8b9851c/src/indexer/segment_writer.rs#L338) and [merging multiple segments](https://github.com/quickwit-inc/tantivy/blob/000d76b11a139a84b16b9b95060a1c93e8b9851c/src/indexer/merger.rs#L1073).
Sorting an index is applied in the serialization step. In general there are two serialization steps: [Finishing a single segment](https://github.com/quickwit-oss/tantivy/blob/000d76b11a139a84b16b9b95060a1c93e8b9851c/src/indexer/segment_writer.rs#L338) and [merging multiple segments](https://github.com/quickwit-oss/tantivy/blob/000d76b11a139a84b16b9b95060a1c93e8b9851c/src/indexer/merger.rs#L1073).
In both cases we generate a docid mapping reflecting the sort. This mapping is used when serializing the different components (doc store, fastfields, posting list, normfield, facets).

128
doc/src/json.md Normal file
View File

@@ -0,0 +1,128 @@
# Json
As of tantivy 0.17, tantivy supports a json object type.
This type can be used to allow for a schema-less search index.
When indexing a json object, we "flatten" the JSON. This operation emits terms that represent a triplet `(json_path, value_type, value)`
For instance, if user is a json field, the following document:
```json
{
"user": {
"name": "Paul Masurel",
"address": {
"city": "Tokyo",
"country": "Japan"
},
"created_at": "2018-11-12T23:20:50.52Z"
}
}
```
emits the following tokens:
- ("name", Text, "Paul")
- ("name", Text, "Masurel")
- ("address.city", Text, "Tokyo")
- ("address.country", Text, "Japan")
- ("created_at", Date, 15420648505)
# Bytes-encoding and lexicographical sort.
Like any other terms, these triplets are encoded into a binary format as follows.
- `json_path`: the json path is a sequence of "segments". In the example above, `address.city`
is just a debug representation of the json path `["address", "city"]`.
Its representation is done by separating segments by a unicode char `\x01`, and ending the path by `\x00`.
- `value type`: One byte represents the `Value` type.
- `value`: The value representation is just the regular Value representation.
This representation is designed to align the natural sort of Terms with the lexicographical sort
of their binary representation (Tantivy's dictionary (whether fst or sstable) is sorted and does prefix encoding).
In the example above, the terms will be sorted as
- ("address.city", Text, "Tokyo")
- ("address.country", Text, "Japan")
- ("name", Text, "Masurel")
- ("name", Text, "Paul")
- ("created_at", Date, 15420648505)
As seen in "pitfalls", we may end up having to search for a value for a same path in several different fields. Putting the field code after the path makes it maximizes compression opportunities but also increases the chances for the two terms to end up in the actual same term dictionary block.
# Pitfalls, limitation and corner cases.
Json gives very little information about the type of the literals it stores.
All numeric types end up mapped as a "Number" and there are no types for dates.
At indexing, tantivy will try to interpret number and strings as different type with a
priority order.
Numbers will be interpreted as u64, i64 and f64 in that order.
Strings will be interpreted as rfc3999 dates or simple strings.
The first working type is picked and is the only term that is emitted for indexing.
Note this interpretation happens on a per-document basis, and there is no effort to try to sniff
a consistent field type at the scale of a segment.
On the query parser side on the other hand, we may end up emitting more than one type.
For instance, we do not even know if the type is a number or string based.
So the query
```
my_path.my_segment:233
```
Will be interpreted as
`(my_path.my_segment, String, 233) or (my_path.my_segment, u64, 233)`
Likewise, we need to emit two tokens if the query contains an rfc3999 date.
Indeed the date could have been actually a single token inside the text of a document at ingestion time. Generally speaking, we will always at least emit a string token in query parsing, and sometimes more.
If one more json field is defined, things get even more complicated.
## Default json field
If the schema contains a text field called "text" and a json field that is set as a default field:
`text:hello` could be reasonably interpreted as targetting the text field or as targetting the json field called `json_dynamic` with the json_path "text".
If there is such an ambiguity, we decide to only search in the "text" field: `text:hello`.
In other words, the parser will not search in default json fields if there is a schema hit.
This is a product decision.
The user can still target the JSON field by specifying its name explicitly:
`json_dynamic.text:hello`.
## Range queries are not supported.
Json field do not support range queries.
## Arrays do not work like nested object.
If json object contains an array, a search query might return more documents
than what might be expected.
Let's take an example.
```json
{
"cart_id": 3234234 ,
"cart": [
{"product_type": "sneakers", "attributes": {"color": "white"} },
{"product_type": "t-shirt", "attributes": {"color": "red"}},
]
}
```
Despite the array structure, a document in tantivy is a bag of terms.
The query:
```
cart.product_type:sneakers AND cart.attributes.color:red
```
Actually match the document above.

129
examples/aggregation.rs Normal file
View File

@@ -0,0 +1,129 @@
// # Aggregation example
//
// This example shows how you can use built-in aggregations.
// We will use range buckets and compute the average in each bucket.
//
use serde_json::Value;
use tantivy::aggregation::agg_req::{
Aggregation, Aggregations, BucketAggregation, BucketAggregationType, MetricAggregation,
RangeAggregation,
};
use tantivy::aggregation::agg_result::AggregationResults;
use tantivy::aggregation::metric::AverageAggregation;
use tantivy::aggregation::AggregationCollector;
use tantivy::query::TermQuery;
use tantivy::schema::{self, Cardinality, IndexRecordOption, Schema, TextFieldIndexing};
use tantivy::{doc, Index, Term};
fn main() -> tantivy::Result<()> {
let mut schema_builder = Schema::builder();
let text_fieldtype = schema::TextOptions::default()
.set_indexing_options(
TextFieldIndexing::default().set_index_option(IndexRecordOption::WithFreqs),
)
.set_stored();
let text_field = schema_builder.add_text_field("text", text_fieldtype);
let score_fieldtype =
crate::schema::NumericOptions::default().set_fast(Cardinality::SingleValue);
let highscore_field = schema_builder.add_f64_field("highscore", score_fieldtype.clone());
let price_field = schema_builder.add_f64_field("price", score_fieldtype.clone());
let schema = schema_builder.build();
// # Indexing documents
//
// Lets index a bunch of documents for this example.
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer(50_000_000)?;
// writing the segment
index_writer.add_document(doc!(
text_field => "cool",
highscore_field => 1f64,
price_field => 0f64,
))?;
index_writer.add_document(doc!(
text_field => "cool",
highscore_field => 3f64,
price_field => 1f64,
))?;
index_writer.add_document(doc!(
text_field => "cool",
highscore_field => 5f64,
price_field => 1f64,
))?;
index_writer.add_document(doc!(
text_field => "nohit",
highscore_field => 6f64,
price_field => 2f64,
))?;
index_writer.add_document(doc!(
text_field => "cool",
highscore_field => 7f64,
price_field => 2f64,
))?;
index_writer.commit()?;
index_writer.add_document(doc!(
text_field => "cool",
highscore_field => 11f64,
price_field => 10f64,
))?;
index_writer.add_document(doc!(
text_field => "cool",
highscore_field => 14f64,
price_field => 15f64,
))?;
index_writer.add_document(doc!(
text_field => "cool",
highscore_field => 15f64,
price_field => 20f64,
))?;
index_writer.commit()?;
let reader = index.reader()?;
let text_field = reader.searcher().schema().get_field("text").unwrap();
let term_query = TermQuery::new(
Term::from_field_text(text_field, "cool"),
IndexRecordOption::Basic,
);
let sub_agg_req_1: Aggregations = vec![(
"average_price".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("price".to_string()),
)),
)]
.into_iter()
.collect();
let agg_req_1: Aggregations = vec![(
"score_ranges".to_string(),
Aggregation::Bucket(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "highscore".to_string(),
ranges: vec![
(-1f64..9f64).into(),
(9f64..14f64).into(),
(14f64..20f64).into(),
],
}),
sub_aggregation: sub_agg_req_1.clone(),
}),
)]
.into_iter()
.collect();
let collector = AggregationCollector::from_aggs(agg_req_1);
let searcher = reader.searcher();
let agg_res: AggregationResults = searcher.search(&term_query, &collector).unwrap();
let res: Value = serde_json::from_str(&serde_json::to_string(&agg_res)?)?;
println!("{}", serde_json::to_string_pretty(&res)?);
Ok(())
}

6
examples/basic_search.rs
View File

@@ -73,7 +73,7 @@ fn main() -> tantivy::Result<()> {
// multithreaded.
//
// Here we give tantivy a budget of `50MB`.
// Using a bigger heap for the indexer may increase
// Using a bigger memory_arena for the indexer may increase
// throughput, but 50 MB is already plenty.
let mut index_writer = index.writer(50_000_000)?;
@@ -91,8 +91,8 @@ fn main() -> tantivy::Result<()> {
old_man_doc.add_text(title, "The Old Man and the Sea");
old_man_doc.add_text(
body,
"He was an old man who fished alone in a skiff in the Gulf Stream and \
he had gone eighty-four days now without taking a fish.",
"He 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.",
);
// ... and add it to the `IndexWriter`.

3
examples/custom_collector.rs
View File

@@ -12,8 +12,7 @@
use tantivy::collector::{Collector, SegmentCollector};
use tantivy::fastfield::{DynamicFastFieldReader, FastFieldReader};
use tantivy::query::QueryParser;
use tantivy::schema::Field;
use tantivy::schema::{Schema, FAST, INDEXED, TEXT};
use tantivy::schema::{Field, Schema, FAST, INDEXED, TEXT};
use tantivy::{doc, Index, Score, SegmentReader};
#[derive(Default)]

2
examples/custom_tokenizer.rs
View File

@@ -62,7 +62,7 @@ fn main() -> tantivy::Result<()> {
// multithreaded.
//
// Here we use a buffer of 50MB per thread. Using a bigger
// heap for the indexer can increase its throughput.
// memory arena for the indexer can increase its throughput.
let mut index_writer = index.writer(50_000_000)?;
index_writer.add_document(doc!(
title => "The Old Man and the Sea",

5
examples/deleting_updating_documents.rs
View File

@@ -56,8 +56,9 @@ fn main() -> tantivy::Result<()> {
// If it is `text`, let's make sure to keep it `raw` and let's avoid
// running any text processing on it.
// This is done by associating this field to the tokenizer named `raw`.
// Rather than building our [`TextOptions`](//docs.rs/tantivy/~0/tantivy/schema/struct.TextOptions.html) manually,
// We use the `STRING` shortcut. `STRING` stands for indexed (without term frequency or positions)
// Rather than building our
// [`TextOptions`](//docs.rs/tantivy/~0/tantivy/schema/struct.TextOptions.html) manually, We
// use the `STRING` shortcut. `STRING` stands for indexed (without term frequency or positions)
// and untokenized.
//
// Because we also want to be able to see this `id` in our returned documents,

6
examples/faceted_search_with_tweaked_score.rs
View File

@@ -1,9 +1,9 @@
use std::collections::HashSet;
use tantivy::collector::TopDocs;
use tantivy::doc;
use tantivy::query::BooleanQuery;
use tantivy::schema::*;
use tantivy::{DocId, Index, Score, SegmentReader};
use tantivy::{doc, DocId, Index, Score, SegmentReader};
fn main() -> tantivy::Result<()> {
let mut schema_builder = Schema::builder();
@@ -87,7 +87,7 @@ fn main() -> tantivy::Result<()> {
.unwrap()
.get_first(title)
.unwrap()
.text()
.as_text()
.unwrap()
.to_owned()
})

16
examples/iterating_docs_and_positions.rs
View File

@@ -52,11 +52,11 @@ fn main() -> tantivy::Result<()> {
let term_the = Term::from_field_text(title, "the");
// This segment posting object is like a cursor over the documents matching the term.
// The `IndexRecordOption` arguments tells tantivy we will be interested in both term frequencies
// and positions.
// The `IndexRecordOption` arguments tells tantivy we will be interested in both term
// frequencies and positions.
//
// If you don't need all this information, you may get better performance by decompressing less
// information.
// If you don't need all this information, you may get better performance by decompressing
// less information.
if let Some(mut segment_postings) =
inverted_index.read_postings(&term_the, IndexRecordOption::WithFreqsAndPositions)?
{
@@ -109,11 +109,11 @@ fn main() -> tantivy::Result<()> {
let inverted_index = segment_reader.inverted_index(title)?;
// This segment posting object is like a cursor over the documents matching the term.
// The `IndexRecordOption` arguments tells tantivy we will be interested in both term frequencies
// and positions.
// The `IndexRecordOption` arguments tells tantivy we will be interested in both term
// frequencies and positions.
//
// If you don't need all this information, you may get better performance by decompressing less
// information.
// If you don't need all this information, you may get better performance by decompressing
// less information.
if let Some(mut block_segment_postings) =
inverted_index.read_block_postings(&term_the, IndexRecordOption::Basic)?
{

80
examples/json_field.rs Normal file
View File

@@ -0,0 +1,80 @@
// # Json field example
//
// This example shows how the json field can be used
// to make tantivy partially schemaless.
use tantivy::collector::{Count, TopDocs};
use tantivy::query::QueryParser;
use tantivy::schema::{Schema, FAST, STORED, STRING, TEXT};
use tantivy::Index;
fn main() -> tantivy::Result<()> {
// # Defining the schema
//
// We need two fields:
// - a timestamp
// - a json object field
let mut schema_builder = Schema::builder();
schema_builder.add_date_field("timestamp", FAST | STORED);
let event_type = schema_builder.add_text_field("event_type", STRING | STORED);
let attributes = schema_builder.add_json_field("attributes", STORED | TEXT);
let schema = schema_builder.build();
// # Indexing documents
let index = Index::create_in_ram(schema.clone());
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",
"attributes": {
"target": "submit-button",
"cart": {"product_id": 103},
"description": "the best vacuum cleaner ever"
}
}"#,
)?;
index_writer.add_document(doc)?;
let doc = schema.parse_document(
r#"{
"timestamp": "2022-02-22T23:20:51.53Z",
"event_type": "click",
"attributes": {
"target": "submit-button",
"cart": {"product_id": 133},
"description": "das keyboard"
}
}"#,
)?;
index_writer.add_document(doc)?;
index_writer.commit()?;
let reader = index.reader()?;
let searcher = reader.searcher();
let query_parser = QueryParser::for_index(&index, vec![event_type, attributes]);
{
let query = query_parser.parse_query("target:submit-button")?;
let count_docs = searcher.search(&*query, &TopDocs::with_limit(2))?;
assert_eq!(count_docs.len(), 2);
}
{
let query = query_parser.parse_query("target:submit")?;
let count_docs = searcher.search(&*query, &TopDocs::with_limit(2))?;
assert_eq!(count_docs.len(), 2);
}
{
let query = query_parser.parse_query("cart.product_id:103")?;
let count_docs = searcher.search(&*query, &Count)?;
assert_eq!(count_docs, 1);
}
{
let query = query_parser
.parse_query("event_type:click AND cart.product_id:133")
.unwrap();
let hits = searcher.search(&*query, &TopDocs::with_limit(2)).unwrap();
assert_eq!(hits.len(), 1);
}
Ok(())
}

4
examples/multiple_producer.rs
View File

@@ -28,6 +28,7 @@
use std::sync::{Arc, RwLock};
use std::thread;
use std::time::Duration;
use tantivy::schema::{Schema, STORED, TEXT};
use tantivy::{doc, Index, IndexWriter, Opstamp, TantivyError};
@@ -90,7 +91,8 @@ fn main() -> tantivy::Result<()> {
// # In the main thread, we commit 10 times, once every 500ms.
for _ in 0..10 {
let opstamp: Opstamp = {
// Committing or rollbacking on the other hand requires write lock. This will block other threads.
// Committing or rollbacking on the other hand requires write lock. This will block
// other threads.
let mut index_writer_wlock = index_writer.write().unwrap();
index_writer_wlock.commit()?
};

5
examples/snippet.rs
View File

@@ -57,7 +57,10 @@ fn main() -> tantivy::Result<()> {
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().text().unwrap());
println!(
"title: {}",
doc.get_first(title).unwrap().as_text().unwrap()
);
println!("snippet: {}", snippet.to_html());
println!("custom highlighting: {}", highlight(snippet));
}

9
examples/warmer.rs
View File

@@ -6,8 +6,10 @@ use tantivy::collector::TopDocs;
use tantivy::fastfield::FastFieldReader;
use tantivy::query::QueryParser;
use tantivy::schema::{Field, Schema, FAST, TEXT};
use tantivy::{doc, DocAddress, DocId, Index, IndexReader, SegmentReader, TrackedObject};
use tantivy::{Opstamp, Searcher, SearcherGeneration, SegmentId, Warmer};
use tantivy::{
doc, DocAddress, DocId, Index, IndexReader, Opstamp, Searcher, SearcherGeneration, SegmentId,
SegmentReader, Warmer,
};
// This example shows how warmers can be used to
// load a values from an external sources using the Warmer API.
@@ -69,7 +71,7 @@ impl Warmer for DynamicPriceColumn {
Ok(())
}
fn garbage_collect(&self, live_generations: &[TrackedObject<SearcherGeneration>]) {
fn garbage_collect(&self, live_generations: &[&SearcherGeneration]) {
let live_segment_id_and_delete_ops: HashSet<(SegmentId, Option<Opstamp>)> =
live_generations
.iter()
@@ -90,7 +92,6 @@ impl Warmer for DynamicPriceColumn {
/// This map represents a map (ProductId -> Price)
///
/// In practise, it could be fetching things from an external service, like a SQL table.
///
#[derive(Default, Clone)]
pub struct ExternalPriceTable {
prices: Arc<RwLock<HashMap<ProductId, Price>>>,

1
fastfield_codecs/.gitignore vendored Normal file
View File

@@ -0,0 +1 @@
datasets/

6
fastfield_codecs/Cargo.toml
View File

@@ -6,10 +6,8 @@ license = "MIT"
edition = "2018"
description = "Fast field codecs used by tantivy"
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
[dependencies]
common = { version = "0.1", path = "../common/", package = "tantivy-common" }
common = { version = "0.2", path = "../common/", package = "tantivy-common" }
tantivy-bitpacker = { version="0.1.1", path = "../bitpacker/" }
prettytable-rs = {version="0.8.0", optional= true}
rand = {version="0.8.3", optional= true}
@@ -19,6 +17,6 @@ more-asserts = "0.2.1"
rand = "0.8.3"
[features]
unstable = [] # useful for benches and experimental codecs.
bin = ["prettytable-rs", "rand"]
default = ["bin"]

6
fastfield_codecs/Makefile Normal file
View File

@@ -0,0 +1,6 @@
DATASETS ?= hdfs_logs_timestamps http_logs_timestamps amazon_reviews_product_ids nooc_temperatures
download:
@echo "--- Downloading datasets ---"
mkdir -p datasets
@for dataset in $(DATASETS); do curl -o - https://quickwit-datasets-public.s3.amazonaws.com/benchmarks/fastfields/$$dataset.txt.gz | gunzip > datasets/$$dataset.txt; done

131
fastfield_codecs/README.md
View File

@@ -13,6 +13,10 @@ A codec needs to implement 2 traits:
- A reader implementing `FastFieldCodecReader` to read the codec.
- A serializer implementing `FastFieldCodecSerializer` for compression estimation and codec name + id.
### Download real world datasets for codecs comparison
Before comparing codecs, you need to execute `make download` to download real world datasets hosted on AWS S3.
To run with the unstable codecs, execute `cargo run --features unstable`.
### Tests
Once the traits are implemented test and benchmark integration is pretty easy (see `test_with_codec_data_sets` and `bench.rs`).
@@ -23,46 +27,101 @@ cargo run --features bin
```
### TODO
- Add real world data sets in comparison
- Add codec to cover sparse data sets
### Codec Comparison
```
+----------------------------------+-------------------+------------------------+
| | Compression Ratio | Compression Estimation |
+----------------------------------+-------------------+------------------------+
| Autoincrement | | |
+----------------------------------+-------------------+------------------------+
| LinearInterpol | 0.000039572664 | 0.000004396963 |
+----------------------------------+-------------------+------------------------+
| MultiLinearInterpol | 0.1477348 | 0.17275847 |
+----------------------------------+-------------------+------------------------+
| Bitpacked | 0.28126493 | 0.28125 |
+----------------------------------+-------------------+------------------------+
| Monotonically increasing concave | | |
+----------------------------------+-------------------+------------------------+
| LinearInterpol | 0.25003937 | 0.26562938 |
+----------------------------------+-------------------+------------------------+
| MultiLinearInterpol | 0.190665 | 0.1883836 |
+----------------------------------+-------------------+------------------------+
| Bitpacked | 0.31251436 | 0.3125 |
+----------------------------------+-------------------+------------------------+
| Monotonically increasing convex | | |
+----------------------------------+-------------------+------------------------+
| LinearInterpol | 0.25003937 | 0.28125438 |
+----------------------------------+-------------------+------------------------+
| MultiLinearInterpol | 0.18676 | 0.2040086 |
+----------------------------------+-------------------+------------------------+
| Bitpacked | 0.31251436 | 0.3125 |
+----------------------------------+-------------------+------------------------+
| Almost monotonically increasing | | |
+----------------------------------+-------------------+------------------------+
| LinearInterpol | 0.14066513 | 0.1562544 |
+----------------------------------+-------------------+------------------------+
| MultiLinearInterpol | 0.16335973 | 0.17275847 |
+----------------------------------+-------------------+------------------------+
| Bitpacked | 0.28126493 | 0.28125 |
+----------------------------------+-------------------+------------------------+
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| | Compression ratio | Compression ratio estimation | Compression time (micro) | Reading time (micro) |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Autoincrement | | | | |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| PiecewiseLinear | 0.0051544965 | 0.17251475 | 960 | 211 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| FOR | 0.118189104 | 0.14172314 | 708 | 212 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Bitpacked | 0.28126493 | 0.28125 | 474 | 112 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Monotonically increasing concave | | | | |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| PiecewiseLinear | 0.005955 | 0.18813984 | 885 | 211 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| FOR | 0.16113 | 0.15734828 | 704 | 212 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Bitpacked | 0.31251436 | 0.3125 | 478 | 113 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Monotonically increasing convex | | | | |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| PiecewiseLinear | 0.00613 | 0.20376484 | 889 | 211 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| FOR | 0.157175 | 0.17297328 | 706 | 212 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Bitpacked | 0.31251436 | 0.3125 | 471 | 113 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Almost monotonically increasing | | | | |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| PiecewiseLinear | 0.14549863 | 0.17251475 | 923 | 210 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| FOR | 0.14943957 | 0.15734814 | 703 | 211 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Bitpacked | 0.28126493 | 0.28125 | 462 | 112 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Random | | | | |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| PiecewiseLinear | 0.14533783 | 0.14126475 | 924 | 211 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| FOR | 0.13381402 | 0.15734814 | 695 | 211 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Bitpacked | 0.12501445 | 0.125 | 422 | 112 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| HDFS logs timestamps | | | | |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| PiecewiseLinear | 0.39826187 | 0.4068908 | 5545 | 1086 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| FOR | 0.39214826 | 0.40734857 | 5082 | 1073 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Bitpacked | 0.39062786 | 0.390625 | 2864 | 567 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| HDFS logs timestamps SORTED | | | | |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| PiecewiseLinear | 0.032736875 | 0.094390824 | 4942 | 1067 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| FOR | 0.02667125 | 0.079223566 | 3626 | 994 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Bitpacked | 0.39062786 | 0.390625 | 2493 | 566 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| HTTP logs timestamps SORTED | | | | |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| PiecewiseLinear | 0.047942877 | 0.20376582 | 5121 | 1065 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| FOR | 0.06637425 | 0.18859856 | 3929 | 1093 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Bitpacked | 0.26562786 | 0.265625 | 2221 | 526 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Amazon review product ids | | | | |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| PiecewiseLinear | 0.41900787 | 0.4225158 | 5239 | 1089 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| FOR | 0.41504425 | 0.43859857 | 4158 | 1052 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Bitpacked | 0.40625286 | 0.40625 | 2603 | 513 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Amazon review product ids SORTED | | | | |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| PiecewiseLinear | 0.18364687 | 0.25064084 | 5036 | 990 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| FOR | 0.21239226 | 0.21984856 | 4087 | 1072 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Bitpacked | 0.40625286 | 0.40625 | 2702 | 525 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Temperatures | | | | |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| PiecewiseLinear | | Codec Disabled | 0 | 0 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| FOR | 1.0088086 | 1.001098 | 1306 | 237 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
| Bitpacked | 1.000012 | 1 | 950 | 108 |
+----------------------------------+-------------------+------------------------------+--------------------------+----------------------+
```

31
fastfield_codecs/benches/bench.rs
View File

@@ -4,14 +4,11 @@ extern crate test;
#[cfg(test)]
mod tests {
use fastfield_codecs::{
bitpacked::{BitpackedFastFieldReader, BitpackedFastFieldSerializer},
linearinterpol::{LinearInterpolFastFieldReader, LinearInterpolFastFieldSerializer},
multilinearinterpol::{
MultiLinearInterpolFastFieldReader, MultiLinearInterpolFastFieldSerializer,
},
*,
use fastfield_codecs::bitpacked::{BitpackedFastFieldReader, BitpackedFastFieldSerializer};
use fastfield_codecs::piecewise_linear::{
PiecewiseLinearFastFieldReader, PiecewiseLinearFastFieldSerializer,
};
use fastfield_codecs::*;
fn get_data() -> Vec<u64> {
let mut data: Vec<_> = (100..55000_u64)
@@ -70,14 +67,9 @@ mod tests {
bench_create::<BitpackedFastFieldSerializer>(b, &data);
}
#[bench]
fn bench_fastfield_linearinterpol_create(b: &mut Bencher) {
fn bench_fastfield_piecewise_linear_create(b: &mut Bencher) {
let data: Vec<_> = get_data();
bench_create::<LinearInterpolFastFieldSerializer>(b, &data);
}
#[bench]
fn bench_fastfield_multilinearinterpol_create(b: &mut Bencher) {
let data: Vec<_> = get_data();
bench_create::<MultiLinearInterpolFastFieldSerializer>(b, &data);
bench_create::<PiecewiseLinearFastFieldSerializer>(b, &data);
}
#[bench]
fn bench_fastfield_bitpack_get(b: &mut Bencher) {
@@ -85,16 +77,9 @@ mod tests {
bench_get::<BitpackedFastFieldSerializer, BitpackedFastFieldReader>(b, &data);
}
#[bench]
fn bench_fastfield_linearinterpol_get(b: &mut Bencher) {
fn bench_fastfield_piecewise_linear_get(b: &mut Bencher) {
let data: Vec<_> = get_data();
bench_get::<LinearInterpolFastFieldSerializer, LinearInterpolFastFieldReader>(b, &data);
}
#[bench]
fn bench_fastfield_multilinearinterpol_get(b: &mut Bencher) {
let data: Vec<_> = get_data();
bench_get::<MultiLinearInterpolFastFieldSerializer, MultiLinearInterpolFastFieldReader>(
b, &data,
);
bench_get::<PiecewiseLinearFastFieldSerializer, PiecewiseLinearFastFieldReader>(b, &data);
}
pub fn stats_from_vec(data: &[u64]) -> FastFieldStats {
let min_value = data.iter().cloned().min().unwrap_or(0);

17
fastfield_codecs/src/bitpacked.rs
View File

@@ -1,13 +1,9 @@
use crate::FastFieldCodecReader;
use crate::FastFieldCodecSerializer;
use crate::FastFieldDataAccess;
use crate::FastFieldStats;
use common::BinarySerializable;
use std::io::{self, Write};
use tantivy_bitpacker::compute_num_bits;
use tantivy_bitpacker::BitPacker;
use tantivy_bitpacker::BitUnpacker;
use common::BinarySerializable;
use tantivy_bitpacker::{compute_num_bits, BitPacker, BitUnpacker};
use crate::{FastFieldCodecReader, FastFieldCodecSerializer, FastFieldDataAccess, FastFieldStats};
/// Depending on the field type, a different
/// fast field is required.
@@ -132,7 +128,10 @@ impl FastFieldCodecSerializer for BitpackedFastFieldSerializer {
) -> bool {
true
}
fn estimate(_fastfield_accessor: &impl FastFieldDataAccess, stats: FastFieldStats) -> f32 {
fn estimate_compression_ratio(
_fastfield_accessor: &impl FastFieldDataAccess,
stats: FastFieldStats,
) -> f32 {
let amplitude = stats.max_value - stats.min_value;
let num_bits = compute_num_bits(amplitude);
let num_bits_uncompressed = 64;

272
fastfield_codecs/src/frame_of_reference.rs Normal file
View File

@@ -0,0 +1,272 @@
use std::io::{self, Read, Write};
use common::{BinarySerializable, DeserializeFrom};
use tantivy_bitpacker::{compute_num_bits, BitPacker, BitUnpacker};
use crate::{FastFieldCodecReader, FastFieldCodecSerializer, FastFieldDataAccess, FastFieldStats};
const BLOCK_SIZE: u64 = 128;
#[derive(Clone)]
pub struct FORFastFieldReader {
num_vals: u64,
min_value: u64,
max_value: u64,
block_readers: Vec<BlockReader>,
}
#[derive(Clone, Debug, Default)]
struct BlockMetadata {
min: u64,
num_bits: u8,
}
#[derive(Clone, Debug, Default)]
struct BlockReader {
metadata: BlockMetadata,
start_offset: u64,
bit_unpacker: BitUnpacker,
}
impl BlockReader {
fn new(metadata: BlockMetadata, start_offset: u64) -> Self {
Self {
bit_unpacker: BitUnpacker::new(metadata.num_bits),
metadata,
start_offset,
}
}
#[inline]
fn get_u64(&self, block_pos: u64, data: &[u8]) -> u64 {
let diff = self
.bit_unpacker
.get(block_pos, &data[self.start_offset as usize..]);
self.metadata.min + diff
}
}
impl BinarySerializable for BlockMetadata {
fn serialize<W: Write>(&self, write: &mut W) -> io::Result<()> {
self.min.serialize(write)?;
self.num_bits.serialize(write)?;
Ok(())
}
fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
let min = u64::deserialize(reader)?;
let num_bits = u8::deserialize(reader)?;
Ok(Self { min, num_bits })
}
}
#[derive(Clone, Debug)]
pub struct FORFooter {
pub num_vals: u64,
pub min_value: u64,
pub max_value: u64,
block_metadatas: Vec<BlockMetadata>,
}
impl BinarySerializable for FORFooter {
fn serialize<W: Write>(&self, write: &mut W) -> io::Result<()> {
let mut out = vec![];
self.num_vals.serialize(&mut out)?;
self.min_value.serialize(&mut out)?;
self.max_value.serialize(&mut out)?;
self.block_metadatas.serialize(&mut out)?;
write.write_all(&out)?;
(out.len() as u32).serialize(write)?;
Ok(())
}
fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
let footer = Self {
num_vals: u64::deserialize(reader)?,
min_value: u64::deserialize(reader)?,
max_value: u64::deserialize(reader)?,
block_metadatas: Vec::<BlockMetadata>::deserialize(reader)?,
};
Ok(footer)
}
}
impl FastFieldCodecReader for FORFastFieldReader {
/// Opens a fast field given a file.
fn open_from_bytes(bytes: &[u8]) -> io::Result<Self> {
let footer_len: u32 = (&bytes[bytes.len() - 4..]).deserialize()?;
let (_, mut footer) = bytes.split_at(bytes.len() - (4 + footer_len) as usize);
let footer = FORFooter::deserialize(&mut footer)?;
let mut block_readers = Vec::with_capacity(footer.block_metadatas.len());
let mut current_data_offset = 0;
for block_metadata in footer.block_metadatas {
let num_bits = block_metadata.num_bits;
block_readers.push(BlockReader::new(block_metadata, current_data_offset));
current_data_offset += num_bits as u64 * BLOCK_SIZE / 8;
}
Ok(Self {
num_vals: footer.num_vals,
min_value: footer.min_value,
max_value: footer.max_value,
block_readers,
})
}
#[inline]
fn get_u64(&self, idx: u64, data: &[u8]) -> u64 {
let block_idx = (idx / BLOCK_SIZE) as usize;
let block_pos = idx - (block_idx as u64) * BLOCK_SIZE;
let block_reader = &self.block_readers[block_idx];
block_reader.get_u64(block_pos, data)
}
#[inline]
fn min_value(&self) -> u64 {
self.min_value
}
#[inline]
fn max_value(&self) -> u64 {
self.max_value
}
}
/// Same as LinearInterpolFastFieldSerializer, but working on chunks of CHUNK_SIZE elements.
pub struct FORFastFieldSerializer {}
impl FastFieldCodecSerializer for FORFastFieldSerializer {
const NAME: &'static str = "FOR";
const ID: u8 = 5;
/// Creates a new fast field serializer.
fn serialize(
write: &mut impl Write,
_: &impl FastFieldDataAccess,
stats: FastFieldStats,
data_iter: impl Iterator<Item = u64>,
_data_iter1: impl Iterator<Item = u64>,
) -> io::Result<()> {
let data = data_iter.collect::<Vec<_>>();
let mut bit_packer = BitPacker::new();
let mut block_metadatas = Vec::new();
for data_pos in (0..data.len() as u64).step_by(BLOCK_SIZE as usize) {
let block_num_vals = BLOCK_SIZE.min(data.len() as u64 - data_pos) as usize;
let block_values = &data[data_pos as usize..data_pos as usize + block_num_vals];
let mut min = block_values[0];
let mut max = block_values[0];
for &current_value in block_values[1..].iter() {
min = min.min(current_value);
max = max.max(current_value);
}
let num_bits = compute_num_bits(max - min);
for current_value in block_values.iter() {
bit_packer.write(current_value - min, num_bits, write)?;
}
bit_packer.flush(write)?;
block_metadatas.push(BlockMetadata { min, num_bits });
}
bit_packer.close(write)?;
let footer = FORFooter {
num_vals: stats.num_vals,
min_value: stats.min_value,
max_value: stats.max_value,
block_metadatas,
};
footer.serialize(write)?;
Ok(())
}
fn is_applicable(
_fastfield_accessor: &impl FastFieldDataAccess,
stats: FastFieldStats,
) -> bool {
stats.num_vals > BLOCK_SIZE
}
/// Estimate compression ratio by compute the ratio of the first block.
fn estimate_compression_ratio(
fastfield_accessor: &impl FastFieldDataAccess,
stats: FastFieldStats,
) -> f32 {
let last_elem_in_first_chunk = BLOCK_SIZE.min(stats.num_vals);
let max_distance = (0..last_elem_in_first_chunk)
.into_iter()
.map(|pos| {
let actual_value = fastfield_accessor.get_val(pos as u64);
actual_value - stats.min_value
})
.max()
.unwrap();
// Estimate one block and multiply by a magic number 3 to select this codec
// when we are almost sure that this is relevant.
let relative_max_value = max_distance as f32 * 3.0;
let num_bits = compute_num_bits(relative_max_value as u64) as u64 * stats.num_vals as u64
// function metadata per block
+ 9 * (stats.num_vals / BLOCK_SIZE);
let num_bits_uncompressed = 64 * stats.num_vals;
num_bits as f32 / num_bits_uncompressed as f32
}
}
#[cfg(test)]
mod tests {
use super::*;
use crate::tests::get_codec_test_data_sets;
fn create_and_validate(data: &[u64], name: &str) -> (f32, f32) {
crate::tests::create_and_validate::<FORFastFieldSerializer, FORFastFieldReader>(data, name)
}
#[test]
fn test_compression() {
let data = (10..=6_000_u64).collect::<Vec<_>>();
let (estimate, actual_compression) =
create_and_validate(&data, "simple monotonically large");
println!("{}", actual_compression);
assert!(actual_compression < 0.2);
assert!(actual_compression > 0.006);
assert!(estimate < 0.20);
assert!(estimate > 0.10);
}
#[test]
fn test_with_codec_data_sets() {
let data_sets = get_codec_test_data_sets();
for (mut data, name) in data_sets {
create_and_validate(&data, name);
data.reverse();
create_and_validate(&data, name);
}
}
#[test]
fn test_simple() {
let data = (10..=20_u64).collect::<Vec<_>>();
create_and_validate(&data, "simple monotonically");
}
#[test]
fn border_cases_1() {
let data = (0..1024).collect::<Vec<_>>();
create_and_validate(&data, "border case");
}
#[test]
fn border_case_2() {
let data = (0..1025).collect::<Vec<_>>();
create_and_validate(&data, "border case");
}
#[test]
fn rand() {
for _ in 0..10 {
let mut data = (5_000..20_000)
.map(|_| rand::random::<u32>() as u64)
.collect::<Vec<_>>();
let (estimate, actual_compression) = create_and_validate(&data, "random");
dbg!(estimate);
dbg!(actual_compression);
data.reverse();
create_and_validate(&data, "random");
}
}
}

87
fastfield_codecs/src/lib.rs
View File

@@ -6,15 +6,20 @@ use std::io;
use std::io::Write;
pub mod bitpacked;
#[cfg(feature = "unstable")]
pub mod frame_of_reference;
pub mod linearinterpol;
pub mod multilinearinterpol;
pub mod piecewise_linear;
pub trait FastFieldCodecReader: Sized {
/// reads the metadata and returns the CodecReader
/// Reads the metadata and returns the CodecReader.
fn open_from_bytes(bytes: &[u8]) -> std::io::Result<Self>;
fn get_u64(&self, doc: u64, data: &[u8]) -> u64;
/// Read u64 value for indice `idx`.
/// `idx` can be either a `DocId` or an index used for
/// `multivalued` fast field.
fn get_u64(&self, idx: u64, data: &[u8]) -> u64;
fn min_value(&self) -> u64;
fn max_value(&self) -> u64;
}
@@ -35,7 +40,10 @@ pub trait FastFieldCodecSerializer {
///
/// It could make sense to also return a value representing
/// computational complexity.
fn estimate(fastfield_accessor: &impl FastFieldDataAccess, stats: FastFieldStats) -> f32;
fn estimate_compression_ratio(
fastfield_accessor: &impl FastFieldDataAccess,
stats: FastFieldStats,
) -> f32;
/// Serializes the data using the serializer into write.
/// There are multiple iterators, in case the codec needs to read the data multiple times.
@@ -53,7 +61,8 @@ pub trait FastFieldCodecSerializer {
pub trait FastFieldDataAccess {
/// Return the value associated to the given position.
///
/// Whenever possible use the Iterator passed to the fastfield creation instead, for performance reasons.
/// Whenever possible use the Iterator passed to the fastfield creation instead, for performance
/// reasons.
///
/// # Panics
///
@@ -62,6 +71,7 @@ pub trait FastFieldDataAccess {
}
#[derive(Debug, Clone)]
/// Statistics are used in codec detection and stored in the fast field footer.
pub struct FastFieldStats {
pub min_value: u64,
pub max_value: u64,
@@ -82,12 +92,9 @@ impl FastFieldDataAccess for Vec<u64> {
#[cfg(test)]
mod tests {
use crate::{
bitpacked::{BitpackedFastFieldReader, BitpackedFastFieldSerializer},
linearinterpol::{LinearInterpolFastFieldReader, LinearInterpolFastFieldSerializer},
multilinearinterpol::{
MultiLinearInterpolFastFieldReader, MultiLinearInterpolFastFieldSerializer,
},
use crate::bitpacked::{BitpackedFastFieldReader, BitpackedFastFieldSerializer};
use crate::piecewise_linear::{
PiecewiseLinearFastFieldReader, PiecewiseLinearFastFieldSerializer,
};
pub fn create_and_validate<S: FastFieldCodecSerializer, R: FastFieldCodecReader>(
@@ -97,7 +104,7 @@ mod tests {
if !S::is_applicable(&data, crate::tests::stats_from_vec(data)) {
return (f32::MAX, 0.0);
}
let estimation = S::estimate(&data, crate::tests::stats_from_vec(data));
let estimation = S::estimate_compression_ratio(&data, crate::tests::stats_from_vec(data));
let mut out = vec![];
S::serialize(
&mut out,
@@ -157,13 +164,10 @@ mod tests {
fn test_codec_bitpacking() {
test_codec::<BitpackedFastFieldSerializer, BitpackedFastFieldReader>();
}
#[test]
fn test_codec_interpolation() {
test_codec::<LinearInterpolFastFieldSerializer, LinearInterpolFastFieldReader>();
}
#[test]
fn test_codec_multi_interpolation() {
test_codec::<MultiLinearInterpolFastFieldSerializer, MultiLinearInterpolFastFieldReader>();
fn test_codec_piecewise_linear() {
test_codec::<PiecewiseLinearFastFieldSerializer, PiecewiseLinearFastFieldReader>();
}
use super::*;
@@ -181,45 +185,50 @@ mod tests {
fn estimation_good_interpolation_case() {
let data = (10..=20000_u64).collect::<Vec<_>>();
let linear_interpol_estimation =
LinearInterpolFastFieldSerializer::estimate(&data, stats_from_vec(&data));
assert_le!(linear_interpol_estimation, 0.01);
let multi_linear_interpol_estimation =
MultiLinearInterpolFastFieldSerializer::estimate(&data, stats_from_vec(&data));
assert_le!(multi_linear_interpol_estimation, 0.2);
assert_le!(linear_interpol_estimation, multi_linear_interpol_estimation);
let piecewise_interpol_estimation =
PiecewiseLinearFastFieldSerializer::estimate_compression_ratio(
&data,
stats_from_vec(&data),
);
assert_le!(piecewise_interpol_estimation, 0.2);
let bitpacked_estimation =
BitpackedFastFieldSerializer::estimate(&data, stats_from_vec(&data));
assert_le!(linear_interpol_estimation, bitpacked_estimation);
BitpackedFastFieldSerializer::estimate_compression_ratio(&data, stats_from_vec(&data));
assert_le!(piecewise_interpol_estimation, bitpacked_estimation);
}
#[test]
fn estimation_test_bad_interpolation_case() {
let data = vec![200, 10, 10, 10, 10, 1000, 20];
let linear_interpol_estimation =
LinearInterpolFastFieldSerializer::estimate(&data, stats_from_vec(&data));
assert_le!(linear_interpol_estimation, 0.32);
let piecewise_interpol_estimation =
PiecewiseLinearFastFieldSerializer::estimate_compression_ratio(
&data,
stats_from_vec(&data),
);
assert_le!(piecewise_interpol_estimation, 0.32);
let bitpacked_estimation =
BitpackedFastFieldSerializer::estimate(&data, stats_from_vec(&data));
assert_le!(bitpacked_estimation, linear_interpol_estimation);
BitpackedFastFieldSerializer::estimate_compression_ratio(&data, stats_from_vec(&data));
assert_le!(bitpacked_estimation, piecewise_interpol_estimation);
}
#[test]
fn estimation_test_bad_interpolation_case_monotonically_increasing() {
fn estimation_test_interpolation_case_monotonically_increasing() {
let mut data = (200..=20000_u64).collect::<Vec<_>>();
data.push(1_000_000);
// in this case the linear interpolation can't in fact not be worse than bitpacking,
// but the estimator adds some threshold, which leads to estimated worse behavior
let linear_interpol_estimation =
LinearInterpolFastFieldSerializer::estimate(&data, stats_from_vec(&data));
assert_le!(linear_interpol_estimation, 0.35);
let piecewise_interpol_estimation =
PiecewiseLinearFastFieldSerializer::estimate_compression_ratio(
&data,
stats_from_vec(&data),
);
assert_le!(piecewise_interpol_estimation, 0.2);
let bitpacked_estimation =
BitpackedFastFieldSerializer::estimate(&data, stats_from_vec(&data));
BitpackedFastFieldSerializer::estimate_compression_ratio(&data, stats_from_vec(&data));
println!("{}", bitpacked_estimation);
assert_le!(bitpacked_estimation, 0.32);
assert_le!(bitpacked_estimation, linear_interpol_estimation);
assert_le!(piecewise_interpol_estimation, bitpacked_estimation);
}
}

49
fastfield_codecs/src/linearinterpol.rs
View File

@@ -1,15 +1,10 @@
use crate::FastFieldCodecReader;
use crate::FastFieldCodecSerializer;
use crate::FastFieldDataAccess;
use crate::FastFieldStats;
use std::io::{self, Read, Write};
use std::ops::Sub;
use tantivy_bitpacker::compute_num_bits;
use tantivy_bitpacker::BitPacker;
use common::BinarySerializable;
use common::FixedSize;
use tantivy_bitpacker::BitUnpacker;
use common::{BinarySerializable, FixedSize};
use tantivy_bitpacker::{compute_num_bits, BitPacker, BitUnpacker};
use crate::{FastFieldCodecReader, FastFieldCodecSerializer, FastFieldDataAccess, FastFieldStats};
/// Depending on the field type, a different
/// fast field is required.
@@ -76,9 +71,9 @@ impl FastFieldCodecReader for LinearInterpolFastFieldReader {
})
}
#[inline]
fn get_u64(&self, doc: u64, data: &[u8]) -> u64 {
let calculated_value = get_calculated_value(self.footer.first_val, doc, self.slope);
(calculated_value + self.bit_unpacker.get(doc, data)) - self.footer.offset
fn get_u64(&self, idx: u64, data: &[u8]) -> u64 {
let calculated_value = get_calculated_value(self.footer.first_val, idx, self.slope);
(calculated_value + self.bit_unpacker.get(idx, data)) - self.footer.offset
}
#[inline]
@@ -93,6 +88,10 @@ impl FastFieldCodecReader for LinearInterpolFastFieldReader {
/// Fastfield serializer, which tries to guess values by linear interpolation
/// and stores the difference bitpacked.
#[deprecated(
note = "Linear interpolation works best only on very rare cases and piecewise linear codec \
already works great on them."
)]
pub struct LinearInterpolFastFieldSerializer {}
#[inline]
@@ -110,6 +109,7 @@ fn get_calculated_value(first_val: u64, pos: u64, slope: f32) -> u64 {
first_val + (pos as f32 * slope) as u64
}
#[allow(deprecated)]
impl FastFieldCodecSerializer for LinearInterpolFastFieldSerializer {
const NAME: &'static str = "LinearInterpol";
const ID: u8 = 2;
@@ -137,7 +137,7 @@ impl FastFieldCodecSerializer for LinearInterpolFastFieldSerializer {
// will be offset to 0
offset = offset.max(calculated_value - actual_value);
} else {
//positive value no offset reuqired
// positive value no offset reuqired
rel_positive_max = rel_positive_max.max(actual_value - calculated_value);
}
}
@@ -171,7 +171,7 @@ impl FastFieldCodecSerializer for LinearInterpolFastFieldSerializer {
stats: FastFieldStats,
) -> bool {
if stats.num_vals < 3 {
return false; //disable compressor for this case
return false; // disable compressor for this case
}
// On serialisation the offset is added to the actual value.
// We need to make sure this won't run into overflow calculation issues.
@@ -187,10 +187,16 @@ impl FastFieldCodecSerializer for LinearInterpolFastFieldSerializer {
}
true
}
/// estimation for linear interpolation is hard because, you don't know
/// Estimation for linear interpolation is hard because, you don't know
/// where the local maxima for the deviation of the calculated value are and
/// the offset to shift all values to >=0 is also unknown.
fn estimate(fastfield_accessor: &impl FastFieldDataAccess, stats: FastFieldStats) -> f32 {
fn estimate_compression_ratio(
fastfield_accessor: &impl FastFieldDataAccess,
stats: FastFieldStats,
) -> f32 {
if stats.num_vals < 3 {
return f32::MAX;
}
let first_val = fastfield_accessor.get_val(0);
let last_val = fastfield_accessor.get_val(stats.num_vals as u64 - 1);
let slope = get_slope(first_val, last_val, stats.num_vals);
@@ -211,8 +217,8 @@ impl FastFieldCodecSerializer for LinearInterpolFastFieldSerializer {
.max()
.unwrap_or(0);
// the theory would be that we don't have the actual max_distance, but we are close within 50%
// threshold.
// the theory would be that we don't have the actual max_distance, but we are close within
// 50% threshold.
// It is multiplied by 2 because in a log case scenario the line would be as much above as
// below. So the offset would = max_distance
//
@@ -234,6 +240,7 @@ fn distance<T: Sub<Output = T> + Ord>(x: T, y: T) -> T {
}
}
#[allow(deprecated)]
#[cfg(test)]
mod tests {
use super::*;
@@ -294,8 +301,10 @@ mod tests {
#[test]
fn linear_interpol_fast_field_rand() {
for _ in 0..5000 {
let mut data = (0..50).map(|_| rand::random::<u64>()).collect::<Vec<_>>();
for _ in 0..10 {
let mut data = (5_000..20_000)
.map(|_| rand::random::<u32>() as u64)
.collect::<Vec<_>>();
create_and_validate(&data, "random");
data.reverse();

141
fastfield_codecs/src/main.rs
View File

@@ -1,33 +1,52 @@
#[macro_use]
extern crate prettytable;
use fastfield_codecs::{
linearinterpol::LinearInterpolFastFieldSerializer,
multilinearinterpol::MultiLinearInterpolFastFieldSerializer, FastFieldCodecSerializer,
FastFieldStats,
use std::fs::File;
use std::io;
use std::io::BufRead;
use std::time::{Duration, Instant};
use common::f64_to_u64;
use fastfield_codecs::bitpacked::BitpackedFastFieldReader;
#[cfg(feature = "unstable")]
use fastfield_codecs::frame_of_reference::{FORFastFieldReader, FORFastFieldSerializer};
use fastfield_codecs::piecewise_linear::{
PiecewiseLinearFastFieldReader, PiecewiseLinearFastFieldSerializer,
};
use fastfield_codecs::{FastFieldCodecReader, FastFieldCodecSerializer, FastFieldStats};
use prettytable::{Cell, Row, Table};
use rand::prelude::StdRng;
use rand::Rng;
fn main() {
let mut table = Table::new();
// Add a row per time
table.add_row(row!["", "Compression Ratio", "Compression Estimation"]);
table.add_row(row![
"",
"Compression ratio",
"Compression ratio estimation",
"Compression time (micro)",
"Reading time (micro)"
]);
for (data, data_set_name) in get_codec_test_data_sets() {
let mut results = vec![];
let res = serialize_with_codec::<LinearInterpolFastFieldSerializer>(&data);
let res = serialize_with_codec::<
PiecewiseLinearFastFieldSerializer,
PiecewiseLinearFastFieldReader,
>(&data);
results.push(res);
let res = serialize_with_codec::<MultiLinearInterpolFastFieldSerializer>(&data);
results.push(res);
let res = serialize_with_codec::<fastfield_codecs::bitpacked::BitpackedFastFieldSerializer>(
&data,
);
#[cfg(feature = "unstable")]
{
let res = serialize_with_codec::<FORFastFieldSerializer, FORFastFieldReader>(&data);
results.push(res);
}
let res = serialize_with_codec::<
fastfield_codecs::bitpacked::BitpackedFastFieldSerializer,
BitpackedFastFieldReader,
>(&data);
results.push(res);
//let best_estimation_codec = results
//.iter()
//.min_by(|res1, res2| res1.partial_cmp(&res2).unwrap())
//.unwrap();
let best_compression_ratio_codec = results
.iter()
.min_by(|res1, res2| res1.partial_cmp(res2).unwrap())
@@ -35,13 +54,12 @@ fn main() {
.unwrap();
table.add_row(Row::new(vec![Cell::new(data_set_name).style_spec("Bbb")]));
for (is_applicable, est, comp, name) in results {
for (is_applicable, est, comp, name, compression_duration, read_duration) in results {
let (est_cell, ratio_cell) = if !is_applicable {
("Codec Disabled".to_string(), "".to_string())
} else {
(est.to_string(), comp.to_string())
};
#[allow(clippy::all)]
let style = if comp == best_compression_ratio_codec.1 {
"Fb"
} else {
@@ -49,9 +67,11 @@ fn main() {
};
table.add_row(Row::new(vec![
Cell::new(&name.to_string()).style_spec("bFg"),
Cell::new(name).style_spec("bFg"),
Cell::new(&ratio_cell).style_spec(style),
Cell::new(&est_cell).style_spec(""),
Cell::new(&compression_duration.as_micros().to_string()),
Cell::new(&read_duration.as_micros().to_string()),
]));
}
}
@@ -73,7 +93,6 @@ pub fn get_codec_test_data_sets() -> Vec<(Vec<u64>, &'static str)> {
current_cumulative
})
.collect::<Vec<_>>();
//let data = (1..=200000_u64).map(|num| num + num).collect::<Vec<_>>();
data_and_names.push((data, "Monotonically increasing concave"));
let mut current_cumulative = 0;
@@ -86,22 +105,79 @@ pub fn get_codec_test_data_sets() -> Vec<(Vec<u64>, &'static str)> {
.collect::<Vec<_>>();
data_and_names.push((data, "Monotonically increasing convex"));
let mut rng: StdRng = rand::SeedableRng::seed_from_u64(1);
let data = (1000..=200_000_u64)
.map(|num| num + rand::random::<u8>() as u64)
.map(|num| num + rng.gen::<u8>() as u64)
.collect::<Vec<_>>();
data_and_names.push((data, "Almost monotonically increasing"));
let data = (1000..=200_000_u64)
.map(|_| rng.gen::<u8>() as u64)
.collect::<Vec<_>>();
data_and_names.push((data, "Random"));
let mut data = load_dataset("datasets/hdfs_logs_timestamps.txt");
data_and_names.push((data.clone(), "HDFS logs timestamps"));
data.sort_unstable();
data_and_names.push((data, "HDFS logs timestamps SORTED"));
let data = load_dataset("datasets/http_logs_timestamps.txt");
data_and_names.push((data, "HTTP logs timestamps SORTED"));
let mut data = load_dataset("datasets/amazon_reviews_product_ids.txt");
data_and_names.push((data.clone(), "Amazon review product ids"));
data.sort_unstable();
data_and_names.push((data, "Amazon review product ids SORTED"));
let data = load_float_dataset("datasets/nooc_temperatures.txt");
data_and_names.push((data, "Temperatures"));
data_and_names
}
pub fn serialize_with_codec<S: FastFieldCodecSerializer>(
pub fn load_dataset(file_path: &str) -> Vec<u64> {
println!("Load dataset from `{}`", file_path);
let file = File::open(file_path).expect("Error when opening file.");
let lines = io::BufReader::new(file).lines();
let mut data = Vec::new();
for line in lines {
let l = line.unwrap();
data.push(l.parse::<u64>().unwrap());
}
data
}
pub fn load_float_dataset(file_path: &str) -> Vec<u64> {
println!("Load float dataset from `{}`", file_path);
let file = File::open(file_path).expect("Error when opening file.");
let lines = io::BufReader::new(file).lines();
let mut data = Vec::new();
for line in lines {
let line_string = line.unwrap();
let value = line_string.parse::<f64>().unwrap();
data.push(f64_to_u64(value));
}
data
}
pub fn serialize_with_codec<S: FastFieldCodecSerializer, R: FastFieldCodecReader>(
data: &[u64],
) -> (bool, f32, f32, &'static str) {
) -> (bool, f32, f32, &'static str, Duration, Duration) {
let is_applicable = S::is_applicable(&data, stats_from_vec(data));
if !is_applicable {
return (false, 0.0, 0.0, S::NAME);
return (
false,
0.0,
0.0,
S::NAME,
Duration::from_secs(0),
Duration::from_secs(0),
);
}
let estimation = S::estimate(&data, stats_from_vec(data));
let start_time_compression = Instant::now();
let estimation = S::estimate_compression_ratio(&data, stats_from_vec(data));
let mut out = vec![];
S::serialize(
&mut out,
@@ -111,9 +187,22 @@ pub fn serialize_with_codec<S: FastFieldCodecSerializer>(
data.iter().cloned(),
)
.unwrap();
let elasped_time_compression = start_time_compression.elapsed();
let actual_compression = out.len() as f32 / (data.len() * 8) as f32;
(true, estimation, actual_compression, S::NAME)
let reader = R::open_from_bytes(&out).unwrap();
let start_time_read = Instant::now();
for doc in 0..data.len() {
reader.get_u64(doc as u64, &out);
}
let elapsed_time_read = start_time_read.elapsed();
(
true,
estimation,
actual_compression,
S::NAME,
elasped_time_compression,
elapsed_time_read,
)
}
pub fn stats_from_vec(data: &[u64]) -> FastFieldStats {

81
fastfield_codecs/src/multilinearinterpol.rs
View File

@@ -1,30 +1,22 @@
/*!
//! MultiLinearInterpol compressor uses linear interpolation to guess a values and stores the
//! offset, but in blocks of 512.
//!
//! With a CHUNK_SIZE of 512 and 29 byte metadata per block, we get a overhead for metadata of 232 /
//! 512 = 0,45 bits per element. The additional space required per element in a block is the the
//! maximum deviation of the linear interpolation estimation function.
//!
//! E.g. if the maximum deviation of an element is 12, all elements cost 4bits.
//!
//! Size per block:
//! Num Elements * Maximum Deviation from Interpolation + 29 Byte Metadata
MultiLinearInterpol compressor uses linear interpolation to guess a values and stores the offset, but in blocks of 512.
With a CHUNK_SIZE of 512 and 29 byte metadata per block, we get a overhead for metadata of 232 / 512 = 0,45 bits per element.
The additional space required per element in a block is the the maximum deviation of the linear interpolation estimation function.
E.g. if the maximum deviation of an element is 12, all elements cost 4bits.
Size per block:
Num Elements * Maximum Deviation from Interpolation + 29 Byte Metadata
*/
use crate::FastFieldCodecReader;
use crate::FastFieldCodecSerializer;
use crate::FastFieldDataAccess;
use crate::FastFieldStats;
use common::CountingWriter;
use std::io::{self, Read, Write};
use std::ops::Sub;
use tantivy_bitpacker::compute_num_bits;
use tantivy_bitpacker::BitPacker;
use common::BinarySerializable;
use common::DeserializeFrom;
use tantivy_bitpacker::BitUnpacker;
use common::{BinarySerializable, CountingWriter, DeserializeFrom};
use tantivy_bitpacker::{compute_num_bits, BitPacker, BitUnpacker};
use crate::{FastFieldCodecReader, FastFieldCodecSerializer, FastFieldDataAccess, FastFieldStats};
const CHUNK_SIZE: u64 = 512;
@@ -163,14 +155,17 @@ impl FastFieldCodecReader for MultiLinearInterpolFastFieldReader {
}
#[inline]
fn get_u64(&self, doc: u64, data: &[u8]) -> u64 {
let interpolation = get_interpolation_function(doc, &self.footer.interpolations);
let doc = doc - interpolation.start_pos;
let calculated_value =
get_calculated_value(interpolation.value_start_pos, doc, interpolation.slope);
fn get_u64(&self, idx: u64, data: &[u8]) -> u64 {
let interpolation = get_interpolation_function(idx, &self.footer.interpolations);
let block_idx = idx - interpolation.start_pos;
let calculated_value = get_calculated_value(
interpolation.value_start_pos,
block_idx,
interpolation.slope,
);
let diff = interpolation
.bit_unpacker
.get(doc, &data[interpolation.data_start_offset as usize..]);
.get(block_idx, &data[interpolation.data_start_offset as usize..]);
(calculated_value + diff) - interpolation.positive_val_offset
}
@@ -195,8 +190,13 @@ fn get_calculated_value(first_val: u64, pos: u64, slope: f32) -> u64 {
}
/// Same as LinearInterpolFastFieldSerializer, but working on chunks of CHUNK_SIZE elements.
#[deprecated(
note = "MultiLinearInterpol is replaced by PiecewiseLinear codec which fixes the slope and is \
a little bit more optimized."
)]
pub struct MultiLinearInterpolFastFieldSerializer {}
#[allow(deprecated)]
impl FastFieldCodecSerializer for MultiLinearInterpolFastFieldSerializer {
const NAME: &'static str = "MultiLinearInterpol";
const ID: u8 = 3;
@@ -252,11 +252,11 @@ impl FastFieldCodecSerializer for MultiLinearInterpolFastFieldSerializer {
);
if calculated_value > actual_value {
// negative value we need to apply an offset
// we ignore negative values in the max value calculation, because negative values
// will be offset to 0
// we ignore negative values in the max value calculation, because negative
// values will be offset to 0
offset = offset.max(calculated_value - actual_value);
} else {
//positive value no offset reuqired
// positive value no offset reuqired
rel_positive_max = rel_positive_max.max(actual_value - calculated_value);
}
}
@@ -319,10 +319,13 @@ impl FastFieldCodecSerializer for MultiLinearInterpolFastFieldSerializer {
}
true
}
/// estimation for linear interpolation is hard because, you don't know
/// Estimation for linear interpolation is hard because, you don't know
/// where the local maxima are for the deviation of the calculated value and
/// the offset is also unknown.
fn estimate(fastfield_accessor: &impl FastFieldDataAccess, stats: FastFieldStats) -> f32 {
fn estimate_compression_ratio(
fastfield_accessor: &impl FastFieldDataAccess,
stats: FastFieldStats,
) -> f32 {
let first_val_in_first_block = fastfield_accessor.get_val(0);
let last_elem_in_first_chunk = CHUNK_SIZE.min(stats.num_vals);
let last_val_in_first_block =
@@ -350,8 +353,8 @@ impl FastFieldCodecSerializer for MultiLinearInterpolFastFieldSerializer {
.unwrap();
// Estimate one block and extrapolate the cost to all blocks.
// the theory would be that we don't have the actual max_distance, but we are close within 50%
// threshold.
// the theory would be that we don't have the actual max_distance, but we are close within
// 50% threshold.
// It is multiplied by 2 because in a log case scenario the line would be as much above as
// below. So the offset would = max_distance
//
@@ -374,6 +377,7 @@ fn distance<T: Sub<Output = T> + Ord>(x: T, y: T) -> T {
}
#[cfg(test)]
#[allow(deprecated)]
mod tests {
use super::*;
use crate::tests::get_codec_test_data_sets;
@@ -427,10 +431,7 @@ mod tests {
let mut data = (5_000..20_000)
.map(|_| rand::random::<u32>() as u64)
.collect::<Vec<_>>();
let (estimate, actual_compression) = create_and_validate(&data, "random");
dbg!(estimate);
dbg!(actual_compression);
let _ = create_and_validate(&data, "random");
data.reverse();
create_and_validate(&data, "random");
}

365
fastfield_codecs/src/piecewise_linear.rs Normal file
View File

@@ -0,0 +1,365 @@
//! PiecewiseLinear codec uses piecewise linear functions for every block of 512 values to predict
//! values and fast field values. The difference with real fast field values is then stored.
//! For every block, the linear function can be expressed as
//! `computed_value = slope * block_position + first_value + positive_offset`
//! where:
//! - `block_position` is the position inside of the block from 0 to 511
//! - `first_value` is the first value on the block
//! - `positive_offset` is computed such that we ensure the diff `real_value - computed_value` is
//! always positive.
//!
//! 21 bytes is needed to store the block metadata, it adds an overhead of 21 * 8 / 512 = 0,33 bits
//! per element.
use std::io::{self, Read, Write};
use std::ops::Sub;
use common::{BinarySerializable, DeserializeFrom};
use tantivy_bitpacker::{compute_num_bits, BitPacker, BitUnpacker};
use crate::{FastFieldCodecReader, FastFieldCodecSerializer, FastFieldDataAccess, FastFieldStats};
const BLOCK_SIZE: u64 = 512;
#[derive(Clone)]
pub struct PiecewiseLinearFastFieldReader {
min_value: u64,
max_value: u64,
block_readers: Vec<BlockReader>,
}
/// Block that stores metadata to predict value with a linear
/// function `predicted_value = slope * position + first_value + positive_offset`
/// where `positive_offset` is comupted such that predicted values
/// are always positive.
#[derive(Clone, Debug, Default)]
struct BlockMetadata {
first_value: u64,
positive_offset: u64,
slope: f32,
num_bits: u8,
}
#[derive(Clone, Debug, Default)]
struct BlockReader {
metadata: BlockMetadata,
start_offset: u64,
bit_unpacker: BitUnpacker,
}
impl BlockReader {
fn new(metadata: BlockMetadata, start_offset: u64) -> Self {
Self {
bit_unpacker: BitUnpacker::new(metadata.num_bits),
metadata,
start_offset,
}
}
#[inline]
fn get_u64(&self, block_pos: u64, data: &[u8]) -> u64 {
let diff = self
.bit_unpacker
.get(block_pos, &data[self.start_offset as usize..]);
let predicted_value =
predict_value(self.metadata.first_value, block_pos, self.metadata.slope);
(predicted_value + diff) - self.metadata.positive_offset
}
}
impl BinarySerializable for BlockMetadata {
fn serialize<W: Write>(&self, write: &mut W) -> io::Result<()> {
self.first_value.serialize(write)?;
self.positive_offset.serialize(write)?;
self.slope.serialize(write)?;
self.num_bits.serialize(write)?;
Ok(())
}
fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
let first_value = u64::deserialize(reader)?;
let positive_offset = u64::deserialize(reader)?;
let slope = f32::deserialize(reader)?;
let num_bits = u8::deserialize(reader)?;
Ok(Self {
first_value,
positive_offset,
slope,
num_bits,
})
}
}
#[derive(Clone, Debug)]
pub struct PiecewiseLinearFooter {
pub num_vals: u64,
pub min_value: u64,
pub max_value: u64,
block_metadatas: Vec<BlockMetadata>,
}
impl BinarySerializable for PiecewiseLinearFooter {
fn serialize<W: Write>(&self, write: &mut W) -> io::Result<()> {
let mut out = vec![];
self.num_vals.serialize(&mut out)?;
self.min_value.serialize(&mut out)?;
self.max_value.serialize(&mut out)?;
self.block_metadatas.serialize(&mut out)?;
write.write_all(&out)?;
(out.len() as u32).serialize(write)?;
Ok(())
}
fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
let footer = Self {
num_vals: u64::deserialize(reader)?,
min_value: u64::deserialize(reader)?,
max_value: u64::deserialize(reader)?,
block_metadatas: Vec::<BlockMetadata>::deserialize(reader)?,
};
Ok(footer)
}
}
impl FastFieldCodecReader for PiecewiseLinearFastFieldReader {
/// Opens a fast field given a file.
fn open_from_bytes(bytes: &[u8]) -> io::Result<Self> {
let footer_len: u32 = (&bytes[bytes.len() - 4..]).deserialize()?;
let (_, mut footer) = bytes.split_at(bytes.len() - (4 + footer_len) as usize);
let footer = PiecewiseLinearFooter::deserialize(&mut footer)?;
let mut block_readers = Vec::with_capacity(footer.block_metadatas.len());
let mut current_data_offset = 0;
for block_metadata in footer.block_metadatas.into_iter() {
let num_bits = block_metadata.num_bits;
block_readers.push(BlockReader::new(block_metadata, current_data_offset));
current_data_offset += num_bits as u64 * BLOCK_SIZE / 8;
}
Ok(Self {
min_value: footer.min_value,
max_value: footer.max_value,
block_readers,
})
}
#[inline]
fn get_u64(&self, idx: u64, data: &[u8]) -> u64 {
let block_idx = (idx / BLOCK_SIZE) as usize;
let block_pos = idx - (block_idx as u64) * BLOCK_SIZE;
let block_reader = &self.block_readers[block_idx];
block_reader.get_u64(block_pos, data)
}
#[inline]
fn min_value(&self) -> u64 {
self.min_value
}
#[inline]
fn max_value(&self) -> u64 {
self.max_value
}
}
#[inline]
fn predict_value(first_val: u64, pos: u64, slope: f32) -> u64 {
(first_val as i64 + (pos as f32 * slope) as i64) as u64
}
pub struct PiecewiseLinearFastFieldSerializer;
impl FastFieldCodecSerializer for PiecewiseLinearFastFieldSerializer {
const NAME: &'static str = "PiecewiseLinear";
const ID: u8 = 4;
/// Creates a new fast field serializer.
fn serialize(
write: &mut impl Write,
_: &impl FastFieldDataAccess,
stats: FastFieldStats,
data_iter: impl Iterator<Item = u64>,
_data_iter1: impl Iterator<Item = u64>,
) -> io::Result<()> {
let mut data = data_iter.collect::<Vec<_>>();
let mut bit_packer = BitPacker::new();
let mut block_metadatas = Vec::new();
for data_pos in (0..data.len() as u64).step_by(BLOCK_SIZE as usize) {
let block_num_vals = BLOCK_SIZE.min(data.len() as u64 - data_pos) as usize;
let block_values = &mut data[data_pos as usize..data_pos as usize + block_num_vals];
let slope = if block_num_vals == 1 {
0f32
} else {
((block_values[block_values.len() - 1] as f64 - block_values[0] as f64)
/ (block_num_vals - 1) as f64) as f32
};
let first_value = block_values[0];
let mut positive_offset = 0;
let mut max_delta = 0;
for (pos, &current_value) in block_values[1..].iter().enumerate() {
let computed_value = predict_value(first_value, pos as u64 + 1, slope);
if computed_value > current_value {
positive_offset = positive_offset.max(computed_value - current_value);
} else {
max_delta = max_delta.max(current_value - computed_value);
}
}
let num_bits = compute_num_bits(max_delta + positive_offset);
for (pos, current_value) in block_values.iter().enumerate() {
let computed_value = predict_value(first_value, pos as u64, slope);
let diff = (current_value + positive_offset) - computed_value;
bit_packer.write(diff, num_bits, write)?;
}
bit_packer.flush(write)?;
block_metadatas.push(BlockMetadata {
first_value,
positive_offset,
slope,
num_bits,
});
}
bit_packer.close(write)?;
let footer = PiecewiseLinearFooter {
num_vals: stats.num_vals,
min_value: stats.min_value,
max_value: stats.max_value,
block_metadatas,
};
footer.serialize(write)?;
Ok(())
}
fn is_applicable(
_fastfield_accessor: &impl FastFieldDataAccess,
stats: FastFieldStats,
) -> bool {
if stats.num_vals < 10 * BLOCK_SIZE {
return false;
}
// On serialization the offset is added to the actual value.
// We need to make sure this won't run into overflow calculation issues.
// For this we take the maximum theroretical offset and add this to the max value.
// If this doesn't overflow the algortihm should be fine
let theorethical_maximum_offset = stats.max_value - stats.min_value;
if stats
.max_value
.checked_add(theorethical_maximum_offset)
.is_none()
{
return false;
}
true
}
/// Estimation for linear interpolation is hard because, you don't know
/// where the local maxima are for the deviation of the calculated value and
/// the offset is also unknown.
fn estimate_compression_ratio(
fastfield_accessor: &impl FastFieldDataAccess,
stats: FastFieldStats,
) -> f32 {
let first_val_in_first_block = fastfield_accessor.get_val(0);
let last_elem_in_first_chunk = BLOCK_SIZE.min(stats.num_vals);
let last_val_in_first_block =
fastfield_accessor.get_val(last_elem_in_first_chunk as u64 - 1);
let slope = ((last_val_in_first_block as f64 - first_val_in_first_block as f64)
/ (stats.num_vals - 1) as f64) as f32;
// let's sample at 0%, 5%, 10% .. 95%, 100%, but for the first block only
let sample_positions = (0..20)
.map(|pos| (last_elem_in_first_chunk as f32 / 100.0 * pos as f32 * 5.0) as usize)
.collect::<Vec<_>>();
let max_distance = sample_positions
.iter()
.map(|&pos| {
let calculated_value = predict_value(first_val_in_first_block, pos as u64, slope);
let actual_value = fastfield_accessor.get_val(pos as u64);
distance(calculated_value, actual_value)
})
.max()
.unwrap();
// Estimate one block and extrapolate the cost to all blocks.
// the theory would be that we don't have the actual max_distance, but we are close within
// 50% threshold.
// It is multiplied by 2 because in a log case scenario the line would be as much above as
// below. So the offset would = max_distance
let relative_max_value = (max_distance as f32 * 1.5) * 2.0;
let num_bits = compute_num_bits(relative_max_value as u64) as u64 * stats.num_vals as u64
// function metadata per block
+ 21 * (stats.num_vals / BLOCK_SIZE);
let num_bits_uncompressed = 64 * stats.num_vals;
num_bits as f32 / num_bits_uncompressed as f32
}
}
fn distance<T: Sub<Output = T> + Ord>(x: T, y: T) -> T {
if x < y {
y - x
} else {
x - y
}
}
#[cfg(test)]
mod tests {
use super::*;
use crate::tests::get_codec_test_data_sets;
fn create_and_validate(data: &[u64], name: &str) -> (f32, f32) {
crate::tests::create_and_validate::<
PiecewiseLinearFastFieldSerializer,
PiecewiseLinearFastFieldReader,
>(data, name)
}
#[test]
fn test_compression() {
let data = (10..=6_000_u64).collect::<Vec<_>>();
let (estimate, actual_compression) =
create_and_validate(&data, "simple monotonically large");
assert!(actual_compression < 0.2);
assert!(estimate < 0.20);
assert!(estimate > 0.15);
assert!(actual_compression > 0.001);
}
#[test]
fn test_with_codec_data_sets() {
let data_sets = get_codec_test_data_sets();
for (mut data, name) in data_sets {
create_and_validate(&data, name);
data.reverse();
create_and_validate(&data, name);
}
}
#[test]
fn test_simple() {
let data = (10..=20_u64).collect::<Vec<_>>();
create_and_validate(&data, "simple monotonically");
}
#[test]
fn border_cases_1() {
let data = (0..1024).collect::<Vec<_>>();
create_and_validate(&data, "border case");
}
#[test]
fn border_case_2() {
let data = (0..1025).collect::<Vec<_>>();
create_and_validate(&data, "border case");
}
#[test]
fn rand() {
for _ in 0..10 {
let mut data = (5_000..20_000)
.map(|_| rand::random::<u32>() as u64)
.collect::<Vec<_>>();
let (estimate, actual_compression) = create_and_validate(&data, "random");
dbg!(estimate);
dbg!(actual_compression);
data.reverse();
create_and_validate(&data, "random");
}
}
}

12
ownedbytes/src/lib.rs
View File

@@ -1,11 +1,9 @@
#![allow(clippy::return_self_not_must_use)]
use stable_deref_trait::StableDeref;
use std::convert::TryInto;
use std::mem;
use std::ops::{Deref, Range};
use std::sync::Arc;
use std::{fmt, io};
use std::{fmt, io, mem};
use stable_deref_trait::StableDeref;
/// An OwnedBytes simply wraps an object that owns a slice of data and exposes
/// this data as a static slice.
@@ -102,7 +100,6 @@ impl OwnedBytes {
}
/// Drops the left most `advance_len` bytes.
///
#[inline]
pub fn advance(&mut self, advance_len: usize) {
self.data = &self.data[advance_len..]
@@ -163,8 +160,7 @@ impl PartialEq<str> for OwnedBytes {
}
impl<'a, T: ?Sized> PartialEq<&'a T> for OwnedBytes
where
OwnedBytes: PartialEq<T>,
where OwnedBytes: PartialEq<T>
{
fn eq(&self, other: &&'a T) -> bool {
*self == **other

5
query-grammar/Cargo.toml
View File

@@ -5,9 +5,8 @@ authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT"
categories = ["database-implementations", "data-structures"]
description = """Search engine library"""
documentation = "https://quickwit-inc.github.io/tantivy/tantivy/index.html"
homepage = "https://github.com/quickwit-inc/tantivy"
repository = "https://github.com/quickwit-inc/tantivy"
homepage = "https://github.com/quickwit-oss/tantivy"
repository = "https://github.com/quickwit-oss/tantivy"
readme = "README.md"
keywords = ["search", "information", "retrieval"]
edition = "2018"

16
query-grammar/src/query_grammar.rs
View File

@@ -1,17 +1,20 @@
use super::user_input_ast::{UserInputAst, UserInputBound, UserInputLeaf, UserInputLiteral};
use crate::Occur;
use combine::error::StringStreamError;
use combine::parser::char::{char, digit, space, spaces, string};
use combine::parser::combinator::recognize;
use combine::parser::range::{take_while, take_while1};
use combine::parser::repeat::escaped;
use combine::parser::Parser;
use combine::{
attempt, choice, eof, many, many1, one_of, optional, parser, satisfy, skip_many1, value,
};
use combine::{error::StringStreamError, parser::combinator::recognize};
use once_cell::sync::Lazy;
use regex::Regex;
// Note: '-' char is only forbidden at the beginning of a field name, would be clearer to add it to special characters.
use super::user_input_ast::{UserInputAst, UserInputBound, UserInputLeaf, UserInputLiteral};
use crate::Occur;
// Note: '-' char is only forbidden at the beginning of a field name, would be clearer to add it to
// special characters.
const SPECIAL_CHARS: &[char] = &[
'+', '^', '`', ':', '{', '}', '"', '[', ']', '(', ')', '~', '!', '\\', '*', ' ',
];
@@ -64,7 +67,7 @@ fn word<'a>() -> impl Parser<&'a str, Output = String> {
///
/// NOTE: also accepts 999999-99-99T99:99:99.266051969+99:99
/// We delegate rejecting such invalid dates to the logical AST compuation code
/// which invokes chrono::DateTime::parse_from_rfc3339 on the value to actually parse
/// which invokes time::OffsetDateTime::parse(..., &Rfc3339) on the value to actually parse
/// it (instead of merely extracting the datetime value as string as done here).
fn date_time<'a>() -> impl Parser<&'a str, Output = String> {
let two_digits = || recognize::<String, _, _>((digit(), digit()));
@@ -363,9 +366,10 @@ mod test {
type TestParseResult = Result<(), StringStreamError>;
use super::*;
use combine::parser::Parser;
use super::*;
pub fn nearly_equals(a: f64, b: f64) -> bool {
(a - b).abs() < 0.0005 * (a + b).abs()
}

4
query-grammar/src/user_input_ast.rs
View File

@@ -59,7 +59,7 @@ pub enum UserInputBound {
}
impl UserInputBound {
fn display_lower(&self, formatter: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
fn display_lower(&self, formatter: &mut fmt::Formatter) -> Result<(), fmt::Error> {
match *self {
UserInputBound::Inclusive(ref word) => write!(formatter, "[\"{}\"", word),
UserInputBound::Exclusive(ref word) => write!(formatter, "{{\"{}\"", word),
@@ -67,7 +67,7 @@ impl UserInputBound {
}
}
fn display_upper(&self, formatter: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
fn display_upper(&self, formatter: &mut fmt::Formatter) -> Result<(), fmt::Error> {
match *self {
UserInputBound::Inclusive(ref word) => write!(formatter, "\"{}\"]", word),
UserInputBound::Exclusive(ref word) => write!(formatter, "\"{}\"}}", word),

8
rustfmt.toml
View File

@@ -1 +1,7 @@
use_try_shorthand = true
comment_width = 120
format_strings = true
group_imports = "StdExternalCrate"
imports_granularity = "Module"
normalize_comments = true
where_single_line = true
wrap_comments = true

36
src/aggregation/README.md Normal file
View File

@@ -0,0 +1,36 @@
# Contributing
When adding new bucket aggregation make sure to extend the "test_aggregation_flushing" test for at least 2 levels.
# Code Organization
Tantivy's aggregations have been designed to mimic the
[aggregations of elasticsearch](https://www.elastic.co/guide/en/elasticsearch/reference/current/search-aggregations.html).
The code is organized in submodules:
## bucket
Contains all bucket aggregations, like range aggregation. These bucket aggregations group documents into buckets and can contain sub-aggegations.
## metric
Contains all metric aggregations, like average aggregation. Metric aggregations do not have sub aggregations.
#### agg_req
agg_req contains the users aggregation request. Deserialization from json is compatible with elasticsearch aggregation requests.
#### agg_req_with_accessor
agg_req_with_accessor contains the users aggregation request enriched with fast field accessors etc, which are
used during collection.
#### segment_agg_result
segment_agg_result contains the aggregation result tree, which is used for collection of a segment.
The tree from agg_req_with_accessor is passed during collection.
#### intermediate_agg_result
intermediate_agg_result contains the aggregation tree for merging with other trees.
#### agg_result
agg_result contains the final aggregation tree.

327
src/aggregation/agg_req.rs Normal file
View File

@@ -0,0 +1,327 @@
//! Contains the aggregation request tree. Used to build an
//! [AggregationCollector](super::AggregationCollector).
//!
//! [Aggregations] is the top level entry point to create a request, which is a `HashMap<String,
//! Aggregation>`.
//!
//! Requests are compatible with the json format of elasticsearch.
//!
//! # Example
//!
//! ```
//! use tantivy::aggregation::bucket::RangeAggregation;
//! use tantivy::aggregation::agg_req::BucketAggregationType;
//! use tantivy::aggregation::agg_req::{Aggregation, Aggregations};
//! use tantivy::aggregation::agg_req::BucketAggregation;
//! let agg_req1: Aggregations = vec![
//! (
//! "range".to_string(),
//! Aggregation::Bucket(BucketAggregation {
//! bucket_agg: BucketAggregationType::Range(RangeAggregation{
//! field: "score".to_string(),
//! ranges: vec![(3f64..7f64).into(), (7f64..20f64).into()],
//! }),
//! sub_aggregation: Default::default(),
//! }),
//! ),
//! ]
//! .into_iter()
//! .collect();
//!
//! let elasticsearch_compatible_json_req = r#"
//! {
//! "range": {
//! "range": {
//! "field": "score",
//! "ranges": [
//! { "from": 3.0, "to": 7.0 },
//! { "from": 7.0, "to": 20.0 }
//! ]
//! }
//! }
//! }"#;
//! let agg_req2: Aggregations = serde_json::from_str(elasticsearch_compatible_json_req).unwrap();
//! assert_eq!(agg_req1, agg_req2);
//! ```
use std::collections::{HashMap, HashSet};
use serde::{Deserialize, Serialize};
use super::bucket::HistogramAggregation;
pub use super::bucket::RangeAggregation;
use super::metric::{AverageAggregation, StatsAggregation};
use super::VecWithNames;
/// The top-level aggregation request structure, which contains [Aggregation] and their user defined
/// names. It is also used in [buckets](BucketAggregation) to define sub-aggregations.
///
/// The key is the user defined name of the aggregation.
pub type Aggregations = HashMap<String, Aggregation>;
/// Like Aggregations, but optimized to work with the aggregation result
#[derive(Clone, Debug)]
pub(crate) struct AggregationsInternal {
pub(crate) metrics: VecWithNames<MetricAggregation>,
pub(crate) buckets: VecWithNames<BucketAggregationInternal>,
}
impl From<Aggregations> for AggregationsInternal {
fn from(aggs: Aggregations) -> Self {
let mut metrics = vec![];
let mut buckets = vec![];
for (key, agg) in aggs {
match agg {
Aggregation::Bucket(bucket) => buckets.push((
key,
BucketAggregationInternal {
bucket_agg: bucket.bucket_agg,
sub_aggregation: bucket.sub_aggregation.into(),
},
)),
Aggregation::Metric(metric) => metrics.push((key, metric)),
}
}
Self {
metrics: VecWithNames::from_entries(metrics),
buckets: VecWithNames::from_entries(buckets),
}
}
}
#[derive(Clone, Debug)]
// Like BucketAggregation, but optimized to work with the result
pub(crate) struct BucketAggregationInternal {
/// Bucket aggregation strategy to group documents.
pub bucket_agg: BucketAggregationType,
/// The sub_aggregations in the buckets. Each bucket will aggregate on the document set in the
/// bucket.
pub sub_aggregation: AggregationsInternal,
}
impl BucketAggregationInternal {
pub(crate) fn as_histogram(&self) -> &HistogramAggregation {
match &self.bucket_agg {
BucketAggregationType::Range(_) => panic!("unexpected aggregation"),
BucketAggregationType::Histogram(histogram) => histogram,
}
}
}
/// Extract all fast field names used in the tree.
pub fn get_fast_field_names(aggs: &Aggregations) -> HashSet<String> {
let mut fast_field_names = Default::default();
for el in aggs.values() {
el.get_fast_field_names(&mut fast_field_names)
}
fast_field_names
}
/// Aggregation request of [BucketAggregation] or [MetricAggregation].
///
/// An aggregation is either a bucket or a metric.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
#[serde(untagged)]
pub enum Aggregation {
/// Bucket aggregation, see [BucketAggregation] for details.
Bucket(BucketAggregation),
/// Metric aggregation, see [MetricAggregation] for details.
Metric(MetricAggregation),
}
impl Aggregation {
fn get_fast_field_names(&self, fast_field_names: &mut HashSet<String>) {
match self {
Aggregation::Bucket(bucket) => bucket.get_fast_field_names(fast_field_names),
Aggregation::Metric(metric) => metric.get_fast_field_names(fast_field_names),
}
}
}
/// BucketAggregations create buckets of documents. Each bucket is associated with a rule which
/// determines whether or not a document in the falls into it. In other words, the buckets
/// effectively define document sets. Buckets are not necessarily disjunct, therefore a document can
/// fall into multiple buckets. In addition to the buckets themselves, the bucket aggregations also
/// compute and return the number of documents for each bucket. Bucket aggregations, as opposed to
/// metric aggregations, can hold sub-aggregations. These sub-aggregations will be aggregated for
/// the buckets created by their "parent" bucket aggregation. There are different bucket
/// aggregators, each with a different "bucketing" strategy. Some define a single bucket, some
/// define fixed number of multiple buckets, and others dynamically create the buckets during the
/// aggregation process.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct BucketAggregation {
/// Bucket aggregation strategy to group documents.
#[serde(flatten)]
pub bucket_agg: BucketAggregationType,
/// The sub_aggregations in the buckets. Each bucket will aggregate on the document set in the
/// bucket.
#[serde(rename = "aggs")]
#[serde(default)]
#[serde(skip_serializing_if = "Aggregations::is_empty")]
pub sub_aggregation: Aggregations,
}
impl BucketAggregation {
fn get_fast_field_names(&self, fast_field_names: &mut HashSet<String>) {
self.bucket_agg.get_fast_field_names(fast_field_names);
fast_field_names.extend(get_fast_field_names(&self.sub_aggregation));
}
}
/// The bucket aggregation types.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub enum BucketAggregationType {
/// Put data into buckets of user-defined ranges.
#[serde(rename = "range")]
Range(RangeAggregation),
/// Put data into buckets of user-defined ranges.
#[serde(rename = "histogram")]
Histogram(HistogramAggregation),
}
impl BucketAggregationType {
fn get_fast_field_names(&self, fast_field_names: &mut HashSet<String>) {
match self {
BucketAggregationType::Range(range) => fast_field_names.insert(range.field.to_string()),
BucketAggregationType::Histogram(histogram) => {
fast_field_names.insert(histogram.field.to_string())
}
};
}
}
/// The aggregations in this family compute metrics based on values extracted
/// from the documents that are being aggregated. Values are extracted from the fast field of
/// the document.
/// Some aggregations output a single numeric metric (e.g. Average) and are called
/// single-value numeric metrics aggregation, others generate multiple metrics (e.g. Stats) and are
/// called multi-value numeric metrics aggregation.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub enum MetricAggregation {
/// Calculates the average.
#[serde(rename = "avg")]
Average(AverageAggregation),
/// Calculates stats sum, average, min, max, standard_deviation on a field.
#[serde(rename = "stats")]
Stats(StatsAggregation),
}
impl MetricAggregation {
fn get_fast_field_names(&self, fast_field_names: &mut HashSet<String>) {
match self {
MetricAggregation::Average(avg) => fast_field_names.insert(avg.field.to_string()),
MetricAggregation::Stats(stats) => fast_field_names.insert(stats.field.to_string()),
};
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn serialize_to_json_test() {
let agg_req1: Aggregations = vec![(
"range".to_string(),
Aggregation::Bucket(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "score".to_string(),
ranges: vec![
(f64::MIN..3f64).into(),
(3f64..7f64).into(),
(7f64..20f64).into(),
(20f64..f64::MAX).into(),
],
}),
sub_aggregation: Default::default(),
}),
)]
.into_iter()
.collect();
let elasticsearch_compatible_json_req = r#"{
"range": {
"range": {
"field": "score",
"ranges": [
{
"to": 3.0
},
{
"from": 3.0,
"to": 7.0
},
{
"from": 7.0,
"to": 20.0
},
{
"from": 20.0
}
]
}
}
}"#;
let agg_req2: String = serde_json::to_string_pretty(&agg_req1).unwrap();
assert_eq!(agg_req2, elasticsearch_compatible_json_req);
}
#[test]
fn test_get_fast_field_names() {
let agg_req2: Aggregations = vec![
(
"range".to_string(),
Aggregation::Bucket(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "score2".to_string(),
ranges: vec![
(f64::MIN..3f64).into(),
(3f64..7f64).into(),
(7f64..20f64).into(),
(20f64..f64::MAX).into(),
],
}),
sub_aggregation: Default::default(),
}),
),
(
"metric".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("field123".to_string()),
)),
),
]
.into_iter()
.collect();
let agg_req1: Aggregations = vec![(
"range".to_string(),
Aggregation::Bucket(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "score".to_string(),
ranges: vec![
(f64::MIN..3f64).into(),
(3f64..7f64).into(),
(7f64..20f64).into(),
(20f64..f64::MAX).into(),
],
}),
sub_aggregation: agg_req2,
}),
)]
.into_iter()
.collect();
assert_eq!(
get_fast_field_names(&agg_req1),
vec![
"score".to_string(),
"score2".to_string(),
"field123".to_string()
]
.into_iter()
.collect()
)
}
}

149
src/aggregation/agg_req_with_accessor.rs Normal file
View File

@@ -0,0 +1,149 @@
//! This will enhance the request tree with access to the fastfield and metadata.
use super::agg_req::{Aggregation, Aggregations, BucketAggregationType, MetricAggregation};
use super::bucket::{HistogramAggregation, RangeAggregation};
use super::metric::{AverageAggregation, StatsAggregation};
use super::VecWithNames;
use crate::fastfield::{type_and_cardinality, DynamicFastFieldReader, FastType};
use crate::schema::{Cardinality, Type};
use crate::{SegmentReader, TantivyError};
#[derive(Clone, Default)]
pub(crate) struct AggregationsWithAccessor {
pub metrics: VecWithNames<MetricAggregationWithAccessor>,
pub buckets: VecWithNames<BucketAggregationWithAccessor>,
}
impl AggregationsWithAccessor {
fn from_data(
metrics: VecWithNames<MetricAggregationWithAccessor>,
buckets: VecWithNames<BucketAggregationWithAccessor>,
) -> Self {
Self { metrics, buckets }
}
pub fn is_empty(&self) -> bool {
self.metrics.is_empty() && self.buckets.is_empty()
}
}
#[derive(Clone)]
pub struct BucketAggregationWithAccessor {
/// In general there can be buckets without fast field access, e.g. buckets that are created
/// based on search terms. So eventually this needs to be Option or moved.
pub(crate) accessor: DynamicFastFieldReader<u64>,
pub(crate) field_type: Type,
pub(crate) bucket_agg: BucketAggregationType,
pub(crate) sub_aggregation: AggregationsWithAccessor,
}
impl BucketAggregationWithAccessor {
fn try_from_bucket(
bucket: &BucketAggregationType,
sub_aggregation: &Aggregations,
reader: &SegmentReader,
) -> crate::Result<BucketAggregationWithAccessor> {
let (accessor, field_type) = match &bucket {
BucketAggregationType::Range(RangeAggregation {
field: field_name,
ranges: _,
}) => get_ff_reader_and_validate(reader, field_name)?,
BucketAggregationType::Histogram(HistogramAggregation {
field: field_name, ..
}) => get_ff_reader_and_validate(reader, field_name)?,
};
let sub_aggregation = sub_aggregation.clone();
Ok(BucketAggregationWithAccessor {
accessor,
field_type,
sub_aggregation: get_aggs_with_accessor_and_validate(&sub_aggregation, reader)?,
bucket_agg: bucket.clone(),
})
}
}
/// Contains the metric request and the fast field accessor.
#[derive(Clone)]
pub struct MetricAggregationWithAccessor {
pub metric: MetricAggregation,
pub field_type: Type,
pub accessor: DynamicFastFieldReader<u64>,
}
impl MetricAggregationWithAccessor {
fn try_from_metric(
metric: &MetricAggregation,
reader: &SegmentReader,
) -> crate::Result<MetricAggregationWithAccessor> {
match &metric {
MetricAggregation::Average(AverageAggregation { field: field_name })
| MetricAggregation::Stats(StatsAggregation { field: field_name }) => {
let (accessor, field_type) = get_ff_reader_and_validate(reader, field_name)?;
Ok(MetricAggregationWithAccessor {
accessor,
field_type,
metric: metric.clone(),
})
}
}
}
}
pub(crate) fn get_aggs_with_accessor_and_validate(
aggs: &Aggregations,
reader: &SegmentReader,
) -> crate::Result<AggregationsWithAccessor> {
let mut metrics = vec![];
let mut buckets = vec![];
for (key, agg) in aggs.iter() {
match agg {
Aggregation::Bucket(bucket) => buckets.push((
key.to_string(),
BucketAggregationWithAccessor::try_from_bucket(
&bucket.bucket_agg,
&bucket.sub_aggregation,
reader,
)?,
)),
Aggregation::Metric(metric) => metrics.push((
key.to_string(),
MetricAggregationWithAccessor::try_from_metric(metric, reader)?,
)),
}
}
Ok(AggregationsWithAccessor::from_data(
VecWithNames::from_entries(metrics),
VecWithNames::from_entries(buckets),
))
}
fn get_ff_reader_and_validate(
reader: &SegmentReader,
field_name: &str,
) -> crate::Result<(DynamicFastFieldReader<u64>, Type)> {
let field = reader
.schema()
.get_field(field_name)
.ok_or_else(|| TantivyError::FieldNotFound(field_name.to_string()))?;
let field_type = reader.schema().get_field_entry(field).field_type();
if let Some((ff_type, cardinality)) = type_and_cardinality(field_type) {
if cardinality == Cardinality::MultiValues || ff_type == FastType::Date {
return Err(TantivyError::InvalidArgument(format!(
"Invalid field type in aggregation {:?}, only Cardinality::SingleValue supported",
field_type.value_type()
)));
}
} else {
return Err(TantivyError::InvalidArgument(format!(
"Only single value fast fields of type f64, u64, i64 are supported, but got {:?} ",
field_type.value_type()
)));
};
let ff_fields = reader.fast_fields();
ff_fields
.u64_lenient(field)
.map(|field| (field, field_type.value_type()))
}

296
src/aggregation/agg_result.rs Normal file
View File

@@ -0,0 +1,296 @@
//! Contains the final aggregation tree.
//! This tree can be converted via the `into()` method from `IntermediateAggregationResults`.
//! This conversion computes the final result. For example: The intermediate result contains
//! intermediate average results, which is the sum and the number of values. The actual average is
//! calculated on the step from intermediate to final aggregation result tree.
use std::cmp::Ordering;
use std::collections::HashMap;
use itertools::Itertools;
use serde::{Deserialize, Serialize};
use super::agg_req::{Aggregations, AggregationsInternal, BucketAggregationInternal};
use super::bucket::intermediate_buckets_to_final_buckets;
use super::intermediate_agg_result::{
IntermediateAggregationResults, IntermediateBucketResult, IntermediateHistogramBucketEntry,
IntermediateMetricResult, IntermediateRangeBucketEntry,
};
use super::metric::{SingleMetricResult, Stats};
use super::Key;
#[derive(Clone, Default, Debug, PartialEq, Serialize, Deserialize)]
/// The final aggegation result.
pub struct AggregationResults(pub HashMap<String, AggregationResult>);
impl AggregationResults {
/// Convert and intermediate result and its aggregation request to the final result
pub fn from_intermediate_and_req(
results: IntermediateAggregationResults,
agg: Aggregations,
) -> Self {
AggregationResults::from_intermediate_and_req_internal(results, &(agg.into()))
}
/// Convert and intermediate result and its aggregation request to the final result
///
/// Internal function, CollectorAggregations is used instead Aggregations, which is optimized
/// for internal processing
fn from_intermediate_and_req_internal(
results: IntermediateAggregationResults,
req: &AggregationsInternal,
) -> Self {
let mut result = HashMap::default();
// Important assumption:
// When the tree contains buckets/metric, we expect it to have all buckets/metrics from the
// request
if let Some(buckets) = results.buckets {
result.extend(buckets.into_iter().zip(req.buckets.values()).map(
|((key, bucket), req)| {
(
key,
AggregationResult::BucketResult(BucketResult::from_intermediate_and_req(
bucket, req,
)),
)
},
));
} else {
result.extend(req.buckets.iter().map(|(key, req)| {
let empty_bucket = IntermediateBucketResult::empty_from_req(&req.bucket_agg);
(
key.to_string(),
AggregationResult::BucketResult(BucketResult::from_intermediate_and_req(
empty_bucket,
req,
)),
)
}));
}
if let Some(metrics) = results.metrics {
result.extend(
metrics
.into_iter()
.map(|(key, metric)| (key, AggregationResult::MetricResult(metric.into()))),
);
} else {
result.extend(req.metrics.iter().map(|(key, req)| {
let empty_bucket = IntermediateMetricResult::empty_from_req(req);
(
key.to_string(),
AggregationResult::MetricResult(empty_bucket.into()),
)
}));
}
Self(result)
}
}
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
#[serde(untagged)]
/// An aggregation is either a bucket or a metric.
pub enum AggregationResult {
/// Bucket result variant.
BucketResult(BucketResult),
/// Metric result variant.
MetricResult(MetricResult),
}
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
#[serde(untagged)]
/// MetricResult
pub enum MetricResult {
/// Average metric result.
Average(SingleMetricResult),
/// Stats metric result.
Stats(Stats),
}
impl From<IntermediateMetricResult> for MetricResult {
fn from(metric: IntermediateMetricResult) -> Self {
match metric {
IntermediateMetricResult::Average(avg_data) => {
MetricResult::Average(avg_data.finalize().into())
}
IntermediateMetricResult::Stats(intermediate_stats) => {
MetricResult::Stats(intermediate_stats.finalize())
}
}
}
}
/// BucketEntry holds bucket aggregation result types.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
#[serde(untagged)]
pub enum BucketResult {
/// This is the range entry for a bucket, which contains a key, count, from, to, and optionally
/// sub_aggregations.
Range {
/// The range buckets sorted by range.
buckets: Vec<RangeBucketEntry>,
},
/// This is the histogram entry for a bucket, which contains a key, count, and optionally
/// sub_aggregations.
Histogram {
/// The buckets.
///
/// If there are holes depends on the request, if min_doc_count is 0, then there are no
/// holes between the first and last bucket.
/// See [HistogramAggregation](super::bucket::HistogramAggregation)
buckets: Vec<BucketEntry>,
},
}
impl BucketResult {
fn from_intermediate_and_req(
bucket_result: IntermediateBucketResult,
req: &BucketAggregationInternal,
) -> Self {
match bucket_result {
IntermediateBucketResult::Range(range_map) => {
let mut buckets: Vec<RangeBucketEntry> = range_map
.into_iter()
.map(|(_, bucket)| {
RangeBucketEntry::from_intermediate_and_req(bucket, &req.sub_aggregation)
})
.collect_vec();
buckets.sort_by(|a, b| {
a.from
.unwrap_or(f64::MIN)
.partial_cmp(&b.from.unwrap_or(f64::MIN))
.unwrap_or(Ordering::Equal)
});
BucketResult::Range { buckets }
}
IntermediateBucketResult::Histogram { buckets } => {
let buckets = intermediate_buckets_to_final_buckets(
buckets,
req.as_histogram(),
&req.sub_aggregation,
);
BucketResult::Histogram { buckets }
}
}
}
}
/// This is the default entry for a bucket, which contains a key, count, and optionally
/// sub_aggregations.
///
/// # JSON Format
/// ```json
/// {
/// ...
/// "my_histogram": {
/// "buckets": [
/// {
/// "key": "2.0",
/// "doc_count": 5
/// },
/// {
/// "key": "4.0",
/// "doc_count": 2
/// },
/// {
/// "key": "6.0",
/// "doc_count": 3
/// }
/// ]
/// }
/// ...
/// }
/// ```
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct BucketEntry {
/// The identifier of the bucket.
pub key: Key,
/// Number of documents in the bucket.
pub doc_count: u64,
#[serde(flatten)]
/// sub-aggregations in this bucket.
pub sub_aggregation: AggregationResults,
}
impl BucketEntry {
pub(crate) fn from_intermediate_and_req(
entry: IntermediateHistogramBucketEntry,
req: &AggregationsInternal,
) -> Self {
BucketEntry {
key: Key::F64(entry.key),
doc_count: entry.doc_count,
sub_aggregation: AggregationResults::from_intermediate_and_req_internal(
entry.sub_aggregation,
req,
),
}
}
}
/// This is the range entry for a bucket, which contains a key, count, and optionally
/// sub_aggregations.
///
/// # JSON Format
/// ```json
/// {
/// ...
/// "my_ranges": {
/// "buckets": [
/// {
/// "key": "*-10",
/// "to": 10,
/// "doc_count": 5
/// },
/// {
/// "key": "10-20",
/// "from": 10,
/// "to": 20,
/// "doc_count": 2
/// },
/// {
/// "key": "20-*",
/// "from": 20,
/// "doc_count": 3
/// }
/// ]
/// }
/// ...
/// }
/// ```
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct RangeBucketEntry {
/// The identifier of the bucket.
pub key: Key,
/// Number of documents in the bucket.
pub doc_count: u64,
#[serde(flatten)]
/// sub-aggregations in this bucket.
pub sub_aggregation: AggregationResults,
/// The from range of the bucket. Equals f64::MIN when None.
#[serde(skip_serializing_if = "Option::is_none")]
pub from: Option<f64>,
/// The to range of the bucket. Equals f64::MAX when None.
#[serde(skip_serializing_if = "Option::is_none")]
pub to: Option<f64>,
}
impl RangeBucketEntry {
fn from_intermediate_and_req(
entry: IntermediateRangeBucketEntry,
req: &AggregationsInternal,
) -> Self {
RangeBucketEntry {
key: entry.key,
doc_count: entry.doc_count,
sub_aggregation: AggregationResults::from_intermediate_and_req_internal(
entry.sub_aggregation,
req,
),
to: entry.to,
from: entry.from,
}
}
}

1350
src/aggregation/bucket/histogram/histogram.rs Normal file
View File

File diff suppressed because it is too large Load Diff

2
src/aggregation/bucket/histogram/mod.rs Normal file
View File

@@ -0,0 +1,2 @@
mod histogram;
pub use histogram::*;

16
src/aggregation/bucket/mod.rs Normal file
View File

@@ -0,0 +1,16 @@
//! Module for all bucket aggregations.
//!
//! BucketAggregations create buckets of documents
//! [BucketAggregation](super::agg_req::BucketAggregation).
//!
//! Results of final buckets are [BucketResult](super::agg_result::BucketResult).
//! Results of intermediate buckets are
//! [IntermediateBucketResult](super::intermediate_agg_result::IntermediateBucketResult)
mod histogram;
mod range;
pub(crate) use histogram::SegmentHistogramCollector;
pub use histogram::*;
pub(crate) use range::SegmentRangeCollector;
pub use range::*;

568
src/aggregation/bucket/range.rs Normal file
View File

@@ -0,0 +1,568 @@
use std::ops::Range;
use serde::{Deserialize, Serialize};
use crate::aggregation::agg_req_with_accessor::{
AggregationsWithAccessor, BucketAggregationWithAccessor,
};
use crate::aggregation::intermediate_agg_result::IntermediateBucketResult;
use crate::aggregation::segment_agg_result::{
SegmentAggregationResultsCollector, SegmentRangeBucketEntry,
};
use crate::aggregation::{f64_from_fastfield_u64, f64_to_fastfield_u64, Key};
use crate::fastfield::FastFieldReader;
use crate::schema::Type;
use crate::{DocId, TantivyError};
/// Provide user-defined buckets to aggregate on.
/// Two special buckets will automatically be created to cover the whole range of values.
/// The provided buckets have to be continous.
/// During the aggregation, the values extracted from the fast_field `field` will be checked
/// against each bucket range. Note that this aggregation includes the from value and excludes the
/// to value for each range.
///
/// Result type is [BucketResult](crate::aggregation::agg_result::BucketResult) with
/// [RangeBucketEntry](crate::aggregation::agg_result::RangeBucketEntry) on the
/// AggregationCollector.
///
/// Result type is
/// [crate::aggregation::intermediate_agg_result::IntermediateBucketResult] with
/// [crate::aggregation::intermediate_agg_result::IntermediateRangeBucketEntry] on the
/// DistributedAggregationCollector.
///
/// # Limitations/Compatibility
/// Overlapping ranges are not yet supported.
///
/// The keyed parameter (elasticsearch) is not yet supported.
///
/// # Request JSON Format
/// ```json
/// {
/// "range": {
/// "field": "score",
/// "ranges": [
/// { "to": 3.0 },
/// { "from": 3.0, "to": 7.0 },
/// { "from": 7.0, "to": 20.0 }
/// { "from": 20.0 }
/// ]
/// }
/// }
/// ```
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct RangeAggregation {
/// The field to aggregate on.
pub field: String,
/// Note that this aggregation includes the from value and excludes the to value for each
/// range. Extra buckets will be created until the first to, and last from, if necessary.
pub ranges: Vec<RangeAggregationRange>,
}
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
/// The range for one range bucket.
pub struct RangeAggregationRange {
/// The from range value, which is inclusive in the range.
/// None equals to an open ended interval.
#[serde(skip_serializing_if = "Option::is_none", default)]
pub from: Option<f64>,
/// The to range value, which is not inclusive in the range.
/// None equals to an open ended interval.
#[serde(skip_serializing_if = "Option::is_none", default)]
pub to: Option<f64>,
}
impl From<Range<f64>> for RangeAggregationRange {
fn from(range: Range<f64>) -> Self {
let from = if range.start == f64::MIN {
None
} else {
Some(range.start)
};
let to = if range.end == f64::MAX {
None
} else {
Some(range.end)
};
RangeAggregationRange { from, to }
}
}
#[derive(Clone, Debug, PartialEq)]
pub(crate) struct SegmentRangeAndBucketEntry {
range: Range<u64>,
bucket: SegmentRangeBucketEntry,
}
/// The collector puts values from the fast field into the correct buckets and does a conversion to
/// the correct datatype.
#[derive(Clone, Debug, PartialEq)]
pub struct SegmentRangeCollector {
/// The buckets containing the aggregation data.
buckets: Vec<SegmentRangeAndBucketEntry>,
field_type: Type,
}
impl SegmentRangeCollector {
pub fn into_intermediate_bucket_result(self) -> IntermediateBucketResult {
let field_type = self.field_type;
let buckets = self
.buckets
.into_iter()
.map(move |range_bucket| {
(
range_to_string(&range_bucket.range, &field_type),
range_bucket.bucket.into(),
)
})
.collect();
IntermediateBucketResult::Range(buckets)
}
pub(crate) fn from_req_and_validate(
req: &RangeAggregation,
sub_aggregation: &AggregationsWithAccessor,
field_type: Type,
) -> crate::Result<Self> {
// The range input on the request is f64.
// We need to convert to u64 ranges, because we read the values as u64.
// The mapping from the conversion is monotonic so ordering is preserved.
let buckets = extend_validate_ranges(&req.ranges, &field_type)?
.iter()
.map(|range| {
let to = if range.end == u64::MAX {
None
} else {
Some(f64_from_fastfield_u64(range.end, &field_type))
};
let from = if range.start == u64::MIN {
None
} else {
Some(f64_from_fastfield_u64(range.start, &field_type))
};
let sub_aggregation = if sub_aggregation.is_empty() {
None
} else {
Some(SegmentAggregationResultsCollector::from_req_and_validate(
sub_aggregation,
)?)
};
Ok(SegmentRangeAndBucketEntry {
range: range.clone(),
bucket: SegmentRangeBucketEntry {
key: range_to_key(range, &field_type),
doc_count: 0,
sub_aggregation,
from,
to,
},
})
})
.collect::<crate::Result<_>>()?;
Ok(SegmentRangeCollector {
buckets,
field_type,
})
}
#[inline]
pub(crate) fn collect_block(
&mut self,
doc: &[DocId],
bucket_with_accessor: &BucketAggregationWithAccessor,
force_flush: bool,
) {
let mut iter = doc.chunks_exact(4);
for docs in iter.by_ref() {
let val1 = bucket_with_accessor.accessor.get(docs[0]);
let val2 = bucket_with_accessor.accessor.get(docs[1]);
let val3 = bucket_with_accessor.accessor.get(docs[2]);
let val4 = bucket_with_accessor.accessor.get(docs[3]);
let bucket_pos1 = self.get_bucket_pos(val1);
let bucket_pos2 = self.get_bucket_pos(val2);
let bucket_pos3 = self.get_bucket_pos(val3);
let bucket_pos4 = self.get_bucket_pos(val4);
self.increment_bucket(bucket_pos1, docs[0], &bucket_with_accessor.sub_aggregation);
self.increment_bucket(bucket_pos2, docs[1], &bucket_with_accessor.sub_aggregation);
self.increment_bucket(bucket_pos3, docs[2], &bucket_with_accessor.sub_aggregation);
self.increment_bucket(bucket_pos4, docs[3], &bucket_with_accessor.sub_aggregation);
}
for doc in iter.remainder() {
let val = bucket_with_accessor.accessor.get(*doc);
let bucket_pos = self.get_bucket_pos(val);
self.increment_bucket(bucket_pos, *doc, &bucket_with_accessor.sub_aggregation);
}
if force_flush {
for bucket in &mut self.buckets {
if let Some(sub_aggregation) = &mut bucket.bucket.sub_aggregation {
sub_aggregation
.flush_staged_docs(&bucket_with_accessor.sub_aggregation, force_flush);
}
}
}
}
#[inline]
fn increment_bucket(
&mut self,
bucket_pos: usize,
doc: DocId,
bucket_with_accessor: &AggregationsWithAccessor,
) {
let bucket = &mut self.buckets[bucket_pos];
bucket.bucket.doc_count += 1;
if let Some(sub_aggregation) = &mut bucket.bucket.sub_aggregation {
sub_aggregation.collect(doc, bucket_with_accessor);
}
}
#[inline]
fn get_bucket_pos(&self, val: u64) -> usize {
let pos = self
.buckets
.binary_search_by_key(&val, |probe| probe.range.start)
.unwrap_or_else(|pos| pos - 1);
debug_assert!(self.buckets[pos].range.contains(&val));
pos
}
}
/// Converts the user provided f64 range value to fast field value space.
///
/// Internally fast field values are always stored as u64.
/// If the fast field has u64 [1,2,5], these values are stored as is in the fast field.
/// A fast field with f64 [1.0, 2.0, 5.0] is converted to u64 space, using a
/// monotonic mapping function, so the order is preserved.
///
/// Consequently, a f64 user range 1.0..3.0 needs to be converted to fast field value space using
/// the same monotonic mapping function, so that the provided ranges contain the u64 values in the
/// fast field.
/// The alternative would be that every value read would be converted to the f64 range, but that is
/// more computational expensive when many documents are hit.
fn to_u64_range(range: &RangeAggregationRange, field_type: &Type) -> crate::Result<Range<u64>> {
let start = if let Some(from) = range.from {
f64_to_fastfield_u64(from, field_type)
.ok_or_else(|| TantivyError::InvalidArgument("invalid field type".to_string()))?
} else {
u64::MIN
};
let end = if let Some(to) = range.to {
f64_to_fastfield_u64(to, field_type)
.ok_or_else(|| TantivyError::InvalidArgument("invalid field type".to_string()))?
} else {
u64::MAX
};
Ok(start..end)
}
/// Extends the provided buckets to contain the whole value range, by inserting buckets at the
/// beginning and end.
fn extend_validate_ranges(
buckets: &[RangeAggregationRange],
field_type: &Type,
) -> crate::Result<Vec<Range<u64>>> {
let mut converted_buckets = buckets
.iter()
.map(|range| to_u64_range(range, field_type))
.collect::<crate::Result<Vec<_>>>()?;
converted_buckets.sort_by_key(|bucket| bucket.start);
if converted_buckets[0].start != u64::MIN {
converted_buckets.insert(0, u64::MIN..converted_buckets[0].start);
}
if converted_buckets[converted_buckets.len() - 1].end != u64::MAX {
converted_buckets.push(converted_buckets[converted_buckets.len() - 1].end..u64::MAX);
}
// fill up holes in the ranges
let find_hole = |converted_buckets: &[Range<u64>]| {
for (pos, ranges) in converted_buckets.windows(2).enumerate() {
if ranges[0].end > ranges[1].start {
return Err(TantivyError::InvalidArgument(format!(
"Overlapping ranges not supported range {:?}, range+1 {:?}",
ranges[0], ranges[1]
)));
}
if ranges[0].end != ranges[1].start {
return Ok(Some(pos));
}
}
Ok(None)
};
while let Some(hole_pos) = find_hole(&converted_buckets)? {
let new_range = converted_buckets[hole_pos].end..converted_buckets[hole_pos + 1].start;
converted_buckets.insert(hole_pos + 1, new_range);
}
Ok(converted_buckets)
}
pub(crate) fn range_to_string(range: &Range<u64>, field_type: &Type) -> String {
// is_start is there for malformed requests, e.g. ig the user passes the range u64::MIN..0.0,
// it should be rendererd as "*-0" and not "*-*"
let to_str = |val: u64, is_start: bool| {
if (is_start && val == u64::MIN) || (!is_start && val == u64::MAX) {
"*".to_string()
} else {
f64_from_fastfield_u64(val, field_type).to_string()
}
};
format!("{}-{}", to_str(range.start, true), to_str(range.end, false))
}
pub(crate) fn range_to_key(range: &Range<u64>, field_type: &Type) -> Key {
Key::Str(range_to_string(range, field_type))
}
#[cfg(test)]
mod tests {
use serde_json::Value;
use super::*;
use crate::aggregation::agg_req::{
Aggregation, Aggregations, BucketAggregation, BucketAggregationType,
};
use crate::aggregation::tests::get_test_index_with_num_docs;
use crate::aggregation::AggregationCollector;
use crate::fastfield::FastValue;
use crate::query::AllQuery;
pub fn get_collector_from_ranges(
ranges: Vec<RangeAggregationRange>,
field_type: Type,
) -> SegmentRangeCollector {
let req = RangeAggregation {
field: "dummy".to_string(),
ranges,
};
SegmentRangeCollector::from_req_and_validate(&req, &Default::default(), field_type).unwrap()
}
#[test]
fn range_fraction_test() -> crate::Result<()> {
let index = get_test_index_with_num_docs(false, 100)?;
let agg_req: Aggregations = vec![(
"range".to_string(),
Aggregation::Bucket(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "fraction_f64".to_string(),
ranges: vec![(0f64..0.1f64).into(), (0.1f64..0.2f64).into()],
}),
sub_aggregation: Default::default(),
}),
)]
.into_iter()
.collect();
let collector = AggregationCollector::from_aggs(agg_req);
let reader = index.reader()?;
let searcher = reader.searcher();
let agg_res = searcher.search(&AllQuery, &collector).unwrap();
let res: Value = serde_json::from_str(&serde_json::to_string(&agg_res)?)?;
assert_eq!(res["range"]["buckets"][0]["key"], "*-0");
assert_eq!(res["range"]["buckets"][0]["doc_count"], 0);
assert_eq!(res["range"]["buckets"][1]["key"], "0-0.1");
assert_eq!(res["range"]["buckets"][1]["doc_count"], 10);
assert_eq!(res["range"]["buckets"][2]["key"], "0.1-0.2");
assert_eq!(res["range"]["buckets"][2]["doc_count"], 10);
assert_eq!(res["range"]["buckets"][3]["key"], "0.2-*");
assert_eq!(res["range"]["buckets"][3]["doc_count"], 80);
Ok(())
}
#[test]
fn bucket_test_extend_range_hole() {
let buckets = vec![(10f64..20f64).into(), (30f64..40f64).into()];
let collector = get_collector_from_ranges(buckets, Type::F64);
let buckets = collector.buckets;
assert_eq!(buckets[0].range.start, u64::MIN);
assert_eq!(buckets[0].range.end, 10f64.to_u64());
assert_eq!(buckets[1].range.start, 10f64.to_u64());
assert_eq!(buckets[1].range.end, 20f64.to_u64());
// Added bucket to fill hole
assert_eq!(buckets[2].range.start, 20f64.to_u64());
assert_eq!(buckets[2].range.end, 30f64.to_u64());
assert_eq!(buckets[3].range.start, 30f64.to_u64());
assert_eq!(buckets[3].range.end, 40f64.to_u64());
}
#[test]
fn bucket_test_range_conversion_special_case() {
// the monotonic conversion between f64 and u64, does not map f64::MIN.to_u64() ==
// u64::MIN, but the into trait converts f64::MIN/MAX to None
let buckets = vec![
(f64::MIN..10f64).into(),
(10f64..20f64).into(),
(20f64..f64::MAX).into(),
];
let collector = get_collector_from_ranges(buckets, Type::F64);
let buckets = collector.buckets;
assert_eq!(buckets[0].range.start, u64::MIN);
assert_eq!(buckets[0].range.end, 10f64.to_u64());
assert_eq!(buckets[1].range.start, 10f64.to_u64());
assert_eq!(buckets[1].range.end, 20f64.to_u64());
assert_eq!(buckets[2].range.start, 20f64.to_u64());
assert_eq!(buckets[2].range.end, u64::MAX);
assert_eq!(buckets.len(), 3);
}
#[test]
fn bucket_range_test_negative_vals() {
let buckets = vec![(-10f64..-1f64).into()];
let collector = get_collector_from_ranges(buckets, Type::F64);
let buckets = collector.buckets;
assert_eq!(&buckets[0].bucket.key.to_string(), "*--10");
assert_eq!(&buckets[buckets.len() - 1].bucket.key.to_string(), "-1-*");
}
#[test]
fn bucket_range_test_positive_vals() {
let buckets = vec![(0f64..10f64).into()];
let collector = get_collector_from_ranges(buckets, Type::F64);
let buckets = collector.buckets;
assert_eq!(&buckets[0].bucket.key.to_string(), "*-0");
assert_eq!(&buckets[buckets.len() - 1].bucket.key.to_string(), "10-*");
}
#[test]
fn range_binary_search_test_u64() {
let check_ranges = |ranges: Vec<RangeAggregationRange>| {
let collector = get_collector_from_ranges(ranges, Type::U64);
let search = |val: u64| collector.get_bucket_pos(val);
assert_eq!(search(u64::MIN), 0);
assert_eq!(search(9), 0);
assert_eq!(search(10), 1);
assert_eq!(search(11), 1);
assert_eq!(search(99), 1);
assert_eq!(search(100), 2);
assert_eq!(search(u64::MAX - 1), 2); // Since the end range is never included, the max
// value
};
let ranges = vec![(10.0..100.0).into()];
check_ranges(ranges);
let ranges = vec![
RangeAggregationRange {
to: Some(10.0),
from: None,
},
(10.0..100.0).into(),
];
check_ranges(ranges);
let ranges = vec![
RangeAggregationRange {
to: Some(10.0),
from: None,
},
(10.0..100.0).into(),
RangeAggregationRange {
to: None,
from: Some(100.0),
},
];
check_ranges(ranges);
}
#[test]
fn range_binary_search_test_f64() {
let ranges = vec![
//(f64::MIN..10.0).into(),
(10.0..100.0).into(),
//(100.0..f64::MAX).into(),
];
let collector = get_collector_from_ranges(ranges, Type::F64);
let search = |val: u64| collector.get_bucket_pos(val);
assert_eq!(search(u64::MIN), 0);
assert_eq!(search(9f64.to_u64()), 0);
assert_eq!(search(10f64.to_u64()), 1);
assert_eq!(search(11f64.to_u64()), 1);
assert_eq!(search(99f64.to_u64()), 1);
assert_eq!(search(100f64.to_u64()), 2);
assert_eq!(search(u64::MAX - 1), 2); // Since the end range is never included,
// the max value
}
}
#[cfg(all(test, feature = "unstable"))]
mod bench {
use itertools::Itertools;
use rand::seq::SliceRandom;
use rand::thread_rng;
use super::*;
use crate::aggregation::bucket::range::tests::get_collector_from_ranges;
const TOTAL_DOCS: u64 = 1_000_000u64;
const NUM_DOCS: u64 = 50_000u64;
fn get_collector_with_buckets(num_buckets: u64, num_docs: u64) -> SegmentRangeCollector {
let bucket_size = num_docs / num_buckets;
let mut buckets: Vec<RangeAggregationRange> = vec![];
for i in 0..num_buckets {
let bucket_start = (i * bucket_size) as f64;
buckets.push((bucket_start..bucket_start + bucket_size as f64).into())
}
get_collector_from_ranges(buckets, Type::U64)
}
fn get_rand_docs(total_docs: u64, num_docs_returned: u64) -> Vec<u64> {
let mut rng = thread_rng();
let all_docs = (0..total_docs - 1).collect_vec();
let mut vals = all_docs
.as_slice()
.choose_multiple(&mut rng, num_docs_returned as usize)
.cloned()
.collect_vec();
vals.sort();
vals
}
fn bench_range_binary_search(b: &mut test::Bencher, num_buckets: u64) {
let collector = get_collector_with_buckets(num_buckets, TOTAL_DOCS);
let vals = get_rand_docs(TOTAL_DOCS, NUM_DOCS);
b.iter(|| {
let mut bucket_pos = 0;
for val in &vals {
bucket_pos = collector.get_bucket_pos(*val);
}
bucket_pos
})
}
#[bench]
fn bench_range_100_buckets(b: &mut test::Bencher) {
bench_range_binary_search(b, 100)
}
#[bench]
fn bench_range_10_buckets(b: &mut test::Bencher) {
bench_range_binary_search(b, 10)
}
}

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

@@ -0,0 +1,142 @@
use super::agg_req::Aggregations;
use super::agg_req_with_accessor::AggregationsWithAccessor;
use super::agg_result::AggregationResults;
use super::intermediate_agg_result::IntermediateAggregationResults;
use super::segment_agg_result::SegmentAggregationResultsCollector;
use crate::aggregation::agg_req_with_accessor::get_aggs_with_accessor_and_validate;
use crate::collector::{Collector, SegmentCollector};
use crate::SegmentReader;
/// Collector for aggregations.
///
/// The collector collects all aggregations by the underlying aggregation request.
pub struct AggregationCollector {
agg: Aggregations,
}
impl AggregationCollector {
/// Create collector from aggregation request.
pub fn from_aggs(agg: Aggregations) -> Self {
Self { agg }
}
}
/// Collector for distributed aggregations.
///
/// The collector collects all aggregations by the underlying aggregation request.
///
/// # Purpose
/// AggregationCollector returns `IntermediateAggregationResults` and not the final
/// `AggregationResults`, so that results from differenct indices can be merged and then converted
/// into the final `AggregationResults` via the `into()` method.
pub struct DistributedAggregationCollector {
agg: Aggregations,
}
impl DistributedAggregationCollector {
/// Create collector from aggregation request.
pub fn from_aggs(agg: Aggregations) -> Self {
Self { agg }
}
}
impl Collector for DistributedAggregationCollector {
type Fruit = IntermediateAggregationResults;
type Child = AggregationSegmentCollector;
fn for_segment(
&self,
_segment_local_id: crate::SegmentOrdinal,
reader: &crate::SegmentReader,
) -> crate::Result<Self::Child> {
AggregationSegmentCollector::from_agg_req_and_reader(&self.agg, reader)
}
fn requires_scoring(&self) -> bool {
false
}
fn merge_fruits(
&self,
segment_fruits: Vec<<Self::Child as SegmentCollector>::Fruit>,
) -> crate::Result<Self::Fruit> {
merge_fruits(segment_fruits)
}
}
impl Collector for AggregationCollector {
type Fruit = AggregationResults;
type Child = AggregationSegmentCollector;
fn for_segment(
&self,
_segment_local_id: crate::SegmentOrdinal,
reader: &crate::SegmentReader,
) -> crate::Result<Self::Child> {
AggregationSegmentCollector::from_agg_req_and_reader(&self.agg, reader)
}
fn requires_scoring(&self) -> bool {
false
}
fn merge_fruits(
&self,
segment_fruits: Vec<<Self::Child as SegmentCollector>::Fruit>,
) -> crate::Result<Self::Fruit> {
merge_fruits(segment_fruits)
.map(|res| AggregationResults::from_intermediate_and_req(res, self.agg.clone()))
}
}
fn merge_fruits(
mut segment_fruits: Vec<IntermediateAggregationResults>,
) -> crate::Result<IntermediateAggregationResults> {
if let Some(mut fruit) = segment_fruits.pop() {
for next_fruit in segment_fruits {
fruit.merge_fruits(next_fruit);
}
Ok(fruit)
} else {
Ok(IntermediateAggregationResults::default())
}
}
/// AggregationSegmentCollector does the aggregation collection on a segment.
pub struct AggregationSegmentCollector {
aggs: AggregationsWithAccessor,
result: SegmentAggregationResultsCollector,
}
impl AggregationSegmentCollector {
/// Creates an AggregationSegmentCollector from an [Aggregations] request and a segment reader.
/// Also includes validation, e.g. checking field types and existence.
pub fn from_agg_req_and_reader(
agg: &Aggregations,
reader: &SegmentReader,
) -> crate::Result<Self> {
let aggs_with_accessor = get_aggs_with_accessor_and_validate(agg, reader)?;
let result =
SegmentAggregationResultsCollector::from_req_and_validate(&aggs_with_accessor)?;
Ok(AggregationSegmentCollector {
aggs: aggs_with_accessor,
result,
})
}
}
impl SegmentCollector for AggregationSegmentCollector {
type Fruit = IntermediateAggregationResults;
#[inline]
fn collect(&mut self, doc: crate::DocId, _score: crate::Score) {
self.result.collect(doc, &self.aggs);
}
fn harvest(mut self) -> Self::Fruit {
self.result.flush_staged_docs(&self.aggs, true);
self.result.into()
}
}

473
src/aggregation/intermediate_agg_result.rs Normal file
View File

@@ -0,0 +1,473 @@
//! Contains the intermediate aggregation tree, that can be merged.
//! Intermediate aggregation results can be used to merge results between segments or between
//! indices.
use std::cmp::Ordering;
use fnv::FnvHashMap;
use itertools::Itertools;
use serde::{Deserialize, Serialize};
use super::agg_req::{AggregationsInternal, BucketAggregationType, MetricAggregation};
use super::metric::{IntermediateAverage, IntermediateStats};
use super::segment_agg_result::{
SegmentAggregationResultsCollector, SegmentBucketResultCollector, SegmentHistogramBucketEntry,
SegmentMetricResultCollector, SegmentRangeBucketEntry,
};
use super::{Key, SerializedKey, VecWithNames};
/// Contains the intermediate aggregation result, which is optimized to be merged with other
/// intermediate results.
#[derive(Default, Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct IntermediateAggregationResults {
pub(crate) metrics: Option<VecWithNames<IntermediateMetricResult>>,
pub(crate) buckets: Option<VecWithNames<IntermediateBucketResult>>,
}
impl From<SegmentAggregationResultsCollector> for IntermediateAggregationResults {
fn from(tree: SegmentAggregationResultsCollector) -> Self {
let metrics = tree.metrics.map(VecWithNames::from_other);
let buckets = tree.buckets.map(VecWithNames::from_other);
Self { metrics, buckets }
}
}
impl IntermediateAggregationResults {
pub(crate) fn empty_from_req(req: &AggregationsInternal) -> Self {
let metrics = if req.metrics.is_empty() {
None
} else {
let metrics = req
.metrics
.iter()
.map(|(key, req)| {
(
key.to_string(),
IntermediateMetricResult::empty_from_req(req),
)
})
.collect();
Some(VecWithNames::from_entries(metrics))
};
let buckets = if req.buckets.is_empty() {
None
} else {
let buckets = req
.buckets
.iter()
.map(|(key, req)| {
(
key.to_string(),
IntermediateBucketResult::empty_from_req(&req.bucket_agg),
)
})
.collect();
Some(VecWithNames::from_entries(buckets))
};
Self { metrics, buckets }
}
/// Merge an other intermediate aggregation result into this result.
///
/// The order of the values need to be the same on both results. This is ensured when the same
/// (key values) are present on the underlying VecWithNames struct.
pub fn merge_fruits(&mut self, other: IntermediateAggregationResults) {
if let (Some(buckets_left), Some(buckets_right)) = (&mut self.buckets, other.buckets) {
for (bucket_left, bucket_right) in
buckets_left.values_mut().zip(buckets_right.into_values())
{
bucket_left.merge_fruits(bucket_right);
}
}
if let (Some(metrics_left), Some(metrics_right)) = (&mut self.metrics, other.metrics) {
for (metric_left, metric_right) in
metrics_left.values_mut().zip(metrics_right.into_values())
{
metric_left.merge_fruits(metric_right);
}
}
}
}
/// An aggregation is either a bucket or a metric.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub enum IntermediateAggregationResult {
/// Bucket variant
Bucket(IntermediateBucketResult),
/// Metric variant
Metric(IntermediateMetricResult),
}
/// Holds the intermediate data for metric results
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub enum IntermediateMetricResult {
/// Average containing intermediate average data result
Average(IntermediateAverage),
/// AverageData variant
Stats(IntermediateStats),
}
impl From<SegmentMetricResultCollector> for IntermediateMetricResult {
fn from(tree: SegmentMetricResultCollector) -> Self {
match tree {
SegmentMetricResultCollector::Average(collector) => {
IntermediateMetricResult::Average(IntermediateAverage::from_collector(collector))
}
SegmentMetricResultCollector::Stats(collector) => {
IntermediateMetricResult::Stats(collector.stats)
}
}
}
}
impl IntermediateMetricResult {
pub(crate) fn empty_from_req(req: &MetricAggregation) -> Self {
match req {
MetricAggregation::Average(_) => {
IntermediateMetricResult::Average(IntermediateAverage::default())
}
MetricAggregation::Stats(_) => {
IntermediateMetricResult::Stats(IntermediateStats::default())
}
}
}
fn merge_fruits(&mut self, other: IntermediateMetricResult) {
match (self, other) {
(
IntermediateMetricResult::Average(avg_data_left),
IntermediateMetricResult::Average(avg_data_right),
) => {
avg_data_left.merge_fruits(avg_data_right);
}
(
IntermediateMetricResult::Stats(stats_left),
IntermediateMetricResult::Stats(stats_right),
) => {
stats_left.merge_fruits(stats_right);
}
_ => {
panic!("incompatible fruit types in tree");
}
}
}
}
/// The intermediate bucket results. Internally they can be easily merged via the keys of the
/// buckets.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub enum IntermediateBucketResult {
/// This is the range entry for a bucket, which contains a key, count, from, to, and optionally
/// sub_aggregations.
Range(FnvHashMap<SerializedKey, IntermediateRangeBucketEntry>),
/// This is the histogram entry for a bucket, which contains a key, count, and optionally
/// sub_aggregations.
Histogram {
/// The buckets
buckets: Vec<IntermediateHistogramBucketEntry>,
},
}
impl From<SegmentBucketResultCollector> for IntermediateBucketResult {
fn from(collector: SegmentBucketResultCollector) -> Self {
match collector {
SegmentBucketResultCollector::Range(range) => range.into_intermediate_bucket_result(),
SegmentBucketResultCollector::Histogram(histogram) => {
histogram.into_intermediate_bucket_result()
}
}
}
}
impl IntermediateBucketResult {
pub(crate) fn empty_from_req(req: &BucketAggregationType) -> Self {
match req {
BucketAggregationType::Range(_) => IntermediateBucketResult::Range(Default::default()),
BucketAggregationType::Histogram(_) => {
IntermediateBucketResult::Histogram { buckets: vec![] }
}
}
}
fn merge_fruits(&mut self, other: IntermediateBucketResult) {
match (self, other) {
(
IntermediateBucketResult::Range(entries_left),
IntermediateBucketResult::Range(entries_right),
) => {
merge_maps(entries_left, entries_right);
}
(
IntermediateBucketResult::Histogram {
buckets: entries_left,
..
},
IntermediateBucketResult::Histogram {
buckets: entries_right,
..
},
) => {
let mut buckets = entries_left
.drain(..)
.merge_join_by(entries_right.into_iter(), |left, right| {
left.key.partial_cmp(&right.key).unwrap_or(Ordering::Equal)
})
.map(|either| match either {
itertools::EitherOrBoth::Both(mut left, right) => {
left.merge_fruits(right);
left
}
itertools::EitherOrBoth::Left(left) => left,
itertools::EitherOrBoth::Right(right) => right,
})
.collect();
std::mem::swap(entries_left, &mut buckets);
}
(IntermediateBucketResult::Range(_), _) => {
panic!("try merge on different types")
}
(IntermediateBucketResult::Histogram { .. }, _) => {
panic!("try merge on different types")
}
}
}
}
trait MergeFruits {
fn merge_fruits(&mut self, other: Self);
}
fn merge_maps<V: MergeFruits + Clone>(
entries_left: &mut FnvHashMap<SerializedKey, V>,
mut entries_right: FnvHashMap<SerializedKey, V>,
) {
for (name, entry_left) in entries_left.iter_mut() {
if let Some(entry_right) = entries_right.remove(name) {
entry_left.merge_fruits(entry_right);
}
}
for (key, res) in entries_right.into_iter() {
entries_left.entry(key).or_insert(res);
}
}
/// This is the histogram entry for a bucket, which contains a key, count, and optionally
/// sub_aggregations.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct IntermediateHistogramBucketEntry {
/// The unique the bucket is identified.
pub key: f64,
/// The number of documents in the bucket.
pub doc_count: u64,
/// The sub_aggregation in this bucket.
pub sub_aggregation: IntermediateAggregationResults,
}
impl From<SegmentHistogramBucketEntry> for IntermediateHistogramBucketEntry {
fn from(entry: SegmentHistogramBucketEntry) -> Self {
IntermediateHistogramBucketEntry {
key: entry.key,
doc_count: entry.doc_count,
sub_aggregation: Default::default(),
}
}
}
impl
From<(
SegmentHistogramBucketEntry,
SegmentAggregationResultsCollector,
)> for IntermediateHistogramBucketEntry
{
fn from(
entry: (
SegmentHistogramBucketEntry,
SegmentAggregationResultsCollector,
),
) -> Self {
IntermediateHistogramBucketEntry {
key: entry.0.key,
doc_count: entry.0.doc_count,
sub_aggregation: entry.1.into(),
}
}
}
/// This is the range entry for a bucket, which contains a key, count, and optionally
/// sub_aggregations.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct IntermediateRangeBucketEntry {
/// The unique the bucket is identified.
pub key: Key,
/// The number of documents in the bucket.
pub doc_count: u64,
pub(crate) values: Option<Vec<u64>>,
/// The sub_aggregation in this bucket.
pub sub_aggregation: IntermediateAggregationResults,
/// The from range of the bucket. Equals f64::MIN when None.
#[serde(skip_serializing_if = "Option::is_none")]
pub from: Option<f64>,
/// The to range of the bucket. Equals f64::MAX when None.
#[serde(skip_serializing_if = "Option::is_none")]
pub to: Option<f64>,
}
impl From<SegmentRangeBucketEntry> for IntermediateRangeBucketEntry {
fn from(entry: SegmentRangeBucketEntry) -> Self {
let sub_aggregation = if let Some(sub_aggregation) = entry.sub_aggregation {
sub_aggregation.into()
} else {
Default::default()
};
IntermediateRangeBucketEntry {
key: entry.key,
doc_count: entry.doc_count,
values: None,
sub_aggregation,
to: entry.to,
from: entry.from,
}
}
}
impl MergeFruits for IntermediateRangeBucketEntry {
fn merge_fruits(&mut self, other: IntermediateRangeBucketEntry) {
self.doc_count += other.doc_count;
self.sub_aggregation.merge_fruits(other.sub_aggregation);
}
}
impl MergeFruits for IntermediateHistogramBucketEntry {
fn merge_fruits(&mut self, other: IntermediateHistogramBucketEntry) {
self.doc_count += other.doc_count;
self.sub_aggregation.merge_fruits(other.sub_aggregation);
}
}
#[cfg(test)]
mod tests {
use std::collections::HashMap;
use pretty_assertions::assert_eq;
use super::*;
fn get_sub_test_tree(data: &[(String, u64)]) -> IntermediateAggregationResults {
let mut map = HashMap::new();
let mut buckets = FnvHashMap::default();
for (key, doc_count) in data {
buckets.insert(
key.to_string(),
IntermediateRangeBucketEntry {
key: Key::Str(key.to_string()),
doc_count: *doc_count,
values: None,
sub_aggregation: Default::default(),
from: None,
to: None,
},
);
}
map.insert(
"my_agg_level2".to_string(),
IntermediateBucketResult::Range(buckets),
);
IntermediateAggregationResults {
buckets: Some(VecWithNames::from_entries(map.into_iter().collect())),
metrics: Default::default(),
}
}
fn get_intermediat_tree_with_ranges(
data: &[(String, u64, String, u64)],
) -> IntermediateAggregationResults {
let mut map = HashMap::new();
let mut buckets: FnvHashMap<_, _> = Default::default();
for (key, doc_count, sub_aggregation_key, sub_aggregation_count) in data {
buckets.insert(
key.to_string(),
IntermediateRangeBucketEntry {
key: Key::Str(key.to_string()),
doc_count: *doc_count,
values: None,
from: None,
to: None,
sub_aggregation: get_sub_test_tree(&[(
sub_aggregation_key.to_string(),
*sub_aggregation_count,
)]),
},
);
}
map.insert(
"my_agg_level1".to_string(),
IntermediateBucketResult::Range(buckets),
);
IntermediateAggregationResults {
buckets: Some(VecWithNames::from_entries(map.into_iter().collect())),
metrics: Default::default(),
}
}
#[test]
fn test_merge_fruits_tree_1() {
let mut tree_left = get_intermediat_tree_with_ranges(&[
("red".to_string(), 50, "1900".to_string(), 25),
("blue".to_string(), 30, "1900".to_string(), 30),
]);
let tree_right = get_intermediat_tree_with_ranges(&[
("red".to_string(), 60, "1900".to_string(), 30),
("blue".to_string(), 25, "1900".to_string(), 50),
]);
tree_left.merge_fruits(tree_right);
let tree_expected = get_intermediat_tree_with_ranges(&[
("red".to_string(), 110, "1900".to_string(), 55),
("blue".to_string(), 55, "1900".to_string(), 80),
]);
assert_eq!(tree_left, tree_expected);
}
#[test]
fn test_merge_fruits_tree_2() {
let mut tree_left = get_intermediat_tree_with_ranges(&[
("red".to_string(), 50, "1900".to_string(), 25),
("blue".to_string(), 30, "1900".to_string(), 30),
]);
let tree_right = get_intermediat_tree_with_ranges(&[
("red".to_string(), 60, "1900".to_string(), 30),
("green".to_string(), 25, "1900".to_string(), 50),
]);
tree_left.merge_fruits(tree_right);
let tree_expected = get_intermediat_tree_with_ranges(&[
("red".to_string(), 110, "1900".to_string(), 55),
("blue".to_string(), 30, "1900".to_string(), 30),
("green".to_string(), 25, "1900".to_string(), 50),
]);
assert_eq!(tree_left, tree_expected);
}
#[test]
fn test_merge_fruits_tree_empty() {
let mut tree_left = get_intermediat_tree_with_ranges(&[
("red".to_string(), 50, "1900".to_string(), 25),
("blue".to_string(), 30, "1900".to_string(), 30),
]);
let orig = tree_left.clone();
tree_left.merge_fruits(IntermediateAggregationResults::default());
assert_eq!(tree_left, orig);
}
}

114
src/aggregation/metric/average.rs Normal file
View File

@@ -0,0 +1,114 @@
use std::fmt::Debug;
use serde::{Deserialize, Serialize};
use crate::aggregation::f64_from_fastfield_u64;
use crate::fastfield::{DynamicFastFieldReader, FastFieldReader};
use crate::schema::Type;
use crate::DocId;
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
/// A single-value metric aggregation that computes the average of numeric values that are
/// extracted from the aggregated documents.
/// Supported field types are u64, i64, and f64.
/// See [super::SingleMetricResult] for return value.
///
/// # JSON Format
/// ```json
/// {
/// "avg": {
/// "field": "score",
/// }
/// }
/// ```
pub struct AverageAggregation {
/// The field name to compute the stats on.
pub field: String,
}
impl AverageAggregation {
/// Create new AverageAggregation from a field.
pub fn from_field_name(field_name: String) -> Self {
AverageAggregation { field: field_name }
}
/// Return the field name.
pub fn field_name(&self) -> &str {
&self.field
}
}
#[derive(Clone, PartialEq)]
pub(crate) struct SegmentAverageCollector {
pub data: IntermediateAverage,
field_type: Type,
}
impl Debug for SegmentAverageCollector {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
f.debug_struct("AverageCollector")
.field("data", &self.data)
.finish()
}
}
impl SegmentAverageCollector {
pub fn from_req(field_type: Type) -> Self {
Self {
field_type,
data: Default::default(),
}
}
pub(crate) fn collect_block(&mut self, doc: &[DocId], field: &DynamicFastFieldReader<u64>) {
let mut iter = doc.chunks_exact(4);
for docs in iter.by_ref() {
let val1 = field.get(docs[0]);
let val2 = field.get(docs[1]);
let val3 = field.get(docs[2]);
let val4 = field.get(docs[3]);
let val1 = f64_from_fastfield_u64(val1, &self.field_type);
let val2 = f64_from_fastfield_u64(val2, &self.field_type);
let val3 = f64_from_fastfield_u64(val3, &self.field_type);
let val4 = f64_from_fastfield_u64(val4, &self.field_type);
self.data.collect(val1);
self.data.collect(val2);
self.data.collect(val3);
self.data.collect(val4);
}
for doc in iter.remainder() {
let val = field.get(*doc);
let val = f64_from_fastfield_u64(val, &self.field_type);
self.data.collect(val);
}
}
}
/// Contains mergeable version of average data.
#[derive(Default, Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct IntermediateAverage {
pub(crate) sum: f64,
pub(crate) doc_count: u64,
}
impl IntermediateAverage {
pub(crate) fn from_collector(collector: SegmentAverageCollector) -> Self {
collector.data
}
/// Merge average data into this instance.
pub fn merge_fruits(&mut self, other: IntermediateAverage) {
self.sum += other.sum;
self.doc_count += other.doc_count;
}
/// compute final result
pub fn finalize(&self) -> Option<f64> {
if self.doc_count == 0 {
None
} else {
Some(self.sum / self.doc_count as f64)
}
}
#[inline]
fn collect(&mut self, val: f64) {
self.doc_count += 1;
self.sum += val;
}
}

30
src/aggregation/metric/mod.rs Normal file
View File

@@ -0,0 +1,30 @@
//! Module for all metric aggregations.
//!
//! The aggregations in this family compute metrics, see [super::agg_req::MetricAggregation] for
//! details.
mod average;
mod stats;
pub use average::*;
use serde::{Deserialize, Serialize};
pub use stats::*;
/// Single-metric aggregations use this common result structure.
///
/// Main reason to wrap it in value is to match elasticsearch output structure.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct SingleMetricResult {
/// The value of the single value metric.
pub value: Option<f64>,
}
impl From<f64> for SingleMetricResult {
fn from(value: f64) -> Self {
Self { value: Some(value) }
}
}
impl From<Option<f64>> for SingleMetricResult {
fn from(value: Option<f64>) -> Self {
Self { value }
}
}

353
src/aggregation/metric/stats.rs Normal file
View File

@@ -0,0 +1,353 @@
use serde::{Deserialize, Serialize};
use crate::aggregation::f64_from_fastfield_u64;
use crate::fastfield::{DynamicFastFieldReader, FastFieldReader};
use crate::schema::Type;
use crate::DocId;
/// A multi-value metric aggregation that computes stats of numeric values that are
/// extracted from the aggregated documents.
/// Supported field types are u64, i64, and f64.
/// See [Stats] for returned statistics.
///
/// # JSON Format
/// ```json
/// {
/// "stats": {
/// "field": "score",
/// }
/// }
/// ```
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct StatsAggregation {
/// The field name to compute the stats on.
pub field: String,
}
impl StatsAggregation {
/// Create new StatsAggregation from a field.
pub fn from_field_name(field_name: String) -> Self {
StatsAggregation { field: field_name }
}
/// Return the field name.
pub fn field_name(&self) -> &str {
&self.field
}
}
/// Stats contains a collection of statistics.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct Stats {
/// The number of documents.
pub count: usize,
/// The sum of the fast field values.
pub sum: f64,
/// The standard deviation of the fast field values. None for count == 0.
pub standard_deviation: Option<f64>,
/// The min value of the fast field values.
pub min: Option<f64>,
/// The max value of the fast field values.
pub max: Option<f64>,
/// The average of the values. None for count == 0.
pub avg: Option<f64>,
}
/// IntermediateStats contains the mergeable version for stats.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct IntermediateStats {
count: usize,
sum: f64,
squared_sum: f64,
min: f64,
max: f64,
}
impl Default for IntermediateStats {
fn default() -> Self {
Self {
count: 0,
sum: 0.0,
squared_sum: 0.0,
min: f64::MAX,
max: f64::MIN,
}
}
}
impl IntermediateStats {
pub(crate) fn avg(&self) -> Option<f64> {
if self.count == 0 {
None
} else {
Some(self.sum / (self.count as f64))
}
}
fn square_mean(&self) -> f64 {
self.squared_sum / (self.count as f64)
}
pub(crate) fn standard_deviation(&self) -> Option<f64> {
self.avg()
.map(|average| (self.square_mean() - average * average).sqrt())
}
/// Merge data from other stats into this instance.
pub fn merge_fruits(&mut self, other: IntermediateStats) {
self.count += other.count;
self.sum += other.sum;
self.squared_sum += other.squared_sum;
self.min = self.min.min(other.min);
self.max = self.max.max(other.max);
}
/// compute final resultimprove_docs
pub fn finalize(&self) -> Stats {
let min = if self.count == 0 {
None
} else {
Some(self.min)
};
let max = if self.count == 0 {
None
} else {
Some(self.max)
};
Stats {
count: self.count,
sum: self.sum,
standard_deviation: self.standard_deviation(),
min,
max,
avg: self.avg(),
}
}
#[inline]
fn collect(&mut self, value: f64) {
self.count += 1;
self.sum += value;
self.squared_sum += value * value;
self.min = self.min.min(value);
self.max = self.max.max(value);
}
}
#[derive(Clone, Debug, PartialEq)]
pub(crate) struct SegmentStatsCollector {
pub(crate) stats: IntermediateStats,
field_type: Type,
}
impl SegmentStatsCollector {
pub fn from_req(field_type: Type) -> Self {
Self {
field_type,
stats: IntermediateStats::default(),
}
}
pub(crate) fn collect_block(&mut self, doc: &[DocId], field: &DynamicFastFieldReader<u64>) {
let mut iter = doc.chunks_exact(4);
for docs in iter.by_ref() {
let val1 = field.get(docs[0]);
let val2 = field.get(docs[1]);
let val3 = field.get(docs[2]);
let val4 = field.get(docs[3]);
let val1 = f64_from_fastfield_u64(val1, &self.field_type);
let val2 = f64_from_fastfield_u64(val2, &self.field_type);
let val3 = f64_from_fastfield_u64(val3, &self.field_type);
let val4 = f64_from_fastfield_u64(val4, &self.field_type);
self.stats.collect(val1);
self.stats.collect(val2);
self.stats.collect(val3);
self.stats.collect(val4);
}
for doc in iter.remainder() {
let val = field.get(*doc);
let val = f64_from_fastfield_u64(val, &self.field_type);
self.stats.collect(val);
}
}
}
#[cfg(test)]
mod tests {
use std::iter;
use serde_json::Value;
use crate::aggregation::agg_req::{
Aggregation, Aggregations, BucketAggregation, BucketAggregationType, MetricAggregation,
RangeAggregation,
};
use crate::aggregation::agg_result::AggregationResults;
use crate::aggregation::metric::StatsAggregation;
use crate::aggregation::tests::{get_test_index_2_segments, get_test_index_from_values};
use crate::aggregation::AggregationCollector;
use crate::query::{AllQuery, TermQuery};
use crate::schema::IndexRecordOption;
use crate::Term;
#[test]
fn test_aggregation_stats_empty_index() -> crate::Result<()> {
// test index without segments
let values = vec![];
let index = get_test_index_from_values(false, &values)?;
let agg_req_1: Aggregations = vec![(
"stats".to_string(),
Aggregation::Metric(MetricAggregation::Stats(StatsAggregation::from_field_name(
"score".to_string(),
))),
)]
.into_iter()
.collect();
let collector = AggregationCollector::from_aggs(agg_req_1);
let reader = index.reader()?;
let searcher = reader.searcher();
let agg_res: AggregationResults = searcher.search(&AllQuery, &collector).unwrap();
let res: Value = serde_json::from_str(&serde_json::to_string(&agg_res)?)?;
assert_eq!(
res["stats"],
json!({
"avg": Value::Null,
"count": 0,
"max": Value::Null,
"min": Value::Null,
"standard_deviation": Value::Null,
"sum": 0.0
})
);
Ok(())
}
#[test]
fn test_aggregation_stats() -> crate::Result<()> {
let index = get_test_index_2_segments(false)?;
let reader = index.reader()?;
let text_field = reader.searcher().schema().get_field("text").unwrap();
let term_query = TermQuery::new(
Term::from_field_text(text_field, "cool"),
IndexRecordOption::Basic,
);
let agg_req_1: Aggregations = vec![
(
"stats_i64".to_string(),
Aggregation::Metric(MetricAggregation::Stats(StatsAggregation::from_field_name(
"score_i64".to_string(),
))),
),
(
"stats_f64".to_string(),
Aggregation::Metric(MetricAggregation::Stats(StatsAggregation::from_field_name(
"score_f64".to_string(),
))),
),
(
"stats".to_string(),
Aggregation::Metric(MetricAggregation::Stats(StatsAggregation::from_field_name(
"score".to_string(),
))),
),
(
"range".to_string(),
Aggregation::Bucket(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "score".to_string(),
ranges: vec![
(3f64..7f64).into(),
(7f64..19f64).into(),
(19f64..20f64).into(),
],
}),
sub_aggregation: iter::once((
"stats".to_string(),
Aggregation::Metric(MetricAggregation::Stats(
StatsAggregation::from_field_name("score".to_string()),
)),
))
.collect(),
}),
),
]
.into_iter()
.collect();
let collector = AggregationCollector::from_aggs(agg_req_1);
let searcher = reader.searcher();
let agg_res: AggregationResults = searcher.search(&term_query, &collector).unwrap();
let res: Value = serde_json::from_str(&serde_json::to_string(&agg_res)?)?;
assert_eq!(
res["stats"],
json!({
"avg": 12.142857142857142,
"count": 7,
"max": 44.0,
"min": 1.0,
"standard_deviation": 13.65313748796613,
"sum": 85.0
})
);
assert_eq!(
res["stats_i64"],
json!({
"avg": 12.142857142857142,
"count": 7,
"max": 44.0,
"min": 1.0,
"standard_deviation": 13.65313748796613,
"sum": 85.0
})
);
assert_eq!(
res["stats_f64"],
json!({
"avg": 12.214285714285714,
"count": 7,
"max": 44.5,
"min": 1.0,
"standard_deviation": 13.819905785437443,
"sum": 85.5
})
);
assert_eq!(
res["range"]["buckets"][2]["stats"],
json!({
"avg": 10.666666666666666,
"count": 3,
"max": 14.0,
"min": 7.0,
"standard_deviation": 2.867441755680877,
"sum": 32.0
})
);
assert_eq!(
res["range"]["buckets"][3]["stats"],
json!({
"avg": serde_json::Value::Null,
"count": 0,
"max": serde_json::Value::Null,
"min": serde_json::Value::Null,
"standard_deviation": serde_json::Value::Null,
"sum": 0.0,
})
);
Ok(())
}
}

1369
src/aggregation/mod.rs Normal file
View File

File diff suppressed because it is too large Load Diff

233
src/aggregation/segment_agg_result.rs Normal file
View File

@@ -0,0 +1,233 @@
//! Contains aggregation trees which is used during collection in a segment.
//! This tree contains datastructrues optimized for fast collection.
//! The tree can be converted to an intermediate tree, which contains datastructrues optimized for
//! merging.
use std::fmt::Debug;
use super::agg_req::MetricAggregation;
use super::agg_req_with_accessor::{
AggregationsWithAccessor, BucketAggregationWithAccessor, MetricAggregationWithAccessor,
};
use super::bucket::{SegmentHistogramCollector, SegmentRangeCollector};
use super::metric::{
AverageAggregation, SegmentAverageCollector, SegmentStatsCollector, StatsAggregation,
};
use super::{Key, VecWithNames};
use crate::aggregation::agg_req::BucketAggregationType;
use crate::DocId;
pub(crate) const DOC_BLOCK_SIZE: usize = 64;
pub(crate) type DocBlock = [DocId; DOC_BLOCK_SIZE];
#[derive(Clone, PartialEq)]
pub(crate) struct SegmentAggregationResultsCollector {
pub(crate) metrics: Option<VecWithNames<SegmentMetricResultCollector>>,
pub(crate) buckets: Option<VecWithNames<SegmentBucketResultCollector>>,
staged_docs: DocBlock,
num_staged_docs: usize,
}
impl Debug for SegmentAggregationResultsCollector {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
f.debug_struct("SegmentAggregationResultsCollector")
.field("metrics", &self.metrics)
.field("buckets", &self.buckets)
.field("staged_docs", &&self.staged_docs[..self.num_staged_docs])
.field("num_staged_docs", &self.num_staged_docs)
.finish()
}
}
impl SegmentAggregationResultsCollector {
pub(crate) fn from_req_and_validate(req: &AggregationsWithAccessor) -> crate::Result<Self> {
let buckets = req
.buckets
.entries()
.map(|(key, req)| {
Ok((
key.to_string(),
SegmentBucketResultCollector::from_req_and_validate(req)?,
))
})
.collect::<crate::Result<Vec<(String, _)>>>()?;
let metrics = req
.metrics
.entries()
.map(|(key, req)| {
Ok((
key.to_string(),
SegmentMetricResultCollector::from_req_and_validate(req)?,
))
})
.collect::<crate::Result<Vec<(String, _)>>>()?;
let metrics = if metrics.is_empty() {
None
} else {
Some(VecWithNames::from_entries(metrics))
};
let buckets = if buckets.is_empty() {
None
} else {
Some(VecWithNames::from_entries(buckets))
};
Ok(SegmentAggregationResultsCollector {
metrics,
buckets,
staged_docs: [0; DOC_BLOCK_SIZE],
num_staged_docs: 0,
})
}
#[inline]
pub(crate) fn collect(
&mut self,
doc: crate::DocId,
agg_with_accessor: &AggregationsWithAccessor,
) {
self.staged_docs[self.num_staged_docs] = doc;
self.num_staged_docs += 1;
if self.num_staged_docs == self.staged_docs.len() {
self.flush_staged_docs(agg_with_accessor, false);
}
}
pub(crate) fn flush_staged_docs(
&mut self,
agg_with_accessor: &AggregationsWithAccessor,
force_flush: bool,
) {
if let Some(metrics) = &mut self.metrics {
for (collector, agg_with_accessor) in
metrics.values_mut().zip(agg_with_accessor.metrics.values())
{
collector
.collect_block(&self.staged_docs[..self.num_staged_docs], agg_with_accessor);
}
}
if let Some(buckets) = &mut self.buckets {
for (collector, agg_with_accessor) in
buckets.values_mut().zip(agg_with_accessor.buckets.values())
{
collector.collect_block(
&self.staged_docs[..self.num_staged_docs],
agg_with_accessor,
force_flush,
);
}
}
self.num_staged_docs = 0;
}
}
#[derive(Clone, Debug, PartialEq)]
pub(crate) enum SegmentMetricResultCollector {
Average(SegmentAverageCollector),
Stats(SegmentStatsCollector),
}
impl SegmentMetricResultCollector {
pub fn from_req_and_validate(req: &MetricAggregationWithAccessor) -> crate::Result<Self> {
match &req.metric {
MetricAggregation::Average(AverageAggregation { field: _ }) => {
Ok(SegmentMetricResultCollector::Average(
SegmentAverageCollector::from_req(req.field_type),
))
}
MetricAggregation::Stats(StatsAggregation { field: _ }) => {
Ok(SegmentMetricResultCollector::Stats(
SegmentStatsCollector::from_req(req.field_type),
))
}
}
}
pub(crate) fn collect_block(&mut self, doc: &[DocId], metric: &MetricAggregationWithAccessor) {
match self {
SegmentMetricResultCollector::Average(avg_collector) => {
avg_collector.collect_block(doc, &metric.accessor);
}
SegmentMetricResultCollector::Stats(stats_collector) => {
stats_collector.collect_block(doc, &metric.accessor);
}
}
}
}
/// SegmentBucketAggregationResultCollectors will have specialized buckets for collection inside
/// segments.
/// The typical structure of Map<Key, Bucket> is not suitable during collection for performance
/// reasons.
#[derive(Clone, Debug, PartialEq)]
pub(crate) enum SegmentBucketResultCollector {
Range(SegmentRangeCollector),
Histogram(SegmentHistogramCollector),
}
impl SegmentBucketResultCollector {
pub fn from_req_and_validate(req: &BucketAggregationWithAccessor) -> crate::Result<Self> {
match &req.bucket_agg {
BucketAggregationType::Range(range_req) => {
Ok(Self::Range(SegmentRangeCollector::from_req_and_validate(
range_req,
&req.sub_aggregation,
req.field_type,
)?))
}
BucketAggregationType::Histogram(histogram) => Ok(Self::Histogram(
SegmentHistogramCollector::from_req_and_validate(
histogram,
&req.sub_aggregation,
req.field_type,
&req.accessor,
)?,
)),
}
}
#[inline]
pub(crate) fn collect_block(
&mut self,
doc: &[DocId],
bucket_with_accessor: &BucketAggregationWithAccessor,
force_flush: bool,
) {
match self {
SegmentBucketResultCollector::Range(range) => {
range.collect_block(doc, bucket_with_accessor, force_flush);
}
SegmentBucketResultCollector::Histogram(histogram) => {
histogram.collect_block(doc, bucket_with_accessor, force_flush)
}
}
}
}
#[derive(Clone, Debug, PartialEq)]
pub(crate) struct SegmentHistogramBucketEntry {
pub key: f64,
pub doc_count: u64,
}
#[derive(Clone, PartialEq)]
pub(crate) struct SegmentRangeBucketEntry {
pub key: Key,
pub doc_count: u64,
pub sub_aggregation: Option<SegmentAggregationResultsCollector>,
/// The from range of the bucket. Equals f64::MIN when None.
pub from: Option<f64>,
/// The to range of the bucket. Equals f64::MAX when None.
pub to: Option<f64>,
}
impl Debug for SegmentRangeBucketEntry {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
f.debug_struct("SegmentRangeBucketEntry")
.field("key", &self.key)
.field("doc_count", &self.doc_count)
.field("from", &self.from)
.field("to", &self.to)
.finish()
}
}

8
src/collector/count_collector.rs
View File

@@ -1,9 +1,6 @@
use super::Collector;
use crate::collector::SegmentCollector;
use crate::DocId;
use crate::Score;
use crate::SegmentOrdinal;
use crate::SegmentReader;
use crate::{DocId, Score, SegmentOrdinal, SegmentReader};
/// `CountCollector` collector only counts how many
/// documents match the query.
@@ -80,8 +77,7 @@ impl SegmentCollector for SegmentCountCollector {
#[cfg(test)]
mod tests {
use super::{Count, SegmentCountCollector};
use crate::collector::Collector;
use crate::collector::SegmentCollector;
use crate::collector::{Collector, SegmentCollector};
#[test]
fn test_count_collect_does_not_requires_scoring() {

6
src/collector/custom_score_top_collector.rs
View File

@@ -8,8 +8,7 @@ pub(crate) struct CustomScoreTopCollector<TCustomScorer, TScore = Score> {
}
impl<TCustomScorer, TScore> CustomScoreTopCollector<TCustomScorer, TScore>
where
TScore: Clone + PartialOrd,
where TScore: Clone + PartialOrd
{
pub(crate) fn new(
custom_scorer: TCustomScorer,
@@ -114,8 +113,7 @@ where
}
impl<F, TScore> CustomSegmentScorer<TScore> for F
where
F: 'static + FnMut(DocId) -> TScore,
where F: 'static + FnMut(DocId) -> TScore
{
fn score(&mut self, doc: DocId) -> TScore {
(self)(doc)

3
src/collector/docset_collector.rs
View File

@@ -1,8 +1,7 @@
use std::collections::HashSet;
use crate::{DocAddress, DocId, Score};
use super::{Collector, SegmentCollector};
use crate::{DocAddress, DocId, Score};
/// Collectors that returns the set of DocAddress that matches the query.
///

53
src/collector/facet_collector.rs
View File

@@ -1,21 +1,14 @@
use crate::collector::Collector;
use crate::collector::SegmentCollector;
use crate::fastfield::FacetReader;
use crate::schema::Facet;
use crate::schema::Field;
use crate::DocId;
use crate::Score;
use crate::SegmentOrdinal;
use crate::SegmentReader;
use std::cmp::Ordering;
use std::collections::btree_map;
use std::collections::BTreeMap;
use std::collections::BTreeSet;
use std::collections::BinaryHeap;
use std::collections::{btree_map, BTreeMap, BTreeSet, BinaryHeap};
use std::iter::Peekable;
use std::ops::Bound;
use std::{u64, usize};
use crate::collector::{Collector, SegmentCollector};
use crate::fastfield::FacetReader;
use crate::schema::{Facet, Field};
use crate::{DocId, Score, SegmentOrdinal, SegmentReader};
struct Hit<'a> {
count: u64,
facet: &'a Facet,
@@ -240,9 +233,7 @@ impl FacetCollector {
/// If you need the correct number of unique documents for two such facets,
/// just add them in separate `FacetCollector`.
pub fn add_facet<T>(&mut self, facet_from: T)
where
Facet: From<T>,
{
where Facet: From<T> {
let facet = Facet::from(facet_from);
for old_facet in &self.facets {
assert!(
@@ -402,9 +393,7 @@ impl FacetCounts {
/// Returns an iterator over all of the facet count pairs inside this result.
/// See the documentation for [FacetCollector] for a usage example.
pub fn get<T>(&self, facet_from: T) -> FacetChildIterator<'_>
where
Facet: From<T>,
{
where Facet: From<T> {
let facet = Facet::from(facet_from);
let left_bound = Bound::Excluded(facet.clone());
let right_bound = if facet.is_root() {
@@ -423,9 +412,7 @@ impl FacetCounts {
/// Returns a vector of top `k` facets with their counts, sorted highest-to-lowest by counts.
/// See the documentation for [FacetCollector] for a usage example.
pub fn top_k<T>(&self, facet: T, k: usize) -> Vec<(&Facet, u64)>
where
Facet: From<T>,
{
where Facet: From<T> {
let mut heap = BinaryHeap::with_capacity(k);
let mut it = self.get(facet);
@@ -458,16 +445,18 @@ impl FacetCounts {
#[cfg(test)]
mod tests {
use std::iter;
use rand::distributions::Uniform;
use rand::prelude::SliceRandom;
use rand::{thread_rng, Rng};
use super::{FacetCollector, FacetCounts};
use crate::collector::Count;
use crate::core::Index;
use crate::query::{AllQuery, QueryParser, TermQuery};
use crate::schema::{Document, Facet, FacetOptions, Field, IndexRecordOption, Schema};
use crate::Term;
use rand::distributions::Uniform;
use rand::prelude::SliceRandom;
use rand::{thread_rng, Rng};
use std::iter;
#[test]
fn test_facet_collector_drilldown() -> crate::Result<()> {
@@ -522,8 +511,9 @@ mod tests {
}
#[test]
#[should_panic(expected = "Tried to add a facet which is a descendant of \
an already added facet.")]
#[should_panic(
expected = "Tried to add a facet which is a descendant of an already added facet."
)]
fn test_misused_facet_collector() {
let mut facet_collector = FacetCollector::for_field(Field::from_field_id(0));
facet_collector.add_facet(Facet::from("/country"));
@@ -700,13 +690,14 @@ mod tests {
#[cfg(all(test, feature = "unstable"))]
mod bench {
use rand::seq::SliceRandom;
use rand::thread_rng;
use test::Bencher;
use crate::collector::FacetCollector;
use crate::query::AllQuery;
use crate::schema::{Facet, Schema, INDEXED};
use crate::Index;
use rand::seq::SliceRandom;
use rand::thread_rng;
use test::Bencher;
#[bench]
fn bench_facet_collector(b: &mut Bencher) {

6
src/collector/filter_collector_wrapper.rs
View File

@@ -17,7 +17,8 @@ use crate::schema::Field;
use crate::{Score, SegmentReader, TantivyError};
/// The `FilterCollector` filters docs using a fast field value and a predicate.
/// Only the documents for which the predicate returned "true" will be passed on to the next collector.
/// Only the documents for which the predicate returned "true" will be passed on to the next
/// collector.
///
/// ```rust
/// use tantivy::collector::{TopDocs, FilterCollector};
@@ -58,8 +59,7 @@ use crate::{Score, SegmentReader, TantivyError};
/// # }
/// ```
pub struct FilterCollector<TCollector, TPredicate, TPredicateValue: FastValue>
where
TPredicate: 'static + Clone,
where TPredicate: 'static + Clone
{
field: Field,
collector: TCollector,

30
src/collector/histogram_collector.rs
View File

@@ -1,8 +1,9 @@
use fastdivide::DividerU64;
use crate::collector::{Collector, SegmentCollector};
use crate::fastfield::{DynamicFastFieldReader, FastFieldReader, FastValue};
use crate::schema::{Field, Type};
use crate::{DocId, Score};
use fastdivide::DividerU64;
/// Histogram builds an histogram of the values of a fastfield for the
/// collected DocSet.
@@ -18,7 +19,7 @@ use fastdivide::DividerU64;
///
/// # Warning
///
/// f64 field. are not supported.
/// f64 fields are not supported.
#[derive(Clone)]
pub struct HistogramCollector {
min_value: u64,
@@ -36,8 +37,8 @@ impl HistogramCollector {
/// - `bucket_width`: the length of the interval that is associated to each buckets.
/// - `num_buckets`: The overall number of buckets.
///
/// Together, this parameters define a partition of `[min_value, min_value + num_buckets * bucket_width)`
/// into `num_buckets` intervals of width bucket that we call `bucket`.
/// Together, this parameters define a partition of `[min_value, min_value + num_buckets *
/// bucket_width)` into `num_buckets` intervals of width bucket that we call `bucket`.
///
/// # Disclaimer
/// This function panics if the field given is of type f64.
@@ -147,13 +148,14 @@ fn add_vecs(mut vals_list: Vec<Vec<u64>>, len: usize) -> Vec<u64> {
#[cfg(test)]
mod tests {
use super::{add_vecs, HistogramCollector, HistogramComputer};
use crate::chrono::{TimeZone, Utc};
use crate::schema::{Schema, FAST};
use crate::{doc, query, Index};
use fastdivide::DividerU64;
use query::AllQuery;
use super::{add_vecs, HistogramCollector, HistogramComputer};
use crate::schema::{Schema, FAST};
use crate::time::{Date, Month};
use crate::{doc, query, DateTime, Index};
#[test]
fn test_add_histograms_simple() {
assert_eq!(
@@ -271,16 +273,20 @@ mod tests {
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut writer = index.writer_with_num_threads(1, 4_000_000)?;
writer.add_document(doc!(date_field=>Utc.ymd(1982, 9, 17).and_hms(0, 0,0)))?;
writer.add_document(doc!(date_field=>Utc.ymd(1986, 3, 9).and_hms(0, 0, 0)))?;
writer.add_document(doc!(date_field=>Utc.ymd(1983, 9, 27).and_hms(0, 0, 0)))?;
writer.add_document(doc!(date_field=>DateTime::new_primitive(Date::from_calendar_date(1982, Month::September, 17)?.with_hms(0, 0, 0)?)))?;
writer.add_document(
doc!(date_field=>DateTime::new_primitive(Date::from_calendar_date(1986, Month::March, 9)?.with_hms(0, 0, 0)?)),
)?;
writer.add_document(doc!(date_field=>DateTime::new_primitive(Date::from_calendar_date(1983, Month::September, 27)?.with_hms(0, 0, 0)?)))?;
writer.commit()?;
let reader = index.reader()?;
let searcher = reader.searcher();
let all_query = AllQuery;
let week_histogram_collector = HistogramCollector::new(
date_field,
Utc.ymd(1980, 1, 1).and_hms(0, 0, 0),
DateTime::new_primitive(
Date::from_calendar_date(1980, Month::January, 1)?.with_hms(0, 0, 0)?,
),
3600 * 24 * 365, // it is just for a unit test... sorry leap years.
10,
);

176
src/collector/mod.rs
View File

@@ -1,95 +1,90 @@
/*!
//! # Collectors
//!
//! Collectors define the information you want to extract from the documents matching the queries.
//! In tantivy jargon, we call this information your search "fruit".
//!
//! Your fruit could for instance be :
//! - [the count of matching documents](./struct.Count.html)
//! - [the top 10 documents, by relevancy or by a fast field](./struct.TopDocs.html)
//! - [facet counts](./struct.FacetCollector.html)
//!
//! At one point in your code, you will trigger the actual search operation by calling
//! [the `search(...)` method of your `Searcher` object](../struct.Searcher.html#method.search).
//! This call will look like this.
//!
//! ```verbatim
//! let fruit = searcher.search(&query, &collector)?;
//! ```
//!
//! Here the type of fruit is actually determined as an associated type of the collector
//! (`Collector::Fruit`).
//!
//!
//! # Combining several collectors
//!
//! A rich search experience often requires to run several collectors on your search query.
//! For instance,
//! - selecting the top-K products matching your query
//! - counting the matching documents
//! - computing several facets
//! - computing statistics about the matching product prices
//!
//! A simple and efficient way to do that is to pass your collectors as one tuple.
//! The resulting `Fruit` will then be a typed tuple with each collector's original fruits
//! in their respective position.
//!
//! ```rust
//! # use tantivy::schema::*;
//! # use tantivy::*;
//! # use tantivy::query::*;
//! use tantivy::collector::{Count, TopDocs};
//! #
//! # fn main() -> tantivy::Result<()> {
//! # let mut schema_builder = Schema::builder();
//! # let title = schema_builder.add_text_field("title", TEXT);
//! # let schema = schema_builder.build();
//! # let index = Index::create_in_ram(schema);
//! # let mut index_writer = index.writer(3_000_000)?;
//! # index_writer.add_document(doc!(
//! # title => "The Name of the Wind",
//! # ))?;
//! # index_writer.add_document(doc!(
//! # title => "The Diary of Muadib",
//! # ))?;
//! # index_writer.commit()?;
//! # let reader = index.reader()?;
//! # let searcher = reader.searcher();
//! # let query_parser = QueryParser::for_index(&index, vec![title]);
//! # let query = query_parser.parse_query("diary")?;
//! let (doc_count, top_docs): (usize, Vec<(Score, DocAddress)>) =
//! searcher.search(&query, &(Count, TopDocs::with_limit(2)))?;
//! # Ok(())
//! # }
//! ```
//!
//! The `Collector` trait is implemented for up to 4 collectors.
//! If you have more than 4 collectors, you can either group them into
//! tuples of tuples `(a,(b,(c,d)))`, or rely on [`MultiCollector`](./struct.MultiCollector.html).
//!
//! # Combining several collectors dynamically
//!
//! Combining collectors into a tuple is a zero-cost abstraction: everything
//! happens as if you had manually implemented a single collector
//! combining all of our features.
//!
//! Unfortunately it requires you to know at compile time your collector types.
//! If on the other hand, the collectors depend on some query parameter,
//! you can rely on `MultiCollector`'s.
//!
//!
//! # Implementing your own collectors.
//!
//! See the `custom_collector` example.
# Collectors
Collectors define the information you want to extract from the documents matching the queries.
In tantivy jargon, we call this information your search "fruit".
Your fruit could for instance be :
- [the count of matching documents](./struct.Count.html)
- [the top 10 documents, by relevancy or by a fast field](./struct.TopDocs.html)
- [facet counts](./struct.FacetCollector.html)
At one point in your code, you will trigger the actual search operation by calling
[the `search(...)` method of your `Searcher` object](../struct.Searcher.html#method.search).
This call will look like this.
```verbatim
let fruit = searcher.search(&query, &collector)?;
```
Here the type of fruit is actually determined as an associated type of the collector (`Collector::Fruit`).
# Combining several collectors
A rich search experience often requires to run several collectors on your search query.
For instance,
- selecting the top-K products matching your query
- counting the matching documents
- computing several facets
- computing statistics about the matching product prices
A simple and efficient way to do that is to pass your collectors as one tuple.
The resulting `Fruit` will then be a typed tuple with each collector's original fruits
in their respective position.
```rust
# use tantivy::schema::*;
# use tantivy::*;
# use tantivy::query::*;
use tantivy::collector::{Count, TopDocs};
#
# fn main() -> tantivy::Result<()> {
# let mut schema_builder = Schema::builder();
# let title = schema_builder.add_text_field("title", TEXT);
# let schema = schema_builder.build();
# let index = Index::create_in_ram(schema);
# let mut index_writer = index.writer(3_000_000)?;
# index_writer.add_document(doc!(
# title => "The Name of the Wind",
# ))?;
# index_writer.add_document(doc!(
# title => "The Diary of Muadib",
# ))?;
# index_writer.commit()?;
# let reader = index.reader()?;
# let searcher = reader.searcher();
# let query_parser = QueryParser::for_index(&index, vec![title]);
# let query = query_parser.parse_query("diary")?;
let (doc_count, top_docs): (usize, Vec<(Score, DocAddress)>) =
searcher.search(&query, &(Count, TopDocs::with_limit(2)))?;
# Ok(())
# }
```
The `Collector` trait is implemented for up to 4 collectors.
If you have more than 4 collectors, you can either group them into
tuples of tuples `(a,(b,(c,d)))`, or rely on [`MultiCollector`](./struct.MultiCollector.html).
# Combining several collectors dynamically
Combining collectors into a tuple is a zero-cost abstraction: everything
happens as if you had manually implemented a single collector
combining all of our features.
Unfortunately it requires you to know at compile time your collector types.
If on the other hand, the collectors depend on some query parameter,
you can rely on `MultiCollector`'s.
# Implementing your own collectors.
See the `custom_collector` example.
*/
use crate::DocId;
use crate::Score;
use crate::SegmentOrdinal;
use crate::SegmentReader;
use downcast_rs::impl_downcast;
use crate::{DocId, Score, SegmentOrdinal, SegmentReader};
mod count_collector;
pub use self::count_collector::Count;
@@ -111,8 +106,7 @@ mod tweak_score_top_collector;
pub use self::tweak_score_top_collector::{ScoreSegmentTweaker, ScoreTweaker};
mod facet_collector;
pub use self::facet_collector::FacetCollector;
pub use self::facet_collector::FacetCounts;
pub use self::facet_collector::{FacetCollector, FacetCounts};
use crate::query::Weight;
mod docset_collector;

21
src/collector/multi_collector.rs
View File

@@ -1,14 +1,10 @@
use super::Collector;
use super::SegmentCollector;
use crate::collector::Fruit;
use crate::DocId;
use crate::Score;
use crate::SegmentOrdinal;
use crate::SegmentReader;
use crate::TantivyError;
use std::marker::PhantomData;
use std::ops::Deref;
use super::{Collector, SegmentCollector};
use crate::collector::Fruit;
use crate::{DocId, Score, SegmentOrdinal, SegmentReader, TantivyError};
pub struct MultiFruit {
sub_fruits: Vec<Option<Box<dyn Fruit>>>,
}
@@ -104,7 +100,8 @@ impl<TFruit: Fruit> FruitHandle<TFruit> {
///
/// If the type of the collectors is known, you can just group yours collectors
/// in a tuple. See the
/// [Combining several collectors section of the collector documentation](./index.html#combining-several-collectors).
/// [Combining several collectors section of the collector
/// documentation](./index.html#combining-several-collectors).
///
/// ```rust
/// use tantivy::collector::{Count, TopDocs, MultiCollector};
@@ -248,10 +245,8 @@ mod tests {
use super::*;
use crate::collector::{Count, TopDocs};
use crate::query::TermQuery;
use crate::schema::IndexRecordOption;
use crate::schema::{Schema, TEXT};
use crate::Index;
use crate::Term;
use crate::schema::{IndexRecordOption, Schema, TEXT};
use crate::{Index, Term};
#[test]
fn test_multi_collector() -> crate::Result<()> {

34
src/collector/tests.rs
View File

@@ -1,20 +1,12 @@
use super::*;
use crate::core::SegmentReader;
use crate::fastfield::BytesFastFieldReader;
use crate::fastfield::DynamicFastFieldReader;
use crate::fastfield::FastFieldReader;
use crate::schema::Field;
use crate::DocId;
use crate::Score;
use crate::SegmentOrdinal;
use crate::{DocAddress, Document, Searcher};
use crate::collector::{Count, FilterCollector, TopDocs};
use crate::core::SegmentReader;
use crate::fastfield::{BytesFastFieldReader, DynamicFastFieldReader, FastFieldReader};
use crate::query::{AllQuery, QueryParser};
use crate::schema::{Schema, FAST, TEXT};
use crate::DateTime;
use crate::{doc, Index};
use std::str::FromStr;
use crate::schema::{Field, Schema, FAST, TEXT};
use crate::time::format_description::well_known::Rfc3339;
use crate::time::OffsetDateTime;
use crate::{doc, DateTime, DocAddress, DocId, Document, Index, Score, Searcher, SegmentOrdinal};
pub const TEST_COLLECTOR_WITH_SCORE: TestCollector = TestCollector {
compute_score: true,
@@ -34,11 +26,11 @@ pub fn test_filter_collector() -> crate::Result<()> {
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_with_num_threads(1, 10_000_000)?;
index_writer.add_document(doc!(title => "The Name of the Wind", price => 30_200u64, date => DateTime::from_str("1898-04-09T00:00:00+00:00").unwrap()))?;
index_writer.add_document(doc!(title => "The Diary of Muadib", price => 29_240u64, date => DateTime::from_str("2020-04-09T00:00:00+00:00").unwrap()))?;
index_writer.add_document(doc!(title => "The Diary of Anne Frank", price => 18_240u64, date => DateTime::from_str("2019-04-20T00:00:00+00:00").unwrap()))?;
index_writer.add_document(doc!(title => "A Dairy Cow", price => 21_240u64, date => DateTime::from_str("2019-04-09T00:00:00+00:00").unwrap()))?;
index_writer.add_document(doc!(title => "The Diary of a Young Girl", price => 20_120u64, date => DateTime::from_str("2018-04-09T00:00:00+00:00").unwrap()))?;
index_writer.add_document(doc!(title => "The Name of the Wind", price => 30_200u64, date => DateTime::new_utc(OffsetDateTime::parse("1898-04-09T00:00:00+00:00", &Rfc3339).unwrap())))?;
index_writer.add_document(doc!(title => "The Diary of Muadib", price => 29_240u64, date => DateTime::new_utc(OffsetDateTime::parse("2020-04-09T00:00:00+00:00", &Rfc3339).unwrap())))?;
index_writer.add_document(doc!(title => "The Diary of Anne Frank", price => 18_240u64, date => DateTime::new_utc(OffsetDateTime::parse("2019-04-20T00:00:00+00:00", &Rfc3339).unwrap())))?;
index_writer.add_document(doc!(title => "A Dairy Cow", price => 21_240u64, date => DateTime::new_utc(OffsetDateTime::parse("2019-04-09T00:00:00+00:00", &Rfc3339).unwrap())))?;
index_writer.add_document(doc!(title => "The Diary of a Young Girl", price => 20_120u64, date => DateTime::new_utc(OffsetDateTime::parse("2018-04-09T00:00:00+00:00", &Rfc3339).unwrap())))?;
index_writer.commit()?;
let reader = index.reader()?;
@@ -63,7 +55,9 @@ pub fn test_filter_collector() -> crate::Result<()> {
assert_eq!(filtered_top_docs.len(), 0);
fn date_filter(value: DateTime) -> bool {
(value - DateTime::from_str("2019-04-09T00:00:00+00:00").unwrap()).num_weeks() > 0
(value.to_utc() - OffsetDateTime::parse("2019-04-09T00:00:00+00:00", &Rfc3339).unwrap())
.whole_weeks()
> 0
}
let filter_dates_collector = FilterCollector::new(date, &date_filter, TopDocs::with_limit(5));

17
src/collector/top_collector.rs
View File

@@ -1,11 +1,9 @@
use crate::DocAddress;
use crate::DocId;
use crate::SegmentOrdinal;
use crate::SegmentReader;
use std::cmp::Ordering;
use std::collections::BinaryHeap;
use std::marker::PhantomData;
use crate::{DocAddress, DocId, SegmentOrdinal, SegmentReader};
/// Contains a feature (field, score, etc.) of a document along with the document address.
///
/// It has a custom implementation of `PartialOrd` that reverses the order. This is because the
@@ -62,8 +60,7 @@ pub(crate) struct TopCollector<T> {
}
impl<T> TopCollector<T>
where
T: PartialOrd + Clone,
where T: PartialOrd + Clone
{
/// Creates a top collector, with a number of documents equal to "limit".
///
@@ -176,8 +173,7 @@ impl<T: PartialOrd + Clone> TopSegmentCollector<T> {
.collect()
}
/// Return true iff at least K documents have gone through
/// the collector.
/// Return true if more documents have been collected than the limit.
#[inline]
pub(crate) fn at_capacity(&self) -> bool {
self.heap.len() >= self.limit
@@ -253,7 +249,7 @@ mod tests {
// when harvesting we have to guarantee stable sorting in case of a tie
// on the score
let doc_ids_collection = [4, 5, 6];
let score = 3.14;
let score = 3.3f32;
let mut top_collector_limit_2 = TopSegmentCollector::new(0, 2);
for id in &doc_ids_collection {
@@ -322,9 +318,10 @@ mod tests {
#[cfg(all(test, feature = "unstable"))]
mod bench {
use super::TopSegmentCollector;
use test::Bencher;
use super::TopSegmentCollector;
#[bench]
fn bench_top_segment_collector_collect_not_at_capacity(b: &mut Bencher) {
let mut top_collector = TopSegmentCollector::new(0, 400);

62
src/collector/top_score_collector.rs
View File

@@ -1,21 +1,18 @@
use std::collections::BinaryHeap;
use std::fmt;
use std::marker::PhantomData;
use super::Collector;
use crate::collector::top_collector::{ComparableDoc, TopCollector};
use crate::collector::custom_score_top_collector::CustomScoreTopCollector;
use crate::collector::top_collector::{ComparableDoc, TopCollector, TopSegmentCollector};
use crate::collector::tweak_score_top_collector::TweakedScoreTopCollector;
use crate::collector::{
CustomScorer, CustomSegmentScorer, ScoreSegmentTweaker, ScoreTweaker, SegmentCollector,
};
use crate::fastfield::{DynamicFastFieldReader, FastFieldReader};
use crate::fastfield::{DynamicFastFieldReader, FastFieldReader, FastValue};
use crate::query::Weight;
use crate::schema::Field;
use crate::DocAddress;
use crate::DocId;
use crate::Score;
use crate::SegmentOrdinal;
use crate::SegmentReader;
use crate::{collector::custom_score_top_collector::CustomScoreTopCollector, fastfield::FastValue};
use crate::{collector::top_collector::TopSegmentCollector, TantivyError};
use std::fmt;
use std::{collections::BinaryHeap, marker::PhantomData};
use crate::{DocAddress, DocId, Score, SegmentOrdinal, SegmentReader, TantivyError};
struct FastFieldConvertCollector<
TCollector: Collector<Fruit = Vec<(u64, DocAddress)>>,
@@ -217,11 +214,12 @@ impl TopDocs {
/// Set top-K to rank documents by a given fast field.
///
/// If the field is not a fast or does not exist, this method returns successfully (it is not aware of any schema).
/// An error will be returned at the moment of search.
/// If the field is not a fast or does not exist, this method returns successfully (it is not
/// aware of any schema). An error will be returned at the moment of search.
///
/// If the field is a FAST field but not a u64 field, search will return successfully but it will return
/// returns a monotonic u64-representation (ie. the order is still correct) of the requested field type.
/// If the field is a FAST field but not a u64 field, search will return successfully but it
/// will return returns a monotonic u64-representation (ie. the order is still correct) of
/// the requested field type.
///
/// # Example
///
@@ -296,14 +294,15 @@ impl TopDocs {
/// Set top-K to rank documents by a given fast field.
///
/// If the field is not a fast field, or its field type does not match the generic type, this method does not panic,
/// but an explicit error will be returned at the moment of collection.
/// If the field is not a fast field, or its field type does not match the generic type, this
/// method does not panic, but an explicit error will be returned at the moment of
/// collection.
///
/// Note that this method is a generic. The requested fast field type will be often
/// inferred in your code by the rust compiler.
///
/// Implementation-wise, for performance reason, tantivy will manipulate the u64 representation of your fast
/// field until the last moment.
/// Implementation-wise, for performance reason, tantivy will manipulate the u64 representation
/// of your fast field until the last moment.
///
/// # Example
///
@@ -715,10 +714,9 @@ mod tests {
use crate::collector::Collector;
use crate::query::{AllQuery, Query, QueryParser};
use crate::schema::{Field, Schema, FAST, STORED, TEXT};
use crate::Index;
use crate::IndexWriter;
use crate::Score;
use crate::{DocAddress, DocId, SegmentReader};
use crate::time::format_description::well_known::Rfc3339;
use crate::time::OffsetDateTime;
use crate::{DateTime, DocAddress, DocId, Index, IndexWriter, Score, SegmentReader};
fn make_index() -> crate::Result<Index> {
let mut schema_builder = Schema::builder();
@@ -894,28 +892,32 @@ mod tests {
#[test]
fn test_top_field_collector_datetime() -> crate::Result<()> {
use std::str::FromStr;
let mut schema_builder = Schema::builder();
let name = schema_builder.add_text_field("name", TEXT);
let birthday = schema_builder.add_date_field("birthday", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests()?;
let pr_birthday = crate::DateTime::from_str("1898-04-09T00:00:00+00:00")?;
let pr_birthday = DateTime::new_utc(OffsetDateTime::parse(
"1898-04-09T00:00:00+00:00",
&Rfc3339,
)?);
index_writer.add_document(doc!(
name => "Paul Robeson",
birthday => pr_birthday
birthday => pr_birthday,
))?;
let mr_birthday = crate::DateTime::from_str("1947-11-08T00:00:00+00:00")?;
let mr_birthday = DateTime::new_utc(OffsetDateTime::parse(
"1947-11-08T00:00:00+00:00",
&Rfc3339,
)?);
index_writer.add_document(doc!(
name => "Minnie Riperton",
birthday => mr_birthday
birthday => mr_birthday,
))?;
index_writer.commit()?;
let searcher = index.reader()?.searcher();
let top_collector = TopDocs::with_limit(3).order_by_fast_field(birthday);
let top_docs: Vec<(crate::DateTime, DocAddress)> =
searcher.search(&AllQuery, &top_collector)?;
let top_docs: Vec<(DateTime, DocAddress)> = searcher.search(&AllQuery, &top_collector)?;
assert_eq!(
&top_docs[..],
&[

9
src/collector/tweak_score_top_collector.rs
View File

@@ -1,7 +1,6 @@
use crate::collector::top_collector::{TopCollector, TopSegmentCollector};
use crate::collector::{Collector, SegmentCollector};
use crate::DocAddress;
use crate::{DocId, Result, Score, SegmentReader};
use crate::{DocAddress, DocId, Result, Score, SegmentReader};
pub(crate) struct TweakedScoreTopCollector<TScoreTweaker, TScore = Score> {
score_tweaker: TScoreTweaker,
@@ -9,8 +8,7 @@ pub(crate) struct TweakedScoreTopCollector<TScoreTweaker, TScore = Score> {
}
impl<TScoreTweaker, TScore> TweakedScoreTopCollector<TScoreTweaker, TScore>
where
TScore: Clone + PartialOrd,
where TScore: Clone + PartialOrd
{
pub fn new(
score_tweaker: TScoreTweaker,
@@ -118,8 +116,7 @@ where
}
impl<F, TScore> ScoreSegmentTweaker<TScore> for F
where
F: 'static + FnMut(DocId, Score) -> TScore,
where F: 'static + FnMut(DocId, Score) -> TScore
{
fn score(&mut self, doc: DocId, score: Score) -> TScore {
(self)(doc, score)

6
src/core/executor.rs
View File

@@ -57,7 +57,11 @@ impl Executor {
let (idx, arg) = arg_with_idx;
let fruit = f(arg);
if let Err(err) = fruit_sender.send((idx, fruit)) {
error!("Failed to send search task. It probably means all search threads have panicked. {:?}", err);
error!(
"Failed to send search task. It probably means all search \
threads have panicked. {:?}",
err
);
}
});
}

123
src/core/index.rs
View File

@@ -1,35 +1,27 @@
use super::{segment::Segment, IndexSettings};
use crate::core::Executor;
use crate::core::IndexMeta;
use crate::core::SegmentId;
use crate::core::SegmentMeta;
use crate::core::SegmentMetaInventory;
use crate::core::META_FILEPATH;
use crate::directory::error::OpenReadError;
use crate::directory::ManagedDirectory;
#[cfg(feature = "mmap")]
use crate::directory::MmapDirectory;
use crate::directory::INDEX_WRITER_LOCK;
use crate::directory::{Directory, RamDirectory};
use crate::error::DataCorruption;
use crate::error::TantivyError;
use crate::indexer::index_writer::{HEAP_SIZE_MIN, MAX_NUM_THREAD};
use crate::indexer::segment_updater::save_new_metas;
use crate::reader::IndexReader;
use crate::reader::IndexReaderBuilder;
use crate::schema::Field;
use crate::schema::FieldType;
use crate::schema::Schema;
use crate::tokenizer::{TextAnalyzer, TokenizerManager};
use crate::IndexWriter;
use std::collections::HashSet;
use std::fmt;
#[cfg(feature = "mmap")]
use std::path::Path;
use std::path::PathBuf;
use std::sync::Arc;
use super::segment::Segment;
use super::IndexSettings;
use crate::core::{
Executor, IndexMeta, SegmentId, SegmentMeta, SegmentMetaInventory, META_FILEPATH,
};
use crate::directory::error::OpenReadError;
#[cfg(feature = "mmap")]
use crate::directory::MmapDirectory;
use crate::directory::{Directory, ManagedDirectory, RamDirectory, INDEX_WRITER_LOCK};
use crate::error::{DataCorruption, TantivyError};
use crate::indexer::index_writer::{MAX_NUM_THREAD, MEMORY_ARENA_NUM_BYTES_MIN};
use crate::indexer::segment_updater::save_new_metas;
use crate::reader::{IndexReader, IndexReaderBuilder};
use crate::schema::{Field, FieldType, Schema};
use crate::tokenizer::{TextAnalyzer, TokenizerManager};
use crate::IndexWriter;
fn load_metas(
directory: &dyn Directory,
inventory: &SegmentMetaInventory,
@@ -72,13 +64,12 @@ fn load_metas(
/// let body_field = schema_builder.add_text_field("body", TEXT);
/// let number_field = schema_builder.add_u64_field(
/// "number",
/// IntOptions::default().set_fast(Cardinality::SingleValue),
/// NumericOptions::default().set_fast(Cardinality::SingleValue),
/// );
///
/// let schema = schema_builder.build();
/// let settings = IndexSettings{sort_by_field: Some(IndexSortByField{field:"number".to_string(), order:Order::Asc}), ..Default::default()};
/// let index = Index::builder().schema(schema).settings(settings).create_in_ram();
///
/// ```
pub struct IndexBuilder {
schema: Option<Schema>,
@@ -97,16 +88,21 @@ impl IndexBuilder {
index_settings: IndexSettings::default(),
}
}
/// Set the settings
#[must_use]
pub fn settings(mut self, settings: IndexSettings) -> Self {
self.index_settings = settings;
self
}
/// Set the schema
#[must_use]
pub fn schema(mut self, schema: Schema) -> Self {
self.schema = Some(schema);
self
}
/// Creates a new index using the `RAMDirectory`.
///
/// The index will be allocated in anonymous memory.
@@ -117,6 +113,7 @@ impl IndexBuilder {
.create(ram_directory)
.expect("Creating a RAMDirectory should never fail"))
}
/// Creates a new index in a given filepath.
/// The index will use the `MMapDirectory`.
///
@@ -129,6 +126,7 @@ impl IndexBuilder {
}
self.create(mmap_directory)
}
/// Creates a new index in a temp directory.
///
/// The index will use the `MMapDirectory` in a newly created directory.
@@ -142,12 +140,14 @@ impl IndexBuilder {
let mmap_directory: Box<dyn Directory> = Box::new(MmapDirectory::create_from_tempdir()?);
self.create(mmap_directory)
}
fn get_expect_schema(&self) -> crate::Result<Schema> {
self.schema
.as_ref()
.cloned()
.ok_or(TantivyError::IndexBuilderMissingArgument("schema"))
}
/// Opens or creates a new index in the provided directory
pub fn open_or_create<T: Into<Box<dyn Directory>>>(self, dir: T) -> crate::Result<Index> {
let dir = dir.into();
@@ -397,17 +397,18 @@ impl Index {
/// - `num_threads` defines the number of indexing workers that
/// should work at the same time.
///
/// - `overall_heap_size_in_bytes` sets the amount of memory
/// - `overall_memory_arena_in_bytes` sets the amount of memory
/// allocated for all indexing thread.
/// Each thread will receive a budget of `overall_heap_size_in_bytes / num_threads`.
/// Each thread will receive a budget of `overall_memory_arena_in_bytes / num_threads`.
///
/// # Errors
/// If the lockfile already exists, returns `Error::DirectoryLockBusy` or an `Error::IoError`.
/// If the heap size per thread is too small or too big, returns `TantivyError::InvalidArgument`
/// If the memory arena per thread is too small or too big, returns
/// `TantivyError::InvalidArgument`
pub fn writer_with_num_threads(
&self,
num_threads: usize,
overall_heap_size_in_bytes: usize,
overall_memory_arena_in_bytes: usize,
) -> crate::Result<IndexWriter> {
let directory_lock = self
.directory
@@ -416,26 +417,25 @@ impl Index {
TantivyError::LockFailure(
err,
Some(
"Failed to acquire index lock. If you are using \
a regular directory, this means there is already an \
`IndexWriter` working on this `Directory`, in this process \
or in a different process."
"Failed to acquire index lock. If you are using a regular directory, this \
means there is already an `IndexWriter` working on this `Directory`, in \
this process or in a different process."
.to_string(),
),
)
})?;
let heap_size_in_bytes_per_thread = overall_heap_size_in_bytes / num_threads;
let memory_arena_in_bytes_per_thread = overall_memory_arena_in_bytes / num_threads;
IndexWriter::new(
self,
num_threads,
heap_size_in_bytes_per_thread,
memory_arena_in_bytes_per_thread,
directory_lock,
)
}
/// Helper to create an index writer for tests.
///
/// That index writer only simply has a single thread and a heap of 10 MB.
/// That index writer only simply has a single thread and a memory arena of 10 MB.
/// Using a single thread gives us a deterministic allocation of DocId.
#[cfg(test)]
pub fn writer_for_tests(&self) -> crate::Result<IndexWriter> {
@@ -446,29 +446,28 @@ impl Index {
///
/// Tantivy will automatically define the number of threads to use, but
/// no more than 8 threads.
/// `overall_heap_size_in_bytes` is the total target memory usage that will be split
/// `overall_memory_arena_in_bytes` is the total target memory usage that will be split
/// between a given number of threads.
///
/// # Errors
/// If the lockfile already exists, returns `Error::FileAlreadyExists`.
/// If the heap size per thread is too small or too big, returns `TantivyError::InvalidArgument`
pub fn writer(&self, overall_heap_size_in_bytes: usize) -> crate::Result<IndexWriter> {
/// If the memory arena per thread is too small or too big, returns
/// `TantivyError::InvalidArgument`
pub fn writer(&self, memory_arena_num_bytes: usize) -> crate::Result<IndexWriter> {
let mut num_threads = std::cmp::min(num_cpus::get(), MAX_NUM_THREAD);
let heap_size_in_bytes_per_thread = overall_heap_size_in_bytes / num_threads;
if heap_size_in_bytes_per_thread < HEAP_SIZE_MIN {
num_threads = (overall_heap_size_in_bytes / HEAP_SIZE_MIN).max(1);
let memory_arena_num_bytes_per_thread = memory_arena_num_bytes / num_threads;
if memory_arena_num_bytes_per_thread < MEMORY_ARENA_NUM_BYTES_MIN {
num_threads = (memory_arena_num_bytes / MEMORY_ARENA_NUM_BYTES_MIN).max(1);
}
self.writer_with_num_threads(num_threads, overall_heap_size_in_bytes)
self.writer_with_num_threads(num_threads, memory_arena_num_bytes)
}
/// Accessor to the index settings
///
pub fn settings(&self) -> &IndexSettings {
&self.settings
}
/// Accessor to the index settings
///
pub fn settings_mut(&mut self) -> &mut IndexSettings {
&mut self.settings
}
@@ -556,15 +555,9 @@ impl fmt::Debug for Index {
#[cfg(test)]
mod tests {
use crate::schema::Field;
use crate::schema::{Schema, INDEXED, TEXT};
use crate::IndexReader;
use crate::ReloadPolicy;
use crate::{
directory::{RamDirectory, WatchCallback},
IndexSettings,
};
use crate::{Directory, Index};
use crate::directory::{RamDirectory, WatchCallback};
use crate::schema::{Field, Schema, INDEXED, TEXT};
use crate::{Directory, Index, IndexReader, IndexSettings, ReloadPolicy};
#[test]
fn test_indexer_for_field() {
@@ -673,10 +666,12 @@ mod tests {
#[cfg(feature = "mmap")]
mod mmap_specific {
use std::path::PathBuf;
use tempfile::TempDir;
use super::*;
use crate::Directory;
use std::path::PathBuf;
use tempfile::TempDir;
#[test]
fn test_index_on_commit_reload_policy_mmap() -> crate::Result<()> {
@@ -786,24 +781,24 @@ mod tests {
for i in 0u64..8_000u64 {
writer.add_document(doc!(field => i))?;
}
let (sender, receiver) = crossbeam::channel::unbounded();
let _handle = directory.watch(WatchCallback::new(move || {
let _ = sender.send(());
}));
writer.commit()?;
let mem_right_after_commit = directory.total_mem_usage();
assert!(receiver.recv().is_ok());
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::Manual)
.try_into()?;
assert_eq!(reader.searcher().num_docs(), 8_000);
assert_eq!(reader.searcher().segment_readers().len(), 8);
writer.wait_merging_threads()?;
let mem_right_after_merge_finished = directory.total_mem_usage();
reader.reload().unwrap();
let searcher = reader.searcher();
assert_eq!(searcher.segment_readers().len(), 1);
assert_eq!(searcher.num_docs(), 8_000);
assert!(
mem_right_after_merge_finished < mem_right_after_commit,

42
src/core/index_meta.rs
View File

@@ -1,12 +1,16 @@
use super::SegmentComponent;
use crate::schema::Schema;
use crate::Opstamp;
use crate::{core::SegmentId, store::Compressor};
use crate::{Inventory, TrackedObject};
use serde::{Deserialize, Serialize};
use std::collections::HashSet;
use std::fmt;
use std::path::PathBuf;
use std::{collections::HashSet, sync::atomic::AtomicBool};
use std::{fmt, sync::Arc};
use std::sync::atomic::AtomicBool;
use std::sync::Arc;
use serde::{Deserialize, Serialize};
use super::SegmentComponent;
use crate::core::SegmentId;
use crate::schema::Schema;
use crate::store::Compressor;
use crate::{Inventory, Opstamp, TrackedObject};
#[derive(Clone, Debug, Serialize, Deserialize)]
struct DeleteMeta {
@@ -188,6 +192,7 @@ impl SegmentMeta {
}
#[doc(hidden)]
#[must_use]
pub fn with_delete_meta(self, num_deleted_docs: u32, opstamp: Opstamp) -> SegmentMeta {
assert!(
num_deleted_docs <= self.max_doc(),
@@ -234,7 +239,7 @@ impl InnerSegmentMeta {
///
/// Contains settings which are applied on the whole
/// index, like presort documents.
#[derive(Clone, Default, Serialize, Deserialize, Eq, PartialEq)]
#[derive(Clone, Debug, Default, Serialize, Deserialize, Eq, PartialEq)]
pub struct IndexSettings {
/// Sorts the documents by information
/// provided in `IndexSortByField`
@@ -249,7 +254,7 @@ pub struct IndexSettings {
/// Presorting documents can greatly performance
/// in some scenarios, by applying top n
/// optimizations.
#[derive(Clone, Serialize, Deserialize, Eq, PartialEq)]
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub struct IndexSortByField {
/// The field to sort the documents by
pub field: String,
@@ -257,7 +262,7 @@ pub struct IndexSortByField {
pub order: Order,
}
/// The order to sort by
#[derive(Clone, Serialize, Deserialize, Eq, PartialEq)]
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub enum Order {
/// Ascending Order
Asc,
@@ -282,7 +287,6 @@ impl Order {
/// * the searchable segments,
/// * the index `docstamp`
/// * the schema
///
#[derive(Clone, Serialize)]
pub struct IndexMeta {
/// `IndexSettings` to configure index options.
@@ -294,12 +298,12 @@ pub struct IndexMeta {
pub schema: Schema,
/// Opstamp associated to the last `commit` operation.
pub opstamp: Opstamp,
#[serde(skip_serializing_if = "Option::is_none")]
/// Payload associated to the last commit.
///
/// Upon commit, clients can optionally add a small `String` payload to their commit
/// to help identify this commit.
/// This payload is entirely unused by tantivy.
#[serde(skip_serializing_if = "Option::is_none")]
pub payload: Option<String>,
}
@@ -370,10 +374,9 @@ impl fmt::Debug for IndexMeta {
mod tests {
use super::IndexMeta;
use crate::{
schema::{Schema, TEXT},
IndexSettings, IndexSortByField, Order,
};
use crate::core::index_meta::UntrackedIndexMeta;
use crate::schema::{Schema, TEXT};
use crate::{IndexSettings, IndexSortByField, Order};
#[test]
fn test_serialize_metas() {
@@ -400,5 +403,10 @@ mod tests {
json,
r#"{"index_settings":{"sort_by_field":{"field":"text","order":"Asc"},"docstore_compression":"lz4"},"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","fieldnorms":true,"tokenizer":"default"},"stored":false}}],"opstamp":0}"#
);
let deser_meta: UntrackedIndexMeta = serde_json::from_str(&json).unwrap();
assert_eq!(index_metas.index_settings, deser_meta.index_settings);
assert_eq!(index_metas.schema, deser_meta.schema);
assert_eq!(index_metas.opstamp, deser_meta.opstamp);
}
}

45
src/core/inverted_index_reader.rs
View File

@@ -1,13 +1,12 @@
use std::io;
use common::BinarySerializable;
use crate::directory::FileSlice;
use crate::positions::PositionReader;
use crate::postings::TermInfo;
use crate::postings::{BlockSegmentPostings, SegmentPostings};
use crate::schema::IndexRecordOption;
use crate::schema::Term;
use crate::postings::{BlockSegmentPostings, SegmentPostings, TermInfo};
use crate::schema::{IndexRecordOption, Term};
use crate::termdict::TermDictionary;
use common::BinarySerializable;
/// The inverted index reader is in charge of accessing
/// the inverted index associated to a specific field.
@@ -89,7 +88,8 @@ impl InvertedIndexReader {
let postings_slice = self
.postings_file_slice
.slice(term_info.postings_range.clone());
block_postings.reset(term_info.doc_freq, postings_slice.read_bytes()?);
let postings_bytes = postings_slice.read_bytes()?;
block_postings.reset(term_info.doc_freq, postings_bytes)?;
Ok(())
}
@@ -198,3 +198,36 @@ impl InvertedIndexReader {
.unwrap_or(0u32))
}
}
#[cfg(feature = "quickwit")]
impl InvertedIndexReader {
pub(crate) async fn get_term_info_async(
&self,
term: &Term,
) -> crate::AsyncIoResult<Option<TermInfo>> {
self.termdict.get_async(term.value_bytes()).await
}
/// Returns a block postings given a `Term`.
/// This method is for an advanced usage only.
///
/// Most user should prefer using `read_postings` instead.
pub async fn warm_postings(
&self,
term: &Term,
with_positions: bool,
) -> crate::AsyncIoResult<()> {
let term_info_opt = self.get_term_info_async(term).await?;
if let Some(term_info) = term_info_opt {
self.postings_file_slice
.read_bytes_slice_async(term_info.postings_range.clone())
.await?;
if with_positions {
self.positions_file_slice
.read_bytes_slice_async(term_info.positions_range.clone())
.await?;
}
}
Ok(())
}
}

7
src/core/mod.rs
View File

@@ -8,6 +8,10 @@ mod segment_component;
mod segment_id;
mod segment_reader;
use std::path::Path;
use once_cell::sync::Lazy;
pub use self::executor::Executor;
pub use self::index::{Index, IndexBuilder};
pub use self::index_meta::{
@@ -20,9 +24,6 @@ pub use self::segment_component::SegmentComponent;
pub use self::segment_id::SegmentId;
pub use self::segment_reader::SegmentReader;
use once_cell::sync::Lazy;
use std::path::Path;
/// The meta file contains all the information about the list of segments and the schema
/// of the index.
pub static META_FILEPATH: Lazy<&'static Path> = Lazy::new(|| Path::new("meta.json"));

31
src/core/searcher.rs
View File

@@ -1,21 +1,14 @@
use crate::collector::Collector;
use crate::core::Executor;
use crate::core::SegmentReader;
use crate::query::Query;
use crate::schema::Document;
use crate::schema::Schema;
use crate::schema::Term;
use crate::space_usage::SearcherSpaceUsage;
use crate::store::StoreReader;
use crate::DocAddress;
use crate::Index;
use crate::Opstamp;
use crate::SegmentId;
use crate::TrackedObject;
use std::collections::BTreeMap;
use std::{fmt, io};
use crate::collector::Collector;
use crate::core::{Executor, SegmentReader};
use crate::query::Query;
use crate::schema::{Document, Schema, Term};
use crate::space_usage::SearcherSpaceUsage;
use crate::store::StoreReader;
use crate::{DocAddress, Index, Opstamp, SegmentId, TrackedObject};
/// Identifies the searcher generation accessed by a [Searcher].
///
/// While this might seem redundant, a [SearcherGeneration] contains
@@ -69,7 +62,6 @@ impl SearcherGeneration {
///
/// It guarantees that the `Segment` will not be removed before
/// the destruction of the `Searcher`.
///
pub struct Searcher {
schema: Schema,
index: Index,
@@ -118,6 +110,13 @@ impl Searcher {
store_reader.get(doc_address.doc_id)
}
/// Fetches a document in an asynchronous manner.
#[cfg(feature = "quickwit")]
pub async fn doc_async(&self, doc_address: DocAddress) -> crate::Result<Document> {
let store_reader = &self.store_readers[doc_address.segment_ord as usize];
store_reader.get_async(doc_address.doc_id).await
}
/// Access the schema associated to the index of this searcher.
pub fn schema(&self) -> &Schema {
&self.schema

17
src/core/segment.rs
View File

@@ -1,15 +1,13 @@
use super::SegmentComponent;
use crate::core::Index;
use crate::core::SegmentId;
use crate::core::SegmentMeta;
use crate::directory::error::{OpenReadError, OpenWriteError};
use crate::directory::Directory;
use crate::directory::{FileSlice, WritePtr};
use crate::schema::Schema;
use crate::Opstamp;
use std::fmt;
use std::path::PathBuf;
use super::SegmentComponent;
use crate::core::{Index, SegmentId, SegmentMeta};
use crate::directory::error::{OpenReadError, OpenWriteError};
use crate::directory::{Directory, FileSlice, WritePtr};
use crate::schema::Schema;
use crate::Opstamp;
/// A segment is a piece of the index.
#[derive(Clone)]
pub struct Segment {
@@ -56,6 +54,7 @@ impl Segment {
}
#[doc(hidden)]
#[must_use]
pub fn with_delete_meta(self, num_deleted_docs: u32, opstamp: Opstamp) -> Segment {
Segment {
index: self.index,

10
src/core/segment_id.rs
View File

@@ -1,14 +1,14 @@
use std::cmp::{Ord, Ordering};
use std::error::Error;
use std::fmt;
use uuid::Uuid;
use std::str::FromStr;
#[cfg(test)]
use std::sync::atomic;
#[cfg(test)]
use once_cell::sync::Lazy;
use serde::{Deserialize, Serialize};
use std::error::Error;
use std::str::FromStr;
#[cfg(test)]
use std::sync::atomic;
use uuid::Uuid;
/// Uuid identifying a segment.
///

66
src/core/segment_reader.rs
View File

@@ -1,28 +1,19 @@
use crate::core::InvertedIndexReader;
use crate::core::Segment;
use crate::core::SegmentComponent;
use crate::core::SegmentId;
use crate::directory::CompositeFile;
use crate::directory::FileSlice;
use std::collections::HashMap;
use std::sync::{Arc, RwLock};
use std::{fmt, io};
use fail::fail_point;
use crate::core::{InvertedIndexReader, Segment, SegmentComponent, SegmentId};
use crate::directory::{CompositeFile, FileSlice};
use crate::error::DataCorruption;
use crate::fastfield::intersect_alive_bitsets;
use crate::fastfield::AliveBitSet;
use crate::fastfield::FacetReader;
use crate::fastfield::FastFieldReaders;
use crate::fastfield::{intersect_alive_bitsets, AliveBitSet, FacetReader, FastFieldReaders};
use crate::fieldnorm::{FieldNormReader, FieldNormReaders};
use crate::schema::FieldType;
use crate::schema::Schema;
use crate::schema::{Field, IndexRecordOption};
use crate::schema::{Field, FieldType, IndexRecordOption, Schema};
use crate::space_usage::SegmentSpaceUsage;
use crate::store::StoreReader;
use crate::termdict::TermDictionary;
use crate::DocId;
use crate::Opstamp;
use fail::fail_point;
use std::fmt;
use std::sync::Arc;
use std::sync::RwLock;
use std::{collections::HashMap, io};
use crate::{DocId, Opstamp};
/// Entry point to access all of the datastructures of the `Segment`
///
@@ -79,7 +70,7 @@ impl SegmentReader {
self.max_doc - self.num_docs
}
/// Returns true iff some of the documents of the segment have been deleted.
/// Returns true if some of the documents of the segment have been deleted.
pub fn has_deletes(&self) -> bool {
self.num_deleted_docs() > 0
}
@@ -130,8 +121,8 @@ impl SegmentReader {
self.fieldnorm_readers.get_field(field)?.ok_or_else(|| {
let field_name = self.schema.get_field_name(field);
let err_msg = format!(
"Field norm not found for field {:?}. Was the field set to record norm during indexing?",
field_name
"Field norm not found for field {field_name:?}. Was the field set to record norm \
during indexing?"
);
crate::TantivyError::SchemaError(err_msg)
})
@@ -259,19 +250,24 @@ impl SegmentReader {
let record_option = record_option_opt.unwrap();
let postings_file = postings_file_opt.unwrap();
let termdict_file: FileSlice = self.termdict_composite.open_read(field)
.ok_or_else(||
DataCorruption::comment_only(format!("Failed to open field {:?}'s term dictionary in the composite file. Has the schema been modified?", field_entry.name()))
)?;
let positions_file = self
.positions_composite
.open_read(field)
.ok_or_else(|| {
let error_msg = format!("Failed to open field {:?}'s positions in the composite file. Has the schema been modified?", field_entry.name());
DataCorruption::comment_only(error_msg)
let termdict_file: FileSlice =
self.termdict_composite.open_read(field).ok_or_else(|| {
DataCorruption::comment_only(format!(
"Failed to open field {:?}'s term dictionary in the composite file. Has the \
schema been modified?",
field_entry.name()
))
})?;
let positions_file = self.positions_composite.open_read(field).ok_or_else(|| {
let error_msg = format!(
"Failed to open field {:?}'s positions in the composite file. Has the schema been \
modified?",
field_entry.name()
);
DataCorruption::comment_only(error_msg)
})?;
let inv_idx_reader = Arc::new(InvertedIndexReader::new(
TermDictionary::open(termdict_file)?,
postings_file,
@@ -305,7 +301,7 @@ impl SegmentReader {
self.alive_bitset_opt.as_ref()
}
/// Returns true iff the `doc` is marked
/// Returns true if the `doc` is marked
/// as deleted.
pub fn is_deleted(&self, doc: DocId) -> bool {
self.alive_bitset()

24
src/directory/composite_file.rs
View File

@@ -1,17 +1,14 @@
use crate::directory::FileSlice;
use crate::directory::{TerminatingWrite, WritePtr};
use crate::schema::Field;
use crate::space_usage::FieldUsage;
use crate::space_usage::PerFieldSpaceUsage;
use common::BinarySerializable;
use common::CountingWriter;
use common::HasLen;
use common::VInt;
use std::collections::HashMap;
use std::io::{self, Read, Write};
use std::iter::ExactSizeIterator;
use std::ops::Range;
use common::{BinarySerializable, CountingWriter, HasLen, VInt};
use crate::directory::{FileSlice, TerminatingWrite, WritePtr};
use crate::schema::Field;
use crate::space_usage::{FieldUsage, PerFieldSpaceUsage};
#[derive(Eq, PartialEq, Hash, Copy, Ord, PartialOrd, Clone, Debug)]
pub struct FileAddr {
field: Field,
@@ -186,13 +183,14 @@ impl CompositeFile {
#[cfg(test)]
mod test {
use std::io::Write;
use std::path::Path;
use common::{BinarySerializable, VInt};
use super::{CompositeFile, CompositeWrite};
use crate::directory::{Directory, RamDirectory};
use crate::schema::Field;
use common::BinarySerializable;
use common::VInt;
use std::io::Write;
use std::path::Path;
#[test]
fn test_composite_file() -> crate::Result<()> {

29
src/directory/directory.rs
View File

@@ -1,18 +1,12 @@
use crate::directory::directory_lock::Lock;
use crate::directory::error::LockError;
use crate::directory::error::{DeleteError, OpenReadError, OpenWriteError};
use crate::directory::WatchHandle;
use crate::directory::{FileHandle, WatchCallback};
use crate::directory::{FileSlice, WritePtr};
use std::fmt;
use std::io;
use std::io::Write;
use std::marker::Send;
use std::marker::Sync;
use std::path::Path;
use std::path::PathBuf;
use std::thread;
use std::marker::{Send, Sync};
use std::path::{Path, PathBuf};
use std::time::Duration;
use std::{fmt, io, thread};
use crate::directory::directory_lock::Lock;
use crate::directory::error::{DeleteError, LockError, OpenReadError, OpenWriteError};
use crate::directory::{FileHandle, FileSlice, WatchCallback, WatchHandle, WritePtr};
/// Retry the logic of acquiring locks is pretty simple.
/// We just retry `n` times after a given `duratio`, both
@@ -102,9 +96,9 @@ fn retry_policy(is_blocking: bool) -> RetryPolicy {
///
/// There are currently two implementations of `Directory`
///
/// - The [`MMapDirectory`](struct.MmapDirectory.html), this
/// - The [`MMapDirectory`][crate::directory::MmapDirectory], this
/// should be your default choice.
/// - The [`RamDirectory`](struct.RamDirectory.html), which
/// - The [`RamDirectory`][crate::directory::RamDirectory], which
/// should be used mostly for tests.
pub trait Directory: DirectoryClone + fmt::Debug + Send + Sync + 'static {
/// Opens a file and returns a boxed `FileHandle`.
@@ -134,7 +128,7 @@ pub trait Directory: DirectoryClone + fmt::Debug + Send + Sync + 'static {
/// `DeleteError::DoesNotExist`.
fn delete(&self, path: &Path) -> Result<(), DeleteError>;
/// Returns true iff the file exists
/// Returns true if and only if the file exists
fn exists(&self, path: &Path) -> Result<bool, OpenReadError>;
/// Opens a writer for the *virtual file* associated with
@@ -233,8 +227,7 @@ pub trait DirectoryClone {
}
impl<T> DirectoryClone for T
where
T: 'static + Directory + Clone,
where T: 'static + Directory + Clone
{
fn box_clone(&self) -> Box<dyn Directory> {
Box::new(self.clone())

4
src/directory/directory_lock.rs
View File

@@ -1,6 +1,7 @@
use once_cell::sync::Lazy;
use std::path::PathBuf;
use once_cell::sync::Lazy;
/// A directory lock.
///
/// A lock is associated to a specific path and some
@@ -11,7 +12,6 @@ use std::path::PathBuf;
/// - [META_LOCK]
///
/// Check out these locks documentation for more information.
///
#[derive(Debug)]
pub struct Lock {
/// The lock needs to be associated with its own file `path`.

12
src/directory/error.rs
View File

@@ -1,15 +1,17 @@
use crate::Version;
use std::fmt;
use std::io;
use std::path::PathBuf;
use std::{fmt, io};
use crate::Version;
/// Error while trying to acquire a directory lock.
#[derive(Debug, Error)]
pub enum LockError {
/// Failed to acquired a lock as it is already held by another
/// client.
/// - In the context of a blocking lock, this means the lock was not released within some `timeout` period.
/// - In the context of a non-blocking lock, this means the lock was busy at the moment of the call.
/// - In the context of a blocking lock, this means the lock was not released within some
/// `timeout` period.
/// - In the context of a non-blocking lock, this means the lock was busy at the moment of the
/// call.
#[error("Could not acquire lock as it is already held, possibly by a different process.")]
LockBusy,
/// Trying to acquire a lock failed with an `IoError`

84
src/directory/file_slice.rs
View File

@@ -1,11 +1,12 @@
use std::ops::{Deref, Range};
use std::sync::{Arc, Weak};
use std::{fmt, io};
use async_trait::async_trait;
use common::HasLen;
use stable_deref_trait::StableDeref;
use crate::directory::OwnedBytes;
use common::HasLen;
use std::fmt;
use std::ops::Range;
use std::sync::{Arc, Weak};
use std::{io, ops::Deref};
pub type ArcBytes = Arc<dyn Deref<Target = [u8]> + Send + Sync + 'static>;
pub type WeakArcBytes = Weak<dyn Deref<Target = [u8]> + Send + Sync + 'static>;
@@ -18,23 +19,39 @@ pub type WeakArcBytes = Weak<dyn Deref<Target = [u8]> + Send + Sync + 'static>;
/// The underlying behavior is therefore specific to the `Directory` that created it.
/// Despite its name, a `FileSlice` may or may not directly map to an actual file
/// on the filesystem.
#[async_trait]
pub trait FileHandle: 'static + Send + Sync + HasLen + fmt::Debug {
/// Reads a slice of bytes.
///
/// This method may panic if the range requested is invalid.
fn read_bytes(&self, range: Range<usize>) -> io::Result<OwnedBytes>;
#[cfg(feature = "quickwit")]
#[doc(hidden)]
async fn read_bytes_async(
&self,
_byte_range: Range<usize>,
) -> crate::AsyncIoResult<OwnedBytes> {
Err(crate::error::AsyncIoError::AsyncUnsupported)
}
}
#[async_trait]
impl FileHandle for &'static [u8] {
fn read_bytes(&self, range: Range<usize>) -> io::Result<OwnedBytes> {
let bytes = &self[range];
Ok(OwnedBytes::new(bytes))
}
#[cfg(feature = "quickwit")]
async fn read_bytes_async(&self, byte_range: Range<usize>) -> crate::AsyncIoResult<OwnedBytes> {
Ok(self.read_bytes(byte_range)?)
}
}
impl<B> From<B> for FileSlice
where
B: StableDeref + Deref<Target = [u8]> + 'static + Send + Sync,
where B: StableDeref + Deref<Target = [u8]> + 'static + Send + Sync
{
fn from(bytes: B) -> FileSlice {
FileSlice::new(Box::new(OwnedBytes::new(bytes)))
@@ -44,7 +61,6 @@ where
/// Logical slice of read only file in tantivy.
///
/// It can be cloned and sliced cheaply.
///
#[derive(Clone)]
pub struct FileSlice {
data: Arc<dyn FileHandle>,
@@ -79,6 +95,7 @@ impl FileSlice {
/// # Panics
///
/// Panics if `byte_range.end` exceeds the filesize.
#[must_use]
pub fn slice(&self, byte_range: Range<usize>) -> FileSlice {
assert!(byte_range.end <= self.len());
FileSlice {
@@ -103,6 +120,12 @@ impl FileSlice {
self.data.read_bytes(self.range.clone())
}
#[cfg(feature = "quickwit")]
#[doc(hidden)]
pub async fn read_bytes_async(&self) -> crate::AsyncIoResult<OwnedBytes> {
self.data.read_bytes_async(self.range.clone()).await
}
/// Reads a specific slice of data.
///
/// This is equivalent to running `file_slice.slice(from, to).read_bytes()`.
@@ -117,6 +140,23 @@ impl FileSlice {
.read_bytes(self.range.start + range.start..self.range.start + range.end)
}
#[cfg(feature = "quickwit")]
#[doc(hidden)]
pub async fn read_bytes_slice_async(
&self,
byte_range: Range<usize>,
) -> crate::AsyncIoResult<OwnedBytes> {
assert!(
self.range.start + byte_range.end <= self.range.end,
"`to` exceeds the fileslice length"
);
self.data
.read_bytes_async(
self.range.start + byte_range.start..self.range.start + byte_range.end,
)
.await
}
/// Splits the FileSlice at the given offset and return two file slices.
/// `file_slice[..split_offset]` and `file_slice[split_offset..]`.
///
@@ -138,6 +178,7 @@ impl FileSlice {
/// boundary.
///
/// Equivalent to `.slice(from_offset, self.len())`
#[must_use]
pub fn slice_from(&self, from_offset: usize) -> FileSlice {
self.slice(from_offset..self.len())
}
@@ -145,6 +186,7 @@ impl FileSlice {
/// Returns a slice from the end.
///
/// Equivalent to `.slice(self.len() - from_offset, self.len())`
#[must_use]
pub fn slice_from_end(&self, from_offset: usize) -> FileSlice {
self.slice(self.len() - from_offset..self.len())
}
@@ -153,15 +195,22 @@ impl FileSlice {
/// boundary.
///
/// Equivalent to `.slice(0, to_offset)`
#[must_use]
pub fn slice_to(&self, to_offset: usize) -> FileSlice {
self.slice(0..to_offset)
}
}
#[async_trait]
impl FileHandle for FileSlice {
fn read_bytes(&self, range: Range<usize>) -> io::Result<OwnedBytes> {
self.read_bytes_slice(range)
}
#[cfg(feature = "quickwit")]
async fn read_bytes_async(&self, byte_range: Range<usize>) -> crate::AsyncIoResult<OwnedBytes> {
self.read_bytes_slice_async(byte_range).await
}
}
impl HasLen for FileSlice {
@@ -170,12 +219,27 @@ impl HasLen for FileSlice {
}
}
#[async_trait]
impl FileHandle for OwnedBytes {
fn read_bytes(&self, range: Range<usize>) -> io::Result<OwnedBytes> {
Ok(self.slice(range))
}
#[cfg(feature = "quickwit")]
async fn read_bytes_async(&self, range: Range<usize>) -> crate::AsyncIoResult<OwnedBytes> {
let bytes = self.read_bytes(range)?;
Ok(bytes)
}
}
#[cfg(test)]
mod tests {
use super::{FileHandle, FileSlice};
use common::HasLen;
use std::io;
use common::HasLen;
use super::{FileHandle, FileSlice};
#[test]
fn test_file_slice() -> io::Result<()> {
let file_slice = FileSlice::new(Box::new(b"abcdef".as_ref()));

17
src/directory/file_watcher.rs
View File

@@ -1,13 +1,13 @@
use crate::directory::{WatchCallback, WatchCallbackList, WatchHandle};
use crc32fast::Hasher;
use std::fs;
use std::io;
use std::io::BufRead;
use std::path::Path;
use std::sync::atomic::{AtomicUsize, Ordering};
use std::sync::Arc;
use std::thread;
use std::time::Duration;
use std::{fs, io, thread};
use crc32fast::Hasher;
use crate::directory::{WatchCallback, WatchCallbackList, WatchHandle};
pub const POLLING_INTERVAL: Duration = Duration::from_millis(if cfg!(test) { 1 } else { 500 });
@@ -53,7 +53,9 @@ impl FileWatcher {
if metafile_has_changed {
info!("Meta file {:?} was modified", path);
current_checksum_opt = Some(checksum);
futures::executor::block_on(callbacks.broadcast());
// We actually ignore callbacks failing here.
// We just wait for the end of their execution.
let _ = callbacks.broadcast().wait();
}
}
@@ -99,9 +101,8 @@ mod tests {
use std::mem;
use crate::directory::mmap_directory::atomic_write;
use super::*;
use crate::directory::mmap_directory::atomic_write;
#[test]
fn test_file_watcher_drop_watcher() -> crate::Result<()> {

39
src/directory/footer.rs
View File

@@ -1,14 +1,13 @@
use crate::directory::error::Incompatibility;
use crate::directory::FileSlice;
use crate::{
directory::{AntiCallToken, TerminatingWrite},
Version, INDEX_FORMAT_VERSION,
};
use std::io;
use std::io::Write;
use common::{BinarySerializable, CountingWriter, DeserializeFrom, FixedSize, HasLen};
use crc32fast::Hasher;
use serde::{Deserialize, Serialize};
use std::io;
use std::io::Write;
use crate::directory::error::Incompatibility;
use crate::directory::{AntiCallToken, FileSlice, TerminatingWrite};
use crate::{Version, INDEX_FORMAT_VERSION};
const FOOTER_MAX_LEN: u32 = 50_000;
@@ -64,7 +63,9 @@ impl Footer {
if footer_magic_byte != FOOTER_MAGIC_NUMBER {
return Err(io::Error::new(
io::ErrorKind::InvalidData,
"Footer magic byte mismatch. File corrupted or index was created using old an tantivy version which is not supported anymore. Please use tantivy 0.15 or above to recreate the index.",
"Footer magic byte mismatch. File corrupted or index was created using old an \
tantivy version which is not supported anymore. Please use tantivy 0.15 or above \
to recreate the index.",
));
}
@@ -73,7 +74,7 @@ impl Footer {
io::ErrorKind::InvalidData,
format!(
"Footer seems invalid as it suggests a footer len of {}. File is corrupted, \
or the index was created with a different & old version of tantivy.",
or the index was created with a different & old version of tantivy.",
footer_len
),
));
@@ -154,12 +155,13 @@ impl<W: TerminatingWrite> TerminatingWrite for FooterProxy<W> {
#[cfg(test)]
mod tests {
use crate::directory::footer::Footer;
use crate::directory::OwnedBytes;
use crate::directory::{footer::FOOTER_MAGIC_NUMBER, FileSlice};
use common::BinarySerializable;
use std::io;
use common::BinarySerializable;
use crate::directory::footer::{Footer, FOOTER_MAGIC_NUMBER};
use crate::directory::{FileSlice, OwnedBytes};
#[test]
fn test_deserialize_footer() {
let mut buf: Vec<u8> = vec![];
@@ -183,8 +185,9 @@ mod tests {
let err = Footer::extract_footer(fileslice).unwrap_err();
assert_eq!(
err.to_string(),
"Footer magic byte mismatch. File corrupted or index was created using old an tantivy version which \
is not supported anymore. Please use tantivy 0.15 or above to recreate the index."
"Footer magic byte mismatch. File corrupted or index was created using old an tantivy \
version which is not supported anymore. Please use tantivy 0.15 or above to recreate \
the index."
);
}
#[test]
@@ -219,8 +222,8 @@ mod tests {
assert_eq!(err.kind(), io::ErrorKind::InvalidData);
assert_eq!(
err.to_string(),
"Footer seems invalid as it suggests a footer len of 50001. File is corrupted, \
or the index was created with a different & old version of tantivy."
"Footer seems invalid as it suggests a footer len of 50001. File is corrupted, or the \
index was created with a different & old version of tantivy."
);
}
}

33
src/directory/managed_directory.rs
View File

@@ -1,25 +1,22 @@
use std::collections::HashSet;
use std::io::Write;
use std::path::{Path, PathBuf};
use std::sync::{Arc, RwLock, RwLockWriteGuard};
use std::{io, result};
use crc32fast::Hasher;
use crate::core::MANAGED_FILEPATH;
use crate::directory::error::{DeleteError, LockError, OpenReadError, OpenWriteError};
use crate::directory::footer::{Footer, FooterProxy};
use crate::directory::GarbageCollectionResult;
use crate::directory::Lock;
use crate::directory::META_LOCK;
use crate::directory::{DirectoryLock, FileHandle};
use crate::directory::{FileSlice, WritePtr};
use crate::directory::{WatchCallback, WatchHandle};
use crate::directory::{
DirectoryLock, FileHandle, FileSlice, GarbageCollectionResult, Lock, WatchCallback,
WatchHandle, WritePtr, META_LOCK,
};
use crate::error::DataCorruption;
use crate::Directory;
use crc32fast::Hasher;
use std::collections::HashSet;
use std::io;
use std::io::Write;
use std::path::{Path, PathBuf};
use std::result;
use std::sync::RwLockWriteGuard;
use std::sync::{Arc, RwLock};
/// Returns true iff the file is "managed".
/// Returns true if the file is "managed".
/// Non-managed file are not subject to garbage collection.
///
/// Filenames that starts by a "." -typically locks-
@@ -344,12 +341,14 @@ impl Clone for ManagedDirectory {
#[cfg(test)]
mod tests_mmap_specific {
use crate::directory::{Directory, ManagedDirectory, MmapDirectory, TerminatingWrite};
use std::collections::HashSet;
use std::io::Write;
use std::path::{Path, PathBuf};
use tempfile::TempDir;
use crate::directory::{Directory, ManagedDirectory, MmapDirectory, TerminatingWrite};
#[test]
fn test_managed_directory() {
let tempdir = TempDir::new().unwrap();

60
src/directory/mmap_directory.rs
View File

@@ -1,32 +1,27 @@
use crate::core::META_FILEPATH;
use crate::directory::error::LockError;
use crate::directory::error::{DeleteError, OpenDirectoryError, OpenReadError, OpenWriteError};
use crate::directory::file_watcher::FileWatcher;
use crate::directory::Directory;
use crate::directory::DirectoryLock;
use crate::directory::Lock;
use crate::directory::WatchCallback;
use crate::directory::WatchHandle;
use crate::directory::{AntiCallToken, FileHandle, OwnedBytes};
use crate::directory::{ArcBytes, WeakArcBytes};
use crate::directory::{TerminatingWrite, WritePtr};
use std::collections::HashMap;
use std::fs::{self, File, OpenOptions};
use std::io::{self, BufWriter, Read, Seek, Write};
use std::ops::Deref;
use std::path::{Path, PathBuf};
use std::sync::{Arc, RwLock};
use std::{fmt, result};
use fs2::FileExt;
use memmap2::Mmap;
use serde::{Deserialize, Serialize};
use stable_deref_trait::StableDeref;
use std::convert::From;
use std::fmt;
use std::fs::OpenOptions;
use std::fs::{self, File};
use std::io::{self, Seek, SeekFrom};
use std::io::{BufWriter, Read, Write};
use std::path::{Path, PathBuf};
use std::result;
use std::sync::Arc;
use std::sync::RwLock;
use std::{collections::HashMap, ops::Deref};
use tempfile::TempDir;
use crate::core::META_FILEPATH;
use crate::directory::error::{
DeleteError, LockError, OpenDirectoryError, OpenReadError, OpenWriteError,
};
use crate::directory::file_watcher::FileWatcher;
use crate::directory::{
AntiCallToken, ArcBytes, Directory, DirectoryLock, FileHandle, Lock, OwnedBytes,
TerminatingWrite, WatchCallback, WatchHandle, WeakArcBytes, WritePtr,
};
/// Create a default io error given a string.
pub(crate) fn make_io_err(msg: String) -> io::Error {
io::Error::new(io::ErrorKind::Other, msg)
@@ -269,7 +264,7 @@ impl Write for SafeFileWriter {
}
impl Seek for SafeFileWriter {
fn seek(&mut self, pos: SeekFrom) -> io::Result<u64> {
fn seek(&mut self, pos: io::SeekFrom) -> io::Result<u64> {
self.0.seek(pos)
}
}
@@ -320,8 +315,7 @@ impl Directory for MmapDirectory {
let mut mmap_cache = self.inner.mmap_cache.write().map_err(|_| {
let msg = format!(
"Failed to acquired write lock \
on mmap cache while reading {:?}",
"Failed to acquired write lock on mmap cache while reading {:?}",
path
);
let io_err = make_io_err(msg);
@@ -457,6 +451,7 @@ impl Directory for MmapDirectory {
#[cfg(windows)]
{
use std::os::windows::fs::OpenOptionsExt;
use winapi::um::winbase;
open_opts
@@ -476,15 +471,12 @@ mod tests {
// There are more tests in directory/mod.rs
// The following tests are specific to the MmapDirectory
use common::HasLen;
use super::*;
use crate::indexer::LogMergePolicy;
use crate::Index;
use crate::ReloadPolicy;
use crate::{
schema::{Schema, SchemaBuilder, TEXT},
IndexSettings,
};
use common::HasLen;
use crate::schema::{Schema, SchemaBuilder, TEXT};
use crate::{Index, IndexSettings, ReloadPolicy};
#[test]
fn test_open_non_existent_path() {
@@ -521,7 +513,7 @@ mod tests {
{
for path in &paths {
let mut w = mmap_directory.open_write(path).unwrap();
w.write(content).unwrap();
w.write_all(content).unwrap();
w.flush().unwrap();
}
}

24
src/directory/mod.rs
View File

@@ -1,8 +1,4 @@
/*!
WORM (Write Once Read Many) directory abstraction.
*/
//! WORM (Write Once Read Many) directory abstraction.
#[cfg(feature = "mmap")]
mod mmap_directory;
@@ -13,7 +9,6 @@ mod file_slice;
mod file_watcher;
mod footer;
mod managed_directory;
mod owned_bytes;
mod ram_directory;
mod watch_event_router;
@@ -22,19 +17,19 @@ pub mod error;
mod composite_file;
use std::io::BufWriter;
use std::path::PathBuf;
pub use common::{AntiCallToken, TerminatingWrite};
pub use ownedbytes::OwnedBytes;
pub(crate) use self::composite_file::{CompositeFile, CompositeWrite};
pub use self::directory::DirectoryLock;
pub use self::directory::{Directory, DirectoryClone};
pub use self::directory::{Directory, DirectoryClone, DirectoryLock};
pub use self::directory_lock::{Lock, INDEX_WRITER_LOCK, META_LOCK};
pub(crate) use self::file_slice::{ArcBytes, WeakArcBytes};
pub use self::file_slice::{FileHandle, FileSlice};
pub use self::owned_bytes::OwnedBytes;
pub use self::ram_directory::RamDirectory;
pub use self::watch_event_router::{WatchCallback, WatchCallbackList, WatchHandle};
pub use common::AntiCallToken;
pub use common::TerminatingWrite;
use std::io::BufWriter;
use std::path::PathBuf;
/// Outcome of the Garbage collection
pub struct GarbageCollectionResult {
@@ -50,11 +45,10 @@ pub struct GarbageCollectionResult {
pub failed_to_delete_files: Vec<PathBuf>,
}
pub use self::managed_directory::ManagedDirectory;
#[cfg(feature = "mmap")]
pub use self::mmap_directory::MmapDirectory;
pub use self::managed_directory::ManagedDirectory;
/// Write object for Directory.
///
/// `WritePtr` are required to implement both Write

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