rust/tantivy
1
0
Fork 0
mirror of https://github.com/quickwit-oss/tantivy.git synced 2025-12-27 20:42:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

Compare commits

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

143 Commits
tokenizer_ ... tracing-ta

Author SHA1 Message Date
Paul Masurel
3a8a83da80 tracing 2023-10-16 19:23:47 +09:00
trinity-1686a
0241a05b90 add support for exists query syntax in query parser (#2170) 
* add support for exists query syntax in query parser

* rustfmt

* make Exists require a field
 2023-09-19 11:10:39 +02:00
PSeitz
e125f3b041 fix test (#2178) 2023-09-19 08:21:50 +02:00
PSeitz
c520ac46fc add support for date in term agg (#2172) 
support DateTime in TermsAggregation
Format dates with Rfc3339
 2023-09-14 09:22:18 +02:00
PSeitz
2d7390341c increase min memory to 15MB for indexing (#2176) 
With tantivy 0.20 the minimum memory consumption per SegmentWriter increased to
12MB. 7MB are for the different fast field collectors types (they could be
lazily created). Increase the minimum memory from 3MB to 15MB.

Change memory variable naming from arena to budget.

closes #2156
 2023-09-13 07:38:34 +02:00
dependabot[bot]
03fcdce016 Bump actions/checkout from 3 to 4 (#2171) 
Bumps [actions/checkout](https://github.com/actions/checkout) from 3 to 4.
- [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/v3...v4)

---
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>
 2023-09-11 10:47:33 +02:00
Ping Xia
e4e416ac42 extend FuzzyTermQuery to support json field (#2173) 
* extend fuzzy search for json field

* comments

* comments

* fmt fix

* comments
 2023-09-11 05:59:40 +02:00
Igor Motov
19325132b7 Fast-field based implementation of ExistsQuery (#2160) 
Adds an implementation of ExistsQuery that takes advantage of fast fields.

Fixes #2159
 2023-09-07 11:51:49 +09:00
Paul Masurel
389d36f760 Added comments 2023-09-04 11:06:56 +09:00
PSeitz
49448b31c6 chore: Release (#2168) 
* chore: Release

* update CHANGELOG
 2023-09-01 13:58:58 +02:00
PSeitz
ebede0bed7 update CHANGELOG (#2167) 2023-08-31 10:01:44 +02:00
PSeitz
b1d8b072db add missing aggregation part 2 (#2149) 
* add missing aggregation part 2

Add missing support for:
- Mixed types columns
- Key of type string on numerical fields

The special aggregation is slower than the integrated one in TermsAggregation and therefore not
chosen by default, although it can cover all use cases.

* simplify, add num_docs to empty
 2023-08-31 07:55:33 +02:00
ethever.eth
ee6a7c2bbb fix a small typo (#2165) 
Co-authored-by: famouscat <onismaa@gmail.com>
 2023-08-30 20:14:26 +02:00
PSeitz
c4e2708901 fix clippy, fmt (#2162) 2023-08-30 08:04:26 +02:00
PSeitz
5c8cfa50eb add missing parameter for percentiles (#2157) 2023-08-29 13:04:24 +02:00
PSeitz
73cb71762f add missing parameter for stats,min,max,count,sum,avg (#2151) 
* add missing parameter for stats,min,max,count,sum,avg

add missing parameter for stats,min,max,count,sum,avg
closes #1913
partially #1789

* Apply suggestions from code review

Co-authored-by: Paul Masurel <paul@quickwit.io>

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-08-28 08:59:51 +02:00
Harrison Burt
267dfe58d7 Fix testing on windows (#2155) 
* Fix missing trait imports

* Fix building tests on windows

* Revert other PR change
 2023-08-27 09:20:44 +09:00
Harrison Burt
131c10d318 Fix missing trait imports (#2154) 2023-08-27 09:20:26 +09:00
Chris Tam
e6cacc40a9 Remove outdated fast field documentation (#2145) 2023-08-24 07:49:49 +02:00
PSeitz
48d4847b38 Improve aggregation error message (#2150) 
* Improve aggregation error message

Improve aggregation error message by wrapping the deserialization with a
custom struct. This deserialization variant is slower, since we need to
keep the deserialized data around twice with this approach.
For now the valid variants list is manually updated. This could be
replaced with a proc macro.
closes #2143

* Simpler implementation

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-08-23 20:52:15 +02:00
PSeitz
59460c767f delayed column opening during merge (#2132) 
* lazy columnar merge

This is the first part of addressing #3633
Instead of loading all Column into memory for the merge, only the current column_name
group is loaded. This can be done since the sstable streams the columns lexicographically.

* refactor

* add rustdoc

* replace iterator with BTreeMap
 2023-08-21 08:55:35 +02:00
Paul Masurel
756156beaf Fix doc 2023-08-17 17:47:45 +09:00
PSeitz
480763db0d track memory arena memory usage (#2148) 2023-08-16 18:19:42 +02:00
PSeitz
62ece86f24 track ff dictionary indexing memory consumption (#2147) 2023-08-16 14:00:08 +02:00
Caleb Hattingh
52d9e6f298 Fix doc typos in count aggregation metric (#2127) 2023-08-15 08:50:23 +02:00
Caleb Hattingh
47b315ff18 doc: escape the backslash (#2144) 2023-08-14 19:10:07 +02:00
PSeitz
ed1deee902 fix sort index by date (#2124) 
closes #2112
 2023-08-14 17:36:52 +02:00
PSeitz
2e109018b7 add missing parameter to term agg (#2103) 
* add missing parameter to term agg

* move missing handling to block accessor

* add multivalue test, fix multivalue case, add comments

* add documentation, deactivate special case

* cargo fmt

* resolve merge conflict
 2023-08-14 14:22:18 +02:00
Adam Reichold
22c35b1e00 Fix explanation of boost queries seeking beyond query result. (#2142) 
* Make current nightly Clippy happy.

* Fix explanation of boost queries seeking beyond query result.
 2023-08-14 11:59:11 +09:00
trinity-1686a
b92082b748 implement lenient parser (#2129) 
* move query parser to nom

* add suupport for term grouping

* initial work on infallible parser

* fmt

* add tests and fix minor parsing bugs

* address review comments

* add support for lenient queries in tantivy

* make lenient parser report errors

* allow mixing occur and bool in query
 2023-08-08 15:41:29 +02:00
PSeitz
c2be6603a2 alternative mixed field aggregation collection (#2135) 
* alternative mixed field aggregation collection

instead of having multiple accessor in one AggregationWithAccessor split it into
multiple independent AggregationWithAccessor

* Update src/aggregation/agg_req_with_accessor.rs

Co-authored-by: Paul Masurel <paul@quickwit.io>

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-07-27 12:25:31 +02:00
Adam Reichold
c805f08ca7 Fix a few more upcoming Clippy lints (#2133) 2023-07-24 17:07:57 +09:00
Adam Reichold
ccc0335158 Minor improvements to OwnedBytes (#2134) 
This makes it obvious where the `StableDerefTrait` is invoked and avoids
`transmute` when only a lifetime needs to be extended. Furthermore, it makes use
of `slice::split_at` where that seemed appropriate.
 2023-07-24 17:06:33 +09:00
Adam Reichold
42acd334f4 Fixes the new deny-by-default incorrect_partial_ord_impl_on_ord_type Clippy lint (#2131) 2023-07-21 11:36:17 +09:00
Adam Reichold
820f126075 Remove support for Brotli and Snappy compression (#2123) 
LZ4 provides fast and simple compression whereas Zstd is exceptionally flexible
so that the additional support for Brotli and Snappy does not really add
any distinct functionality on top of those two algorithms.

Removing them reduces our maintenance burden and reduces the number of choices
users have to make when setting up their project based on Tantivy.
 2023-07-14 16:54:59 +09:00
Adam Reichold
7e6c4a1856 Include only built-in compression algorithms as enum variants (#2121) 
* Include only built-in compression algorithms as enum variants

This enables compile-time errors when a compression algorithm is requested which
is not actually enabled for the current Cargo project. The cost is that indexes
using other compression algorithms cannot even be loaded (even though they
are not fully accessible in any case).

As a drive-by, this also fixes `--no-default-features` on `cfg(unix)`.

* Provide more instructive error messages for unsupported, but not unknown compression variants.
 2023-07-14 11:02:49 +09:00
Adam Reichold
5fafe4b1ab Add missing query_terms impl for TermSetQuery. (#2120) 2023-07-13 14:54:29 +02:00
PSeitz
1e7cd48cfa remove allocations in split compound words (#2080) 
* remove allocations in split compound words

* clear reused data
 2023-07-13 09:43:02 +09:00
dependabot[bot]
7f51d85bbd Update lru requirement from 0.10.0 to 0.11.0 (#2117) 
Updates the requirements on [lru](https://github.com/jeromefroe/lru-rs) to permit the latest version.
- [Changelog](https://github.com/jeromefroe/lru-rs/blob/master/CHANGELOG.md)
- [Commits](https://github.com/jeromefroe/lru-rs/compare/0.10.0...0.11.0)

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

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2023-07-13 09:42:21 +09:00
PSeitz
ad76e32398 Update CHANGELOG.md (#2091) 
* Update CHANGELOG.md

* Update CHANGELOG.md
 2023-07-11 13:58:49 +08:00
dependabot[bot]
7575f9bf1c Update itertools requirement from 0.10.3 to 0.11.0 (#2098) 
Updates the requirements on [itertools](https://github.com/rust-itertools/itertools) to permit the latest version.
- [Changelog](https://github.com/rust-itertools/itertools/blob/master/CHANGELOG.md)
- [Commits](https://github.com/rust-itertools/itertools/compare/v0.10.5...v0.11.0)

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

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2023-07-07 11:14:46 +02:00
Naveen Aiathurai
67bdf3f5f6 fixes order_by_u64_field and order_by_fast_field should allow sorting in ascending order #1676 (#2111) 
* feat: order_by_fast_field allows sorting using parameter order

* chore: change the corresponding values to original one

* chore: fix formatting issues

* fix: first_or_default_col should also sort by order

* chore: empty doc to testcase and docstest fixes

* chore: fix failure tests

* core: add empty document without fastfield

* chore: fix fmt

* chore: change variable name
 2023-07-06 05:10:10 +02:00
François Massot
3c300666ad Merge pull request #2110 from quickwit-oss/fulmicoton/dynamic-follow-up 
Add dynamic filters to text analyzer builder.
 2023-07-03 21:49:24 +02:00
François Massot
b91d3f6be4 Clean comment on 'TextAnalyzerBuilder::filter_dynamic' method. 2023-07-03 18:45:59 +02:00
François Massot
a8e76513bb Remove useless clone. 2023-07-03 22:05:11 +09:00
François Massot
0a23201338 Fix stackoverflow and add docs. 2023-07-03 22:05:11 +09:00
François Massot
81330aaf89 WIP 2023-07-03 22:05:10 +09:00
Paul Masurel
98a3b01992 Removing the BoxedTokenizer 2023-07-03 22:05:10 +09:00
Paul Masurel
d341520938 Dynamic follow up 2023-07-03 22:05:10 +09:00
François Massot
5c9af73e41 Followup fulmicoton poc. 2023-07-03 22:05:10 +09:00
Paul Masurel
ad4c940fa3 proof of concept for dynamic tokenizer. 2023-07-03 22:05:10 +09:00
Paul Masurel
910b0b0c61 Cargo fmt 2023-07-03 22:03:31 +09:00
PSeitz
3fef052bf1 fix flaky test (#2107) 
closes #2099
 2023-06-29 14:30:56 +08:00
PSeitz
040554f2f9 Update to lz4_flex 0.11 (#2106) 2023-06-29 14:16:00 +08:00
PSeitz
17186ca9c9 improve docs (#2105) 2023-06-27 13:37:14 +08:00
François Massot
212d59c9ab Merge pull request #2102 from quickwit-oss/fmassot/ngram-new-should-return-error 
Ngram tokenizer now returns an error with invalid arguments.
 2023-06-27 05:36:09 +02:00
dependabot[bot]
1a1f252a3f Update memmap2 requirement from 0.6.0 to 0.7.1 (#2104) 
Updates the requirements on [memmap2](https://github.com/RazrFalcon/memmap2-rs) to permit the latest version.
- [Changelog](https://github.com/RazrFalcon/memmap2-rs/blob/master/CHANGELOG.md)
- [Commits](https://github.com/RazrFalcon/memmap2-rs/compare/v0.6.0...v0.7.1)

---
updated-dependencies:
- dependency-name: memmap2
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2023-06-27 05:15:43 +02:00
François Massot
d73706dede Ngram tokenizer now returns an error with invalid arguments. 2023-06-25 20:13:24 +02:00
PSeitz
44850e1036 move fail dep to dev only (#2094) 
wasm compilation fails with dep only
 2023-06-22 06:59:11 +02:00
Adam Reichold
3b0cbf8102 Cosmetic updates to the warmer example. (#2095) 
Just some cosmetic tweaks to make the example easier on the eyes as a colleague
was staring at this for quite some time this week.
 2023-06-22 11:25:01 +09:00
Adam Reichold
4aa131c3db Make TextAnalyzerBuilder publically accessible (#2097) 
This way, client code can name the type to e.g. store it inside structs without
resorting to generics and it means that its documentation is part of the crate
documentation generated by `cargo doc`.
 2023-06-22 11:24:21 +09:00
Naveen Aiathurai
59962097d0 fix: #2078 return error when tokenizer not found while indexing (#2093) 
* fix: #2078 return error when tokenizer not found while indexing

* chore: formatting issues

* chore: fix review comments
 2023-06-16 04:33:55 +02:00
Adam Reichold
ebc78127f3 Add BytesFilterCollector to support filtering based on a bytes fast field (#2075) 
* Do some Clippy- and Cargo-related boy-scouting.

* Add BytesFilterCollector to support filtering based on a bytes fast field

This is basically a copy of the existing FilterCollector but modified and
specialised to work on a bytes fast field.

* Changed semantics of filter collectors to consider multi-valued fields
 2023-06-13 14:19:58 +09:00
PSeitz
8199aa7de7 bump version to 0.20.2 (#2089) 2023-06-12 18:56:54 +08:00
PSeitz
657f0cd3bd add missing Bytes validation to term_agg (#2077) 
returns empty for now instead of failing like before
 2023-06-12 16:38:07 +08:00
Adam Reichold
3a82ef2560 Fix is_child_of function not considering the root facet. (#2086) 2023-06-12 08:35:18 +02:00
PSeitz
3546e7fc63 small agg limit docs improvement (#2073) 
small docs improvement as follow up on bug https://github.com/quickwit-oss/quickwit/issues/3503
 2023-06-12 10:55:24 +09:00
PSeitz
862f367f9e release without Alice in Wonderland, bump version to 0.20.1 (#2087) 
* Release without Alice in Wonderland

* bump version to 0.20.1
 2023-06-12 10:54:03 +09:00
PSeitz
14137d91c4 Update CHANGELOG.md (#2081) 2023-06-12 10:53:40 +09:00
François Massot
924fc70cb5 Merge pull request #2088 from quickwit-oss/fmassot/align-type-priorities-for-json-numbers 
Align numerical type priority order on the search side.
 2023-06-11 22:04:54 +02:00
François Massot
07023948aa Add test that indexes and searches a JSON field. 2023-06-11 21:47:52 +02:00
François Massot
0cb53207ec Fix tests. 2023-06-11 12:13:35 +02:00
François Massot
17c783b4db Align numerical type priority order on the search side. 2023-06-11 11:49:27 +02:00
Harrison Burt
7220df8a09 Fix building on windows with mmap (#2070) 
* Fix windows build

* Make pub

* Update docs

* Re arrange

* Fix compilation error on unix

* Fix unix borrows

* Revert "Fix unix borrows"

This reverts commit c1d94fd12b.

* Fix unix borrows and revert original change

* Fix warning

* Cleaner code.

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-06-10 18:32:39 +02:00
PSeitz
e3eacb4388 release tantivy (#2083) 
* prerelease

* chore: Release
 2023-06-09 10:47:46 +02:00
PSeitz
fdecb79273 tokenizer-api: reduce Tokenizer overhead (#2062) 
* tokenizer-api: reduce Tokenizer overhead

Previously a new `Token` for each text encountered was created, which
contains `String::with_capacity(200)`
In the new API the token_stream gets mutable access to the tokenizer,
this allows state to be shared (in this PR Token is shared).
Ideally the allocation for the BoxTokenStream would also be removed, but
this may require some lifetime tricks.

* simplify api

* move lowercase and ascii folding buffer to global

* empty Token text as default
 2023-06-08 18:37:58 +08:00
PSeitz
27f202083c Improve Termmap Indexing Performance +~30% (#2058) 
* update benchmark

* Improve Termmap Indexing Performance +~30%

This contains many small changes to improve Termmap performance.
Most notably:
* Specialized byte compare and equality versions, instead of glibc calls.
* ExpUnrolledLinkedList to not contain inline items.

Allow compare hash only via a feature flag compare_hash_only:
64bits should be enough with a good hash function to compare strings by
their hashes instead of comparing the strings. Disabled by default

CreateHashMap/alice/174693
                        time:   [642.23 µs 643.80 µs 645.24 µs]
                        thrpt:  [258.20 MiB/s 258.78 MiB/s 259.41 MiB/s]
                 change:
                        time:   [-14.429% -13.303% -12.348%] (p = 0.00 < 0.05)
                        thrpt:  [+14.088% +15.344% +16.862%]
                        Performance has improved.
CreateHashMap/alice_expull/174693
                        time:   [877.03 µs 880.44 µs 884.67 µs]
                        thrpt:  [188.32 MiB/s 189.22 MiB/s 189.96 MiB/s]
                 change:
                        time:   [-26.460% -26.274% -26.091%] (p = 0.00 < 0.05)
                        thrpt:  [+35.301% +35.637% +35.981%]
                        Performance has improved.
CreateHashMap/numbers_zipf/8000000
                        time:   [9.1198 ms 9.1573 ms 9.1961 ms]
                        thrpt:  [829.64 MiB/s 833.15 MiB/s 836.57 MiB/s]
                 change:
                        time:   [-35.229% -34.828% -34.384%] (p = 0.00 < 0.05)
                        thrpt:  [+52.403% +53.440% +54.390%]
                        Performance has improved.

* clippy

* add bench for ids

* inline(always) to inline whole block with bounds checks

* cleanup
 2023-06-08 11:13:52 +02:00
PSeitz
ccb09aaa83 allow histogram bounds to be passed as Rfc3339 (#2076) 2023-06-08 09:07:08 +02:00
Valerii
4b7c485a08 feat: add stop words for Hungarian language (#2069) 2023-06-02 07:26:03 +02:00
PSeitz
3942fc6d2b update CHANGELOG (#2068) 2023-06-02 05:00:12 +02:00
Adam Reichold
b325d569ad Expose phrase-prefix queries via the built-in query parser (#2044) 
* Expose phrase-prefix queries via the built-in query parser

This proposes the less-than-imaginative syntax `field:"phrase ter"*` to
perform a phrase prefix query against `field` using `phrase` and `ter` as the
terms. The aim of this is to make this type of query more discoverable and
simplify manual testing.

I did consider exposing the `max_expansions` parameter similar to how slop is
handled, but I think that this is rather something that should be configured via
the querser parser (similar to `set_field_boost` and `set_field_fuzzy`) as
choosing it requires rather intimiate knowledge of the backing index.

* Prevent construction of zero or one term phrase-prefix queries via the query parser.

* Add example using phrase-prefix search via surface API to improve feature discoverability.
 2023-06-01 13:03:16 +02:00
Paul Masurel
7ee78bda52 Readding s in datetime precision variant names (#2065) 
There is no clear win and it change some serialization in quickwit.
 2023-06-01 06:39:46 +02:00
Paul Masurel
184a9daa8a Cancels concurrently running actions for the same PR. (#2067) 2023-06-01 12:57:38 +09:00
Paul Masurel
47e01b345b Simplified linear probing code (#2066) 2023-06-01 04:58:42 +02:00
PSeitz
3af456972e Fix min doc_count empty merge bug (#2057) 
This fixes an issue when min_doc==0 loads terms from the dictionary from
one segment and merges the same term with a subaggregation from another
segment.
Previously the empty structure was not correctly initialized to contain
the subaggregation so the merge was incorrect.
 2023-05-29 14:20:50 +08:00
PSeitz
e56addc63e enable tokenizer on json fields (#2053) 
* enable tokenizer on json fields

enable tokenizer on json fields for type text

* Avoid making the tokenizer within the TextAnalyzer pub(crate)

* Moving BoxableTokenizer to tantivy.

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-05-24 10:47:39 +02:00
dependabot[bot]
4be6f83b0a Update criterion requirement from 0.4 to 0.5 (#2056) 
Updates the requirements on [criterion](https://github.com/bheisler/criterion.rs) to permit the latest version.
- [Changelog](https://github.com/bheisler/criterion.rs/blob/master/CHANGELOG.md)
- [Commits](https://github.com/bheisler/criterion.rs/compare/0.4.0...0.5.0)

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

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2023-05-24 15:59:51 +09:00
Adrien Guillo
a789ad9aee Rename DatePrecision to DateTimePrecision (#2051) 2023-05-23 17:09:11 +02:00
Sergei Lavrentev
8cf26da4b2 Add possibility to set up highlighten prefix and postfix for snippet (#1422) 
* add possibility to change highlight prefix and postfix

* add comment to Snippet::new

* add test for highlighten elements

* add default highlight prefix and postfix constants

* fix spelling

* fix tests

* fix spelling

* do fixes after code review

* reduce test_snippet_generator_custom_highlighted_elements code

* fix fmt

* change names to more convenient

---------

Co-authored-by: Sergei Lavrentev <23312691+lavrxxx@users.noreply.github.com>
 2023-05-23 15:09:24 +02:00
trinity-1686a
a3f001360f add support for warming up range of terms (#2042) 
* add support for warming up range of terms

* simplify handling of limit
 2023-05-22 14:29:35 +02:00
trinity-1686a
6564e0c467 fix phrase prefix query (#2043) 
* fix phrase prefix query

it would fail spectacularly when no doc in the segment would match the phrase part of the query

* clippy
 2023-05-22 12:36:20 +02:00
Paul Masurel
d7e97331e5 Minor refactoring find field (#2055) 
* Minor refactoring

Moving find_field_with_default to Schema.

* Clippy comments
 2023-05-22 15:00:48 +09:00
Paul Masurel
4417be165d Minor refactoring (#2054) 
Moving find_field_with_default to Schema.
 2023-05-22 14:56:38 +09:00
PSeitz
6239697a02 switch to ms in histogram for date type (#2045) 
* switch to ms in histogram for date type

switch to ms in histogram, by adding a normalization step that converts
to nanoseconds precision when creating the collector.

closes #2028
related to #2026

* add missing unit long variants

* use single thread to avoid handling test case

* fix docs

* revert CI

* cleanup

* improve docs

* Update src/aggregation/bucket/histogram/histogram.rs

Co-authored-by: Paul Masurel <paul@quickwit.io>

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-05-19 08:15:44 +02:00
Paul Masurel
62709b8094 Change in the query grammar. (#2050) 
* Change in the query grammar.

Quotation mark can now be used for phrase queries.
The delimiter is part of the `UserInputLeaf`.
That information is meant to be used in Quickwit to solve #3364.

This PR also adds support for quotation marks escaping in phrase
queries.

* Apply suggestions from code review
 2023-05-19 12:07:10 +09:00
PSeitz
04562c0318 add fastfield tokenizer to IndexBuilder (#2046) 2023-05-18 04:33:42 +02:00
PSeitz
2dfe37940d handle multiple types in term aggregation (#2041) 2023-05-15 11:57:38 +02:00
Denis Bazhenov
e248a4959f Enforcing "NOT" and "-" queries consistency in UserInputAst (#1609) 
* Enforcing "NOT" and "-" queries consistency in UserInputAst

* Mutable implementation if rewrite_ast_clause()
 2023-05-13 00:27:48 +09:00
PSeitz
00c5df610c update termmap benchmark (#2040) 2023-05-12 07:35:06 +02:00
Adam Reichold
fedd9559e7 Expose create a query from a user input AST. (#2039) 2023-05-11 21:53:18 +09:00
Paul Masurel
fe3ecf9567 Added support for madvise (#2036) 
Added support for madvise
 2023-05-11 05:39:17 +02:00
PSeitz
ba3a885a3b handle multiple agg results (#2035) 
handle multiple intermediate aggregation results with the same name.
 2023-05-10 15:00:38 +02:00
PSeitz
d1988be8e9 fix and extend benchmark (#2030) 
* add benchmark, add missing inlines

* fix stacker bench

* add wiki benchmark

* move line split out of bench
 2023-05-10 13:01:56 +02:00
PSeitz
0eafbaab8e fix slop (#2031) 
Fix slop by carrying slop so far for multiterms.
Define slop contract in the API
 2023-05-10 11:45:14 +02:00
PSeitz
d3357a8426 fix ArenaHashMap default (#2034) 
an empty ArenaHashMap is invalid and causes a panic when combined with `get`
 2023-05-10 11:39:47 +02:00
Yuri Astrakhan
74275b76a6 Inline format arguments where makes sense (#2038) 
Applied this command to the code, making it a bit shorter and slightly
more readable.

```
cargo +nightly clippy --all-features --benches --tests --workspace --fix -- -A clippy::all -W clippy::uninlined_format_args
cargo +nightly fmt --all
```
 2023-05-10 18:03:59 +09:00
dependabot[bot]
f479840a1b Update memmap2 requirement from 0.5.3 to 0.6.0 (#2033) 
Updates the requirements on [memmap2](https://github.com/RazrFalcon/memmap2-rs) to permit the latest version.
- [Changelog](https://github.com/RazrFalcon/memmap2-rs/blob/master/CHANGELOG.md)
- [Commits](https://github.com/RazrFalcon/memmap2-rs/compare/v0.5.3...v0.6.0)

---
updated-dependencies:
- dependency-name: memmap2
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2023-05-10 03:50:14 +02:00
PSeitz
4ee1b5cda0 add seperate tokenizer manager for fast fields (#2019) 
* add seperate tokenizer manager for fast fields

* rename
 2023-05-08 11:22:31 +02:00
PSeitz
45ff0e3c5c clear memory consumption in AggregationLimits (#2022) 
* clear memory consumption in AggregationLimits

clear memory consumption in AggregationLimits at the end of segment collection

* switch to ResourceLimitGuard

* unduplicate code

* merge methods

* Apply suggestions from code review

Co-authored-by: Paul Masurel <paul@quickwit.io>

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-05-08 10:15:09 +02:00
PSeitz
4c58b0086d allow slop in both directions (#2020) 
* allow slop in both directions

allow slop in both directions
so "big wolf"~3 can also match "wolf big"

This also fixes #1934, when the docsets were reordered by size and didn't
match the terms.

* remove count

* add test for repeating tokens, unduplicate tests
 2023-05-07 12:05:21 +09:00
Tomoko Uchida
85df322ceb fix typo in the architecture doc (#2009) 2023-05-07 12:04:07 +09:00
François Massot
38c863830f Merge pull request #2027 from quickwit-oss/fmassot/fix-date-histogram 
Fix date histogram bounds and field name.
 2023-05-05 13:03:25 +02:00
François Massot
992f755298 Fix clippy. 2023-05-05 10:51:29 +02:00
François Massot
c8df843f96 Fix date histogram bounds and field name. 2023-05-05 00:52:55 +02:00
Paul Masurel
f28ddb711e Exposing u64-based FastFieldRangeWeight (#2024) 2023-05-03 18:32:00 +09:00
tottoto
73452284ae Remove unused crates from dependencies (#2018) 
* Remove unused crates from dependencies

* Revert rand to columnar

* Revert criterion to stacker
 2023-05-02 12:34:20 +02:00
PSeitz
ba309e18a1 switch to nanosecond precision (#2016) 2023-05-01 03:32:20 +02:00
PSeitz
cbf2bdc75b change bucket count type (#2013) 
* change bucket count type

closes #2012

* Update src/aggregation/agg_limits.rs

Co-authored-by: Paul Masurel <paul@quickwit.io>

* Update src/directory/managed_directory.rs

Co-authored-by: Paul Masurel <paul@quickwit.io>

* fix test

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-04-27 15:47:31 +08:00
PSeitz
1f06997d04 fix single collector special case (#2014) 2023-04-27 09:30:19 +02:00
PSeitz
c599bf3b6c chore!:drop JSON support on intermediate agg result (#1992) 
* chore!:drop JSON support on intermediate agg result

add support for other formats by removing skip_serialize and untagged
JSON support is broken anyway due it's lack on f64::INF etc. handling

* Update src/aggregation/intermediate_agg_result.rs

Co-authored-by: Paul Masurel <paul@quickwit.io>

* move from impl

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-04-26 13:05:16 +02:00
PSeitz
80df1d9835 Handle error for exists on MMapDirectory (#1988) 
`exists` will return false in case of other io errors, like permission denied
 2023-04-25 09:20:33 +02:00
PSeitz
2e369db936 switch to Aggregation without serde_untagged (#2003) 
* refactor result handling

* remove Internal stuff

* merge different accessors

* switch to Aggregation without serde_untagged

* fix doctests
 2023-04-25 08:54:51 +02:00
PSeitz
7b31100208 refactor vint (#2010) 
- improve performance of vint
vint serialization shows up in performance profiles during indexing.
It would also make sense to limit the value space to u29 and operate on 4 bytes only.
- remove unused code
- add missing inlines
- fix regex test
 2023-04-25 08:49:36 +02:00
trinity-1686a
9c93bfeb51 optimise warmup code path (#2007) 
* optimise warmup code path

* better function naming
 2023-04-21 11:23:09 +02:00
PSeitz
74f9eafefc refactor Term (#2006) 
* refactor Term

add ValueBytes for serialized term values
add missing debug for ip
skip unnecessary json path validation
remove code duplication
add DATE_TIME_PRECISION_INDEXED constant
add missing Term clarification
remove weird value_bytes_mut() API

* fix naming
 2023-04-20 15:31:43 +02:00
RT_Enzyme
ff3d3313c4 fix BooleanQuery document (#1999) 
* fix BooleanQuery document

* Update src/query/boolean_query/boolean_query.rs

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-04-20 11:37:20 +02:00
Paul Masurel
fbda511a1a Making more things public for quickwit. (#2005) 2023-04-20 11:37:45 +09:00
Adam Reichold
c1defdda05 Bump aho-corasick dependency to version 1.0 and adjust to API changes (#2002) 
* Drop additional Arc-layer as the automaton itself is now cheap-to-clone.
* Drop state ID type parameter as it is not exposed by the library any more.
 2023-04-18 07:34:30 +02:00
PSeitz
e522163a1c use json in agg tests (#1998) 
* switch to JSON in tests, add flat aggregation types

* use method

* clippy

* remove commented file
 2023-04-17 14:08:48 +02:00
PSeitz
e83abbfe4a perf: faster term hash map (#1940) 
* add term hashmap benchmark

* refactor arena hashmap

add inlines
remove occupied array and use table_entry.is_empty instead (saves 4 bytes per entry)
reduce saturation threshold from 1/3 to 1/2 to reduce memory
use u32 for UnorderedId (we have the 4billion limit anyways on the Columnar stuff)
fix naming LinearProbing
remove byteorder dependency

memory consumption went down from 2Gb to 1.8GB on indexing wikipedia dataset in tantivy

* Update stacker/src/arena_hashmap.rs

Co-authored-by: Paul Masurel <paul@quickwit.io>

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-04-17 09:07:33 +02:00
trinity-1686a
780e26331d sstable compression (#1946) 
* compress sstable with zstd

* add some details to sstable readme

* compress only block which benefit from it

* multiple changes to sstable

make compression optional
use OwnedBytes instead of impl Read in sstable, required for next point
use zstd bulk api, which is much faster on small records

* cleanup and use bulk api for compression

* use dedicated byte for compression

* switch block len and compression flag

* change default zstd level in sstable
 2023-04-14 16:25:50 +02:00
trinity-1686a
0286ecea09 re-export a few sstable functions on dicitonary (#1996) 
* re-export a few sstable functions on dicitonary

* Update documentation

Co-authored-by: François Massot <francois.massot@gmail.com>

---------

Co-authored-by: François Massot <francois.massot@gmail.com>
 2023-04-14 11:13:48 +02:00
PSeitz
b0ef9a6252 use crates.io dependency (#1990) 2023-04-14 09:35:20 +08:00
François Massot
36138c493b Merge pull request #1994 from quickwit-oss/fmassot/expose-simple-token-stream 
Expose `SimpleTokenStream` to use it in quickwit for the multilanguage tokenizer
 2023-04-13 18:55:02 +02:00
François Massot
64bce340b2 Expose to use it in quickwit. 2023-04-13 18:28:53 +02:00
trinity-1686a
205e8a0a92 encode dictionary type in fst footer (#1968) 
* encode additional footer for dictionary kind in fst
 2023-04-12 09:43:01 +02:00
Paul Masurel
4b01cc4c49 Made BooleanWeight and BoostWeight public (#1991) 2023-04-12 10:26:30 +09:00
PSeitz
0ed13eeea8 add sparse to agg benchmark (#1986) 
* add sparse to agg benchmark

* Update src/aggregation/agg_bench.rs

Co-authored-by: Paul Masurel <paul@quickwit.io>

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-04-11 08:13:32 +02:00
Tony-X
91a38058fe Fix typo in READEME.md (#1989) 2023-04-11 12:07:20 +09:00
PSeitz
41af70799d add percentiles aggregations (#1984) 
* add percentiles aggregations

add percentiles aggregation
fix disabled agg benchmark

* Update src/aggregation/metric/percentiles.rs

Co-authored-by: Paul Masurel <paul@quickwit.io>

* Apply suggestions from code review

Co-authored-by: Paul Masurel <paul@quickwit.io>

* fix import

* fix import

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-04-07 07:18:28 +02:00
Paul Masurel
f853bf204b Align the numerical type priority order with columnar. (#1978) 
Closes #1956
 2023-04-07 10:07:54 +09:00
Tony-X
11ae48d3bc Update benchmarks section in READEME.md to link to the bench repo (#1985) 
* Update benchmarks section in READEME.md to link to the bench repo

* Apply suggestions from code review

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-04-07 10:07:06 +09:00
Paul Masurel
5eb12173d6 Proptest merge columnar (#1976) 
* Added proptest on columnar merge with a shuffle

Made column serialization more explicit.
Bugfix when a bytes column is missing, and with a shuffle.
Improved the cardinality detection logic / column detection.

* Code review

* CR comments

* Following CR
 2023-04-04 11:28:42 +09:00
240 changed files with 15579 additions and 6215 deletions
Expand all files Collapse all files

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

@@ -6,11 +6,16 @@ on:
pull_request:
branches: [main]
# Ensures that we cancel running jobs for the same PR / same workflow.
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
jobs:
coverage:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v3
- uses: actions/checkout@v4
- name: Install Rust
run: rustup toolchain install nightly --profile minimal --component llvm-tools-preview
- uses: Swatinem/rust-cache@v2

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

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

11
.github/workflows/test.yml vendored
View File

@@ -9,13 +9,18 @@ on:
env:
CARGO_TERM_COLOR: always
# Ensures that we cancel running jobs for the same PR / same workflow.
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
jobs:
check:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v3
- uses: actions/checkout@v4
- name: Install nightly
uses: actions-rs/toolchain@v1
@@ -48,14 +53,14 @@ jobs:
strategy:
matrix:
features: [
{ label: "all", flags: "mmap,stopwords,brotli-compression,lz4-compression,snappy-compression,zstd-compression,failpoints" },
{ label: "all", flags: "mmap,stopwords,lz4-compression,zstd-compression,failpoints" },
{ label: "quickwit", flags: "mmap,quickwit,failpoints" }
]
name: test-${{ matrix.features.label}}
steps:
- uses: actions/checkout@v3
- uses: actions/checkout@v4
- name: Install stable
uses: actions-rs/toolchain@v1

2
ARCHITECTURE.md
View File

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

116
CHANGELOG.md
View File

@@ -1,3 +1,119 @@
Tantivy 0.21
================================
#### Bugfixes
- Fix track fast field memory consumption, which led to higher memory consumption than the budget allowed during indexing [#2148](https://github.com/quickwit-oss/tantivy/issues/2148)[#2147](https://github.com/quickwit-oss/tantivy/issues/2147)(@PSeitz)
- Fix a regression from 0.20 where sort index by date wasn't working anymore [#2124](https://github.com/quickwit-oss/tantivy/issues/2124)(@PSeitz)
- Fix getting the root facet on the `FacetCollector`. [#2086](https://github.com/quickwit-oss/tantivy/issues/2086)(@adamreichold)
- Align numerical type priority order of columnar and query. [#2088](https://github.com/quickwit-oss/tantivy/issues/2088)(@fmassot)
#### Breaking Changes
- Remove support for Brotli and Snappy compression [#2123](https://github.com/quickwit-oss/tantivy/issues/2123)(@adamreichold)
#### Features/Improvements
- Implement lenient query parser [#2129](https://github.com/quickwit-oss/tantivy/pull/2129)(@trinity-1686a)
- order_by_u64_field and order_by_fast_field allow sorting in ascending and descending order [#2111](https://github.com/quickwit-oss/tantivy/issues/2111)(@naveenann)
- Allow dynamic filters in text analyzer builder [#2110](https://github.com/quickwit-oss/tantivy/issues/2110)(@fulmicoton @fmassot)
- **Aggregation**
- Add missing parameter for term aggregation [#2149](https://github.com/quickwit-oss/tantivy/issues/2149)[#2103](https://github.com/quickwit-oss/tantivy/issues/2103)(@PSeitz)
- Add missing parameter for percentiles [#2157](https://github.com/quickwit-oss/tantivy/issues/2157)(@PSeitz)
- Add missing parameter for stats,min,max,count,sum,avg [#2151](https://github.com/quickwit-oss/tantivy/issues/2151)(@PSeitz)
- Improve aggregation deserialization error message [#2150](https://github.com/quickwit-oss/tantivy/issues/2150)(@PSeitz)
- Add validation for type Bytes to term_agg [#2077](https://github.com/quickwit-oss/tantivy/issues/2077)(@PSeitz)
- Alternative mixed field collection [#2135](https://github.com/quickwit-oss/tantivy/issues/2135)(@PSeitz)
- Add missing query_terms impl for TermSetQuery. [#2120](https://github.com/quickwit-oss/tantivy/issues/2120)(@adamreichold)
- Minor improvements to OwnedBytes [#2134](https://github.com/quickwit-oss/tantivy/issues/2134)(@adamreichold)
- Remove allocations in split compound words [#2080](https://github.com/quickwit-oss/tantivy/issues/2080)(@PSeitz)
- Ngram tokenizer now returns an error with invalid arguments [#2102](https://github.com/quickwit-oss/tantivy/issues/2102)(@fmassot)
- Make TextAnalyzerBuilder public [#2097](https://github.com/quickwit-oss/tantivy/issues/2097)(@adamreichold)
- Return an error when tokenizer is not found while indexing [#2093](https://github.com/quickwit-oss/tantivy/issues/2093)(@naveenann)
- Delayed column opening during merge [#2132](https://github.com/quickwit-oss/tantivy/issues/2132)(@PSeitz)
Tantivy 0.20.2
================================
- Align numerical type priority order on the search side. [#2088](https://github.com/quickwit-oss/tantivy/issues/2088) (@fmassot)
- Fix is_child_of function not considering the root facet. [#2086](https://github.com/quickwit-oss/tantivy/issues/2086) (@adamreichhold)
Tantivy 0.20.1
================================
- Fix building on windows with mmap [#2070](https://github.com/quickwit-oss/tantivy/issues/2070) (@ChillFish8)
Tantivy 0.20
================================
#### Bugfixes
- Fix phrase queries with slop (slop supports now transpositions, algorithm that carries slop so far for num terms > 2) [#2031](https://github.com/quickwit-oss/tantivy/issues/2031)[#2020](https://github.com/quickwit-oss/tantivy/issues/2020)(@PSeitz)
- Handle error for exists on MMapDirectory [#1988](https://github.com/quickwit-oss/tantivy/issues/1988) (@PSeitz)
- Aggregation
- Fix min doc_count empty merge bug [#2057](https://github.com/quickwit-oss/tantivy/issues/2057) (@PSeitz)
- Fix: Sort order for term aggregations (sort order on key was inverted) [#1858](https://github.com/quickwit-oss/tantivy/issues/1858) (@PSeitz)
#### Features/Improvements
- Add PhrasePrefixQuery [#1842](https://github.com/quickwit-oss/tantivy/issues/1842) (@trinity-1686a)
- Add `coerce` option for text and numbers types (convert the value instead of returning an error during indexing) [#1904](https://github.com/quickwit-oss/tantivy/issues/1904) (@PSeitz)
- Add regex tokenizer [#1759](https://github.com/quickwit-oss/tantivy/issues/1759)(@mkleen)
- Move tokenizer API to seperate crate. Having a seperate crate with a stable API will allow us to use tokenizers with different tantivy versions. [#1767](https://github.com/quickwit-oss/tantivy/issues/1767) (@PSeitz)
- **Columnar crate**: New fast field handling (@fulmicoton @PSeitz) [#1806](https://github.com/quickwit-oss/tantivy/issues/1806)[#1809](https://github.com/quickwit-oss/tantivy/issues/1809)
- Support for fast fields with optional values. Previously tantivy supported only single-valued and multi-value fast fields. The encoding of optional fast fields is now very compact.
- Fast field Support for JSON (schemaless fast fields). Support multiple types on the same column. [#1876](https://github.com/quickwit-oss/tantivy/issues/1876) (@fulmicoton)
- Unified access for fast fields over different cardinalities.
- Unified storage for typed and untyped fields.
- Move fastfield codecs into columnar. [#1782](https://github.com/quickwit-oss/tantivy/issues/1782) (@fulmicoton)
- Sparse dense index for optional values [#1716](https://github.com/quickwit-oss/tantivy/issues/1716) (@PSeitz)
- Switch to nanosecond precision in DateTime fastfield [#2016](https://github.com/quickwit-oss/tantivy/issues/2016) (@PSeitz)
- **Aggregation**
- Add `date_histogram` aggregation (only `fixed_interval` for now) [#1900](https://github.com/quickwit-oss/tantivy/issues/1900) (@PSeitz)
- Add `percentiles` aggregations [#1984](https://github.com/quickwit-oss/tantivy/issues/1984) (@PSeitz)
- [**breaking**] Drop JSON support on intermediate agg result (we use postcard as format in `quickwit` to send intermediate results) [#1992](https://github.com/quickwit-oss/tantivy/issues/1992) (@PSeitz)
- Set memory limit in bytes for aggregations after which they abort (Previously there was only the bucket limit) [#1942](https://github.com/quickwit-oss/tantivy/issues/1942)[#1957](https://github.com/quickwit-oss/tantivy/issues/1957)(@PSeitz)
- Add support for u64,i64,f64 fields in term aggregation [#1883](https://github.com/quickwit-oss/tantivy/issues/1883) (@PSeitz)
- Allow histogram bounds to be passed as Rfc3339 [#2076](https://github.com/quickwit-oss/tantivy/issues/2076) (@PSeitz)
- Add count, min, max, and sum aggregations [#1794](https://github.com/quickwit-oss/tantivy/issues/1794) (@guilload)
- Switch to Aggregation without serde_untagged => better deserialization errors. [#2003](https://github.com/quickwit-oss/tantivy/issues/2003) (@PSeitz)
- Switch to ms in histogram for date type (ES compatibility) [#2045](https://github.com/quickwit-oss/tantivy/issues/2045) (@PSeitz)
- Reduce term aggregation memory consumption [#2013](https://github.com/quickwit-oss/tantivy/issues/2013) (@PSeitz)
- Reduce agg memory consumption: Replace generic aggregation collector (which has a high memory requirement per instance) in aggregation tree with optimized versions behind a trait.
- Split term collection count and sub_agg (Faster term agg with less memory consumption for cases without sub-aggs) [#1921](https://github.com/quickwit-oss/tantivy/issues/1921) (@PSeitz)
- Schemaless aggregations: In combination with stacker tantivy supports now schemaless aggregations via the JSON type.
- Add aggregation support for JSON type [#1888](https://github.com/quickwit-oss/tantivy/issues/1888) (@PSeitz)
- Mixed types support on JSON fields in aggs [#1971](https://github.com/quickwit-oss/tantivy/issues/1971) (@PSeitz)
- Perf: Fetch blocks of vals in aggregation for all cardinality [#1950](https://github.com/quickwit-oss/tantivy/issues/1950) (@PSeitz)
- Allow histogram bounds to be passed as Rfc3339 [#2076](https://github.com/quickwit-oss/tantivy/issues/2076) (@PSeitz)
- `Searcher` with disabled scoring via `EnableScoring::Disabled` [#1780](https://github.com/quickwit-oss/tantivy/issues/1780) (@shikhar)
- Enable tokenizer on json fields [#2053](https://github.com/quickwit-oss/tantivy/issues/2053) (@PSeitz)
- Enforcing "NOT" and "-" queries consistency in UserInputAst [#1609](https://github.com/quickwit-oss/tantivy/issues/1609) (@bazhenov)
- Faster indexing
- Refactor tokenization pipeline to use GATs [#1924](https://github.com/quickwit-oss/tantivy/issues/1924) (@trinity-1686a)
- Faster term hash map [#2058](https://github.com/quickwit-oss/tantivy/issues/2058)[#1940](https://github.com/quickwit-oss/tantivy/issues/1940) (@PSeitz)
- tokenizer-api: reduce Tokenizer allocation overhead [#2062](https://github.com/quickwit-oss/tantivy/issues/2062) (@PSeitz)
- Refactor vint [#2010](https://github.com/quickwit-oss/tantivy/issues/2010) (@PSeitz)
- Faster search
- Work in batches of docs on the SegmentCollector (Only for cases without score for now) [#1937](https://github.com/quickwit-oss/tantivy/issues/1937) (@PSeitz)
- Faster fast field range queries using SIMD [#1954](https://github.com/quickwit-oss/tantivy/issues/1954) (@fulmicoton)
- Improve fast field range query performance [#1864](https://github.com/quickwit-oss/tantivy/issues/1864) (@PSeitz)
- Make BM25 scoring more flexible [#1855](https://github.com/quickwit-oss/tantivy/issues/1855) (@alexcole)
- Switch fs2 to fs4 as it is now unmaintained and does not support illumos [#1944](https://github.com/quickwit-oss/tantivy/issues/1944) (@Toasterson)
- Made BooleanWeight and BoostWeight public [#1991](https://github.com/quickwit-oss/tantivy/issues/1991) (@fulmicoton)
- Make index compatible with virtual drives on Windows [#1843](https://github.com/quickwit-oss/tantivy/issues/1843) (@gyk)
- Add stop words for Hungarian language [#2069](https://github.com/quickwit-oss/tantivy/issues/2069) (@tnxbutno)
- Auto downgrade index record option, instead of vint error [#1857](https://github.com/quickwit-oss/tantivy/issues/1857) (@PSeitz)
- Enable range query on fast field for u64 compatible types [#1762](https://github.com/quickwit-oss/tantivy/issues/1762) (@PSeitz) [#1876]
- sstable
- Isolating sstable and stacker in independant crates. [#1718](https://github.com/quickwit-oss/tantivy/issues/1718) (@fulmicoton)
- New sstable format [#1943](https://github.com/quickwit-oss/tantivy/issues/1943)[#1953](https://github.com/quickwit-oss/tantivy/issues/1953) (@trinity-1686a)
- Use DeltaReader directly to implement Dictionnary::ord_to_term [#1928](https://github.com/quickwit-oss/tantivy/issues/1928) (@trinity-1686a)
- Use DeltaReader directly to implement Dictionnary::term_ord [#1925](https://github.com/quickwit-oss/tantivy/issues/1925) (@trinity-1686a)
- Add seperate tokenizer manager for fast fields [#2019](https://github.com/quickwit-oss/tantivy/issues/2019) (@PSeitz)
- Make construction of LevenshteinAutomatonBuilder for FuzzyTermQuery instances lazy. [#1756](https://github.com/quickwit-oss/tantivy/issues/1756) (@adamreichold)
- Added support for madvise when opening an mmaped Index [#2036](https://github.com/quickwit-oss/tantivy/issues/2036) (@fulmicoton)
- Rename `DatePrecision` to `DateTimePrecision` [#2051](https://github.com/quickwit-oss/tantivy/issues/2051) (@guilload)
- Query Parser
- Quotation mark can now be used for phrase queries. [#2050](https://github.com/quickwit-oss/tantivy/issues/2050) (@fulmicoton)
- PhrasePrefixQuery is supported in the query parser via: `field:"phrase ter"*` [#2044](https://github.com/quickwit-oss/tantivy/issues/2044) (@adamreichold)
- Docs
- Update examples for literate docs [#1880](https://github.com/quickwit-oss/tantivy/issues/1880) (@PSeitz)
- Add ip field example [#1775](https://github.com/quickwit-oss/tantivy/issues/1775) (@PSeitz)
- Fix doc store cache documentation [#1821](https://github.com/quickwit-oss/tantivy/issues/1821) (@PSeitz)
- Fix BooleanQuery document [#1999](https://github.com/quickwit-oss/tantivy/issues/1999) (@RT_Enzyme)
- Update comments in the faceted search example [#1737](https://github.com/quickwit-oss/tantivy/issues/1737) (@DawChihLiou)
Tantivy 0.19
================================
#### Bugfixes

56
Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy"
version = "0.19.0"
version = "0.21.0"
authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT"
categories = ["database-implementations", "data-structures"]
@@ -12,21 +12,21 @@ readme = "README.md"
keywords = ["search", "information", "retrieval"]
edition = "2021"
rust-version = "1.62"
exclude = ["benches/*.json", "benches/*.txt"]
[dependencies]
oneshot = "0.1.5"
base64 = "0.21.0"
byteorder = "1.4.3"
crc32fast = "1.3.2"
tracing = "0.1"
once_cell = "1.10.0"
regex = { version = "1.5.5", default-features = false, features = ["std", "unicode"] }
aho-corasick = "0.7"
aho-corasick = "1.0"
tantivy-fst = "0.4.0"
memmap2 = { version = "0.5.3", optional = true }
lz4_flex = { version = "0.10", default-features = false, features = ["checked-decode"], optional = true }
brotli = { version = "3.3.4", optional = true }
memmap2 = { version = "0.7.1", optional = true }
lz4_flex = { version = "0.11", default-features = false, optional = true }
zstd = { version = "0.12", optional = true, default-features = false }
snap = { version = "1.0.5", optional = true }
tempfile = { version = "3.3.0", optional = true }
log = "0.4.16"
serde = { version = "1.0.136", features = ["derive"] }
@@ -43,25 +43,27 @@ census = "0.4.0"
rustc-hash = "1.1.0"
thiserror = "1.0.30"
htmlescape = "0.3.1"
fail = "0.5.0"
fail = { version = "0.5.0", optional = true }
murmurhash32 = "0.3.0"
time = { version = "0.3.10", features = ["serde-well-known"] }
smallvec = "1.8.0"
rayon = "1.5.2"
lru = "0.10.0"
lru = "0.11.0"
fastdivide = "0.4.0"
itertools = "0.10.3"
itertools = "0.11.0"
measure_time = "0.8.2"
async-trait = "0.1.53"
arc-swap = "1.5.0"
columnar = { version="0.1", path="./columnar", package ="tantivy-columnar" }
sstable = { version="0.1", path="./sstable", package ="tantivy-sstable", optional = true }
stacker = { version="0.1", path="./stacker", package ="tantivy-stacker" }
query-grammar = { version= "0.19.0", path="./query-grammar", package = "tantivy-query-grammar" }
tantivy-bitpacker = { version= "0.3", path="./bitpacker" }
common = { version= "0.5", path = "./common/", package = "tantivy-common" }
tokenizer-api = { version="0.1", path="./tokenizer-api", package="tantivy-tokenizer-api" }
columnar = { version= "0.2", path="./columnar", package ="tantivy-columnar" }
sstable = { version= "0.2", path="./sstable", package ="tantivy-sstable", optional = true }
stacker = { version= "0.2", path="./stacker", package ="tantivy-stacker" }
query-grammar = { version= "0.21.0", path="./query-grammar", package = "tantivy-query-grammar" }
tantivy-bitpacker = { version= "0.5", path="./bitpacker" }
common = { version= "0.6", path = "./common/", package = "tantivy-common" }
tokenizer-api = { version= "0.2", path="./tokenizer-api", package="tantivy-tokenizer-api" }
sketches-ddsketch = { version = "0.2.1", features = ["use_serde"] }
futures-util = { version = "0.3.28", optional = true }
[target.'cfg(windows)'.dependencies]
winapi = "0.3.9"
@@ -72,12 +74,16 @@ maplit = "1.0.2"
matches = "0.1.9"
pretty_assertions = "1.2.1"
proptest = "1.0.0"
criterion = "0.4"
test-log = "0.2.10"
env_logger = "0.10.0"
pprof = { version = "0.11.0", features = ["flamegraph", "criterion"] }
futures = "0.3.21"
paste = "1.0.11"
more-asserts = "0.3.1"
rand_distr = "0.4.3"
[target.'cfg(not(windows))'.dev-dependencies]
criterion = "0.5"
pprof = { git = "https://github.com/PSeitz/pprof-rs/", rev = "53af24b", features = ["flamegraph", "criterion"] } # temp fork that works with criterion 0.5
[dev-dependencies.fail]
version = "0.5.0"
@@ -88,6 +94,11 @@ opt-level = 3
debug = false
debug-assertions = false
[profile.bench]
opt-level = 3
debug = true
debug-assertions = false
[profile.test]
debug-assertions = true
overflow-checks = true
@@ -97,15 +108,13 @@ default = ["mmap", "stopwords", "lz4-compression"]
mmap = ["fs4", "tempfile", "memmap2"]
stopwords = []
brotli-compression = ["brotli"]
lz4-compression = ["lz4_flex"]
snappy-compression = ["snap"]
zstd-compression = ["zstd"]
failpoints = ["fail/failpoints"]
failpoints = ["fail", "fail/failpoints"]
unstable = [] # useful for benches.
quickwit = ["sstable"]
quickwit = ["sstable", "futures-util"]
[workspace]
members = ["query-grammar", "bitpacker", "common", "ownedbytes", "stacker", "sstable", "tokenizer-api", "columnar"]
@@ -120,7 +129,7 @@ members = ["query-grammar", "bitpacker", "common", "ownedbytes", "stacker", "sst
[[test]]
name = "failpoints"
path = "tests/failpoints/mod.rs"
required-features = ["fail/failpoints"]
required-features = ["failpoints"]
[[bench]]
name = "analyzer"
@@ -129,4 +138,3 @@ harness = false
[[bench]]
name = "index-bench"
harness = false

2
Makefile
View File

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

4
README.md
View File

@@ -26,6 +26,8 @@ Your mileage WILL vary depending on the nature of queries and their load.
<img src="doc/assets/images/searchbenchmark.png">
Details about the benchmark can be found at this [repository](https://github.com/quickwit-oss/search-benchmark-game).
# Features
- Full-text search
@@ -42,7 +44,7 @@ Your mileage WILL vary depending on the nature of queries and their load.
- Single valued and multivalued u64, i64, and f64 fast fields (equivalent of doc values in Lucene)
- `&[u8]` fast fields
- Text, i64, u64, f64, dates, ip, bool, and hierarchical facet fields
- Compressed document store (LZ4, Zstd, None, Brotli, Snap)
- Compressed document store (LZ4, Zstd, None)
- Range queries
- Faceted search
- Configurable indexing (optional term frequency and position indexing)

21
RELEASE.md Normal file
View File

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

23
appveyor.yml
View File

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

27
benches/analyzer.rs
View File

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

1000
benches/gh.json Normal file
View File

File diff suppressed because one or more lines are too long

159
benches/index-bench.rs
View File

@@ -1,10 +1,15 @@
use criterion::{criterion_group, criterion_main, Criterion};
use criterion::{criterion_group, criterion_main, Criterion, Throughput};
use pprof::criterion::{Output, PProfProfiler};
use tantivy::schema::{INDEXED, STORED, STRING, TEXT};
use tantivy::schema::{FAST, INDEXED, STORED, STRING, TEXT};
use tantivy::Index;
const HDFS_LOGS: &str = include_str!("hdfs.json");
const NUM_REPEATS: usize = 2;
const GH_LOGS: &str = include_str!("gh.json");
const WIKI: &str = include_str!("wiki.json");
fn get_lines(input: &str) -> Vec<&str> {
input.trim().split('\n').collect()
}
pub fn hdfs_index_benchmark(c: &mut Criterion) {
let schema = {
@@ -28,85 +33,147 @@ pub fn hdfs_index_benchmark(c: &mut Criterion) {
};
let mut group = c.benchmark_group("index-hdfs");
group.throughput(Throughput::Bytes(HDFS_LOGS.len() as u64));
group.sample_size(20);
group.bench_function("index-hdfs-no-commit", |b| {
let lines = get_lines(HDFS_LOGS);
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();
}
for doc_json in &lines {
let doc = schema.parse_document(doc_json).unwrap();
index_writer.add_document(doc).unwrap();
}
})
});
group.bench_function("index-hdfs-with-commit", |b| {
let lines = get_lines(HDFS_LOGS);
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();
}
for doc_json in &lines {
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| {
let lines = get_lines(HDFS_LOGS);
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();
}
for doc_json in &lines {
let doc = schema.parse_document(doc_json).unwrap();
index_writer.add_document(doc).unwrap();
}
})
});
group.bench_function("index-hdfs-with-commit-with-docstore", |b| {
let lines = get_lines(HDFS_LOGS);
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();
}
for doc_json in &lines {
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| {
let lines = get_lines(HDFS_LOGS);
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();
}
for doc_json in &lines {
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| {
}
pub fn gh_index_benchmark(c: &mut Criterion) {
let dynamic_schema = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_json_field("json", TEXT | FAST);
schema_builder.build()
};
let mut group = c.benchmark_group("index-gh");
group.throughput(Throughput::Bytes(GH_LOGS.len() as u64));
group.bench_function("index-gh-no-commit", |b| {
let lines = get_lines(GH_LOGS);
b.iter(|| {
let index = Index::create_in_ram(dynamic_schema.clone());
let json_field = dynamic_schema.get_field("json").unwrap();
let index = Index::create_in_ram(dynamic_schema.clone());
let index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
for doc_json in &lines {
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();
}
})
});
group.bench_function("index-gh-with-commit", |b| {
let lines = get_lines(GH_LOGS);
b.iter(|| {
let json_field = dynamic_schema.get_field("json").unwrap();
let index = Index::create_in_ram(dynamic_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 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();
}
for doc_json in &lines {
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();
})
});
}
pub fn wiki_index_benchmark(c: &mut Criterion) {
let dynamic_schema = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_json_field("json", TEXT | FAST);
schema_builder.build()
};
let mut group = c.benchmark_group("index-wiki");
group.throughput(Throughput::Bytes(WIKI.len() as u64));
group.bench_function("index-wiki-no-commit", |b| {
let lines = get_lines(WIKI);
b.iter(|| {
let json_field = dynamic_schema.get_field("json").unwrap();
let index = Index::create_in_ram(dynamic_schema.clone());
let index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
for doc_json in &lines {
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();
}
})
});
group.bench_function("index-wiki-with-commit", |b| {
let lines = get_lines(WIKI);
b.iter(|| {
let json_field = dynamic_schema.get_field("json").unwrap();
let index = Index::create_in_ram(dynamic_schema.clone());
let mut index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
for doc_json in &lines {
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();
})
@@ -115,7 +182,17 @@ pub fn hdfs_index_benchmark(c: &mut Criterion) {
criterion_group! {
name = benches;
config = Criterion::default().with_profiler(PProfProfiler::new(100, Output::Flamegraph(None)));
config = Criterion::default();
targets = hdfs_index_benchmark
}
criterion_main!(benches);
criterion_group! {
name = gh_benches;
config = Criterion::default().with_profiler(PProfProfiler::new(100, Output::Flamegraph(None)));
targets = gh_index_benchmark
}
criterion_group! {
name = wiki_benches;
config = Criterion::default().with_profiler(PProfProfiler::new(100, Output::Flamegraph(None)));
targets = wiki_index_benchmark
}
criterion_main!(benches, gh_benches, wiki_benches);

1000
benches/wiki.json Normal file
View File

File diff suppressed because one or more lines are too long

2
bitpacker/Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy-bitpacker"
version = "0.3.0"
version = "0.5.0"
edition = "2021"
authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT"

4
bitpacker/src/blocked_bitpacker.rs
View File

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

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

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

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

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

90
cliff.toml Normal file
View File

@@ -0,0 +1,90 @@
# configuration file for git-cliff{ pattern = "foo", replace = "bar"}
# see https://github.com/orhun/git-cliff#configuration-file
[changelog]
# changelog header
header = """
"""
# template for the changelog body
# https://tera.netlify.app/docs/#introduction
body = """
{% if version %}\
{{ version | trim_start_matches(pat="v") }} ({{ timestamp | date(format="%Y-%m-%d") }})
==================
{% else %}\
## [unreleased]
{% endif %}\
{% for commit in commits %}
- {% if commit.breaking %}[**breaking**] {% endif %}{{ commit.message | split(pat="\n") | first | trim | upper_first }}(@{{ commit.author.name }})\
{% endfor %}
"""
# remove the leading and trailing whitespace from the template
trim = true
# changelog footer
footer = """
"""
postprocessors = [
{ pattern = 'Paul Masurel', replace = "fulmicoton"}, # replace with github user
{ pattern = 'PSeitz', replace = "PSeitz"}, # replace with github user
{ pattern = 'Adam Reichold', replace = "adamreichold"}, # replace with github user
{ pattern = 'trinity-1686a', replace = "trinity-1686a"}, # replace with github user
{ pattern = 'Michael Kleen', replace = "mkleen"}, # replace with github user
{ pattern = 'Adrien Guillo', replace = "guilload"}, # replace with github user
{ pattern = 'François Massot', replace = "fmassot"}, # replace with github user
{ pattern = 'Naveen Aiathurai', replace = "naveenann"}, # replace with github user
{ pattern = '', replace = ""}, # replace with github user
]
[git]
# parse the commits based on https://www.conventionalcommits.org
# This is required or commit.message contains the whole commit message and not just the title
conventional_commits = true
# filter out the commits that are not conventional
filter_unconventional = false
# process each line of a commit as an individual commit
split_commits = false
# regex for preprocessing the commit messages
commit_preprocessors = [
{ pattern = '\((\w+\s)?#([0-9]+)\)', replace = "[#${2}](https://github.com/quickwit-oss/tantivy/issues/${2})"}, # replace issue numbers
]
#link_parsers = [
#{ pattern = "#(\\d+)", href = "https://github.com/quickwit-oss/tantivy/pulls/$1"},
#]
# regex for parsing and grouping commits
commit_parsers = [
{ message = "^feat", group = "Features"},
{ message = "^fix", group = "Bug Fixes"},
{ message = "^doc", group = "Documentation"},
{ message = "^perf", group = "Performance"},
{ message = "^refactor", group = "Refactor"},
{ message = "^style", group = "Styling"},
{ message = "^test", group = "Testing"},
{ message = "^chore\\(release\\): prepare for", skip = true},
{ message = "(?i)clippy", skip = true},
{ message = "(?i)dependabot", skip = true},
{ message = "(?i)fmt", skip = true},
{ message = "(?i)bump", skip = true},
{ message = "(?i)readme", skip = true},
{ message = "(?i)comment", skip = true},
{ message = "(?i)spelling", skip = true},
{ message = "^chore", group = "Miscellaneous Tasks"},
{ body = ".*security", group = "Security"},
{ message = ".*", group = "Other", default_scope = "other"},
]
# protect breaking changes from being skipped due to matching a skipping commit_parser
protect_breaking_commits = false
# filter out the commits that are not matched by commit parsers
filter_commits = false
# glob pattern for matching git tags
tag_pattern = "v[0-9]*"
# regex for skipping tags
skip_tags = "v0.1.0-beta.1"
# regex for ignoring tags
ignore_tags = ""
# sort the tags topologically
topo_order = false
# sort the commits inside sections by oldest/newest order
sort_commits = "newest"
# limit the number of commits included in the changelog.
# limit_commits = 42

22
columnar/Cargo.toml
View File

@@ -1,28 +1,28 @@
[package]
name = "tantivy-columnar"
version = "0.1.0"
version = "0.2.0"
edition = "2021"
license = "MIT"
homepage = "https://github.com/quickwit-oss/tantivy"
repository = "https://github.com/quickwit-oss/tantivy"
description = "column oriented storage for tantivy"
categories = ["database-implementations", "data-structures", "compression"]
[dependencies]
itertools = "0.10.5"
log = "0.4.17"
itertools = "0.11.0"
fnv = "1.0.7"
fastdivide = "0.4.0"
rand = { version = "0.8.5", optional = true }
measure_time = { version = "0.8.2", optional = true }
prettytable-rs = { version = "0.10.0", optional = true }
stacker = { path = "../stacker", package="tantivy-stacker"}
sstable = { path = "../sstable", package = "tantivy-sstable" }
common = { path = "../common", package = "tantivy-common" }
tantivy-bitpacker = { version= "0.3", path = "../bitpacker/" }
stacker = { version= "0.2", path = "../stacker", package="tantivy-stacker"}
sstable = { version= "0.2", path = "../sstable", package = "tantivy-sstable" }
common = { version= "0.6", path = "../common", package = "tantivy-common" }
tantivy-bitpacker = { version= "0.5", path = "../bitpacker/" }
serde = "1.0.152"
[dev-dependencies]
proptest = "1"
more-asserts = "0.3.1"
rand = "0.8.5"
rand = "0.8"
[features]
unstable = []

96
columnar/src/block_accessor.rs
View File

@@ -1,9 +1,12 @@
use std::cmp::Ordering;
use crate::{Column, DocId, RowId};
#[derive(Debug, Default, Clone)]
pub struct ColumnBlockAccessor<T> {
val_cache: Vec<T>,
docid_cache: Vec<DocId>,
missing_docids_cache: Vec<DocId>,
row_id_cache: Vec<RowId>,
}
@@ -20,6 +23,20 @@ impl<T: PartialOrd + Copy + std::fmt::Debug + Send + Sync + 'static + Default>
.values
.get_vals(&self.row_id_cache, &mut self.val_cache);
}
#[inline]
pub fn fetch_block_with_missing(&mut self, docs: &[u32], accessor: &Column<T>, missing: T) {
self.fetch_block(docs, accessor);
// We can compare docid_cache with docs to find missing docs
if docs.len() != self.docid_cache.len() || accessor.index.is_multivalue() {
self.missing_docids_cache.clear();
find_missing_docs(docs, &self.docid_cache, |doc| {
self.missing_docids_cache.push(doc);
self.val_cache.push(missing);
});
self.docid_cache
.extend_from_slice(&self.missing_docids_cache);
}
}
#[inline]
pub fn iter_vals(&self) -> impl Iterator<Item = T> + '_ {
@@ -34,3 +51,82 @@ impl<T: PartialOrd + Copy + std::fmt::Debug + Send + Sync + 'static + Default>
.zip(self.val_cache.iter().cloned())
}
}
/// Given two sorted lists of docids `docs` and `hits`, hits is a subset of `docs`.
/// Return all docs that are not in `hits`.
fn find_missing_docs<F>(docs: &[u32], hits: &[u32], mut callback: F)
where F: FnMut(u32) {
let mut docs_iter = docs.iter();
let mut hits_iter = hits.iter();
let mut doc = docs_iter.next();
let mut hit = hits_iter.next();
while let (Some(&current_doc), Some(&current_hit)) = (doc, hit) {
match current_doc.cmp(&current_hit) {
Ordering::Less => {
callback(current_doc);
doc = docs_iter.next();
}
Ordering::Equal => {
doc = docs_iter.next();
hit = hits_iter.next();
}
Ordering::Greater => {
hit = hits_iter.next();
}
}
}
while let Some(&current_doc) = doc {
callback(current_doc);
doc = docs_iter.next();
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn test_find_missing_docs() {
let docs: Vec<u32> = vec![1, 2, 3, 4, 5, 6, 7, 8, 9, 10];
let hits: Vec<u32> = vec![2, 4, 6, 8, 10];
let mut missing_docs: Vec<u32> = Vec::new();
find_missing_docs(&docs, &hits, |missing_doc| {
missing_docs.push(missing_doc);
});
assert_eq!(missing_docs, vec![1, 3, 5, 7, 9]);
}
#[test]
fn test_find_missing_docs_empty() {
let docs: Vec<u32> = Vec::new();
let hits: Vec<u32> = vec![2, 4, 6, 8, 10];
let mut missing_docs: Vec<u32> = Vec::new();
find_missing_docs(&docs, &hits, |missing_doc| {
missing_docs.push(missing_doc);
});
assert_eq!(missing_docs, vec![]);
}
#[test]
fn test_find_missing_docs_all_missing() {
let docs: Vec<u32> = vec![1, 2, 3, 4, 5];
let hits: Vec<u32> = Vec::new();
let mut missing_docs: Vec<u32> = Vec::new();
find_missing_docs(&docs, &hits, |missing_doc| {
missing_docs.push(missing_doc);
});
assert_eq!(missing_docs, vec![1, 2, 3, 4, 5]);
}
}

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

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

2
columnar/src/column/mod.rs
View File

@@ -130,7 +130,7 @@ impl<T: PartialOrd + Copy + Debug + Send + Sync + 'static> Column<T> {
.select_batch_in_place(selected_docid_range.start, doc_ids);
}
/// Fils the output vector with the (possibly multiple values that are associated_with
/// Fills the output vector with the (possibly multiple values that are associated_with
/// `row_id`.
///
/// This method clears the `output` vector.

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

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

8
columnar/src/column_index/merge/shuffled.rs
View File

@@ -157,7 +157,13 @@ mod tests {
Cardinality::Optional,
&shuffle_merge_order,
);
let SerializableColumnIndex::Optional { non_null_row_ids, num_rows } = serializable_index else { panic!() };
let SerializableColumnIndex::Optional {
non_null_row_ids,
num_rows,
} = serializable_index
else {
panic!()
};
assert_eq!(num_rows, 2);
let non_null_rows: Vec<RowId> = non_null_row_ids.boxed_iter().collect();
assert_eq!(&non_null_rows, &[1]);

4
columnar/src/column_index/mod.rs
View File

@@ -37,6 +37,10 @@ impl From<MultiValueIndex> for ColumnIndex {
}
impl ColumnIndex {
#[inline]
pub fn is_multivalue(&self) -> bool {
matches!(self, ColumnIndex::Multivalued(_))
}
// Returns the cardinality of the column index.
//
// By convention, if the column contains no docs, we consider that it is

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

@@ -2,7 +2,7 @@
//! # `fastfield_codecs`
//!
//! - Columnar storage of data for tantivy [`Column`].
//! - Columnar storage of data for tantivy [`crate::Column`].
//! - Encode data in different codecs.
//! - Monotonically map values to u64/u128

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

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

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

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

3
columnar/src/column_values/u64_based/bitpacked.rs
View File

@@ -83,7 +83,8 @@ impl ColumnValues for BitpackedReader {
doc_id_range: Range<u32>,
positions: &mut Vec<u32>,
) {
let Some(transformed_range) = transform_range_before_linear_transformation(&self.stats, range)
let Some(transformed_range) =
transform_range_before_linear_transformation(&self.stats, range)
else {
positions.clear();
return;

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

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

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

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

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

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

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

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

241
columnar/src/columnar/merge/mod.rs
View File

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

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

@@ -1,3 +1,5 @@
use std::collections::BTreeMap;
use itertools::Itertools;
use super::*;
@@ -25,70 +27,73 @@ fn test_column_coercion_to_u64() {
let columnar1 = make_columnar("numbers", &[1i64]);
// u64 type
let columnar2 = make_columnar("numbers", &[u64::MAX]);
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(&[&columnar1, &columnar2], &[]).unwrap();
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(columnars, &[], &merge_order).unwrap();
assert_eq!(column_map.len(), 1);
assert!(column_map.contains_key(&("numbers".to_string(), ColumnType::U64)));
}
#[test]
fn test_column_no_coercion_if_all_the_same() {
let columnar1 = make_columnar("numbers", &[1u64]);
let columnar2 = make_columnar("numbers", &[2u64]);
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(&[&columnar1, &columnar2], &[]).unwrap();
assert_eq!(column_map.len(), 1);
assert!(column_map.contains_key(&("numbers".to_string(), ColumnType::U64)));
assert!(column_map.contains_key(&("numbers".to_string(), ColumnTypeCategory::Numerical)));
}
#[test]
fn test_column_coercion_to_i64() {
let columnar1 = make_columnar("numbers", &[-1i64]);
let columnar2 = make_columnar("numbers", &[2u64]);
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(&[&columnar1, &columnar2], &[]).unwrap();
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(columnars, &[], &merge_order).unwrap();
assert_eq!(column_map.len(), 1);
assert!(column_map.contains_key(&("numbers".to_string(), ColumnType::I64)));
assert!(column_map.contains_key(&("numbers".to_string(), ColumnTypeCategory::Numerical)));
}
#[test]
fn test_impossible_coercion_returns_an_error() {
let columnar1 = make_columnar("numbers", &[u64::MAX]);
let group_error =
group_columns_for_merge(&[&columnar1], &[("numbers".to_string(), ColumnType::I64)])
.map(|_| ())
.unwrap_err();
assert_eq!(group_error.kind(), io::ErrorKind::InvalidInput);
}
//#[test]
// fn test_impossible_coercion_returns_an_error() {
// let columnar1 = make_columnar("numbers", &[u64::MAX]);
// let merge_order = StackMergeOrder::stack(&[&columnar1]).into();
// let group_error = group_columns_for_merge_iter(
//&[&columnar1],
//&[("numbers".to_string(), ColumnType::I64)],
//&merge_order,
//)
//.unwrap_err();
// assert_eq!(group_error.kind(), io::ErrorKind::InvalidInput);
//}
#[test]
fn test_group_columns_with_required_column() {
let columnar1 = make_columnar("numbers", &[1i64]);
let columnar2 = make_columnar("numbers", &[2u64]);
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(
&[&columnar1, &columnar2],
&[("numbers".to_string(), ColumnType::U64)],
&merge_order,
)
.unwrap();
assert_eq!(column_map.len(), 1);
assert!(column_map.contains_key(&("numbers".to_string(), ColumnType::U64)));
assert!(column_map.contains_key(&("numbers".to_string(), ColumnTypeCategory::Numerical)));
}
#[test]
fn test_group_columns_required_column_with_no_existing_columns() {
let columnar1 = make_columnar("numbers", &[2u64]);
let columnar2 = make_columnar("numbers", &[2u64]);
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(
&[&columnar1, &columnar2],
&[("required_col".to_string(), ColumnType::Str)],
)
.unwrap();
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<_, _> = group_columns_for_merge(
columnars,
&[("required_col".to_string(), ColumnType::Str)],
&merge_order,
)
.unwrap();
assert_eq!(column_map.len(), 2);
let columns = column_map
.get(&("required_col".to_string(), ColumnType::Str))
.unwrap();
let columns = &column_map
.get(&("required_col".to_string(), ColumnTypeCategory::Str))
.unwrap()
.columns;
assert_eq!(columns.len(), 2);
assert!(columns[0].is_none());
assert!(columns[1].is_none());
@@ -98,35 +103,42 @@ fn test_group_columns_required_column_with_no_existing_columns() {
fn test_group_columns_required_column_is_above_all_columns_have_the_same_type_rule() {
let columnar1 = make_columnar("numbers", &[2i64]);
let columnar2 = make_columnar("numbers", &[2i64]);
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(
&[&columnar1, &columnar2],
columnars,
&[("numbers".to_string(), ColumnType::U64)],
&merge_order,
)
.unwrap();
assert_eq!(column_map.len(), 1);
assert!(column_map.contains_key(&("numbers".to_string(), ColumnType::U64)));
assert!(column_map.contains_key(&("numbers".to_string(), ColumnTypeCategory::Numerical)));
}
#[test]
fn test_missing_column() {
let columnar1 = make_columnar("numbers", &[-1i64]);
let columnar2 = make_columnar("numbers2", &[2u64]);
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(&[&columnar1, &columnar2], &[]).unwrap();
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(columnars, &[], &merge_order).unwrap();
assert_eq!(column_map.len(), 2);
assert!(column_map.contains_key(&("numbers".to_string(), ColumnType::I64)));
assert!(column_map.contains_key(&("numbers".to_string(), ColumnTypeCategory::Numerical)));
{
let columns = column_map
.get(&("numbers".to_string(), ColumnType::I64))
.unwrap();
let columns = &column_map
.get(&("numbers".to_string(), ColumnTypeCategory::Numerical))
.unwrap()
.columns;
assert!(columns[0].is_some());
assert!(columns[1].is_none());
}
{
let columns = column_map
.get(&("numbers2".to_string(), ColumnType::U64))
.unwrap();
let columns = &column_map
.get(&("numbers2".to_string(), ColumnTypeCategory::Numerical))
.unwrap()
.columns;
assert!(columns[0].is_none());
assert!(columns[1].is_some());
}
@@ -224,7 +236,9 @@ fn test_merge_columnar_numbers() {
assert_eq!(columnar_reader.num_columns(), 1);
let cols = columnar_reader.read_columns("numbers").unwrap();
let dynamic_column = cols[0].open().unwrap();
let DynamicColumn::F64(vals) = dynamic_column else { panic!() };
let DynamicColumn::F64(vals) = dynamic_column else {
panic!()
};
assert_eq!(vals.get_cardinality(), Cardinality::Optional);
assert_eq!(vals.first(0u32), Some(-1f64));
assert_eq!(vals.first(1u32), None);
@@ -250,7 +264,9 @@ fn test_merge_columnar_texts() {
assert_eq!(columnar_reader.num_columns(), 1);
let cols = columnar_reader.read_columns("texts").unwrap();
let dynamic_column = cols[0].open().unwrap();
let DynamicColumn::Str(vals) = dynamic_column else { panic!() };
let DynamicColumn::Str(vals) = dynamic_column else {
panic!()
};
assert_eq!(vals.ords().get_cardinality(), Cardinality::Optional);
let get_str_for_ord = |ord| {
@@ -297,7 +313,9 @@ fn test_merge_columnar_byte() {
assert_eq!(columnar_reader.num_columns(), 1);
let cols = columnar_reader.read_columns("bytes").unwrap();
let dynamic_column = cols[0].open().unwrap();
let DynamicColumn::Bytes(vals) = dynamic_column else { panic!() };
let DynamicColumn::Bytes(vals) = dynamic_column else {
panic!()
};
let get_bytes_for_ord = |ord| {
let mut out = Vec::new();
vals.ord_to_bytes(ord, &mut out).unwrap();
@@ -351,7 +369,9 @@ fn test_merge_columnar_byte_with_missing() {
assert_eq!(columnar_reader.num_columns(), 2);
let cols = columnar_reader.read_columns("col").unwrap();
let dynamic_column = cols[0].open().unwrap();
let DynamicColumn::Bytes(vals) = dynamic_column else { panic!() };
let DynamicColumn::Bytes(vals) = dynamic_column else {
panic!()
};
let get_bytes_for_ord = |ord| {
let mut out = Vec::new();
vals.ord_to_bytes(ord, &mut out).unwrap();
@@ -403,7 +423,9 @@ fn test_merge_columnar_different_types() {
// numeric column
let dynamic_column = cols[0].open().unwrap();
let DynamicColumn::I64(vals) = dynamic_column else { panic!() };
let DynamicColumn::I64(vals) = dynamic_column else {
panic!()
};
assert_eq!(vals.get_cardinality(), Cardinality::Optional);
assert_eq!(vals.values_for_doc(0).collect_vec(), vec![]);
assert_eq!(vals.values_for_doc(1).collect_vec(), vec![]);
@@ -413,7 +435,9 @@ fn test_merge_columnar_different_types() {
// text column
let dynamic_column = cols[1].open().unwrap();
let DynamicColumn::Str(vals) = dynamic_column else { panic!() };
let DynamicColumn::Str(vals) = dynamic_column else {
panic!()
};
assert_eq!(vals.ords().get_cardinality(), Cardinality::Optional);
let get_str_for_ord = |ord| {
let mut out = String::new();

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

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

35
columnar/src/columnar/writer/mod.rs
View File

@@ -79,7 +79,6 @@ fn mutate_or_create_column<V, TMutator>(
impl ColumnarWriter {
pub fn mem_usage(&self) -> usize {
// TODO add dictionary builders.
self.arena.mem_usage()
+ self.numerical_field_hash_map.mem_usage()
+ self.bool_field_hash_map.mem_usage()
@@ -87,6 +86,11 @@ impl ColumnarWriter {
+ self.str_field_hash_map.mem_usage()
+ self.ip_addr_field_hash_map.mem_usage()
+ self.datetime_field_hash_map.mem_usage()
+ self
.dictionaries
.iter()
.map(|dict| dict.mem_usage())
.sum::<usize>()
}
/// Returns the list of doc ids from 0..num_docs sorted by the `sort_field`
@@ -98,9 +102,15 @@ impl ColumnarWriter {
///
/// The sort applied is stable.
pub fn sort_order(&self, sort_field: &str, num_docs: RowId, reversed: bool) -> Vec<u32> {
let Some(numerical_col_writer) =
self.numerical_field_hash_map.get::<NumericalColumnWriter>(sort_field.as_bytes()) else {
return Vec::new();
let Some(numerical_col_writer) = self
.numerical_field_hash_map
.get::<NumericalColumnWriter>(sort_field.as_bytes())
.or_else(|| {
self.datetime_field_hash_map
.get::<NumericalColumnWriter>(sort_field.as_bytes())
})
else {
return Vec::new();
};
let mut symbols_buffer = Vec::new();
let mut values = Vec::new();
@@ -266,7 +276,7 @@ impl ColumnarWriter {
let mut column: ColumnWriter = column_opt.unwrap_or_default();
column.record(
doc,
NumericalValue::I64(datetime.into_timestamp_micros()),
NumericalValue::I64(datetime.into_timestamp_nanos()),
arena,
);
column
@@ -370,7 +380,7 @@ impl ColumnarWriter {
let column_writer: ColumnWriter = self.bool_field_hash_map.read(addr);
let cardinality = column_writer.get_cardinality(num_docs);
let mut column_serializer =
serializer.serialize_column(column_name, column_type);
serializer.start_serialize_column(column_name, column_type);
serialize_bool_column(
cardinality,
num_docs,
@@ -382,12 +392,13 @@ impl ColumnarWriter {
buffers,
&mut column_serializer,
)?;
column_serializer.finalize()?;
}
ColumnType::IpAddr => {
let column_writer: ColumnWriter = self.ip_addr_field_hash_map.read(addr);
let cardinality = column_writer.get_cardinality(num_docs);
let mut column_serializer =
serializer.serialize_column(column_name, ColumnType::IpAddr);
serializer.start_serialize_column(column_name, ColumnType::IpAddr);
serialize_ip_addr_column(
cardinality,
num_docs,
@@ -399,6 +410,7 @@ impl ColumnarWriter {
buffers,
&mut column_serializer,
)?;
column_serializer.finalize()?;
}
ColumnType::Bytes | ColumnType::Str => {
let str_or_bytes_column_writer: StrOrBytesColumnWriter =
@@ -413,7 +425,7 @@ impl ColumnarWriter {
.column_writer
.get_cardinality(num_docs);
let mut column_serializer =
serializer.serialize_column(column_name, column_type);
serializer.start_serialize_column(column_name, column_type);
serialize_bytes_or_str_column(
cardinality,
num_docs,
@@ -427,13 +439,14 @@ impl ColumnarWriter {
buffers,
&mut column_serializer,
)?;
column_serializer.finalize()?;
}
ColumnType::F64 | ColumnType::I64 | ColumnType::U64 => {
let numerical_column_writer: NumericalColumnWriter =
self.numerical_field_hash_map.read(addr);
let cardinality = numerical_column_writer.cardinality(num_docs);
let mut column_serializer =
serializer.serialize_column(column_name, column_type);
serializer.start_serialize_column(column_name, column_type);
let numerical_type = column_type.numerical_type().unwrap();
serialize_numerical_column(
cardinality,
@@ -447,12 +460,13 @@ impl ColumnarWriter {
buffers,
&mut column_serializer,
)?;
column_serializer.finalize()?;
}
ColumnType::DateTime => {
let column_writer: ColumnWriter = self.datetime_field_hash_map.read(addr);
let cardinality = column_writer.get_cardinality(num_docs);
let mut column_serializer =
serializer.serialize_column(column_name, ColumnType::DateTime);
serializer.start_serialize_column(column_name, ColumnType::DateTime);
serialize_numerical_column(
cardinality,
num_docs,
@@ -465,6 +479,7 @@ impl ColumnarWriter {
buffers,
&mut column_serializer,
)?;
column_serializer.finalize()?;
}
};
}

16
columnar/src/columnar/writer/serializer.rs
View File

@@ -34,11 +34,12 @@ impl<W: io::Write> ColumnarSerializer<W> {
}
}
pub fn serialize_column<'a>(
/// Creates a ColumnSerializer.
pub fn start_serialize_column<'a>(
&'a mut self,
column_name: &[u8],
column_type: ColumnType,
) -> impl io::Write + 'a {
) -> ColumnSerializer<'a, W> {
let start_offset = self.wrt.written_bytes();
prepare_key(column_name, column_type, &mut self.prepare_key_buffer);
ColumnSerializer {
@@ -60,20 +61,21 @@ impl<W: io::Write> ColumnarSerializer<W> {
}
}
struct ColumnSerializer<'a, W: io::Write> {
pub struct ColumnSerializer<'a, W: io::Write> {
columnar_serializer: &'a mut ColumnarSerializer<W>,
start_offset: u64,
}
impl<'a, W: io::Write> Drop for ColumnSerializer<'a, W> {
fn drop(&mut self) {
impl<'a, W: io::Write> ColumnSerializer<'a, W> {
pub fn finalize(self) -> io::Result<()> {
let end_offset: u64 = self.columnar_serializer.wrt.written_bytes();
let byte_range = self.start_offset..end_offset;
self.columnar_serializer.sstable_range.insert_cannot_fail(
self.columnar_serializer.sstable_range.insert(
&self.columnar_serializer.prepare_key_buffer[..],
&byte_range,
);
)?;
self.columnar_serializer.prepare_key_buffer.clear();
Ok(())
}
}

7
columnar/src/dictionary.rs
View File

@@ -32,6 +32,7 @@ pub struct OrderedId(pub u32);
#[derive(Default)]
pub(crate) struct DictionaryBuilder {
dict: FnvHashMap<Vec<u8>, UnorderedId>,
memory_consumption: usize,
}
impl DictionaryBuilder {
@@ -43,6 +44,8 @@ impl DictionaryBuilder {
}
let new_id = UnorderedId(self.dict.len() as u32);
self.dict.insert(term.to_vec(), new_id);
self.memory_consumption += term.len();
self.memory_consumption += 40; // Term Metadata + HashMap overhead
new_id
}
@@ -63,6 +66,10 @@ impl DictionaryBuilder {
sstable_builder.finish()?;
Ok(TermIdMapping { unordered_to_ord })
}
pub(crate) fn mem_usage(&self) -> usize {
self.memory_consumption
}
}
#[cfg(test)]

27
columnar/src/dynamic_column.rs
View File

@@ -26,14 +26,14 @@ impl fmt::Debug for DynamicColumn {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "[{} {} |", self.get_cardinality(), self.column_type())?;
match self {
DynamicColumn::Bool(col) => write!(f, " {:?}", col)?,
DynamicColumn::I64(col) => write!(f, " {:?}", col)?,
DynamicColumn::U64(col) => write!(f, " {:?}", col)?,
DynamicColumn::F64(col) => write!(f, "{:?}", col)?,
DynamicColumn::IpAddr(col) => write!(f, "{:?}", col)?,
DynamicColumn::DateTime(col) => write!(f, "{:?}", col)?,
DynamicColumn::Bytes(col) => write!(f, "{:?}", col)?,
DynamicColumn::Str(col) => write!(f, "{:?}", col)?,
DynamicColumn::Bool(col) => write!(f, " {col:?}")?,
DynamicColumn::I64(col) => write!(f, " {col:?}")?,
DynamicColumn::U64(col) => write!(f, " {col:?}")?,
DynamicColumn::F64(col) => write!(f, "{col:?}")?,
DynamicColumn::IpAddr(col) => write!(f, "{col:?}")?,
DynamicColumn::DateTime(col) => write!(f, "{col:?}")?,
DynamicColumn::Bytes(col) => write!(f, "{col:?}")?,
DynamicColumn::Str(col) => write!(f, "{col:?}")?,
}
write!(f, "]")
}
@@ -228,7 +228,7 @@ static_dynamic_conversions!(StrColumn, Str);
static_dynamic_conversions!(BytesColumn, Bytes);
static_dynamic_conversions!(Column<Ipv6Addr>, IpAddr);
#[derive(Clone)]
#[derive(Clone, Debug)]
pub struct DynamicColumnHandle {
pub(crate) file_slice: FileSlice,
pub(crate) column_type: ColumnType,
@@ -247,7 +247,7 @@ impl DynamicColumnHandle {
}
/// Returns the `u64` fast field reader reader associated with `fields` of types
/// Str, u64, i64, f64, or datetime.
/// Str, u64, i64, f64, bool, or datetime.
///
/// If not, the fastfield reader will returns the u64-value associated with the original
/// FastValue.
@@ -258,9 +258,12 @@ impl DynamicColumnHandle {
let column: BytesColumn = crate::column::open_column_bytes(column_bytes)?;
Ok(Some(column.term_ord_column))
}
ColumnType::Bool => Ok(None),
ColumnType::IpAddr => Ok(None),
ColumnType::I64 | ColumnType::U64 | ColumnType::F64 | ColumnType::DateTime => {
ColumnType::Bool
| ColumnType::I64
| ColumnType::U64
| ColumnType::F64
| ColumnType::DateTime => {
let column = crate::column::open_column_u64::<u64>(column_bytes)?;
Ok(Some(column))
}

2
columnar/src/lib.rs
View File

@@ -39,7 +39,7 @@ pub use self::dynamic_column::{DynamicColumn, DynamicColumnHandle};
pub type RowId = u32;
pub type DocId = u32;
#[derive(Clone, Copy)]
#[derive(Clone, Copy, Debug)]
pub struct RowAddr {
pub segment_ord: u32,
pub row_id: RowId,

258
columnar/src/tests.rs
View File

@@ -4,13 +4,15 @@ use std::net::Ipv6Addr;
use common::DateTime;
use proptest::prelude::*;
use proptest::sample::subsequence;
use crate::column_values::MonotonicallyMappableToU128;
use crate::columnar::{ColumnType, ColumnTypeCategory};
use crate::dynamic_column::{DynamicColumn, DynamicColumnHandle};
use crate::value::{Coerce, NumericalValue};
use crate::{
BytesColumn, Cardinality, Column, ColumnarReader, ColumnarWriter, RowId, StackMergeOrder,
BytesColumn, Cardinality, Column, ColumnarReader, ColumnarWriter, RowAddr, RowId,
ShuffleMergeOrder, StackMergeOrder,
};
#[test]
@@ -24,7 +26,7 @@ fn test_dataframe_writer_str() {
assert_eq!(columnar.num_columns(), 1);
let cols: Vec<DynamicColumnHandle> = columnar.read_columns("my_string").unwrap();
assert_eq!(cols.len(), 1);
assert_eq!(cols[0].num_bytes(), 89);
assert_eq!(cols[0].num_bytes(), 87);
}
#[test]
@@ -38,7 +40,7 @@ fn test_dataframe_writer_bytes() {
assert_eq!(columnar.num_columns(), 1);
let cols: Vec<DynamicColumnHandle> = columnar.read_columns("my_string").unwrap();
assert_eq!(cols.len(), 1);
assert_eq!(cols[0].num_bytes(), 89);
assert_eq!(cols[0].num_bytes(), 87);
}
#[test]
@@ -55,7 +57,9 @@ fn test_dataframe_writer_bool() {
assert_eq!(cols[0].num_bytes(), 22);
assert_eq!(cols[0].column_type(), ColumnType::Bool);
let dyn_bool_col = cols[0].open().unwrap();
let DynamicColumn::Bool(bool_col) = dyn_bool_col else { panic!(); };
let DynamicColumn::Bool(bool_col) = dyn_bool_col else {
panic!();
};
let vals: Vec<Option<bool>> = (0..5).map(|row_id| bool_col.first(row_id)).collect();
assert_eq!(&vals, &[None, Some(false), None, Some(true), None,]);
}
@@ -77,7 +81,9 @@ fn test_dataframe_writer_u64_multivalued() {
assert_eq!(cols.len(), 1);
assert_eq!(cols[0].num_bytes(), 29);
let dyn_i64_col = cols[0].open().unwrap();
let DynamicColumn::I64(divisor_col) = dyn_i64_col else { panic!(); };
let DynamicColumn::I64(divisor_col) = dyn_i64_col else {
panic!();
};
assert_eq!(
divisor_col.get_cardinality(),
crate::Cardinality::Multivalued
@@ -99,7 +105,9 @@ fn test_dataframe_writer_ip_addr() {
assert_eq!(cols[0].num_bytes(), 42);
assert_eq!(cols[0].column_type(), ColumnType::IpAddr);
let dyn_bool_col = cols[0].open().unwrap();
let DynamicColumn::IpAddr(ip_col) = dyn_bool_col else { panic!(); };
let DynamicColumn::IpAddr(ip_col) = dyn_bool_col else {
panic!();
};
let vals: Vec<Option<Ipv6Addr>> = (0..5).map(|row_id| ip_col.first(row_id)).collect();
assert_eq!(
&vals,
@@ -132,7 +140,9 @@ fn test_dataframe_writer_numerical() {
// - null footer 6 bytes
assert_eq!(cols[0].num_bytes(), 33);
let column = cols[0].open().unwrap();
let DynamicColumn::I64(column_i64) = column else { panic!(); };
let DynamicColumn::I64(column_i64) = column else {
panic!();
};
assert_eq!(column_i64.index.get_cardinality(), Cardinality::Optional);
assert_eq!(column_i64.first(0), None);
assert_eq!(column_i64.first(1), Some(12i64));
@@ -196,7 +206,9 @@ fn test_dictionary_encoded_str() {
assert_eq!(columnar_reader.num_columns(), 2);
let col_handles = columnar_reader.read_columns("my.column").unwrap();
assert_eq!(col_handles.len(), 1);
let DynamicColumn::Str(str_col) = col_handles[0].open().unwrap() else { panic!(); };
let DynamicColumn::Str(str_col) = col_handles[0].open().unwrap() else {
panic!();
};
let index: Vec<Option<u64>> = (0..5).map(|row_id| str_col.ords().first(row_id)).collect();
assert_eq!(index, &[None, Some(0), None, Some(2), Some(1)]);
assert_eq!(str_col.num_rows(), 5);
@@ -228,7 +240,9 @@ fn test_dictionary_encoded_bytes() {
assert_eq!(columnar_reader.num_columns(), 2);
let col_handles = columnar_reader.read_columns("my.column").unwrap();
assert_eq!(col_handles.len(), 1);
let DynamicColumn::Bytes(bytes_col) = col_handles[0].open().unwrap() else { panic!(); };
let DynamicColumn::Bytes(bytes_col) = col_handles[0].open().unwrap() else {
panic!();
};
let index: Vec<Option<u64>> = (0..5)
.map(|row_id| bytes_col.ords().first(row_id))
.collect();
@@ -260,12 +274,15 @@ fn test_dictionary_encoded_bytes() {
fn num_strategy() -> impl Strategy<Value = NumericalValue> {
prop_oneof![
Just(NumericalValue::U64(0u64)),
Just(NumericalValue::U64(u64::MAX)),
Just(NumericalValue::I64(0i64)),
Just(NumericalValue::I64(i64::MIN)),
Just(NumericalValue::I64(i64::MAX)),
Just(NumericalValue::F64(1.2f64)),
3 => Just(NumericalValue::U64(0u64)),
3 => Just(NumericalValue::U64(u64::MAX)),
3 => Just(NumericalValue::I64(0i64)),
3 => Just(NumericalValue::I64(i64::MIN)),
3 => Just(NumericalValue::I64(i64::MAX)),
3 => Just(NumericalValue::F64(1.2f64)),
1 => any::<f64>().prop_map(NumericalValue::from),
1 => any::<u64>().prop_map(NumericalValue::from),
1 => any::<i64>().prop_map(NumericalValue::from),
]
}
@@ -279,6 +296,12 @@ enum ColumnValue {
DateTime(DateTime),
}
impl<T: Into<NumericalValue>> From<T> for ColumnValue {
fn from(val: T) -> ColumnValue {
ColumnValue::Numerical(val.into())
}
}
impl ColumnValue {
pub(crate) fn column_type_category(&self) -> ColumnTypeCategory {
match self {
@@ -328,12 +351,22 @@ fn column_value_strategy() -> impl Strategy<Value = ColumnValue> {
// A document contains up to 4 values.
fn doc_strategy() -> impl Strategy<Value = Vec<(&'static str, ColumnValue)>> {
proptest::collection::vec((column_name_strategy(), column_value_strategy()), 0..4)
proptest::collection::vec((column_name_strategy(), column_value_strategy()), 0..=4)
}
fn num_docs_strategy() -> impl Strategy<Value = usize> {
prop_oneof!(
// We focus heavily on the 0..2 case as we assume it is sufficient to cover all edge cases.
0usize..=3usize,
// We leave 50% of the effort exploring more defensively.
3usize..=12usize
)
}
// A columnar contains up to 2 docs.
fn columnar_docs_strategy() -> impl Strategy<Value = Vec<Vec<(&'static str, ColumnValue)>>> {
proptest::collection::vec(doc_strategy(), 0..=2)
num_docs_strategy()
.prop_flat_map(|num_docs| proptest::collection::vec(doc_strategy(), num_docs))
}
fn columnar_docs_and_mapping_strategy(
@@ -347,6 +380,11 @@ fn permutation_strategy(n: usize) -> impl Strategy<Value = Vec<RowId>> {
Just((0u32..n as RowId).collect()).prop_shuffle()
}
fn permutation_and_subset_strategy(n: usize) -> impl Strategy<Value = Vec<usize>> {
let vals: Vec<usize> = (0..n).collect();
subsequence(vals, 0..=n).prop_shuffle()
}
fn build_columnar_with_mapping(
docs: &[Vec<(&'static str, ColumnValue)>],
old_to_new_row_ids_opt: Option<&[RowId]>,
@@ -389,7 +427,15 @@ fn build_columnar(docs: &[Vec<(&'static str, ColumnValue)>]) -> ColumnarReader {
build_columnar_with_mapping(docs, None)
}
fn assert_columnar_eq(left: &ColumnarReader, right: &ColumnarReader) {
fn assert_columnar_eq_strict(left: &ColumnarReader, right: &ColumnarReader) {
assert_columnar_eq(left, right, false);
}
fn assert_columnar_eq(
left: &ColumnarReader,
right: &ColumnarReader,
lenient_on_numerical_value: bool,
) {
assert_eq!(left.num_rows(), right.num_rows());
let left_columns = left.list_columns().unwrap();
let right_columns = right.list_columns().unwrap();
@@ -398,7 +444,7 @@ fn assert_columnar_eq(left: &ColumnarReader, right: &ColumnarReader) {
assert_eq!(left_columns[i].0, right_columns[i].0);
let left_column = left_columns[i].1.open().unwrap();
let right_column = right_columns[i].1.open().unwrap();
assert_dyn_column_eq(&left_column, &right_column);
assert_dyn_column_eq(&left_column, &right_column, lenient_on_numerical_value);
}
}
@@ -442,11 +488,11 @@ fn assert_bytes_column_eq(left: &BytesColumn, right: &BytesColumn) {
assert!(!right_terms.advance());
}
fn assert_dyn_column_eq(left_dyn_column: &DynamicColumn, right_dyn_column: &DynamicColumn) {
assert_eq!(
&left_dyn_column.column_type(),
&right_dyn_column.column_type()
);
fn assert_dyn_column_eq(
left_dyn_column: &DynamicColumn,
right_dyn_column: &DynamicColumn,
lenient_on_numerical_value: bool,
) {
assert_eq!(
&left_dyn_column.get_cardinality(),
&right_dyn_column.get_cardinality()
@@ -476,8 +522,19 @@ fn assert_dyn_column_eq(left_dyn_column: &DynamicColumn, right_dyn_column: &Dyna
(DynamicColumn::Str(left_col), DynamicColumn::Str(right_col)) => {
assert_bytes_column_eq(left_col, right_col);
}
_ => {
unreachable!()
(left, right) => {
if lenient_on_numerical_value {
assert_eq!(
ColumnTypeCategory::from(left.column_type()),
ColumnTypeCategory::from(right.column_type())
);
} else {
panic!(
"Column type are not the same: {:?} vs {:?}",
left.column_type(),
right.column_type()
);
}
}
}
}
@@ -488,28 +545,36 @@ trait AssertEqualToColumnValue {
impl AssertEqualToColumnValue for bool {
fn assert_equal_to_column_value(&self, column_value: &ColumnValue) {
let ColumnValue::Bool(val) = column_value else { panic!() };
let ColumnValue::Bool(val) = column_value else {
panic!()
};
assert_eq!(self, val);
}
}
impl AssertEqualToColumnValue for Ipv6Addr {
fn assert_equal_to_column_value(&self, column_value: &ColumnValue) {
let ColumnValue::IpAddr(val) = column_value else { panic!() };
let ColumnValue::IpAddr(val) = column_value else {
panic!()
};
assert_eq!(self, val);
}
}
impl<T: Coerce + PartialEq + Debug + Into<NumericalValue>> AssertEqualToColumnValue for T {
fn assert_equal_to_column_value(&self, column_value: &ColumnValue) {
let ColumnValue::Numerical(num) = column_value else { panic!() };
let ColumnValue::Numerical(num) = column_value else {
panic!()
};
assert_eq!(self, &T::coerce(*num));
}
}
impl AssertEqualToColumnValue for DateTime {
fn assert_equal_to_column_value(&self, column_value: &ColumnValue) {
let ColumnValue::DateTime(dt) = column_value else { panic!() };
let ColumnValue::DateTime(dt) = column_value else {
panic!()
};
assert_eq!(self, dt);
}
}
@@ -683,7 +748,7 @@ proptest! {
let merged_columnar = ColumnarReader::open(output).unwrap();
let concat_rows: Vec<Vec<(&'static str, ColumnValue)>> = columnar_docs.iter().cloned().flatten().collect();
let expected_merged_columnar = build_columnar(&concat_rows[..]);
assert_columnar_eq(&merged_columnar, &expected_merged_columnar);
assert_columnar_eq_strict(&merged_columnar, &expected_merged_columnar);
}
}
@@ -709,7 +774,7 @@ fn test_columnar_merging_empty_columnar() {
let concat_rows: Vec<Vec<(&'static str, ColumnValue)>> =
columnar_docs.iter().cloned().flatten().collect();
let expected_merged_columnar = build_columnar(&concat_rows[..]);
assert_columnar_eq(&merged_columnar, &expected_merged_columnar);
assert_columnar_eq_strict(&merged_columnar, &expected_merged_columnar);
}
#[test]
@@ -746,8 +811,135 @@ fn test_columnar_merging_number_columns() {
let concat_rows: Vec<Vec<(&'static str, ColumnValue)>> =
columnar_docs.iter().cloned().flatten().collect();
let expected_merged_columnar = build_columnar(&concat_rows[..]);
assert_columnar_eq(&merged_columnar, &expected_merged_columnar);
assert_columnar_eq_strict(&merged_columnar, &expected_merged_columnar);
}
// TODO add non trivial remap and merge
// TODO test required_columns
// TODO document edge case: required_columns incompatible with values.
fn columnar_docs_and_remap(
) -> impl Strategy<Value = (Vec<Vec<Vec<(&'static str, ColumnValue)>>>, Vec<RowAddr>)> {
proptest::collection::vec(columnar_docs_strategy(), 2..=3).prop_flat_map(
|columnars_docs: Vec<Vec<Vec<(&str, ColumnValue)>>>| {
let row_addrs: Vec<RowAddr> = columnars_docs
.iter()
.enumerate()
.flat_map(|(segment_ord, columnar_docs)| {
(0u32..columnar_docs.len() as u32).map(move |row_id| RowAddr {
segment_ord: segment_ord as u32,
row_id,
})
})
.collect();
permutation_and_subset_strategy(row_addrs.len()).prop_map(move |shuffled_subset| {
let shuffled_row_addr_subset: Vec<RowAddr> =
shuffled_subset.iter().map(|ord| row_addrs[*ord]).collect();
(columnars_docs.clone(), shuffled_row_addr_subset)
})
},
)
}
proptest! {
#![proptest_config(ProptestConfig::with_cases(1000))]
#[test]
fn test_columnar_merge_and_remap_proptest((columnar_docs, shuffle_merge_order) in columnar_docs_and_remap()) {
let shuffled_rows: Vec<Vec<(&'static str, ColumnValue)>> = shuffle_merge_order.iter()
.map(|row_addr| columnar_docs[row_addr.segment_ord as usize][row_addr.row_id as usize].clone())
.collect();
let expected_merged_columnar = build_columnar(&shuffled_rows[..]);
let columnar_readers: Vec<ColumnarReader> = columnar_docs.iter()
.map(|docs| build_columnar(&docs[..]))
.collect::<Vec<_>>();
let columnar_readers_arr: Vec<&ColumnarReader> = columnar_readers.iter().collect();
let mut output: Vec<u8> = Vec::new();
let segment_num_rows: Vec<RowId> = columnar_docs.iter().map(|docs| docs.len() as RowId).collect();
let shuffle_merge_order = ShuffleMergeOrder::for_test(&segment_num_rows, shuffle_merge_order);
crate::merge_columnar(&columnar_readers_arr[..], &[], shuffle_merge_order.into(), &mut output).unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
assert_columnar_eq(&merged_columnar, &expected_merged_columnar, true);
}
}
#[test]
fn test_columnar_merge_empty() {
let columnar_reader_1 = build_columnar(&[]);
let rows: &[Vec<_>] = &[vec![("c1", ColumnValue::Str("a"))]][..];
let columnar_reader_2 = build_columnar(rows);
let mut output: Vec<u8> = Vec::new();
let segment_num_rows: Vec<RowId> = vec![0, 0];
let shuffle_merge_order = ShuffleMergeOrder::for_test(&segment_num_rows, vec![]);
crate::merge_columnar(
&[&columnar_reader_1, &columnar_reader_2],
&[],
shuffle_merge_order.into(),
&mut output,
)
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
assert_eq!(merged_columnar.num_rows(), 0);
assert_eq!(merged_columnar.num_columns(), 0);
}
#[test]
fn test_columnar_merge_single_str_column() {
let columnar_reader_1 = build_columnar(&[]);
let rows: &[Vec<_>] = &[vec![("c1", ColumnValue::Str("a"))]][..];
let columnar_reader_2 = build_columnar(rows);
let mut output: Vec<u8> = Vec::new();
let segment_num_rows: Vec<RowId> = vec![0, 1];
let shuffle_merge_order = ShuffleMergeOrder::for_test(
&segment_num_rows,
vec![RowAddr {
segment_ord: 1u32,
row_id: 0u32,
}],
);
crate::merge_columnar(
&[&columnar_reader_1, &columnar_reader_2],
&[],
shuffle_merge_order.into(),
&mut output,
)
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
assert_eq!(merged_columnar.num_rows(), 1);
assert_eq!(merged_columnar.num_columns(), 1);
}
#[test]
fn test_delete_decrease_cardinality() {
let columnar_reader_1 = build_columnar(&[]);
let rows: &[Vec<_>] = &[
vec![
("c", ColumnValue::from(0i64)),
("c", ColumnValue::from(0i64)),
],
vec![("c", ColumnValue::from(0i64))],
][..];
// c is multivalued here
let columnar_reader_2 = build_columnar(rows);
let mut output: Vec<u8> = Vec::new();
let shuffle_merge_order = ShuffleMergeOrder::for_test(
&[0, 2],
vec![RowAddr {
segment_ord: 1u32,
row_id: 1u32,
}],
);
crate::merge_columnar(
&[&columnar_reader_1, &columnar_reader_2],
&[],
shuffle_merge_order.into(),
&mut output,
)
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
assert_eq!(merged_columnar.num_rows(), 1);
assert_eq!(merged_columnar.num_columns(), 1);
let cols = merged_columnar.read_columns("c").unwrap();
assert_eq!(cols.len(), 1);
assert_eq!(cols[0].column_type(), ColumnType::I64);
assert_eq!(cols[0].open().unwrap().get_cardinality(), Cardinality::Full);
}

2
columnar/src/value.rs
View File

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

4
common/Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy-common"
version = "0.5.0"
version = "0.6.0"
authors = ["Paul Masurel <paul@quickwit.io>", "Pascal Seitz <pascal@quickwit.io>"]
license = "MIT"
edition = "2021"
@@ -14,7 +14,7 @@ repository = "https://github.com/quickwit-oss/tantivy"
[dependencies]
byteorder = "1.4.3"
ownedbytes = { version= "0.5", path="../ownedbytes" }
ownedbytes = { version= "0.6", path="../ownedbytes" }
async-trait = "0.1"
time = { version = "0.3.10", features = ["serde-well-known"] }
serde = { version = "1.0.136", features = ["derive"] }

39
common/benches/bench.rs Normal file
View File

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

2
common/src/byte_count.rs
View File

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

74
common/src/datetime.rs
View File

@@ -1,25 +1,33 @@
#![allow(deprecated)]
use std::fmt;
use serde::{Deserialize, Serialize};
use time::format_description::well_known::Rfc3339;
use time::{OffsetDateTime, PrimitiveDateTime, UtcOffset};
/// DateTime Precision
/// Precision with which datetimes are truncated when stored in fast fields. This setting is only
/// relevant for fast fields. In the docstore, datetimes are always saved with nanosecond precision.
#[derive(
Clone, Copy, Debug, Hash, PartialEq, Eq, PartialOrd, Ord, Serialize, Deserialize, Default,
)]
#[serde(rename_all = "lowercase")]
pub enum DatePrecision {
/// Seconds precision
pub enum DateTimePrecision {
/// Second precision.
#[default]
Seconds,
/// Milli-seconds precision.
/// Millisecond precision.
Milliseconds,
/// Micro-seconds precision.
/// Microsecond precision.
Microseconds,
/// Nanosecond precision.
Nanoseconds,
}
/// A date/time value with microsecond precision.
#[deprecated(since = "0.20.0", note = "Use `DateTimePrecision` instead")]
pub type DatePrecision = DateTimePrecision;
/// A date/time value with nanoseconds precision.
///
/// This timestamp does not carry any explicit time zone information.
/// Users are responsible for applying the provided conversion
@@ -31,39 +39,46 @@ pub enum DatePrecision {
/// to prevent unintended usage.
#[derive(Clone, Default, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
pub struct DateTime {
// Timestamp in microseconds.
pub(crate) timestamp_micros: i64,
// Timestamp in nanoseconds.
pub(crate) timestamp_nanos: i64,
}
impl DateTime {
/// Minimum possible `DateTime` value.
pub const MIN: DateTime = DateTime {
timestamp_micros: i64::MIN,
timestamp_nanos: i64::MIN,
};
/// Maximum possible `DateTime` value.
pub const MAX: DateTime = DateTime {
timestamp_micros: i64::MAX,
timestamp_nanos: i64::MAX,
};
/// Create new from UNIX timestamp in seconds
pub const fn from_timestamp_secs(seconds: i64) -> Self {
Self {
timestamp_micros: seconds * 1_000_000,
timestamp_nanos: seconds * 1_000_000_000,
}
}
/// Create new from UNIX timestamp in milliseconds
pub const fn from_timestamp_millis(milliseconds: i64) -> Self {
Self {
timestamp_micros: milliseconds * 1_000,
timestamp_nanos: milliseconds * 1_000_000,
}
}
/// Create new from UNIX timestamp in microseconds.
pub const fn from_timestamp_micros(microseconds: i64) -> Self {
Self {
timestamp_micros: microseconds,
timestamp_nanos: microseconds * 1_000,
}
}
/// Create new from UNIX timestamp in nanoseconds.
pub const fn from_timestamp_nanos(nanoseconds: i64) -> Self {
Self {
timestamp_nanos: nanoseconds,
}
}
@@ -71,9 +86,9 @@ impl DateTime {
///
/// The given date/time is converted to UTC and the actual
/// time zone is discarded.
pub const fn from_utc(dt: OffsetDateTime) -> Self {
let timestamp_micros = dt.unix_timestamp() * 1_000_000 + dt.microsecond() as i64;
Self { timestamp_micros }
pub fn from_utc(dt: OffsetDateTime) -> Self {
let timestamp_nanos = dt.unix_timestamp_nanos() as i64;
Self { timestamp_nanos }
}
/// Create new from `PrimitiveDateTime`
@@ -87,23 +102,27 @@ impl DateTime {
/// Convert to UNIX timestamp in seconds.
pub const fn into_timestamp_secs(self) -> i64 {
self.timestamp_micros / 1_000_000
self.timestamp_nanos / 1_000_000_000
}
/// Convert to UNIX timestamp in milliseconds.
pub const fn into_timestamp_millis(self) -> i64 {
self.timestamp_micros / 1_000
self.timestamp_nanos / 1_000_000
}
/// Convert to UNIX timestamp in microseconds.
pub const fn into_timestamp_micros(self) -> i64 {
self.timestamp_micros
self.timestamp_nanos / 1_000
}
/// Convert to UNIX timestamp in nanoseconds.
pub const fn into_timestamp_nanos(self) -> i64 {
self.timestamp_nanos
}
/// Convert to UTC `OffsetDateTime`
pub fn into_utc(self) -> OffsetDateTime {
let timestamp_nanos = self.timestamp_micros as i128 * 1000;
let utc_datetime = OffsetDateTime::from_unix_timestamp_nanos(timestamp_nanos)
let utc_datetime = OffsetDateTime::from_unix_timestamp_nanos(self.timestamp_nanos as i128)
.expect("valid UNIX timestamp");
debug_assert_eq!(UtcOffset::UTC, utc_datetime.offset());
utc_datetime
@@ -126,20 +145,21 @@ impl DateTime {
}
/// Truncates the microseconds value to the corresponding precision.
pub fn truncate(self, precision: DatePrecision) -> Self {
pub fn truncate(self, precision: DateTimePrecision) -> Self {
let truncated_timestamp_micros = match precision {
DatePrecision::Seconds => (self.timestamp_micros / 1_000_000) * 1_000_000,
DatePrecision::Milliseconds => (self.timestamp_micros / 1_000) * 1_000,
DatePrecision::Microseconds => self.timestamp_micros,
DateTimePrecision::Seconds => (self.timestamp_nanos / 1_000_000_000) * 1_000_000_000,
DateTimePrecision::Milliseconds => (self.timestamp_nanos / 1_000_000) * 1_000_000,
DateTimePrecision::Microseconds => (self.timestamp_nanos / 1_000) * 1_000,
DateTimePrecision::Nanoseconds => self.timestamp_nanos,
};
Self {
timestamp_micros: truncated_timestamp_micros,
timestamp_nanos: truncated_timestamp_micros,
}
}
}
impl fmt::Debug for DateTime {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
let utc_rfc3339 = self.into_utc().format(&Rfc3339).map_err(|_| fmt::Error)?;
f.write_str(&utc_rfc3339)
}

63
common/src/dictionary_footer.rs
View File

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

81
common/src/file_slice.rs
View File

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

26
common/src/group_by.rs
View File

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

9
common/src/lib.rs
View File

@@ -7,7 +7,6 @@ pub use byteorder::LittleEndian as Endianness;
mod bitset;
mod byte_count;
mod datetime;
mod dictionary_footer;
pub mod file_slice;
mod group_by;
mod serialize;
@@ -15,14 +14,14 @@ mod vint;
mod writer;
pub use bitset::*;
pub use byte_count::ByteCount;
pub use datetime::{DatePrecision, DateTime};
pub use dictionary_footer::*;
#[allow(deprecated)]
pub use datetime::DatePrecision;
pub use datetime::{DateTime, DateTimePrecision};
pub use group_by::GroupByIteratorExtended;
pub use ownedbytes::{OwnedBytes, StableDeref};
pub use serialize::{BinarySerializable, DeserializeFrom, FixedSize};
pub use vint::{
deserialize_vint_u128, read_u32_vint, read_u32_vint_no_advance, serialize_vint_u128,
serialize_vint_u32, write_u32_vint, VInt, VIntU128,
read_u32_vint, read_u32_vint_no_advance, serialize_vint_u32, write_u32_vint, VInt, VIntU128,
};
pub use writer::{AntiCallToken, CountingWriter, TerminatingWrite};

77
common/src/vint.rs
View File

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

59
examples/aggregation.rs
View File

@@ -7,13 +7,8 @@
// ---
use serde_json::{Deserializer, Value};
use tantivy::aggregation::agg_req::{
Aggregation, Aggregations, BucketAggregation, BucketAggregationType, MetricAggregation,
RangeAggregation,
};
use tantivy::aggregation::agg_req::Aggregations;
use tantivy::aggregation::agg_result::AggregationResults;
use tantivy::aggregation::bucket::RangeAggregationRange;
use tantivy::aggregation::metric::AverageAggregation;
use tantivy::aggregation::AggregationCollector;
use tantivy::query::AllQuery;
use tantivy::schema::{self, IndexRecordOption, Schema, TextFieldIndexing, FAST};
@@ -194,56 +189,9 @@ fn main() -> tantivy::Result<()> {
let agg_req: Aggregations = serde_json::from_str(agg_req_str)?;
let collector = AggregationCollector::from_aggs(agg_req, Default::default());
let agg_res: AggregationResults = searcher.search(&AllQuery, &collector).unwrap();
let res2: Value = serde_json::to_value(agg_res)?;
// ### Request Rust API
//
// This is exactly the same request as above, but via the rust structures.
//
let agg_req: Aggregations = vec![(
"group_by_stock".to_string(),
Aggregation::Bucket(Box::new(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "stock".to_string(),
ranges: vec![
RangeAggregationRange {
key: Some("few".into()),
from: None,
to: Some(1f64),
},
RangeAggregationRange {
key: Some("some".into()),
from: Some(1f64),
to: Some(10f64),
},
RangeAggregationRange {
key: Some("many".into()),
from: Some(10f64),
to: None,
},
],
..Default::default()
}),
sub_aggregation: vec![(
"average_price".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("price".to_string()),
)),
)]
.into_iter()
.collect(),
})),
)]
.into_iter()
.collect();
let collector = AggregationCollector::from_aggs(agg_req, Default::default());
// We use the `AllQuery` which will pass all documents to the AggregationCollector.
let agg_res: AggregationResults = searcher.search(&AllQuery, &collector).unwrap();
let res1: Value = serde_json::to_value(agg_res)?;
let res: Value = serde_json::to_value(agg_res)?;
// ### Aggregation Result
//
@@ -261,8 +209,7 @@ fn main() -> tantivy::Result<()> {
}
"#;
let expected_json: Value = serde_json::from_str(expected_res)?;
assert_eq!(expected_json, res1);
assert_eq!(expected_json, res2);
assert_eq!(expected_json, res);
// ### Request 2
//

14
examples/basic_search.rs
View File

@@ -221,5 +221,19 @@ fn main() -> tantivy::Result<()> {
println!("{}", schema.to_json(&retrieved_doc));
}
// We can also get an explanation to understand
// how a found document got its score.
let query = query_parser.parse_query("title:sea^20 body:whale^70")?;
let (_score, doc_address) = searcher
.search(&query, &TopDocs::with_limit(1))?
.into_iter()
.next()
.unwrap();
let explanation = query.explain(&searcher, doc_address)?;
println!("{}", explanation.to_pretty_json());
Ok(())
}

2
examples/custom_tokenizer.rs
View File

@@ -53,7 +53,7 @@ fn main() -> tantivy::Result<()> {
// this will store tokens of 3 characters each
index
.tokenizers()
.register("ngram3", NgramTokenizer::new(3, 3, false));
.register("ngram3", NgramTokenizer::new(3, 3, false).unwrap());
// To insert document we need an index writer.
// There must be only one writer at a time.

2
examples/date_time_field.rs
View File

@@ -13,7 +13,7 @@ fn main() -> tantivy::Result<()> {
let opts = DateOptions::from(INDEXED)
.set_stored()
.set_fast()
.set_precision(tantivy::DatePrecision::Seconds);
.set_precision(tantivy::DateTimePrecision::Seconds);
// Add `occurred_at` date field type
let occurred_at = schema_builder.add_date_field("occurred_at", opts);
let event_type = schema_builder.add_text_field("event", STRING | STORED);

2
examples/index_from_multiple_threads.rs
View File

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

4
examples/iterating_docs_and_positions.rs
View File

@@ -84,7 +84,7 @@ fn main() -> tantivy::Result<()> {
// Doc 0: TermFreq 2: [0, 4]
// Doc 2: TermFreq 1: [0]
// ```
println!("Doc {}: TermFreq {}: {:?}", doc_id, term_freq, positions);
println!("Doc {doc_id}: TermFreq {term_freq}: {positions:?}");
doc_id = segment_postings.advance();
}
}
@@ -125,7 +125,7 @@ fn main() -> tantivy::Result<()> {
// Once again these docs MAY contains deleted documents as well.
let docs = block_segment_postings.docs();
// Prints `Docs [0, 2].`
println!("Docs {:?}", docs);
println!("Docs {docs:?}");
block_segment_postings.advance();
}
}

79
examples/phrase_prefix_search.rs Normal file
View File

@@ -0,0 +1,79 @@
use tantivy::collector::TopDocs;
use tantivy::query::QueryParser;
use tantivy::schema::*;
use tantivy::{doc, Index, ReloadPolicy, Result};
use tempfile::TempDir;
fn main() -> Result<()> {
let index_path = TempDir::new()?;
let mut schema_builder = Schema::builder();
schema_builder.add_text_field("title", TEXT | STORED);
schema_builder.add_text_field("body", TEXT);
let schema = schema_builder.build();
let title = schema.get_field("title").unwrap();
let body = schema.get_field("body").unwrap();
let index = Index::create_in_dir(&index_path, schema)?;
let mut index_writer = index.writer(50_000_000)?;
index_writer.add_document(doc!(
title => "The Old Man and the Sea",
body => "He was an old man who fished alone in a skiff in the Gulf Stream and he had gone \
eighty-four days now without taking a fish.",
))?;
index_writer.add_document(doc!(
title => "Of Mice and Men",
body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
bank and runs deep and green. The water is warm too, for it has slipped twinkling \
over the yellow sands in the sunlight before reaching the narrow pool. On one \
side of the river the golden foothill slopes curve up to the strong and rocky \
Gabilan Mountains, but on the valley side the water is lined with trees—willows \
fresh and green with every spring, carrying in their lower leaf junctures the \
debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
limbs and branches that arch over the pool"
))?;
// Multivalued field just need to be repeated.
index_writer.add_document(doc!(
title => "Frankenstein",
title => "The Modern Prometheus",
body => "You will rejoice to hear that no disaster has accompanied the commencement of an \
enterprise which you have regarded with such evil forebodings. I arrived here \
yesterday, and my first task is to assure my dear sister of my welfare and \
increasing confidence in the success of my undertaking."
))?;
index_writer.commit()?;
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::OnCommit)
.try_into()?;
let searcher = reader.searcher();
let query_parser = QueryParser::for_index(&index, vec![title, body]);
// This will match documents containing the phrase "in the"
// followed by some word starting with "su",
// i.e. it will match "in the sunlight" and "in the success",
// but not "in the Gulf Stream".
let query = query_parser.parse_query("\"in the su\"*")?;
let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
let mut titles = top_docs
.into_iter()
.map(|(_score, doc_address)| {
let doc = searcher.doc(doc_address)?;
let title = doc.get_first(title).unwrap().as_text().unwrap().to_owned();
Ok(title)
})
.collect::<Result<Vec<_>>>()?;
titles.sort_unstable();
assert_eq!(titles, ["Frankenstein", "Of Mice and Men"]);
Ok(())
}

3
examples/pre_tokenized_text.rs
View File

@@ -17,7 +17,8 @@ use tantivy::{doc, Index, ReloadPolicy};
use tempfile::TempDir;
fn pre_tokenize_text(text: &str) -> Vec<Token> {
let mut token_stream = SimpleTokenizer.token_stream(text);
let mut tokenizer = SimpleTokenizer::default();
let mut token_stream = tokenizer.token_stream(text);
let mut tokens = vec![];
while token_stream.advance() {
tokens.push(token_stream.token().clone());

2
examples/snippet.rs
View File

@@ -56,7 +56,7 @@ fn main() -> tantivy::Result<()> {
for (score, doc_address) in top_docs {
let doc = searcher.doc(doc_address)?;
let snippet = snippet_generator.snippet_from_doc(&doc);
println!("Document score {}:", score);
println!("Document score {score}:");
println!(
"title: {}",
doc.get_first(title).unwrap().as_text().unwrap()

4
examples/stop_words.rs
View File

@@ -50,7 +50,7 @@ fn main() -> tantivy::Result<()> {
// This tokenizer lowers all of the text (to help with stop word matching)
// then removes all instances of `the` and `and` from the corpus
let tokenizer = TextAnalyzer::builder(SimpleTokenizer)
let tokenizer = TextAnalyzer::builder(SimpleTokenizer::default())
.filter(LowerCaser)
.filter(StopWordFilter::remove(vec![
"the".to_string(),
@@ -106,7 +106,7 @@ fn main() -> tantivy::Result<()> {
for (score, doc_address) in top_docs {
let retrieved_doc = searcher.doc(doc_address)?;
println!("\n==\nDocument score {}:", score);
println!("\n==\nDocument score {score}:");
println!("{}", schema.to_json(&retrieved_doc));
}

77
examples/warmer.rs
View File

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

2
ownedbytes/Cargo.toml
View File

@@ -1,7 +1,7 @@
[package]
authors = ["Paul Masurel <paul@quickwit.io>", "Pascal Seitz <pascal@quickwit.io>"]
name = "ownedbytes"
version = "0.5.0"
version = "0.6.0"
edition = "2021"
description = "Expose data as static slice"
license = "MIT"

109
ownedbytes/src/lib.rs
View File

@@ -1,7 +1,7 @@
use std::convert::TryInto;
use std::ops::{Deref, Range};
use std::sync::Arc;
use std::{fmt, io, mem};
use std::{fmt, io};
pub use stable_deref_trait::StableDeref;
@@ -26,8 +26,8 @@ impl OwnedBytes {
data_holder: T,
) -> OwnedBytes {
let box_stable_deref = Arc::new(data_holder);
let bytes: &[u8] = box_stable_deref.as_ref();
let data = unsafe { mem::transmute::<_, &'static [u8]>(bytes.deref()) };
let bytes: &[u8] = box_stable_deref.deref();
let data = unsafe { &*(bytes as *const [u8]) };
OwnedBytes {
data,
box_stable_deref,
@@ -57,6 +57,12 @@ impl OwnedBytes {
self.data.len()
}
/// Returns true iff this `OwnedBytes` is empty.
#[inline]
pub fn is_empty(&self) -> bool {
self.data.is_empty()
}
/// Splits the OwnedBytes into two OwnedBytes `(left, right)`.
///
/// Left will hold `split_len` bytes.
@@ -68,13 +74,14 @@ impl OwnedBytes {
#[inline]
#[must_use]
pub fn split(self, split_len: usize) -> (OwnedBytes, OwnedBytes) {
let (left_data, right_data) = self.data.split_at(split_len);
let right_box_stable_deref = self.box_stable_deref.clone();
let left = OwnedBytes {
data: &self.data[..split_len],
data: left_data,
box_stable_deref: self.box_stable_deref,
};
let right = OwnedBytes {
data: &self.data[split_len..],
data: right_data,
box_stable_deref: right_box_stable_deref,
};
(left, right)
@@ -99,45 +106,45 @@ impl OwnedBytes {
///
/// `self` is truncated to `split_len`, left with the remaining bytes.
pub fn split_off(&mut self, split_len: usize) -> OwnedBytes {
let (left, right) = self.data.split_at(split_len);
let right_box_stable_deref = self.box_stable_deref.clone();
let right_piece = OwnedBytes {
data: &self.data[split_len..],
data: right,
box_stable_deref: right_box_stable_deref,
};
self.data = &self.data[..split_len];
self.data = left;
right_piece
}
/// Returns true iff this `OwnedBytes` is empty.
#[inline]
pub fn is_empty(&self) -> bool {
self.as_slice().is_empty()
}
/// Drops the left most `advance_len` bytes.
#[inline]
pub fn advance(&mut self, advance_len: usize) {
self.data = &self.data[advance_len..]
pub fn advance(&mut self, advance_len: usize) -> &[u8] {
let (data, rest) = self.data.split_at(advance_len);
self.data = rest;
data
}
/// Reads an `u8` from the `OwnedBytes` and advance by one byte.
#[inline]
pub fn read_u8(&mut self) -> u8 {
assert!(!self.is_empty());
self.advance(1)[0]
}
let byte = self.as_slice()[0];
self.advance(1);
byte
#[inline]
fn read_n<const N: usize>(&mut self) -> [u8; N] {
self.advance(N).try_into().unwrap()
}
/// Reads an `u32` encoded as little-endian from the `OwnedBytes` and advance by 4 bytes.
#[inline]
pub fn read_u32(&mut self) -> u32 {
u32::from_le_bytes(self.read_n())
}
/// Reads an `u64` encoded as little-endian from the `OwnedBytes` and advance by 8 bytes.
#[inline]
pub fn read_u64(&mut self) -> u64 {
assert!(self.len() > 7);
let octlet: [u8; 8] = self.as_slice()[..8].try_into().unwrap();
self.advance(8);
u64::from_le_bytes(octlet)
u64::from_le_bytes(self.read_n())
}
}
@@ -150,7 +157,7 @@ impl fmt::Debug for OwnedBytes {
} else {
self.as_slice()
};
write!(f, "OwnedBytes({:?}, len={})", bytes_truncated, self.len())
write!(f, "OwnedBytes({bytes_truncated:?}, len={})", self.len())
}
}
@@ -191,32 +198,33 @@ impl Deref for OwnedBytes {
}
}
impl AsRef<[u8]> for OwnedBytes {
#[inline]
fn as_ref(&self) -> &[u8] {
self.as_slice()
}
}
impl io::Read for OwnedBytes {
#[inline]
fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
let read_len = {
let data = self.as_slice();
if data.len() >= buf.len() {
let buf_len = buf.len();
buf.copy_from_slice(&data[..buf_len]);
buf.len()
} else {
let data_len = data.len();
buf[..data_len].copy_from_slice(data);
data_len
}
};
self.advance(read_len);
Ok(read_len)
let data_len = self.data.len();
let buf_len = buf.len();
if data_len >= buf_len {
let data = self.advance(buf_len);
buf.copy_from_slice(data);
Ok(buf_len)
} else {
buf[..data_len].copy_from_slice(self.data);
self.data = &[];
Ok(data_len)
}
}
#[inline]
fn read_to_end(&mut self, buf: &mut Vec<u8>) -> io::Result<usize> {
let read_len = {
let data = self.as_slice();
buf.extend(data);
data.len()
};
self.advance(read_len);
buf.extend(self.data);
let read_len = self.data.len();
self.data = &[];
Ok(read_len)
}
#[inline]
@@ -232,13 +240,6 @@ impl io::Read for OwnedBytes {
}
}
impl AsRef<[u8]> for OwnedBytes {
#[inline]
fn as_ref(&self) -> &[u8] {
self.as_slice()
}
}
#[cfg(test)]
mod tests {
use std::io::{self, Read};
@@ -249,12 +250,12 @@ mod tests {
fn test_owned_bytes_debug() {
let short_bytes = OwnedBytes::new(b"abcd".as_ref());
assert_eq!(
format!("{:?}", short_bytes),
format!("{short_bytes:?}"),
"OwnedBytes([97, 98, 99, 100], len=4)"
);
let long_bytes = OwnedBytes::new(b"abcdefghijklmnopq".as_ref());
assert_eq!(
format!("{:?}", long_bytes),
format!("{long_bytes:?}"),
"OwnedBytes([97, 98, 99, 100, 101, 102, 103, 104, 105, 106], len=17)"
);
}

6
query-grammar/Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy-query-grammar"
version = "0.19.0"
version = "0.21.0"
authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT"
categories = ["database-implementations", "data-structures"]
@@ -12,6 +12,4 @@ keywords = ["search", "information", "retrieval"]
edition = "2021"
[dependencies]
combine = {version="4", default-features=false, features=[] }
once_cell = "1.7.2"
regex ={ version = "1.5.4", default-features = false, features = ["std", "unicode"] }
nom = "7"

353
query-grammar/src/infallible.rs Normal file
View File

@@ -0,0 +1,353 @@
//! nom combinators for infallible operations
use std::convert::Infallible;
use nom::{AsChar, IResult, InputLength, InputTakeAtPosition};
pub(crate) type ErrorList = Vec<LenientErrorInternal>;
pub(crate) type JResult<I, O> = IResult<I, (O, ErrorList), Infallible>;
/// An error, with an end-of-string based offset
#[derive(Debug)]
pub(crate) struct LenientErrorInternal {
pub pos: usize,
pub message: String,
}
/// A recoverable error and the position it happened at
#[derive(Debug, PartialEq)]
pub struct LenientError {
pub pos: usize,
pub message: String,
}
impl LenientError {
pub(crate) fn from_internal(internal: LenientErrorInternal, str_len: usize) -> LenientError {
LenientError {
pos: str_len - internal.pos,
message: internal.message,
}
}
}
fn unwrap_infallible<T>(res: Result<T, nom::Err<Infallible>>) -> T {
match res {
Ok(val) => val,
Err(_) => unreachable!(),
}
}
// when rfcs#1733 get stabilized, this can make things clearer
// trait InfallibleParser<I, O> = nom::Parser<I, (O, ErrorList), std::convert::Infallible>;
/// A variant of the classical `opt` parser, except it returns an infallible error type.
///
/// It's less generic than the original to ease type resolution in the rest of the code.
pub(crate) fn opt_i<I: Clone, O, F>(mut f: F) -> impl FnMut(I) -> JResult<I, Option<O>>
where F: nom::Parser<I, O, nom::error::Error<I>> {
move |input: I| {
let i = input.clone();
match f.parse(input) {
Ok((i, o)) => Ok((i, (Some(o), Vec::new()))),
Err(_) => Ok((i, (None, Vec::new()))),
}
}
}
pub(crate) fn opt_i_err<'a, I: Clone + InputLength, O, F>(
mut f: F,
message: impl ToString + 'a,
) -> impl FnMut(I) -> JResult<I, Option<O>> + 'a
where
F: nom::Parser<I, O, nom::error::Error<I>> + 'a,
{
move |input: I| {
let i = input.clone();
match f.parse(input) {
Ok((i, o)) => Ok((i, (Some(o), Vec::new()))),
Err(_) => {
let errs = vec![LenientErrorInternal {
pos: i.input_len(),
message: message.to_string(),
}];
Ok((i, (None, errs)))
}
}
}
}
pub(crate) fn space0_infallible<T>(input: T) -> JResult<T, T>
where
T: InputTakeAtPosition + Clone,
<T as InputTakeAtPosition>::Item: AsChar + Clone,
{
opt_i(nom::character::complete::space0)(input)
.map(|(left, (spaces, errors))| (left, (spaces.expect("space0 can't fail"), errors)))
}
pub(crate) fn space1_infallible<T>(input: T) -> JResult<T, Option<T>>
where
T: InputTakeAtPosition + Clone + InputLength,
<T as InputTakeAtPosition>::Item: AsChar + Clone,
{
opt_i(nom::character::complete::space1)(input).map(|(left, (spaces, mut errors))| {
if spaces.is_none() {
errors.push(LenientErrorInternal {
pos: left.input_len(),
message: "missing space".to_string(),
})
}
(left, (spaces, errors))
})
}
pub(crate) fn fallible<I, O, E: nom::error::ParseError<I>, F>(
mut f: F,
) -> impl FnMut(I) -> IResult<I, O, E>
where F: nom::Parser<I, (O, ErrorList), Infallible> {
use nom::Err;
move |input: I| match f.parse(input) {
Ok((input, (output, _err))) => Ok((input, output)),
Err(Err::Incomplete(needed)) => Err(Err::Incomplete(needed)),
Err(Err::Error(val)) | Err(Err::Failure(val)) => match val {},
}
}
pub(crate) fn delimited_infallible<I, O1, O2, O3, F, G, H>(
mut first: F,
mut second: G,
mut third: H,
) -> impl FnMut(I) -> JResult<I, O2>
where
F: nom::Parser<I, (O1, ErrorList), Infallible>,
G: nom::Parser<I, (O2, ErrorList), Infallible>,
H: nom::Parser<I, (O3, ErrorList), Infallible>,
{
move |input: I| {
let (input, (_, mut err)) = first.parse(input)?;
let (input, (o2, mut err2)) = second.parse(input)?;
err.append(&mut err2);
let (input, (_, mut err3)) = third.parse(input)?;
err.append(&mut err3);
Ok((input, (o2, err)))
}
}
// Parse nothing. Just a lazy way to not implement terminated/preceded and use delimited instead
pub(crate) fn nothing(i: &str) -> JResult<&str, ()> {
Ok((i, ((), Vec::new())))
}
pub(crate) trait TupleInfallible<I, O> {
/// Parses the input and returns a tuple of results of each parser.
fn parse(&mut self, input: I) -> JResult<I, O>;
}
impl<Input, Output, F: nom::Parser<Input, (Output, ErrorList), Infallible>>
TupleInfallible<Input, (Output,)> for (F,)
{
fn parse(&mut self, input: Input) -> JResult<Input, (Output,)> {
self.0.parse(input).map(|(i, (o, e))| (i, ((o,), e)))
}
}
// these macros are heavily copied from nom, with some minor adaptations for our type
macro_rules! tuple_trait(
($name1:ident $ty1:ident, $name2: ident $ty2:ident, $($name:ident $ty:ident),*) => (
tuple_trait!(__impl $name1 $ty1, $name2 $ty2; $($name $ty),*);
);
(__impl $($name:ident $ty: ident),+; $name1:ident $ty1:ident, $($name2:ident $ty2:ident),*) => (
tuple_trait_impl!($($name $ty),+);
tuple_trait!(__impl $($name $ty),+ , $name1 $ty1; $($name2 $ty2),*);
);
(__impl $($name:ident $ty: ident),+; $name1:ident $ty1:ident) => (
tuple_trait_impl!($($name $ty),+);
tuple_trait_impl!($($name $ty),+, $name1 $ty1);
);
);
macro_rules! tuple_trait_impl(
($($name:ident $ty: ident),+) => (
impl<
Input: Clone, $($ty),+ ,
$($name: nom::Parser<Input, ($ty, ErrorList), Infallible>),+
> TupleInfallible<Input, ( $($ty),+ )> for ( $($name),+ ) {
fn parse(&mut self, input: Input) -> JResult<Input, ( $($ty),+ )> {
let mut error_list = Vec::new();
tuple_trait_inner!(0, self, input, (), error_list, $($name)+)
}
}
);
);
macro_rules! tuple_trait_inner(
($it:tt, $self:expr, $input:expr, (), $error_list:expr, $head:ident $($id:ident)+) => ({
let (i, (o, mut err)) = $self.$it.parse($input.clone())?;
$error_list.append(&mut err);
succ!($it, tuple_trait_inner!($self, i, ( o ), $error_list, $($id)+))
});
($it:tt, $self:expr, $input:expr, ($($parsed:tt)*), $error_list:expr, $head:ident $($id:ident)+) => ({
let (i, (o, mut err)) = $self.$it.parse($input.clone())?;
$error_list.append(&mut err);
succ!($it, tuple_trait_inner!($self, i, ($($parsed)* , o), $error_list, $($id)+))
});
($it:tt, $self:expr, $input:expr, ($($parsed:tt)*), $error_list:expr, $head:ident) => ({
let (i, (o, mut err)) = $self.$it.parse($input.clone())?;
$error_list.append(&mut err);
Ok((i, (($($parsed)* , o), $error_list)))
});
);
macro_rules! succ (
(0, $submac:ident ! ($($rest:tt)*)) => ($submac!(1, $($rest)*));
(1, $submac:ident ! ($($rest:tt)*)) => ($submac!(2, $($rest)*));
(2, $submac:ident ! ($($rest:tt)*)) => ($submac!(3, $($rest)*));
(3, $submac:ident ! ($($rest:tt)*)) => ($submac!(4, $($rest)*));
(4, $submac:ident ! ($($rest:tt)*)) => ($submac!(5, $($rest)*));
(5, $submac:ident ! ($($rest:tt)*)) => ($submac!(6, $($rest)*));
(6, $submac:ident ! ($($rest:tt)*)) => ($submac!(7, $($rest)*));
(7, $submac:ident ! ($($rest:tt)*)) => ($submac!(8, $($rest)*));
(8, $submac:ident ! ($($rest:tt)*)) => ($submac!(9, $($rest)*));
(9, $submac:ident ! ($($rest:tt)*)) => ($submac!(10, $($rest)*));
(10, $submac:ident ! ($($rest:tt)*)) => ($submac!(11, $($rest)*));
(11, $submac:ident ! ($($rest:tt)*)) => ($submac!(12, $($rest)*));
(12, $submac:ident ! ($($rest:tt)*)) => ($submac!(13, $($rest)*));
(13, $submac:ident ! ($($rest:tt)*)) => ($submac!(14, $($rest)*));
(14, $submac:ident ! ($($rest:tt)*)) => ($submac!(15, $($rest)*));
(15, $submac:ident ! ($($rest:tt)*)) => ($submac!(16, $($rest)*));
(16, $submac:ident ! ($($rest:tt)*)) => ($submac!(17, $($rest)*));
(17, $submac:ident ! ($($rest:tt)*)) => ($submac!(18, $($rest)*));
(18, $submac:ident ! ($($rest:tt)*)) => ($submac!(19, $($rest)*));
(19, $submac:ident ! ($($rest:tt)*)) => ($submac!(20, $($rest)*));
(20, $submac:ident ! ($($rest:tt)*)) => ($submac!(21, $($rest)*));
);
tuple_trait!(FnA A, FnB B, FnC C, FnD D, FnE E, FnF F, FnG G, FnH H, FnI I, FnJ J, FnK K, FnL L,
FnM M, FnN N, FnO O, FnP P, FnQ Q, FnR R, FnS S, FnT T, FnU U);
// Special case: implement `TupleInfallible` for `()`, the unit type.
// This can come up in macros which accept a variable number of arguments.
// Literally, `()` is an empty tuple, so it should simply parse nothing.
impl<I> TupleInfallible<I, ()> for () {
fn parse(&mut self, input: I) -> JResult<I, ()> {
Ok((input, ((), Vec::new())))
}
}
pub(crate) fn tuple_infallible<I, O, List: TupleInfallible<I, O>>(
mut l: List,
) -> impl FnMut(I) -> JResult<I, O> {
move |i: I| l.parse(i)
}
pub(crate) fn separated_list_infallible<I, O, O2, F, G>(
mut sep: G,
mut f: F,
) -> impl FnMut(I) -> JResult<I, Vec<O>>
where
I: Clone + InputLength,
F: nom::Parser<I, (O, ErrorList), Infallible>,
G: nom::Parser<I, (O2, ErrorList), Infallible>,
{
move |i: I| {
let mut res: Vec<O> = Vec::new();
let mut errors: ErrorList = Vec::new();
let (mut i, (o, mut err)) = unwrap_infallible(f.parse(i.clone()));
errors.append(&mut err);
res.push(o);
loop {
let (i_sep_parsed, (_, mut err_sep)) = unwrap_infallible(sep.parse(i.clone()));
let len_before = i_sep_parsed.input_len();
let (i_elem_parsed, (o, mut err_elem)) =
unwrap_infallible(f.parse(i_sep_parsed.clone()));
// infinite loop check: the parser must always consume
// if we consumed nothing here, don't produce an element.
if i_elem_parsed.input_len() == len_before {
return Ok((i, (res, errors)));
}
res.push(o);
errors.append(&mut err_sep);
errors.append(&mut err_elem);
i = i_elem_parsed;
}
}
}
pub(crate) trait Alt<I, O> {
/// Tests each parser in the tuple and returns the result of the first one that succeeds
fn choice(&mut self, input: I) -> Option<JResult<I, O>>;
}
macro_rules! alt_trait(
($first_cond:ident $first:ident, $($id_cond:ident $id: ident),+) => (
alt_trait!(__impl $first_cond $first; $($id_cond $id),+);
);
(__impl $($current_cond:ident $current:ident),*; $head_cond:ident $head:ident, $($id_cond:ident $id:ident),+) => (
alt_trait_impl!($($current_cond $current),*);
alt_trait!(__impl $($current_cond $current,)* $head_cond $head; $($id_cond $id),+);
);
(__impl $($current_cond:ident $current:ident),*; $head_cond:ident $head:ident) => (
alt_trait_impl!($($current_cond $current),*);
alt_trait_impl!($($current_cond $current,)* $head_cond $head);
);
);
macro_rules! alt_trait_impl(
($($id_cond:ident $id:ident),+) => (
impl<
Input: Clone, Output,
$(
// () are to make things easier on me, but I'm not entirely sure whether we can do better
// with rule E0207
$id_cond: nom::Parser<Input, (), ()>,
$id: nom::Parser<Input, (Output, ErrorList), Infallible>
),+
> Alt<Input, Output> for ( $(($id_cond, $id),)+ ) {
fn choice(&mut self, input: Input) -> Option<JResult<Input, Output>> {
match self.0.0.parse(input.clone()) {
Err(_) => alt_trait_inner!(1, self, input, $($id_cond $id),+),
Ok((input_left, _)) => Some(self.0.1.parse(input_left)),
}
}
}
);
);
macro_rules! alt_trait_inner(
($it:tt, $self:expr, $input:expr, $head_cond:ident $head:ident, $($id_cond:ident $id:ident),+) => (
match $self.$it.0.parse($input.clone()) {
Err(_) => succ!($it, alt_trait_inner!($self, $input, $($id_cond $id),+)),
Ok((input_left, _)) => Some($self.$it.1.parse(input_left)),
}
);
($it:tt, $self:expr, $input:expr, $head_cond:ident $head:ident) => (
None
);
);
alt_trait!(A1 A, B1 B, C1 C, D1 D, E1 E, F1 F, G1 G, H1 H, I1 I, J1 J, K1 K,
L1 L, M1 M, N1 N, O1 O, P1 P, Q1 Q, R1 R, S1 S, T1 T, U1 U);
/// An alt() like combinator. For each branch, it first tries a fallible parser, which commits to
/// this branch, or tells to check next branch, and the execute the infallible parser which follow.
///
/// In case no branch match, the default (fallible) parser is executed.
pub(crate) fn alt_infallible<I: Clone, O, F, List: Alt<I, O>>(
mut l: List,
mut default: F,
) -> impl FnMut(I) -> JResult<I, O>
where
F: nom::Parser<I, (O, ErrorList), Infallible>,
{
move |i: I| l.choice(i.clone()).unwrap_or_else(|| default.parse(i))
}

17
query-grammar/src/lib.rs
View File

@@ -1,17 +1,26 @@
#![allow(clippy::derive_partial_eq_without_eq)]
mod infallible;
mod occur;
mod query_grammar;
mod user_input_ast;
use combine::parser::Parser;
pub use crate::infallible::LenientError;
pub use crate::occur::Occur;
use crate::query_grammar::parse_to_ast;
pub use crate::user_input_ast::{UserInputAst, UserInputBound, UserInputLeaf, UserInputLiteral};
use crate::query_grammar::{parse_to_ast, parse_to_ast_lenient};
pub use crate::user_input_ast::{
Delimiter, UserInputAst, UserInputBound, UserInputLeaf, UserInputLiteral,
};
pub struct Error;
/// Parse a query
pub fn parse_query(query: &str) -> Result<UserInputAst, Error> {
let (user_input_ast, _remaining) = parse_to_ast().parse(query).map_err(|_| Error)?;
let (_remaining, user_input_ast) = parse_to_ast(query).map_err(|_| Error)?;
Ok(user_input_ast)
}
/// Parse a query, trying to recover from syntax errors, and giving hints toward fixing errors.
pub fn parse_query_lenient(query: &str) -> (UserInputAst, Vec<LenientError>) {
parse_to_ast_lenient(query)
}

1711
query-grammar/src/query_grammar.rs
View File

File diff suppressed because it is too large Load Diff

101
query-grammar/src/user_input_ast.rs
View File

@@ -3,7 +3,7 @@ use std::fmt::{Debug, Formatter};
use crate::Occur;
#[derive(PartialEq)]
#[derive(PartialEq, Clone)]
pub enum UserInputLeaf {
Literal(UserInputLiteral),
All,
@@ -16,10 +16,38 @@ pub enum UserInputLeaf {
field: Option<String>,
elements: Vec<String>,
},
Exists {
field: String,
},
}
impl UserInputLeaf {
pub(crate) fn set_field(self, field: Option<String>) -> Self {
match self {
UserInputLeaf::Literal(mut literal) => {
literal.field_name = field;
UserInputLeaf::Literal(literal)
}
UserInputLeaf::All => UserInputLeaf::All,
UserInputLeaf::Range {
field: _,
lower,
upper,
} => UserInputLeaf::Range {
field,
lower,
upper,
},
UserInputLeaf::Set { field: _, elements } => UserInputLeaf::Set { field, elements },
UserInputLeaf::Exists { field: _ } => UserInputLeaf::Exists {
field: field.expect("Exist query without a field isn't allowed"),
},
}
}
}
impl Debug for UserInputLeaf {
fn fmt(&self, formatter: &mut Formatter<'_>) -> Result<(), fmt::Error> {
fn fmt(&self, formatter: &mut Formatter) -> Result<(), fmt::Error> {
match self {
UserInputLeaf::Literal(literal) => literal.fmt(formatter),
UserInputLeaf::Range {
@@ -28,7 +56,8 @@ impl Debug for UserInputLeaf {
ref upper,
} => {
if let Some(ref field) = field {
write!(formatter, "\"{}\":", field)?;
// TODO properly escape field (in case of \")
write!(formatter, "\"{field}\":")?;
}
lower.display_lower(formatter)?;
write!(formatter, " TO ")?;
@@ -37,43 +66,73 @@ impl Debug for UserInputLeaf {
}
UserInputLeaf::Set { field, elements } => {
if let Some(ref field) = field {
write!(formatter, "\"{}\": ", field)?;
// TODO properly escape field (in case of \")
write!(formatter, "\"{field}\": ")?;
}
write!(formatter, "IN [")?;
for (i, element) in elements.iter().enumerate() {
for (i, text) in elements.iter().enumerate() {
if i != 0 {
write!(formatter, " ")?;
}
write!(formatter, "\"{}\"", element)?;
// TODO properly escape element
write!(formatter, "\"{text}\"")?;
}
write!(formatter, "]")
}
UserInputLeaf::All => write!(formatter, "*"),
UserInputLeaf::Exists { field } => {
write!(formatter, "\"{field}\":*")
}
}
}
}
#[derive(PartialEq)]
#[derive(Copy, Clone, Eq, PartialEq, Debug)]
pub enum Delimiter {
SingleQuotes,
DoubleQuotes,
None,
}
#[derive(PartialEq, Clone)]
pub struct UserInputLiteral {
pub field_name: Option<String>,
pub phrase: String,
pub delimiter: Delimiter,
pub slop: u32,
pub prefix: bool,
}
impl fmt::Debug for UserInputLiteral {
fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
fn fmt(&self, formatter: &mut fmt::Formatter) -> Result<(), fmt::Error> {
if let Some(ref field) = self.field_name {
write!(formatter, "\"{}\":", field)?;
// TODO properly escape field (in case of \")
write!(formatter, "\"{field}\":")?;
}
match self.delimiter {
Delimiter::SingleQuotes => {
// TODO properly escape element (in case of \')
write!(formatter, "'{}'", self.phrase)?;
}
Delimiter::DoubleQuotes => {
// TODO properly escape element (in case of \")
write!(formatter, "\"{}\"", self.phrase)?;
}
Delimiter::None => {
// TODO properly escape element
write!(formatter, "{}", self.phrase)?;
}
}
write!(formatter, "\"{}\"", self.phrase)?;
if self.slop > 0 {
write!(formatter, "~{}", self.slop)?;
} else if self.prefix {
write!(formatter, "*")?;
}
Ok(())
}
}
#[derive(PartialEq)]
#[derive(PartialEq, Debug, Clone)]
pub enum UserInputBound {
Inclusive(String),
Exclusive(String),
@@ -83,16 +142,18 @@ pub enum UserInputBound {
impl UserInputBound {
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),
// TODO properly escape word if required
UserInputBound::Inclusive(ref word) => write!(formatter, "[\"{word}\""),
UserInputBound::Exclusive(ref word) => write!(formatter, "{{\"{word}\""),
UserInputBound::Unbounded => write!(formatter, "{{\"*\""),
}
}
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),
// TODO properly escape word if required
UserInputBound::Inclusive(ref word) => write!(formatter, "\"{word}\"]"),
UserInputBound::Exclusive(ref word) => write!(formatter, "\"{word}\"}}"),
UserInputBound::Unbounded => write!(formatter, "\"*\"}}"),
}
}
@@ -106,6 +167,7 @@ impl UserInputBound {
}
}
#[derive(PartialEq, Clone)]
pub enum UserInputAst {
Clause(Vec<(Option<Occur>, UserInputAst)>),
Leaf(Box<UserInputLeaf>),
@@ -163,9 +225,9 @@ fn print_occur_ast(
formatter: &mut fmt::Formatter,
) -> fmt::Result {
if let Some(occur) = occur_opt {
write!(formatter, "{}{:?}", occur, ast)?;
write!(formatter, "{occur}{ast:?}")?;
} else {
write!(formatter, "*{:?}", ast)?;
write!(formatter, "*{ast:?}")?;
}
Ok(())
}
@@ -175,6 +237,7 @@ impl fmt::Debug for UserInputAst {
match *self {
UserInputAst::Clause(ref subqueries) => {
if subqueries.is_empty() {
// TODO this will break ast reserialization, is writing "( )" enought?
write!(formatter, "<emptyclause>")?;
} else {
write!(formatter, "(")?;
@@ -187,8 +250,8 @@ impl fmt::Debug for UserInputAst {
}
Ok(())
}
UserInputAst::Leaf(ref subquery) => write!(formatter, "{:?}", subquery),
UserInputAst::Boost(ref leaf, boost) => write!(formatter, "({:?})^{}", leaf, boost),
UserInputAst::Leaf(ref subquery) => write!(formatter, "{subquery:?}"),
UserInputAst::Boost(ref leaf, boost) => write!(formatter, "({leaf:?})^{boost}"),
}
}
}

497
src/aggregation/agg_bench.rs
View File

@@ -1,19 +1,36 @@
#[cfg(all(test, feature = "unstable"))]
mod bench {
use columnar::Cardinality;
use rand::prelude::SliceRandom;
use rand::{thread_rng, Rng};
use rand::rngs::StdRng;
use rand::{Rng, SeedableRng};
use rand_distr::Distribution;
use serde_json::json;
use test::{self, Bencher};
use super::*;
use crate::aggregation::bucket::{
CustomOrder, HistogramAggregation, HistogramBounds, Order, OrderTarget, TermsAggregation,
};
use crate::aggregation::metric::StatsAggregation;
use crate::query::AllQuery;
use crate::schema::{Schema, TextFieldIndexing, FAST, STRING};
use crate::Index;
use crate::aggregation::agg_req::Aggregations;
use crate::aggregation::AggregationCollector;
use crate::query::{AllQuery, TermQuery};
use crate::schema::{IndexRecordOption, Schema, TextFieldIndexing, FAST, STRING};
use crate::{Index, Term};
#[derive(Clone, Copy, Hash, Default, Debug, PartialEq, Eq, PartialOrd, Ord)]
enum Cardinality {
/// All documents contain exactly one value.
/// `Full` is the default for auto-detecting the Cardinality, since it is the most strict.
#[default]
Full = 0,
/// All documents contain at most one value.
Optional = 1,
/// All documents may contain any number of values.
Multivalued = 2,
/// 1 / 20 documents has a value
Sparse = 3,
}
fn get_collector(agg_req: Aggregations) -> AggregationCollector {
AggregationCollector::from_aggs(agg_req, Default::default())
}
fn get_test_index_bench(cardinality: Cardinality) -> crate::Result<Index> {
let mut schema_builder = Schema::builder();
@@ -23,6 +40,7 @@ mod bench {
)
.set_stored();
let text_field = schema_builder.add_text_field("text", text_fieldtype);
let json_field = schema_builder.add_json_field("json", FAST);
let text_field_many_terms = schema_builder.add_text_field("text_many_terms", STRING | FAST);
let text_field_few_terms = schema_builder.add_text_field("text_few_terms", STRING | FAST);
let score_fieldtype = crate::schema::NumericOptions::default().set_fast();
@@ -31,12 +49,15 @@ mod bench {
let score_field_i64 = schema_builder.add_i64_field("score_i64", score_fieldtype);
let index = Index::create_from_tempdir(schema_builder.build())?;
let few_terms_data = vec!["INFO", "ERROR", "WARN", "DEBUG"];
let lg_norm = rand_distr::LogNormal::new(2.996f64, 0.979f64).unwrap();
let many_terms_data = (0..150_000)
.map(|num| format!("author{}", num))
.collect::<Vec<_>>();
{
let mut rng = thread_rng();
let mut index_writer = index.writer_with_num_threads(1, 100_000_000)?;
let mut rng = StdRng::from_seed([1u8; 32]);
let mut index_writer = index.writer_with_num_threads(1, 200_000_000)?;
// To make the different test cases comparable we just change one doc to force the
// cardinality
if cardinality == Cardinality::Optional {
@@ -44,6 +65,8 @@ mod bench {
}
if cardinality == Cardinality::Multivalued {
index_writer.add_document(doc!(
json_field => json!({"mixed_type": 10.0}),
json_field => json!({"mixed_type": 10.0}),
text_field => "cool",
text_field => "cool",
text_field_many_terms => "cool",
@@ -52,22 +75,39 @@ mod bench {
text_field_few_terms => "cool",
score_field => 1u64,
score_field => 1u64,
score_field_f64 => 1.0,
score_field_f64 => 1.0,
score_field_f64 => lg_norm.sample(&mut rng),
score_field_f64 => lg_norm.sample(&mut rng),
score_field_i64 => 1i64,
score_field_i64 => 1i64,
))?;
}
for _ in 0..1_000_000 {
let mut doc_with_value = 1_000_000;
if cardinality == Cardinality::Sparse {
doc_with_value /= 20;
}
let val_max = 1_000_000.0;
for _ in 0..doc_with_value {
let val: f64 = rng.gen_range(0.0..1_000_000.0);
let json = if rng.gen_bool(0.1) {
// 10% are numeric values
json!({ "mixed_type": val })
} else {
json!({"mixed_type": many_terms_data.choose(&mut rng).unwrap().to_string()})
};
index_writer.add_document(doc!(
text_field => "cool",
json_field => json,
text_field_many_terms => many_terms_data.choose(&mut rng).unwrap().to_string(),
text_field_few_terms => few_terms_data.choose(&mut rng).unwrap().to_string(),
score_field => val as u64,
score_field_f64 => val,
score_field_f64 => lg_norm.sample(&mut rng),
score_field_i64 => val as i64,
))?;
if cardinality == Cardinality::Sparse {
for _ in 0..20 {
index_writer.add_document(doc!(text_field => "cool"))?;
}
}
}
// writing the segment
index_writer.commit()?;
@@ -95,6 +135,12 @@ mod bench {
fn [<$x _multi>](b: &mut Bencher) {
[<$x _card>](b, Cardinality::Multivalued)
}
#[bench]
fn [<$x _sparse>](b: &mut Bencher) {
[<$x _card>](b, Cardinality::Sparse)
}
}
};
}
@@ -112,14 +158,10 @@ mod bench {
IndexRecordOption::Basic,
);
let agg_req_1: Aggregations = vec![(
"average".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("score".to_string()),
)),
)]
.into_iter()
.collect();
let agg_req_1: Aggregations = serde_json::from_value(json!({
"average": { "avg": { "field": "score", } }
}))
.unwrap();
let collector = get_collector(agg_req_1);
@@ -141,14 +183,10 @@ mod bench {
IndexRecordOption::Basic,
);
let agg_req_1: Aggregations = vec![(
"average_f64".to_string(),
Aggregation::Metric(MetricAggregation::Stats(StatsAggregation::from_field_name(
"score_f64".to_string(),
))),
)]
.into_iter()
.collect();
let agg_req_1: Aggregations = serde_json::from_value(json!({
"average_f64": { "stats": { "field": "score_f64", } }
}))
.unwrap();
let collector = get_collector(agg_req_1);
@@ -170,14 +208,10 @@ mod bench {
IndexRecordOption::Basic,
);
let agg_req_1: Aggregations = vec![(
"average_f64".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("score_f64".to_string()),
)),
)]
.into_iter()
.collect();
let agg_req_1: Aggregations = serde_json::from_value(json!({
"average_f64": { "avg": { "field": "score_f64", } }
}))
.unwrap();
let collector = get_collector(agg_req_1);
@@ -186,6 +220,31 @@ mod bench {
});
}
bench_all_cardinalities!(bench_aggregation_percentiles_f64);
fn bench_aggregation_percentiles_f64_card(b: &mut Bencher, cardinality: Cardinality) {
let index = get_test_index_bench(cardinality).unwrap();
let reader = index.reader().unwrap();
b.iter(|| {
let agg_req_str = r#"
{
"mypercentiles": {
"percentiles": {
"field": "score_f64",
"percents": [ 95, 99, 99.9 ]
}
}
} "#;
let agg_req_1: Aggregations = serde_json::from_str(agg_req_str).unwrap();
let collector = get_collector(agg_req_1);
let searcher = reader.searcher();
searcher.search(&AllQuery, &collector).unwrap()
});
}
bench_all_cardinalities!(bench_aggregation_average_u64_and_f64);
fn bench_aggregation_average_u64_and_f64_card(b: &mut Bencher, cardinality: Cardinality) {
@@ -199,22 +258,11 @@ mod bench {
IndexRecordOption::Basic,
);
let agg_req_1: Aggregations = vec![
(
"average_f64".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("score_f64".to_string()),
)),
),
(
"average".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("score".to_string()),
)),
),
]
.into_iter()
.collect();
let agg_req_1: Aggregations = serde_json::from_value(json!({
"average_f64": { "avg": { "field": "score_f64" } },
"average": { "avg": { "field": "score" } },
}))
.unwrap();
let collector = get_collector(agg_req_1);
@@ -230,21 +278,10 @@ mod bench {
let reader = index.reader().unwrap();
b.iter(|| {
let agg_req: Aggregations = vec![(
"my_texts".to_string(),
Aggregation::Bucket(
BucketAggregation {
bucket_agg: BucketAggregationType::Terms(TermsAggregation {
field: "text_few_terms".to_string(),
..Default::default()
}),
sub_aggregation: Default::default(),
}
.into(),
),
)]
.into_iter()
.collect();
let agg_req: Aggregations = serde_json::from_value(json!({
"my_texts": { "terms": { "field": "text_few_terms" } },
}))
.unwrap();
let collector = get_collector(agg_req);
@@ -260,30 +297,42 @@ mod bench {
let reader = index.reader().unwrap();
b.iter(|| {
let sub_agg_req: Aggregations = vec![(
"average_f64".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("score_f64".to_string()),
)),
)]
.into_iter()
.collect();
let agg_req: Aggregations = vec![(
"my_texts".to_string(),
Aggregation::Bucket(
BucketAggregation {
bucket_agg: BucketAggregationType::Terms(TermsAggregation {
field: "text_many_terms".to_string(),
..Default::default()
}),
sub_aggregation: sub_agg_req,
let agg_req: Aggregations = serde_json::from_value(json!({
"my_texts": {
"terms": { "field": "text_many_terms" },
"aggs": {
"average_f64": { "avg": { "field": "score_f64" } }
}
.into(),
),
)]
.into_iter()
.collect();
},
}))
.unwrap();
let collector = get_collector(agg_req);
let searcher = reader.searcher();
searcher.search(&AllQuery, &collector).unwrap()
});
}
bench_all_cardinalities!(bench_aggregation_terms_many_json_mixed_type_with_sub_agg);
fn bench_aggregation_terms_many_json_mixed_type_with_sub_agg_card(
b: &mut Bencher,
cardinality: Cardinality,
) {
let index = get_test_index_bench(cardinality).unwrap();
let reader = index.reader().unwrap();
b.iter(|| {
let agg_req: Aggregations = serde_json::from_value(json!({
"my_texts": {
"terms": { "field": "json.mixed_type" },
"aggs": {
"average_f64": { "avg": { "field": "score_f64" } }
}
},
}))
.unwrap();
let collector = get_collector(agg_req);
@@ -299,21 +348,10 @@ mod bench {
let reader = index.reader().unwrap();
b.iter(|| {
let agg_req: Aggregations = vec![(
"my_texts".to_string(),
Aggregation::Bucket(
BucketAggregation {
bucket_agg: BucketAggregationType::Terms(TermsAggregation {
field: "text_many_terms".to_string(),
..Default::default()
}),
sub_aggregation: Default::default(),
}
.into(),
),
)]
.into_iter()
.collect();
let agg_req: Aggregations = serde_json::from_value(json!({
"my_texts": { "terms": { "field": "text_many_terms" } },
}))
.unwrap();
let collector = get_collector(agg_req);
@@ -329,25 +367,10 @@ mod bench {
let reader = index.reader().unwrap();
b.iter(|| {
let agg_req: Aggregations = vec![(
"my_texts".to_string(),
Aggregation::Bucket(
BucketAggregation {
bucket_agg: BucketAggregationType::Terms(TermsAggregation {
field: "text_many_terms".to_string(),
order: Some(CustomOrder {
order: Order::Desc,
target: OrderTarget::Key,
}),
..Default::default()
}),
sub_aggregation: Default::default(),
}
.into(),
),
)]
.into_iter()
.collect();
let agg_req: Aggregations = serde_json::from_value(json!({
"my_texts": { "terms": { "field": "text_many_terms", "order": { "_key": "desc" } } },
}))
.unwrap();
let collector = get_collector(agg_req);
@@ -363,29 +386,17 @@ mod bench {
let reader = index.reader().unwrap();
b.iter(|| {
let agg_req_1: Aggregations = vec![(
"rangef64".to_string(),
Aggregation::Bucket(
BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "score_f64".to_string(),
ranges: vec![
(3f64..7000f64).into(),
(7000f64..20000f64).into(),
(20000f64..30000f64).into(),
(30000f64..40000f64).into(),
(40000f64..50000f64).into(),
(50000f64..60000f64).into(),
],
..Default::default()
}),
sub_aggregation: Default::default(),
}
.into(),
),
)]
.into_iter()
.collect();
let agg_req_1: Aggregations = serde_json::from_value(json!({
"range_f64": { "range": { "field": "score_f64", "ranges": [
{ "from": 3, "to": 7000 },
{ "from": 7000, "to": 20000 },
{ "from": 20000, "to": 30000 },
{ "from": 30000, "to": 40000 },
{ "from": 40000, "to": 50000 },
{ "from": 50000, "to": 60000 }
] } },
}))
.unwrap();
let collector = get_collector(agg_req_1);
@@ -401,38 +412,25 @@ mod bench {
let reader = index.reader().unwrap();
b.iter(|| {
let sub_agg_req: Aggregations = vec![(
"average_f64".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("score_f64".to_string()),
)),
)]
.into_iter()
.collect();
let agg_req_1: Aggregations = vec![(
"rangef64".to_string(),
Aggregation::Bucket(
BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "score_f64".to_string(),
ranges: vec![
(3f64..7000f64).into(),
(7000f64..20000f64).into(),
(20000f64..30000f64).into(),
(30000f64..40000f64).into(),
(40000f64..50000f64).into(),
(50000f64..60000f64).into(),
],
..Default::default()
}),
sub_aggregation: sub_agg_req,
let agg_req_1: Aggregations = serde_json::from_value(json!({
"rangef64": {
"range": {
"field": "score_f64",
"ranges": [
{ "from": 3, "to": 7000 },
{ "from": 7000, "to": 20000 },
{ "from": 20000, "to": 30000 },
{ "from": 30000, "to": 40000 },
{ "from": 40000, "to": 50000 },
{ "from": 50000, "to": 60000 }
]
},
"aggs": {
"average_f64": { "avg": { "field": "score_f64" } }
}
.into(),
),
)]
.into_iter()
.collect();
},
}))
.unwrap();
let collector = get_collector(agg_req_1);
@@ -453,26 +451,10 @@ mod bench {
let reader = index.reader().unwrap();
b.iter(|| {
let agg_req_1: Aggregations = vec![(
"rangef64".to_string(),
Aggregation::Bucket(
BucketAggregation {
bucket_agg: BucketAggregationType::Histogram(HistogramAggregation {
field: "score_f64".to_string(),
interval: 100f64,
hard_bounds: Some(HistogramBounds {
min: 1000.0,
max: 300_000.0,
}),
..Default::default()
}),
sub_aggregation: Default::default(),
}
.into(),
),
)]
.into_iter()
.collect();
let agg_req_1: Aggregations = serde_json::from_value(json!({
"rangef64": { "histogram": { "field": "score_f64", "interval": 100, "hard_bounds": { "min": 1000, "max": 300000 } } },
}))
.unwrap();
let collector = get_collector(agg_req_1);
let searcher = reader.searcher();
@@ -487,31 +469,15 @@ mod bench {
let reader = index.reader().unwrap();
b.iter(|| {
let sub_agg_req: Aggregations = vec![(
"average_f64".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("score_f64".to_string()),
)),
)]
.into_iter()
.collect();
let agg_req_1: Aggregations = vec![(
"rangef64".to_string(),
Aggregation::Bucket(
BucketAggregation {
bucket_agg: BucketAggregationType::Histogram(HistogramAggregation {
field: "score_f64".to_string(),
interval: 100f64, // 1000 buckets
..Default::default()
}),
sub_aggregation: sub_agg_req,
let agg_req_1: Aggregations = serde_json::from_value(json!({
"rangef64": {
"histogram": { "field": "score_f64", "interval": 100 },
"aggs": {
"average_f64": { "avg": { "field": "score_f64" } }
}
.into(),
),
)]
.into_iter()
.collect();
}
}))
.unwrap();
let collector = get_collector(agg_req_1);
@@ -527,22 +493,15 @@ mod bench {
let reader = index.reader().unwrap();
b.iter(|| {
let agg_req_1: Aggregations = vec![(
"rangef64".to_string(),
Aggregation::Bucket(
BucketAggregation {
bucket_agg: BucketAggregationType::Histogram(HistogramAggregation {
field: "score_f64".to_string(),
interval: 100f64, // 1000 buckets
..Default::default()
}),
sub_aggregation: Default::default(),
}
.into(),
),
)]
.into_iter()
.collect();
let agg_req_1: Aggregations = serde_json::from_value(json!({
"rangef64": {
"histogram": {
"field": "score_f64",
"interval": 100 // 1000 buckets
},
}
}))
.unwrap();
let collector = get_collector(agg_req_1);
@@ -564,43 +523,23 @@ mod bench {
IndexRecordOption::Basic,
);
let sub_agg_req_1: Aggregations = vec![(
"average_in_range".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("score".to_string()),
)),
)]
.into_iter()
.collect();
let agg_req_1: Aggregations = vec![
(
"average".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("score".to_string()),
)),
),
(
"rangef64".to_string(),
Aggregation::Bucket(
BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "score_f64".to_string(),
ranges: vec![
(3f64..7000f64).into(),
(7000f64..20000f64).into(),
(20000f64..60000f64).into(),
],
..Default::default()
}),
sub_aggregation: sub_agg_req_1,
}
.into(),
),
),
]
.into_iter()
.collect();
let agg_req_1: Aggregations = serde_json::from_value(json!({
"rangef64": {
"range": {
"field": "score_f64",
"ranges": [
{ "from": 3, "to": 7000 },
{ "from": 7000, "to": 20000 },
{ "from": 20000, "to": 60000 }
]
},
"aggs": {
"average_in_range": { "avg": { "field": "score" } }
}
},
"average": { "avg": { "field": "score" } }
}))
.unwrap();
let collector = get_collector(agg_req_1);

85
src/aggregation/agg_limits.rs
View File

@@ -1,12 +1,11 @@
use std::collections::HashMap;
use std::sync::atomic::AtomicU64;
use std::sync::atomic::{AtomicU64, Ordering};
use std::sync::Arc;
use common::ByteCount;
use super::collector::DEFAULT_MEMORY_LIMIT;
use super::{AggregationError, DEFAULT_BUCKET_LIMIT};
use crate::TantivyError;
/// An estimate for memory consumption. Non recursive
pub trait MemoryConsumption {
@@ -15,8 +14,8 @@ pub trait MemoryConsumption {
impl<K, V, S> MemoryConsumption for HashMap<K, V, S> {
fn memory_consumption(&self) -> usize {
let num_items = self.capacity();
(std::mem::size_of::<K>() + std::mem::size_of::<V>()) * num_items
let capacity = self.capacity();
(std::mem::size_of::<K>() + std::mem::size_of::<V>() + 1) * capacity
}
}
@@ -61,6 +60,8 @@ impl AggregationLimits {
/// *bucket_limit*
/// Limits the maximum number of buckets returned from an aggregation request.
/// bucket_limit will default to `DEFAULT_BUCKET_LIMIT` (65000)
///
/// Note: The returned instance contains a Arc shared counter to track memory consumption.
pub fn new(memory_limit: Option<u64>, bucket_limit: Option<u32>) -> Self {
Self {
memory_consumption: Default::default(),
@@ -68,28 +69,68 @@ impl AggregationLimits {
bucket_limit: bucket_limit.unwrap_or(DEFAULT_BUCKET_LIMIT),
}
}
pub(crate) fn validate_memory_consumption(&self) -> crate::Result<()> {
if self.get_memory_consumed() > self.memory_limit {
return Err(TantivyError::AggregationError(
AggregationError::MemoryExceeded {
limit: self.memory_limit,
current: self.get_memory_consumed(),
},
));
/// Create a new ResourceLimitGuard, that will release the memory when dropped.
pub fn new_guard(&self) -> ResourceLimitGuard {
ResourceLimitGuard {
/// The counter which is shared between the aggregations for one request.
memory_consumption: Arc::clone(&self.memory_consumption),
/// The memory_limit in bytes
memory_limit: self.memory_limit,
allocated_with_the_guard: 0,
}
}
pub(crate) fn add_memory_consumed(&self, num_bytes: u64) -> crate::Result<()> {
self.memory_consumption
.fetch_add(num_bytes, Ordering::Relaxed);
validate_memory_consumption(&self.memory_consumption, self.memory_limit)?;
Ok(())
}
pub(crate) fn add_memory_consumed(&self, num_bytes: u64) {
self.memory_consumption
.fetch_add(num_bytes, std::sync::atomic::Ordering::Relaxed);
}
/// Returns the estimated memory consumed by the aggregations
pub fn get_memory_consumed(&self) -> ByteCount {
self.memory_consumption
.load(std::sync::atomic::Ordering::Relaxed)
.into()
}
pub(crate) fn get_bucket_limit(&self) -> u32 {
self.bucket_limit
}
}
fn validate_memory_consumption(
memory_consumption: &AtomicU64,
memory_limit: ByteCount,
) -> Result<(), AggregationError> {
// Load the estimated memory consumed by the aggregations
let memory_consumed: ByteCount = memory_consumption.load(Ordering::Relaxed).into();
if memory_consumed > memory_limit {
return Err(AggregationError::MemoryExceeded {
limit: memory_limit,
current: memory_consumed,
});
}
Ok(())
}
pub struct ResourceLimitGuard {
/// The counter which is shared between the aggregations for one request.
memory_consumption: Arc<AtomicU64>,
/// The memory_limit in bytes
memory_limit: ByteCount,
/// Allocated memory with this guard.
allocated_with_the_guard: u64,
}
impl ResourceLimitGuard {
pub(crate) fn add_memory_consumed(&self, num_bytes: u64) -> crate::Result<()> {
self.memory_consumption
.fetch_add(num_bytes, Ordering::Relaxed);
validate_memory_consumption(&self.memory_consumption, self.memory_limit)?;
Ok(())
}
}
impl Drop for ResourceLimitGuard {
/// Removes the memory consumed tracked by this _instance_ of AggregationLimits.
/// This is used to clear the segment specific memory consumption all at once.
fn drop(&mut self) {
self.memory_consumption
.fetch_sub(self.allocated_with_the_guard, Ordering::Relaxed);
}
}

439
src/aggregation/agg_req.rs
View File

@@ -9,25 +9,7 @@
//! # 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(Box::new(BucketAggregation {
//! bucket_agg: BucketAggregationType::Range(RangeAggregation{
//! field: "score".to_string(),
//! ranges: vec![(3f64..7f64).into(), (7f64..20f64).into()],
//! keyed: false,
//! }),
//! sub_aggregation: Default::default(),
//! })),
//! ),
//! ]
//! .into_iter()
//! .collect();
//! use tantivy::aggregation::agg_req::Aggregations;
//!
//! let elasticsearch_compatible_json_req = r#"
//! {
@@ -41,89 +23,78 @@
//! }
//! }
//! }"#;
//! let agg_req2: Aggregations = serde_json::from_str(elasticsearch_compatible_json_req).unwrap();
//! assert_eq!(agg_req1, agg_req2);
//! let _agg_req: Aggregations = serde_json::from_str(elasticsearch_compatible_json_req).unwrap();
//! ```
use std::collections::{HashMap, HashSet};
use serde::{Deserialize, Serialize};
pub use super::bucket::RangeAggregation;
use super::bucket::{DateHistogramAggregationReq, HistogramAggregation, TermsAggregation};
use super::metric::{
AverageAggregation, CountAggregation, MaxAggregation, MinAggregation, StatsAggregation,
SumAggregation,
use super::bucket::{
DateHistogramAggregationReq, HistogramAggregation, RangeAggregation, TermsAggregation,
};
use super::metric::{
AverageAggregation, CountAggregation, MaxAggregation, MinAggregation,
PercentilesAggregationReq, StatsAggregation, SumAggregation,
};
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.
/// defined names. It is also used in buckets aggregations 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>,
/// Aggregation request.
///
/// An aggregation is either a bucket or a metric.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
#[serde(try_from = "AggregationForDeserialization")]
pub struct Aggregation {
/// The aggregation variant, which can be either a bucket or a metric.
#[serde(flatten)]
pub agg: AggregationVariants,
/// on the document set in the bucket.
#[serde(rename = "aggs")]
#[serde(skip_serializing_if = "Aggregations::is_empty")]
pub sub_aggregation: Aggregations,
}
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),
}
/// In order to display proper error message, we cannot rely on flattening
/// the json enum. Instead we introduce an intermediary struct to separate
/// the aggregation from the subaggregation.
#[derive(Deserialize)]
struct AggregationForDeserialization {
#[serde(flatten)]
pub aggs_remaining_json: serde_json::Value,
#[serde(rename = "aggs")]
#[serde(default)]
pub sub_aggregation: Aggregations,
}
impl TryFrom<AggregationForDeserialization> for Aggregation {
type Error = serde_json::Error;
fn try_from(value: AggregationForDeserialization) -> serde_json::Result<Self> {
let AggregationForDeserialization {
aggs_remaining_json,
sub_aggregation,
} = value;
let agg: AggregationVariants = serde_json::from_value(aggs_remaining_json)?;
Ok(Aggregation {
agg,
sub_aggregation,
})
}
}
#[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 Aggregation {
pub(crate) fn sub_aggregation(&self) -> &Aggregations {
&self.sub_aggregation
}
impl BucketAggregationInternal {
pub(crate) fn as_range(&self) -> Option<&RangeAggregation> {
match &self.bucket_agg {
BucketAggregationType::Range(range) => Some(range),
_ => None,
}
}
pub(crate) fn as_histogram(&self) -> crate::Result<Option<HistogramAggregation>> {
match &self.bucket_agg {
BucketAggregationType::Histogram(histogram) => Ok(Some(histogram.clone())),
BucketAggregationType::DateHistogram(histogram) => {
Ok(Some(histogram.to_histogram_req()?))
}
_ => Ok(None),
}
}
pub(crate) fn as_term(&self) -> Option<&TermsAggregation> {
match &self.bucket_agg {
BucketAggregationType::Terms(terms) => Some(terms),
_ => None,
}
fn get_fast_field_names(&self, fast_field_names: &mut HashSet<String>) {
fast_field_names.insert(self.agg.get_fast_field_name().to_string());
fast_field_names.extend(get_fast_field_names(&self.sub_aggregation));
}
}
@@ -136,97 +107,24 @@ pub fn get_fast_field_names(aggs: &Aggregations) -> HashSet<String> {
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(Box<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) => {
fast_field_names.insert(metric.get_fast_field_name().to_string());
}
}
}
}
/// 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>) {
let fast_field_name = self.bucket_agg.get_fast_field_name();
fast_field_names.insert(fast_field_name.to_string());
fast_field_names.extend(get_fast_field_names(&self.sub_aggregation));
}
}
/// The bucket aggregation types.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub enum BucketAggregationType {
/// All aggregation types.
pub enum AggregationVariants {
// Bucket aggregation types
/// Put data into buckets of user-defined ranges.
#[serde(rename = "range")]
Range(RangeAggregation),
/// Put data into buckets of user-defined ranges.
/// Put data into a histogram.
#[serde(rename = "histogram")]
Histogram(HistogramAggregation),
/// Put data into buckets of user-defined ranges.
/// Put data into a date histogram.
#[serde(rename = "date_histogram")]
DateHistogram(DateHistogramAggregationReq),
/// Put data into buckets of terms.
#[serde(rename = "terms")]
Terms(TermsAggregation),
}
impl BucketAggregationType {
fn get_fast_field_name(&self) -> &str {
match self {
BucketAggregationType::Terms(terms) => terms.field.as_str(),
BucketAggregationType::Range(range) => range.field.as_str(),
BucketAggregationType::Histogram(histogram) => histogram.field.as_str(),
BucketAggregationType::DateHistogram(histogram) => histogram.field.as_str(),
}
}
}
/// 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 {
// Metric aggregation types
/// Computes the average of the extracted values.
#[serde(rename = "avg")]
Average(AverageAggregation),
@@ -246,25 +144,108 @@ pub enum MetricAggregation {
/// Computes the sum of the extracted values.
#[serde(rename = "sum")]
Sum(SumAggregation),
/// Computes the sum of the extracted values.
#[serde(rename = "percentiles")]
Percentiles(PercentilesAggregationReq),
}
impl MetricAggregation {
fn get_fast_field_name(&self) -> &str {
impl AggregationVariants {
/// Returns the name of the field used by the aggregation.
pub fn get_fast_field_name(&self) -> &str {
match self {
MetricAggregation::Average(avg) => avg.field_name(),
MetricAggregation::Count(count) => count.field_name(),
MetricAggregation::Max(max) => max.field_name(),
MetricAggregation::Min(min) => min.field_name(),
MetricAggregation::Stats(stats) => stats.field_name(),
MetricAggregation::Sum(sum) => sum.field_name(),
AggregationVariants::Terms(terms) => terms.field.as_str(),
AggregationVariants::Range(range) => range.field.as_str(),
AggregationVariants::Histogram(histogram) => histogram.field.as_str(),
AggregationVariants::DateHistogram(histogram) => histogram.field.as_str(),
AggregationVariants::Average(avg) => avg.field_name(),
AggregationVariants::Count(count) => count.field_name(),
AggregationVariants::Max(max) => max.field_name(),
AggregationVariants::Min(min) => min.field_name(),
AggregationVariants::Stats(stats) => stats.field_name(),
AggregationVariants::Sum(sum) => sum.field_name(),
AggregationVariants::Percentiles(per) => per.field_name(),
}
}
pub(crate) fn as_range(&self) -> Option<&RangeAggregation> {
match &self {
AggregationVariants::Range(range) => Some(range),
_ => None,
}
}
pub(crate) fn as_histogram(&self) -> crate::Result<Option<HistogramAggregation>> {
match &self {
AggregationVariants::Histogram(histogram) => Ok(Some(histogram.clone())),
AggregationVariants::DateHistogram(histogram) => {
Ok(Some(histogram.to_histogram_req()?))
}
_ => Ok(None),
}
}
pub(crate) fn as_term(&self) -> Option<&TermsAggregation> {
match &self {
AggregationVariants::Terms(terms) => Some(terms),
_ => None,
}
}
pub(crate) fn as_percentile(&self) -> Option<&PercentilesAggregationReq> {
match &self {
AggregationVariants::Percentiles(percentile_req) => Some(percentile_req),
_ => None,
}
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn deser_json_test() {
let agg_req_json = r#"{
"price_avg": { "avg": { "field": "price" } },
"price_count": { "value_count": { "field": "price" } },
"price_max": { "max": { "field": "price" } },
"price_min": { "min": { "field": "price" } },
"price_stats": { "stats": { "field": "price" } },
"price_sum": { "sum": { "field": "price" } }
}"#;
let _agg_req: Aggregations = serde_json::from_str(agg_req_json).unwrap();
}
#[test]
fn deser_json_test_bucket() {
let agg_req_json = r#"
{
"termagg": {
"terms": {
"field": "json.mixed_type",
"order": { "min_price": "desc" }
},
"aggs": {
"min_price": { "min": { "field": "json.mixed_type" } }
}
},
"rangeagg": {
"range": {
"field": "json.mixed_type",
"ranges": [
{ "to": 3.0 },
{ "from": 19.0, "to": 20.0 },
{ "from": 20.0 }
]
},
"aggs": {
"average_in_range": { "avg": { "field": "json.mixed_type" } }
}
}
} "#;
let _agg_req: Aggregations = serde_json::from_str(agg_req_json).unwrap();
}
#[test]
fn test_metric_aggregations_deser() {
let agg_req_json = r#"{
@@ -278,46 +259,27 @@ mod tests {
let agg_req: Aggregations = serde_json::from_str(agg_req_json).unwrap();
assert!(
matches!(agg_req.get("price_avg").unwrap(), Aggregation::Metric(MetricAggregation::Average(avg)) if avg.field == "price")
matches!(&agg_req.get("price_avg").unwrap().agg, AggregationVariants::Average(avg) if avg.field == "price")
);
assert!(
matches!(agg_req.get("price_count").unwrap(), Aggregation::Metric(MetricAggregation::Count(count)) if count.field == "price")
matches!(&agg_req.get("price_count").unwrap().agg, AggregationVariants::Count(count) if count.field == "price")
);
assert!(
matches!(agg_req.get("price_max").unwrap(), Aggregation::Metric(MetricAggregation::Max(max)) if max.field == "price")
matches!(&agg_req.get("price_max").unwrap().agg, AggregationVariants::Max(max) if max.field == "price")
);
assert!(
matches!(agg_req.get("price_min").unwrap(), Aggregation::Metric(MetricAggregation::Min(min)) if min.field == "price")
matches!(&agg_req.get("price_min").unwrap().agg, AggregationVariants::Min(min) if min.field == "price")
);
assert!(
matches!(agg_req.get("price_stats").unwrap(), Aggregation::Metric(MetricAggregation::Stats(stats)) if stats.field == "price")
matches!(&agg_req.get("price_stats").unwrap().agg, AggregationVariants::Stats(stats) if stats.field == "price")
);
assert!(
matches!(agg_req.get("price_sum").unwrap(), Aggregation::Metric(MetricAggregation::Sum(sum)) if sum.field == "price")
matches!(&agg_req.get("price_sum").unwrap().agg, AggregationVariants::Sum(sum) if sum.field == "price")
);
}
#[test]
fn serialize_to_json_test() {
let agg_req1: Aggregations = vec![(
"range".to_string(),
Aggregation::Bucket(Box::new(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(),
],
keyed: true,
}),
sub_aggregation: Default::default(),
})),
)]
.into_iter()
.collect();
let elasticsearch_compatible_json_req = r#"{
"range": {
"range": {
@@ -342,57 +304,56 @@ mod tests {
}
}
}"#;
let agg_req1: Aggregations =
{ serde_json::from_str(elasticsearch_compatible_json_req).unwrap() };
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(Box::new(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(),
],
..Default::default()
}),
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(Box::new(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(),
let range_agg: Aggregation = {
serde_json::from_value(json!({
"range": {
"field": "score",
"ranges": [
{ "to": 3.0 },
{ "from": 3.0, "to": 7.0 },
{ "from": 7.0, "to": 20.0 },
{ "from": 20.0 }
],
..Default::default()
}),
sub_aggregation: agg_req2,
})),
)]
.into_iter()
.collect();
}
}))
.unwrap()
};
let agg_req1: Aggregations = {
serde_json::from_value(json!({
"range1": range_agg,
"range2":{
"range": {
"field": "score2",
"ranges": [
{ "to": 3.0 },
{ "from": 3.0, "to": 7.0 },
{ "from": 7.0, "to": 20.0 },
{ "from": 20.0 }
],
},
"aggs": {
"metric": {
"avg": {
"field": "field123"
}
}
}
}
}))
.unwrap()
};
assert_eq!(
get_fast_field_names(&agg_req1),

400
src/aggregation/agg_req_with_accessor.rs
View File

@@ -2,7 +2,8 @@
use columnar::{Column, ColumnBlockAccessor, ColumnType, StrColumn};
use super::agg_req::{Aggregation, Aggregations, BucketAggregationType, MetricAggregation};
use super::agg_limits::ResourceLimitGuard;
use super::agg_req::{Aggregation, AggregationVariants, Aggregations};
use super::bucket::{
DateHistogramAggregationReq, HistogramAggregation, RangeAggregation, TermsAggregation,
};
@@ -12,38 +13,262 @@ use super::metric::{
};
use super::segment_agg_result::AggregationLimits;
use super::VecWithNames;
use crate::aggregation::{f64_to_fastfield_u64, Key};
use crate::SegmentReader;
#[derive(Clone, Default)]
#[derive(Default)]
pub(crate) struct AggregationsWithAccessor {
pub metrics: VecWithNames<MetricAggregationWithAccessor>,
pub buckets: VecWithNames<BucketAggregationWithAccessor>,
pub aggs: VecWithNames<AggregationWithAccessor>,
}
impl AggregationsWithAccessor {
fn from_data(
metrics: VecWithNames<MetricAggregationWithAccessor>,
buckets: VecWithNames<BucketAggregationWithAccessor>,
) -> Self {
Self { metrics, buckets }
fn from_data(aggs: VecWithNames<AggregationWithAccessor>) -> Self {
Self { aggs }
}
pub fn is_empty(&self) -> bool {
self.metrics.is_empty() && self.buckets.is_empty()
self.aggs.is_empty()
}
}
#[derive(Clone)]
pub struct BucketAggregationWithAccessor {
pub struct AggregationWithAccessor {
/// 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.
/// based on search terms. That is not that case currently, but eventually this needs to be
/// Option or moved.
pub(crate) accessor: Column<u64>,
/// Load insert u64 for missing use case
pub(crate) missing_value_for_accessor: Option<u64>,
pub(crate) str_dict_column: Option<StrColumn>,
pub(crate) field_type: ColumnType,
pub(crate) bucket_agg: BucketAggregationType,
pub(crate) sub_aggregation: AggregationsWithAccessor,
pub(crate) limits: AggregationLimits,
pub(crate) limits: ResourceLimitGuard,
pub(crate) column_block_accessor: ColumnBlockAccessor<u64>,
/// Used for missing term aggregation, which checks all columns for existence.
/// By convention the missing aggregation is chosen, when this property is set
/// (instead bein set in `agg`).
/// If this needs to used by other aggregations, we need to refactor this.
pub(crate) accessors: Vec<Column<u64>>,
pub(crate) agg: Aggregation,
}
impl AggregationWithAccessor {
/// May return multiple accessors if the aggregation is e.g. on mixed field types.
fn try_from_agg(
agg: &Aggregation,
sub_aggregation: &Aggregations,
reader: &SegmentReader,
limits: AggregationLimits,
) -> crate::Result<Vec<AggregationWithAccessor>> {
let add_agg_with_accessor = |accessor: Column<u64>,
column_type: ColumnType,
aggs: &mut Vec<AggregationWithAccessor>|
-> crate::Result<()> {
let res = AggregationWithAccessor {
accessor,
accessors: Vec::new(),
field_type: column_type,
sub_aggregation: get_aggs_with_segment_accessor_and_validate(
sub_aggregation,
reader,
&limits,
)?,
agg: agg.clone(),
limits: limits.new_guard(),
missing_value_for_accessor: None,
str_dict_column: None,
column_block_accessor: Default::default(),
};
aggs.push(res);
Ok(())
};
let mut res: Vec<AggregationWithAccessor> = Vec::new();
use AggregationVariants::*;
match &agg.agg {
Range(RangeAggregation {
field: field_name, ..
}) => {
let (accessor, column_type) =
get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?;
add_agg_with_accessor(accessor, column_type, &mut res)?;
}
Histogram(HistogramAggregation {
field: field_name, ..
}) => {
let (accessor, column_type) =
get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?;
add_agg_with_accessor(accessor, column_type, &mut res)?;
}
DateHistogram(DateHistogramAggregationReq {
field: field_name, ..
}) => {
let (accessor, column_type) =
get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?;
add_agg_with_accessor(accessor, column_type, &mut res)?;
}
Terms(TermsAggregation {
field: field_name,
missing,
..
}) => {
let str_dict_column = reader.fast_fields().str(field_name)?;
let allowed_column_types = [
ColumnType::I64,
ColumnType::U64,
ColumnType::F64,
ColumnType::Str,
ColumnType::DateTime,
// ColumnType::Bytes Unsupported
// ColumnType::Bool Unsupported
// ColumnType::IpAddr Unsupported
];
// In case the column is empty we want the shim column to match the missing type
let fallback_type = missing
.as_ref()
.map(|missing| match missing {
Key::Str(_) => ColumnType::Str,
Key::F64(_) => ColumnType::F64,
})
.unwrap_or(ColumnType::U64);
let column_and_types = get_all_ff_reader_or_empty(
reader,
field_name,
Some(&allowed_column_types),
fallback_type,
)?;
let missing_and_more_than_one_col = column_and_types.len() > 1 && missing.is_some();
let text_on_non_text_col = column_and_types.len() == 1
&& column_and_types[0].1.numerical_type().is_some()
&& missing
.as_ref()
.map(|m| matches!(m, Key::Str(_)))
.unwrap_or(false);
// Actually we could convert the text to a number and have the fast path, if it is
// provided in Rfc3339 format. But this use case is probably common
// enough to justify the effort.
let text_on_date_col = column_and_types.len() == 1
&& column_and_types[0].1 == ColumnType::DateTime
&& missing
.as_ref()
.map(|m| matches!(m, Key::Str(_)))
.unwrap_or(false);
let use_special_missing_agg =
missing_and_more_than_one_col || text_on_non_text_col || text_on_date_col;
if use_special_missing_agg {
let column_and_types =
get_all_ff_reader_or_empty(reader, field_name, None, fallback_type)?;
let accessors: Vec<Column> =
column_and_types.iter().map(|(a, _)| a.clone()).collect();
let agg_wit_acc = AggregationWithAccessor {
missing_value_for_accessor: None,
accessor: accessors[0].clone(),
accessors,
field_type: ColumnType::U64,
sub_aggregation: get_aggs_with_segment_accessor_and_validate(
sub_aggregation,
reader,
&limits,
)?,
agg: agg.clone(),
str_dict_column: str_dict_column.clone(),
limits: limits.new_guard(),
column_block_accessor: Default::default(),
};
res.push(agg_wit_acc);
}
for (accessor, column_type) in column_and_types {
let missing_value_term_agg = if use_special_missing_agg {
None
} else {
missing.clone()
};
let missing_value_for_accessor =
if let Some(missing) = missing_value_term_agg.as_ref() {
get_missing_val(column_type, missing, agg.agg.get_fast_field_name())?
} else {
None
};
let agg = AggregationWithAccessor {
missing_value_for_accessor,
accessor,
accessors: Vec::new(),
field_type: column_type,
sub_aggregation: get_aggs_with_segment_accessor_and_validate(
sub_aggregation,
reader,
&limits,
)?,
agg: agg.clone(),
str_dict_column: str_dict_column.clone(),
limits: limits.new_guard(),
column_block_accessor: Default::default(),
};
res.push(agg);
}
}
Average(AverageAggregation {
field: field_name, ..
})
| Count(CountAggregation {
field: field_name, ..
})
| Max(MaxAggregation {
field: field_name, ..
})
| Min(MinAggregation {
field: field_name, ..
})
| Stats(StatsAggregation {
field: field_name, ..
})
| Sum(SumAggregation {
field: field_name, ..
}) => {
let (accessor, column_type) =
get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?;
add_agg_with_accessor(accessor, column_type, &mut res)?;
}
Percentiles(percentiles) => {
let (accessor, column_type) = get_ff_reader(
reader,
percentiles.field_name(),
Some(get_numeric_or_date_column_types()),
)?;
add_agg_with_accessor(accessor, column_type, &mut res)?;
}
};
Ok(res)
}
}
fn get_missing_val(
column_type: ColumnType,
missing: &Key,
field_name: &str,
) -> crate::Result<Option<u64>> {
let missing_val = match missing {
Key::Str(_) if column_type == ColumnType::Str => Some(u64::MAX),
// Allow fallback to number on text fields
Key::F64(_) if column_type == ColumnType::Str => Some(u64::MAX),
Key::F64(val) if column_type.numerical_type().is_some() => {
f64_to_fastfield_u64(*val, &column_type)
}
_ => {
return Err(crate::TantivyError::InvalidArgument(format!(
"Missing value {:?} for field {} is not supported for column type {:?}",
missing, field_name, column_type
)));
}
};
Ok(missing_val)
}
fn get_numeric_or_date_column_types() -> &'static [ColumnType] {
@@ -55,131 +280,30 @@ fn get_numeric_or_date_column_types() -> &'static [ColumnType] {
]
}
impl BucketAggregationWithAccessor {
fn try_from_bucket(
bucket: &BucketAggregationType,
sub_aggregation: &Aggregations,
reader: &SegmentReader,
limits: AggregationLimits,
) -> crate::Result<BucketAggregationWithAccessor> {
let mut str_dict_column = None;
let (accessor, field_type) = match &bucket {
BucketAggregationType::Range(RangeAggregation {
field: field_name, ..
}) => get_ff_reader_and_validate(
reader,
field_name,
Some(get_numeric_or_date_column_types()),
)?,
BucketAggregationType::Histogram(HistogramAggregation {
field: field_name, ..
}) => get_ff_reader_and_validate(
reader,
field_name,
Some(get_numeric_or_date_column_types()),
)?,
BucketAggregationType::DateHistogram(DateHistogramAggregationReq {
field: field_name,
..
}) => get_ff_reader_and_validate(
reader,
field_name,
Some(get_numeric_or_date_column_types()),
)?,
BucketAggregationType::Terms(TermsAggregation {
field: field_name, ..
}) => {
str_dict_column = reader.fast_fields().str(field_name)?;
get_ff_reader_and_validate(reader, field_name, None)?
}
};
let sub_aggregation = sub_aggregation.clone();
Ok(BucketAggregationWithAccessor {
accessor,
field_type,
sub_aggregation: get_aggs_with_accessor_and_validate(
&sub_aggregation,
reader,
&limits.clone(),
)?,
bucket_agg: bucket.clone(),
str_dict_column,
limits,
column_block_accessor: Default::default(),
})
}
}
/// Contains the metric request and the fast field accessor.
#[derive(Clone)]
pub struct MetricAggregationWithAccessor {
pub metric: MetricAggregation,
pub field_type: ColumnType,
pub accessor: Column<u64>,
pub column_block_accessor: ColumnBlockAccessor<u64>,
}
impl MetricAggregationWithAccessor {
fn try_from_metric(
metric: &MetricAggregation,
reader: &SegmentReader,
) -> crate::Result<MetricAggregationWithAccessor> {
match &metric {
MetricAggregation::Average(AverageAggregation { field: field_name })
| MetricAggregation::Count(CountAggregation { field: field_name })
| MetricAggregation::Max(MaxAggregation { field: field_name })
| MetricAggregation::Min(MinAggregation { field: field_name })
| MetricAggregation::Stats(StatsAggregation { field: field_name })
| MetricAggregation::Sum(SumAggregation { field: field_name }) => {
let (accessor, field_type) = get_ff_reader_and_validate(
reader,
field_name,
Some(get_numeric_or_date_column_types()),
)?;
Ok(MetricAggregationWithAccessor {
accessor,
field_type,
metric: metric.clone(),
column_block_accessor: Default::default(),
})
}
}
}
}
pub(crate) fn get_aggs_with_accessor_and_validate(
pub(crate) fn get_aggs_with_segment_accessor_and_validate(
aggs: &Aggregations,
reader: &SegmentReader,
limits: &AggregationLimits,
) -> crate::Result<AggregationsWithAccessor> {
let mut metrics = vec![];
let mut buckets = vec![];
let mut aggss = Vec::new();
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,
limits.clone(),
)?,
)),
Aggregation::Metric(metric) => metrics.push((
key.to_string(),
MetricAggregationWithAccessor::try_from_metric(metric, reader)?,
)),
let aggs = AggregationWithAccessor::try_from_agg(
agg,
agg.sub_aggregation(),
reader,
limits.clone(),
)?;
for agg in aggs {
aggss.push((key.to_string(), agg));
}
}
Ok(AggregationsWithAccessor::from_data(
VecWithNames::from_entries(metrics),
VecWithNames::from_entries(buckets),
VecWithNames::from_entries(aggss),
))
}
/// Get fast field reader with given cardinatility.
fn get_ff_reader_and_validate(
/// Get fast field reader or empty as default.
fn get_ff_reader(
reader: &SegmentReader,
field_name: &str,
allowed_column_types: Option<&[ColumnType]>,
@@ -195,3 +319,21 @@ fn get_ff_reader_and_validate(
});
Ok(ff_field_with_type)
}
/// Get all fast field reader or empty as default.
///
/// Is guaranteed to return at least one column.
fn get_all_ff_reader_or_empty(
reader: &SegmentReader,
field_name: &str,
allowed_column_types: Option<&[ColumnType]>,
fallback_type: ColumnType,
) -> crate::Result<Vec<(columnar::Column<u64>, ColumnType)>> {
let ff_fields = reader.fast_fields();
let mut ff_field_with_type =
ff_fields.u64_lenient_for_type_all(allowed_column_types, field_name)?;
if ff_field_with_type.is_empty() {
ff_field_with_type.push((Column::build_empty_column(reader.num_docs()), fallback_type));
}
Ok(ff_field_with_type)
}

47
src/aggregation/agg_result.rs
View File

@@ -7,12 +7,9 @@
use rustc_hash::FxHashMap;
use serde::{Deserialize, Serialize};
use super::agg_req::BucketAggregationInternal;
use super::bucket::GetDocCount;
use super::intermediate_agg_result::{IntermediateBucketResult, IntermediateMetricResult};
use super::metric::{SingleMetricResult, Stats};
use super::segment_agg_result::AggregationLimits;
use super::Key;
use super::metric::{PercentilesMetricResult, SingleMetricResult, Stats};
use super::{AggregationError, Key};
use crate::TantivyError;
#[derive(Clone, Default, Debug, PartialEq, Serialize, Deserialize)]
@@ -37,8 +34,7 @@ impl AggregationResults {
} else {
// Validation is be done during request parsing, so we can't reach this state.
Err(TantivyError::InternalError(format!(
"Can't find aggregation {:?} in sub-aggregations",
name
"Can't find aggregation {name:?} in sub-aggregations"
)))
}
}
@@ -94,6 +90,8 @@ pub enum MetricResult {
Stats(Stats),
/// Sum metric result.
Sum(SingleMetricResult),
/// Sum metric result.
Percentiles(PercentilesMetricResult),
}
impl MetricResult {
@@ -105,30 +103,9 @@ impl MetricResult {
MetricResult::Min(min) => Ok(min.value),
MetricResult::Stats(stats) => stats.get_value(agg_property),
MetricResult::Sum(sum) => Ok(sum.value),
}
}
}
impl From<IntermediateMetricResult> for MetricResult {
fn from(metric: IntermediateMetricResult) -> Self {
match metric {
IntermediateMetricResult::Average(intermediate_avg) => {
MetricResult::Average(intermediate_avg.finalize().into())
}
IntermediateMetricResult::Count(intermediate_count) => {
MetricResult::Count(intermediate_count.finalize().into())
}
IntermediateMetricResult::Max(intermediate_max) => {
MetricResult::Max(intermediate_max.finalize().into())
}
IntermediateMetricResult::Min(intermediate_min) => {
MetricResult::Min(intermediate_min.finalize().into())
}
IntermediateMetricResult::Stats(intermediate_stats) => {
MetricResult::Stats(intermediate_stats.finalize())
}
IntermediateMetricResult::Sum(intermediate_sum) => {
MetricResult::Sum(intermediate_sum.finalize().into())
}
MetricResult::Percentiles(_) => Err(TantivyError::AggregationError(
AggregationError::InvalidRequest("percentiles can't be used to order".to_string()),
)),
}
}
}
@@ -183,14 +160,6 @@ impl BucketResult {
} => buckets.iter().map(|bucket| bucket.get_bucket_count()).sum(),
}
}
pub(crate) fn empty_from_req(
req: &BucketAggregationInternal,
limits: &AggregationLimits,
) -> crate::Result<Self> {
let empty_bucket = IntermediateBucketResult::empty_from_req(&req.bucket_agg);
empty_bucket.into_final_bucket_result(req, limits)
}
}
/// This is the wrapper of buckets entries, which can be vector or hashmap

332
src/aggregation/agg_tests.rs
View File

@@ -1,14 +1,9 @@
use serde_json::Value;
use crate::aggregation::agg_req::{
Aggregation, Aggregations, BucketAggregation, BucketAggregationType, MetricAggregation,
};
use crate::aggregation::agg_req::{Aggregation, Aggregations};
use crate::aggregation::agg_result::AggregationResults;
use crate::aggregation::bucket::{RangeAggregation, TermsAggregation};
use crate::aggregation::buf_collector::DOC_BLOCK_SIZE;
use crate::aggregation::collector::AggregationCollector;
use crate::aggregation::intermediate_agg_result::IntermediateAggregationResults;
use crate::aggregation::metric::AverageAggregation;
use crate::aggregation::segment_agg_result::AggregationLimits;
use crate::aggregation::tests::{get_test_index_2_segments, get_test_index_from_values_and_terms};
use crate::aggregation::DistributedAggregationCollector;
@@ -17,9 +12,12 @@ use crate::schema::{IndexRecordOption, Schema, FAST};
use crate::{Index, Term};
fn get_avg_req(field_name: &str) -> Aggregation {
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name(field_name.to_string()),
))
serde_json::from_value(json!({
"avg": {
"field": field_name,
}
}))
.unwrap()
}
fn get_collector(agg_req: Aggregations) -> AggregationCollector {
@@ -111,7 +109,7 @@ fn test_aggregation_flushing(
let searcher = reader.searcher();
let intermediate_agg_result = searcher.search(&AllQuery, &collector).unwrap();
intermediate_agg_result
.into_final_bucket_result(agg_req, &Default::default())
.into_final_result(agg_req, &Default::default())
.unwrap()
} else {
let collector = get_collector(agg_req);
@@ -198,6 +196,74 @@ fn test_aggregation_flushing_variants() {
test_aggregation_flushing(true, true).unwrap();
}
#[test]
fn test_aggregation_level1_simple() -> crate::Result<()> {
let index = get_test_index_2_segments(true)?;
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 range_agg = |field_name: &str| -> Aggregation {
serde_json::from_value(json!({
"range": {
"field": field_name,
"ranges": [ { "from": 3.0f64, "to": 7.0f64 }, { "from": 7.0f64, "to": 20.0f64 } ]
}
}))
.unwrap()
};
let agg_req_1: Aggregations = vec![
("average".to_string(), get_avg_req("score")),
("range".to_string(), range_agg("score")),
]
.into_iter()
.collect();
let collector = get_collector(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["average"]["value"], 12.142857142857142);
assert_eq!(
res["range"]["buckets"],
json!(
[
{
"key": "*-3",
"doc_count": 1,
"to": 3.0
},
{
"key": "3-7",
"doc_count": 2,
"from": 3.0,
"to": 7.0
},
{
"key": "7-20",
"doc_count": 3,
"from": 7.0,
"to": 20.0
},
{
"key": "20-*",
"doc_count": 1,
"from": 20.0
}
])
);
Ok(())
}
#[test]
fn test_aggregation_level1() -> crate::Result<()> {
let index = get_test_index_2_segments(true)?;
@@ -210,43 +276,23 @@ fn test_aggregation_level1() -> crate::Result<()> {
IndexRecordOption::Basic,
);
let range_agg = |field_name: &str| -> Aggregation {
serde_json::from_value(json!({
"range": {
"field": field_name,
"ranges": [ { "from": 3.0f64, "to": 7.0f64 }, { "from": 7.0f64, "to": 20.0f64 } ]
}
}))
.unwrap()
};
let agg_req_1: Aggregations = vec![
("average_i64".to_string(), get_avg_req("score_i64")),
("average_f64".to_string(), get_avg_req("score_f64")),
("average".to_string(), get_avg_req("score")),
(
"range".to_string(),
Aggregation::Bucket(Box::new(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "score".to_string(),
ranges: vec![(3f64..7f64).into(), (7f64..20f64).into()],
..Default::default()
}),
sub_aggregation: Default::default(),
})),
),
(
"rangef64".to_string(),
Aggregation::Bucket(Box::new(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "score_f64".to_string(),
ranges: vec![(3f64..7f64).into(), (7f64..20f64).into()],
..Default::default()
}),
sub_aggregation: Default::default(),
})),
),
(
"rangei64".to_string(),
Aggregation::Bucket(Box::new(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "score_i64".to_string(),
ranges: vec![(3f64..7f64).into(), (7f64..20f64).into()],
..Default::default()
}),
sub_aggregation: Default::default(),
})),
),
("range".to_string(), range_agg("score")),
("rangef64".to_string(), range_agg("score_f64")),
("rangei64".to_string(), range_agg("score_i64")),
]
.into_iter()
.collect();
@@ -295,7 +341,6 @@ fn test_aggregation_level1() -> crate::Result<()> {
fn test_aggregation_level2(
merge_segments: bool,
use_distributed_collector: bool,
use_elastic_json_req: bool,
) -> crate::Result<()> {
let index = get_test_index_2_segments(merge_segments)?;
@@ -312,23 +357,7 @@ fn test_aggregation_level2(
IndexRecordOption::Basic,
);
let sub_agg_req: Aggregations = vec![
("average_in_range".to_string(), get_avg_req("score")),
(
"term_agg".to_string(),
Aggregation::Bucket(Box::new(BucketAggregation {
bucket_agg: BucketAggregationType::Terms(TermsAggregation {
field: "text".to_string(),
..Default::default()
}),
sub_aggregation: Default::default(),
})),
),
]
.into_iter()
.collect();
let agg_req: Aggregations = if use_elastic_json_req {
let elasticsearch_compatible_json_req = r#"
let elasticsearch_compatible_json_req = r#"
{
"rangef64": {
"range": {
@@ -383,61 +412,7 @@ fn test_aggregation_level2(
}
}
"#;
let value: Aggregations = serde_json::from_str(elasticsearch_compatible_json_req).unwrap();
value
} else {
let agg_req: Aggregations = vec![
("average".to_string(), get_avg_req("score")),
(
"range".to_string(),
Aggregation::Bucket(Box::new(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "score".to_string(),
ranges: vec![
(3f64..7f64).into(),
(7f64..19f64).into(),
(19f64..20f64).into(),
],
..Default::default()
}),
sub_aggregation: sub_agg_req.clone(),
})),
),
(
"rangef64".to_string(),
Aggregation::Bucket(Box::new(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "score_f64".to_string(),
ranges: vec![
(3f64..7f64).into(),
(7f64..19f64).into(),
(19f64..20f64).into(),
],
..Default::default()
}),
sub_aggregation: sub_agg_req.clone(),
})),
),
(
"rangei64".to_string(),
Aggregation::Bucket(Box::new(BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "score_i64".to_string(),
ranges: vec![
(3f64..7f64).into(),
(7f64..19f64).into(),
(19f64..20f64).into(),
],
..Default::default()
}),
sub_aggregation: sub_agg_req,
})),
),
]
.into_iter()
.collect();
agg_req
};
let agg_req: Aggregations = serde_json::from_str(elasticsearch_compatible_json_req).unwrap();
let agg_res: AggregationResults = if use_distributed_collector {
let collector =
@@ -445,10 +420,7 @@ fn test_aggregation_level2(
let searcher = reader.searcher();
let res = searcher.search(&term_query, &collector).unwrap();
// Test de/serialization roundtrip on intermediate_agg_result
let res: IntermediateAggregationResults =
serde_json::from_str(&serde_json::to_string(&res).unwrap()).unwrap();
res.into_final_bucket_result(agg_req.clone(), &Default::default())
res.into_final_result(agg_req.clone(), &Default::default())
.unwrap()
} else {
let collector = get_collector(agg_req.clone());
@@ -518,42 +490,22 @@ fn test_aggregation_level2(
#[test]
fn test_aggregation_level2_multi_segments() -> crate::Result<()> {
test_aggregation_level2(false, false, false)
test_aggregation_level2(false, false)
}
#[test]
fn test_aggregation_level2_single_segment() -> crate::Result<()> {
test_aggregation_level2(true, false, false)
test_aggregation_level2(true, false)
}
#[test]
fn test_aggregation_level2_multi_segments_distributed_collector() -> crate::Result<()> {
test_aggregation_level2(false, true, false)
test_aggregation_level2(false, true)
}
#[test]
fn test_aggregation_level2_single_segment_distributed_collector() -> crate::Result<()> {
test_aggregation_level2(true, true, false)
}
#[test]
fn test_aggregation_level2_multi_segments_use_json() -> crate::Result<()> {
test_aggregation_level2(false, false, true)
}
#[test]
fn test_aggregation_level2_single_segment_use_json() -> crate::Result<()> {
test_aggregation_level2(true, false, true)
}
#[test]
fn test_aggregation_level2_multi_segments_distributed_collector_use_json() -> crate::Result<()> {
test_aggregation_level2(false, true, true)
}
#[test]
fn test_aggregation_level2_single_segment_distributed_collector_use_json() -> crate::Result<()> {
test_aggregation_level2(true, true, true)
test_aggregation_level2(true, true)
}
#[test]
@@ -563,14 +515,14 @@ fn test_aggregation_invalid_requests() -> crate::Result<()> {
let reader = index.reader()?;
let avg_on_field = |field_name: &str| {
let agg_req_1: Aggregations = vec![(
"average".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name(field_name.to_string()),
)),
)]
.into_iter()
.collect();
let agg_req_1: Aggregations = serde_json::from_value(json!({
"average": {
"avg": {
"field": field_name,
},
}
}))
.unwrap();
let collector = get_collector(agg_req_1);
@@ -581,10 +533,36 @@ fn test_aggregation_invalid_requests() -> crate::Result<()> {
let agg_res = avg_on_field("dummy_text").unwrap_err();
assert_eq!(
format!("{:?}", agg_res),
format!("{agg_res:?}"),
r#"InvalidArgument("Field \"dummy_text\" is not configured as fast field")"#
);
let agg_req_1: Result<Aggregations, serde_json::Error> = serde_json::from_value(json!({
"average": {
"avg": {
"fieldd": "a",
},
}
}));
assert_eq!(agg_req_1.is_err(), true);
assert_eq!(agg_req_1.unwrap_err().to_string(), "missing field `field`");
let agg_req_1: Result<Aggregations, serde_json::Error> = serde_json::from_value(json!({
"average": {
"doesnotmatchanyagg": {
"field": "a",
},
}
}));
assert_eq!(agg_req_1.is_err(), true);
// TODO: This should list valid values
assert!(agg_req_1
.unwrap_err()
.to_string()
.contains("unknown variant `doesnotmatchanyagg`, expected one of"));
// TODO: This should return an error
// let agg_res = avg_on_field("not_exist_field").unwrap_err();
// assert_eq!(
@@ -618,18 +596,16 @@ fn test_aggregation_on_json_object() {
index_writer.commit().unwrap();
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let agg: Aggregations = vec![(
"jsonagg".to_string(),
Aggregation::Bucket(Box::new(BucketAggregation {
bucket_agg: BucketAggregationType::Terms(TermsAggregation {
field: "json.color".to_string(),
..Default::default()
}),
sub_aggregation: Default::default(),
})),
)]
.into_iter()
.collect();
let agg: Aggregations = serde_json::from_value(json!({
"jsonagg": {
"terms": {
"field": "json.color",
}
}
}))
.unwrap();
let aggregation_collector = get_collector(agg);
let aggregation_results = searcher.search(&AllQuery, &aggregation_collector).unwrap();
let aggregation_res_json = serde_json::to_value(aggregation_results).unwrap();
@@ -687,18 +663,15 @@ fn test_aggregation_on_json_object_empty_columns() {
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let agg: Aggregations = vec![(
"jsonagg".to_string(),
Aggregation::Bucket(Box::new(BucketAggregation {
bucket_agg: BucketAggregationType::Terms(TermsAggregation {
field: "json.color".to_string(),
..Default::default()
}),
sub_aggregation: Default::default(),
})),
)]
.into_iter()
.collect();
let agg: Aggregations = serde_json::from_value(json!({
"jsonagg": {
"terms": {
"field": "json.color",
}
}
}))
.unwrap();
let aggregation_collector = get_collector(agg);
let aggregation_results = searcher.search(&AllQuery, &aggregation_collector).unwrap();
@@ -853,10 +826,9 @@ fn test_aggregation_on_json_object_mixed_types() {
"buckets": [
{ "doc_count": 1, "key": 10.0, "min_price": { "value": 10.0 } },
{ "doc_count": 1, "key": -20.5, "min_price": { "value": -20.5 } },
// TODO red is missing since there is no multi aggregation within one
// segment for multiple types
// TODO bool is also not yet handled in aggregation
{ "doc_count": 1, "key": "blue", "min_price": { "value": null } }
{ "doc_count": 1, "key": "blue", "min_price": { "value": null } },
{ "doc_count": 1, "key": "red", "min_price": { "value": null } },
],
"sum_other_doc_count": 0
}

528
src/aggregation/bucket/histogram/date_histogram.rs
View File

@@ -37,10 +37,10 @@ pub struct DateHistogramAggregationReq {
interval: Option<String>,
#[doc(hidden)]
/// Only for validation
date_interval: Option<String>,
calendar_interval: Option<String>,
/// The field to aggregate on.
pub field: String,
/// The format to format dates.
/// The format to format dates. Unsupported currently.
pub format: Option<String>,
/// The interval to chunk your data range. Each bucket spans a value range of
/// [0..fixed_interval). Accepted values
@@ -67,6 +67,13 @@ pub struct DateHistogramAggregationReq {
pub fixed_interval: Option<String>,
/// Intervals implicitly defines an absolute grid of buckets `[interval * k, interval * (k +
/// 1))`.
///
/// Offset makes it possible to shift this grid into
/// `[offset + interval * k, offset + interval * (k + 1))`. Offset has to be in the range [0,
/// interval).
///
/// The `offset` parameter is has the same syntax as the `fixed_interval` parameter, but
/// also allows for negative values.
pub offset: Option<String>,
/// The minimum number of documents in a bucket to be returned. Defaults to 0.
pub min_doc_count: Option<u64>,
@@ -77,7 +84,7 @@ pub struct DateHistogramAggregationReq {
/// hard_bounds only limits the buckets, to force a range set both extended_bounds and
/// hard_bounds to the same range.
///
/// Needs to be provided as timestamp in microseconds precision.
/// Needs to be provided as timestamp in millisecond precision.
///
/// ## Example
/// ```json
@@ -88,7 +95,7 @@ pub struct DateHistogramAggregationReq {
/// "interval": "1d",
/// "hard_bounds": {
/// "min": 0,
/// "max": 1420502400000000
/// "max": 1420502400000
/// }
/// }
/// }
@@ -114,16 +121,16 @@ impl DateHistogramAggregationReq {
self.validate()?;
Ok(HistogramAggregation {
field: self.field.to_string(),
interval: parse_into_microseconds(self.fixed_interval.as_ref().unwrap())? as f64,
interval: parse_into_milliseconds(self.fixed_interval.as_ref().unwrap())? as f64,
offset: self
.offset
.as_ref()
.map(|offset| parse_offset_into_microseconds(offset))
.map(|offset| parse_offset_into_milliseconds(offset))
.transpose()?
.map(|el| el as f64),
min_doc_count: self.min_doc_count,
hard_bounds: None,
extended_bounds: None,
hard_bounds: self.hard_bounds,
extended_bounds: self.extended_bounds,
keyed: self.keyed,
})
}
@@ -131,16 +138,14 @@ impl DateHistogramAggregationReq {
fn validate(&self) -> crate::Result<()> {
if let Some(interval) = self.interval.as_ref() {
return Err(crate::TantivyError::InvalidArgument(format!(
"`interval` parameter {:?} in date histogram is unsupported, only \
`fixed_interval` is supported",
interval
"`interval` parameter {interval:?} in date histogram is unsupported, only \
`fixed_interval` is supported"
)));
}
if let Some(interval) = self.date_interval.as_ref() {
if let Some(interval) = self.calendar_interval.as_ref() {
return Err(crate::TantivyError::InvalidArgument(format!(
"`date_interval` parameter {:?} in date histogram is unsupported, only \
`fixed_interval` is supported",
interval
"`calendar_interval` parameter {interval:?} in date histogram is unsupported, \
only `fixed_interval` is supported"
)));
}
if self.format.is_some() {
@@ -155,7 +160,7 @@ impl DateHistogramAggregationReq {
));
}
parse_into_microseconds(self.fixed_interval.as_ref().unwrap())?;
parse_into_milliseconds(self.fixed_interval.as_ref().unwrap())?;
Ok(())
}
@@ -176,9 +181,12 @@ pub enum DateHistogramParseError {
/// Offset invalid
#[error("passed offset is invalid {0:?}")]
InvalidOffset(String),
/// Value out of bounds
#[error("passed value is out of bounds: {0:?}")]
OutOfBounds(String),
}
fn parse_offset_into_microseconds(input: &str) -> Result<i64, AggregationError> {
fn parse_offset_into_milliseconds(input: &str) -> Result<i64, AggregationError> {
let is_sign = |byte| &[byte] == b"-" || &[byte] == b"+";
if input.is_empty() {
return Err(DateHistogramParseError::InvalidOffset(input.to_string()).into());
@@ -187,18 +195,18 @@ fn parse_offset_into_microseconds(input: &str) -> Result<i64, AggregationError>
let has_sign = is_sign(input.as_bytes()[0]);
if has_sign {
let (sign, input) = input.split_at(1);
let val = parse_into_microseconds(input)?;
let val = parse_into_milliseconds(input)?;
if sign == "-" {
Ok(-val)
} else {
Ok(val)
}
} else {
parse_into_microseconds(input)
parse_into_milliseconds(input)
}
}
fn parse_into_microseconds(input: &str) -> Result<i64, AggregationError> {
fn parse_into_milliseconds(input: &str) -> Result<i64, AggregationError> {
let split_boundary = input
.as_bytes()
.iter()
@@ -217,16 +225,21 @@ fn parse_into_microseconds(input: &str) -> Result<i64, AggregationError> {
// here and being defensive does not hurt.
.map_err(|_err| DateHistogramParseError::NumberMissing(input.to_string()))?;
let multiplier_from_unit = match unit {
"ms" => 1,
"s" => 1000,
"m" => 60 * 1000,
"h" => 60 * 60 * 1000,
"d" => 24 * 60 * 60 * 1000,
let unit_in_ms = match unit {
"ms" | "milliseconds" => 1,
"s" | "seconds" => 1000,
"m" | "minutes" => 60 * 1000,
"h" | "hours" => 60 * 60 * 1000,
"d" | "days" => 24 * 60 * 60 * 1000,
_ => return Err(DateHistogramParseError::UnitNotRecognized(unit.to_string()).into()),
};
Ok(number * multiplier_from_unit * 1000)
let val = number * unit_in_ms;
// The field type is in nanoseconds precision, so validate the value to fit the range
val.checked_mul(1_000_000)
.ok_or_else(|| DateHistogramParseError::OutOfBounds(input.to_string()))?;
Ok(val)
}
#[cfg(test)]
@@ -241,49 +254,50 @@ mod tests {
use crate::Index;
#[test]
fn test_parse_into_microseconds() {
assert_eq!(parse_into_microseconds("1m").unwrap(), 60_000_000);
assert_eq!(parse_into_microseconds("2m").unwrap(), 120_000_000);
fn test_parse_into_millisecs() {
assert_eq!(parse_into_milliseconds("1m").unwrap(), 60_000);
assert_eq!(parse_into_milliseconds("2m").unwrap(), 120_000);
assert_eq!(parse_into_milliseconds("2minutes").unwrap(), 120_000);
assert_eq!(
parse_into_microseconds("2y").unwrap_err(),
parse_into_milliseconds("2y").unwrap_err(),
DateHistogramParseError::UnitNotRecognized("y".to_string()).into()
);
assert_eq!(
parse_into_microseconds("2000").unwrap_err(),
parse_into_milliseconds("2000").unwrap_err(),
DateHistogramParseError::UnitMissing("2000".to_string()).into()
);
assert_eq!(
parse_into_microseconds("ms").unwrap_err(),
parse_into_milliseconds("ms").unwrap_err(),
DateHistogramParseError::NumberMissing("ms".to_string()).into()
);
}
#[test]
fn test_parse_offset_into_microseconds() {
assert_eq!(parse_offset_into_microseconds("1m").unwrap(), 60_000_000);
assert_eq!(parse_offset_into_microseconds("+1m").unwrap(), 60_000_000);
assert_eq!(parse_offset_into_microseconds("-1m").unwrap(), -60_000_000);
assert_eq!(parse_offset_into_microseconds("2m").unwrap(), 120_000_000);
assert_eq!(parse_offset_into_microseconds("+2m").unwrap(), 120_000_000);
assert_eq!(parse_offset_into_microseconds("-2m").unwrap(), -120_000_000);
assert_eq!(parse_offset_into_microseconds("-2ms").unwrap(), -2_000);
fn test_parse_offset_into_milliseconds() {
assert_eq!(parse_offset_into_milliseconds("1m").unwrap(), 60_000);
assert_eq!(parse_offset_into_milliseconds("+1m").unwrap(), 60_000);
assert_eq!(parse_offset_into_milliseconds("-1m").unwrap(), -60_000);
assert_eq!(parse_offset_into_milliseconds("2m").unwrap(), 120_000);
assert_eq!(parse_offset_into_milliseconds("+2m").unwrap(), 120_000);
assert_eq!(parse_offset_into_milliseconds("-2m").unwrap(), -120_000);
assert_eq!(parse_offset_into_milliseconds("-2ms").unwrap(), -2);
assert_eq!(
parse_offset_into_microseconds("2y").unwrap_err(),
parse_offset_into_milliseconds("2y").unwrap_err(),
DateHistogramParseError::UnitNotRecognized("y".to_string()).into()
);
assert_eq!(
parse_offset_into_microseconds("2000").unwrap_err(),
parse_offset_into_milliseconds("2000").unwrap_err(),
DateHistogramParseError::UnitMissing("2000".to_string()).into()
);
assert_eq!(
parse_offset_into_microseconds("ms").unwrap_err(),
parse_offset_into_milliseconds("ms").unwrap_err(),
DateHistogramParseError::NumberMissing("ms".to_string()).into()
);
}
#[test]
fn test_parse_into_milliseconds_do_not_accept_non_ascii() {
assert!(parse_into_microseconds("１m").is_err());
assert!(parse_into_milliseconds("１m").is_err());
}
pub fn get_test_index_from_docs(
@@ -322,168 +336,316 @@ mod tests {
}
#[test]
fn histogram_test_date_force_merge_segments() -> crate::Result<()> {
fn histogram_test_date_force_merge_segments() {
histogram_test_date_merge_segments(true)
}
#[test]
fn histogram_test_date() -> crate::Result<()> {
fn histogram_test_date() {
histogram_test_date_merge_segments(false)
}
fn histogram_test_date_merge_segments(merge_segments: bool) -> crate::Result<()> {
fn histogram_test_date_merge_segments(merge_segments: bool) {
let docs = vec![
vec![r#"{ "date": "2015-01-01T12:10:30Z", "text": "aaa" }"#],
vec![r#"{ "date": "2015-01-01T11:11:30Z", "text": "bbb" }"#],
vec![r#"{ "date": "2015-01-02T00:00:00Z", "text": "bbb" }"#],
vec![r#"{ "date": "2015-01-06T00:00:00Z", "text": "ccc" }"#],
];
let index = get_test_index_from_docs(merge_segments, &docs).unwrap();
let index = get_test_index_from_docs(merge_segments, &docs)?;
// 30day + offset
let elasticsearch_compatible_json = json!(
{
"sales_over_time": {
"date_histogram": {
"field": "date",
"fixed_interval": "30d",
"offset": "-4d"
}
}
}
);
let agg_req: Aggregations =
serde_json::from_str(&serde_json::to_string(&elasticsearch_compatible_json).unwrap())
.unwrap();
let res = exec_request(agg_req, &index)?;
let expected_res = json!({
"sales_over_time" : {
"buckets" : [
{
"key_as_string" : "2015-01-01T00:00:00Z",
"key" : 1420070400000000.0,
"doc_count" : 4
}
]
}
});
assert_eq!(res, expected_res);
// 30day + offset + sub_agg
let elasticsearch_compatible_json = json!(
{
"sales_over_time": {
"date_histogram": {
"field": "date",
"fixed_interval": "30d",
"offset": "-4d"
},
"aggs": {
"texts": {
"terms": {"field": "text"}
{
// 30day + offset
let elasticsearch_compatible_json = json!(
{
"sales_over_time": {
"date_histogram": {
"field": "date",
"fixed_interval": "30d",
"offset": "-4d"
}
}
}
}
);
let agg_req: Aggregations =
serde_json::from_str(&serde_json::to_string(&elasticsearch_compatible_json).unwrap())
.unwrap();
let res = exec_request(agg_req, &index)?;
println!("{}", serde_json::to_string_pretty(&res).unwrap());
let expected_res = json!({
"sales_over_time" : {
"buckets" : [
{
"key_as_string" : "2015-01-01T00:00:00Z",
"key" : 1420070400000000.0,
"doc_count" : 4,
"texts": {
"buckets": [
{
"doc_count": 2,
"key": "bbb"
},
{
"doc_count": 1,
"key": "ccc"
},
{
"doc_count": 1,
"key": "aaa"
}
],
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0
}
}
]
);
let agg_req: Aggregations = serde_json::from_str(
&serde_json::to_string(&elasticsearch_compatible_json).unwrap(),
)
.unwrap();
let res = exec_request(agg_req, &index).unwrap();
let expected_res = json!({
"sales_over_time" : {
"buckets" : [
{
"key_as_string" : "2015-01-01T00:00:00Z",
"key" : 1420070400000.0,
"doc_count" : 4
}
]
}
});
assert_eq!(res, expected_res);
}
});
assert_eq!(res, expected_res);
// 1day
let elasticsearch_compatible_json = json!(
{
"sales_over_time": {
"date_histogram": {
"field": "date",
"fixed_interval": "1d"
{
// 30day + offset + sub_agg
let elasticsearch_compatible_json = json!(
{
"sales_over_time": {
"date_histogram": {
"field": "date",
"fixed_interval": "30d",
"offset": "-4d"
},
"aggs": {
"texts": {
"terms": {"field": "text"}
}
}
}
}
}
}
);
);
let agg_req: Aggregations =
serde_json::from_str(&serde_json::to_string(&elasticsearch_compatible_json).unwrap())
.unwrap();
let res = exec_request(agg_req, &index)?;
let expected_res = json!( {
"sales_over_time": {
"buckets": [
{
"doc_count": 2,
"key": 1420070400000000.0,
"key_as_string": "2015-01-01T00:00:00Z"
},
{
"doc_count": 1,
"key": 1420156800000000.0,
"key_as_string": "2015-01-02T00:00:00Z"
},
{
"doc_count": 0,
"key": 1420243200000000.0,
"key_as_string": "2015-01-03T00:00:00Z"
},
{
"doc_count": 0,
"key": 1420329600000000.0,
"key_as_string": "2015-01-04T00:00:00Z"
},
{
"doc_count": 0,
"key": 1420416000000000.0,
"key_as_string": "2015-01-05T00:00:00Z"
},
{
"doc_count": 1,
"key": 1420502400000000.0,
"key_as_string": "2015-01-06T00:00:00Z"
}
]
}
});
assert_eq!(res, expected_res);
let agg_req: Aggregations = serde_json::from_str(
&serde_json::to_string(&elasticsearch_compatible_json).unwrap(),
)
.unwrap();
let res = exec_request(agg_req, &index).unwrap();
let expected_res = json!({
"sales_over_time" : {
"buckets" : [
{
"key_as_string" : "2015-01-01T00:00:00Z",
"key" : 1420070400000.0,
"doc_count" : 4,
"texts": {
"buckets": [
{
"doc_count": 2,
"key": "bbb"
},
{
"doc_count": 1,
"key": "ccc"
},
{
"doc_count": 1,
"key": "aaa"
}
],
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0
}
}
]
}
});
assert_eq!(res, expected_res);
}
{
// 1day
let elasticsearch_compatible_json = json!(
{
"sales_over_time": {
"date_histogram": {
"field": "date",
"fixed_interval": "1d"
}
}
}
);
Ok(())
let agg_req: Aggregations = serde_json::from_str(
&serde_json::to_string(&elasticsearch_compatible_json).unwrap(),
)
.unwrap();
let res = exec_request(agg_req, &index).unwrap();
let expected_res = json!( {
"sales_over_time": {
"buckets": [
{
"doc_count": 2,
"key": 1420070400000.0,
"key_as_string": "2015-01-01T00:00:00Z"
},
{
"doc_count": 1,
"key": 1420156800000.0,
"key_as_string": "2015-01-02T00:00:00Z"
},
{
"doc_count": 0,
"key": 1420243200000.0,
"key_as_string": "2015-01-03T00:00:00Z"
},
{
"doc_count": 0,
"key": 1420329600000.0,
"key_as_string": "2015-01-04T00:00:00Z"
},
{
"doc_count": 0,
"key": 1420416000000.0,
"key_as_string": "2015-01-05T00:00:00Z"
},
{
"doc_count": 1,
"key": 1420502400000.0,
"key_as_string": "2015-01-06T00:00:00Z"
}
]
}
});
assert_eq!(res, expected_res);
}
{
// 1day + extended_bounds
let elasticsearch_compatible_json = json!(
{
"sales_over_time": {
"date_histogram": {
"field": "date",
"fixed_interval": "1d",
"extended_bounds": {
"min": 1419984000000.0,
"max": 1420588800000.0
}
}
}
}
);
let agg_req: Aggregations = serde_json::from_str(
&serde_json::to_string(&elasticsearch_compatible_json).unwrap(),
)
.unwrap();
let res = exec_request(agg_req, &index).unwrap();
let expected_res = json!({
"sales_over_time" : {
"buckets": [
{
"doc_count": 0,
"key": 1419984000000.0,
"key_as_string": "2014-12-31T00:00:00Z"
},
{
"doc_count": 2,
"key": 1420070400000.0,
"key_as_string": "2015-01-01T00:00:00Z"
},
{
"doc_count": 1,
"key": 1420156800000.0,
"key_as_string": "2015-01-02T00:00:00Z"
},
{
"doc_count": 0,
"key": 1420243200000.0,
"key_as_string": "2015-01-03T00:00:00Z"
},
{
"doc_count": 0,
"key": 1420329600000.0,
"key_as_string": "2015-01-04T00:00:00Z"
},
{
"doc_count": 0,
"key": 1420416000000.0,
"key_as_string": "2015-01-05T00:00:00Z"
},
{
"doc_count": 1,
"key": 1420502400000.0,
"key_as_string": "2015-01-06T00:00:00Z"
},
{
"doc_count": 0,
"key": 1420588800000.0,
"key_as_string": "2015-01-07T00:00:00Z"
}
]
}
});
assert_eq!(res, expected_res);
}
{
// 1day + hard_bounds + extended_bounds
let elasticsearch_compatible_json = json!(
{
"sales_over_time": {
"date_histogram": {
"field": "date",
"fixed_interval": "1d",
"hard_bounds": {
"min": 1420156800000.0,
"max": 1420243200000.0
}
}
}
}
);
let agg_req: Aggregations = serde_json::from_str(
&serde_json::to_string(&elasticsearch_compatible_json).unwrap(),
)
.unwrap();
let res = exec_request(agg_req, &index).unwrap();
let expected_res = json!({
"sales_over_time" : {
"buckets": [
{
"doc_count": 1,
"key": 1420156800000.0,
"key_as_string": "2015-01-02T00:00:00Z"
}
]
}
});
assert_eq!(res, expected_res);
}
{
// 1day + hard_bounds as Rfc3339
let elasticsearch_compatible_json = json!(
{
"sales_over_time": {
"date_histogram": {
"field": "date",
"fixed_interval": "1d",
"hard_bounds": {
"min": "2015-01-02T00:00:00Z",
"max": "2015-01-02T12:00:00Z"
}
}
}
}
);
let agg_req: Aggregations = serde_json::from_str(
&serde_json::to_string(&elasticsearch_compatible_json).unwrap(),
)
.unwrap();
let res = exec_request(agg_req, &index).unwrap();
let expected_res = json!({
"sales_over_time" : {
"buckets": [
{
"doc_count": 1,
"key": 1420156800000.0,
"key_as_string": "2015-01-02T00:00:00Z"
}
]
}
});
assert_eq!(res, expected_res);
}
}
#[test]
fn histogram_test_invalid_req() -> crate::Result<()> {
fn histogram_test_invalid_req() {
let docs = vec![];
let index = get_test_index_from_docs(false, &docs)?;
let index = get_test_index_from_docs(false, &docs).unwrap();
let elasticsearch_compatible_json = json!(
{
"sales_over_time": {
@@ -504,7 +666,5 @@ mod tests {
err.to_string(),
r#"An invalid argument was passed: '`interval` parameter "30d" in date histogram is unsupported, only `fixed_interval` is supported'"#
);
Ok(())
}
}

834
src/aggregation/bucket/histogram/histogram.rs
View File

File diff suppressed because it is too large Load Diff

23
src/aggregation/bucket/mod.rs
View File

@@ -1,24 +1,39 @@
//! Module for all bucket aggregations.
//!
//! BucketAggregations create buckets of documents
//! [`BucketAggregation`](super::agg_req::BucketAggregation).
//! 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.
//!
//! Results of final buckets are [`BucketResult`](super::agg_result::BucketResult).
//! Results of intermediate buckets are
//! [`IntermediateBucketResult`](super::intermediate_agg_result::IntermediateBucketResult)
//!
//! ## Supported Bucket Aggregations
//! - [Histogram](HistogramAggregation)
//! - [DateHistogram](DateHistogramAggregationReq)
//! - [Range](RangeAggregation)
//! - [Terms](TermsAggregation)
mod histogram;
mod range;
mod term_agg;
mod term_missing_agg;
use std::collections::HashMap;
pub(crate) use histogram::SegmentHistogramCollector;
pub use histogram::*;
pub(crate) use range::SegmentRangeCollector;
pub use range::*;
use serde::{de, Deserialize, Deserializer, Serialize, Serializer};
pub use term_agg::*;
pub use term_missing_agg::*;
/// Order for buckets in a bucket aggregation.
#[derive(Clone, Copy, Debug, PartialEq, Serialize, Deserialize, Default)]

267
src/aggregation/bucket/range.rs
View File

@@ -5,16 +5,17 @@ use columnar::{ColumnType, MonotonicallyMappableToU64};
use rustc_hash::FxHashMap;
use serde::{Deserialize, Serialize};
use crate::aggregation::agg_limits::ResourceLimitGuard;
use crate::aggregation::agg_req_with_accessor::AggregationsWithAccessor;
use crate::aggregation::intermediate_agg_result::{
IntermediateAggregationResults, IntermediateBucketResult, IntermediateRangeBucketEntry,
IntermediateRangeBucketResult,
IntermediateAggregationResult, IntermediateAggregationResults, IntermediateBucketResult,
IntermediateRangeBucketEntry, IntermediateRangeBucketResult,
};
use crate::aggregation::segment_agg_result::{
build_segment_agg_collector, AggregationLimits, SegmentAggregationCollector,
build_segment_agg_collector, SegmentAggregationCollector,
};
use crate::aggregation::{
f64_from_fastfield_u64, f64_to_fastfield_u64, format_date, Key, SerializedKey, VecWithNames,
f64_from_fastfield_u64, f64_to_fastfield_u64, format_date, Key, SerializedKey,
};
use crate::TantivyError;
@@ -157,16 +158,18 @@ impl SegmentRangeBucketEntry {
self,
agg_with_accessor: &AggregationsWithAccessor,
) -> crate::Result<IntermediateRangeBucketEntry> {
let sub_aggregation = if let Some(sub_aggregation) = self.sub_aggregation {
sub_aggregation.into_intermediate_aggregations_result(agg_with_accessor)?
let mut sub_aggregation_res = IntermediateAggregationResults::default();
if let Some(sub_aggregation) = self.sub_aggregation {
sub_aggregation
.add_intermediate_aggregation_result(agg_with_accessor, &mut sub_aggregation_res)?
} else {
Default::default()
};
Ok(IntermediateRangeBucketEntry {
key: self.key,
key: self.key.into(),
doc_count: self.doc_count,
sub_aggregation,
sub_aggregation: sub_aggregation_res,
from: self.from,
to: self.to,
})
@@ -174,13 +177,14 @@ impl SegmentRangeBucketEntry {
}
impl SegmentAggregationCollector for SegmentRangeCollector {
fn into_intermediate_aggregations_result(
fn add_intermediate_aggregation_result(
self: Box<Self>,
agg_with_accessor: &AggregationsWithAccessor,
) -> crate::Result<IntermediateAggregationResults> {
results: &mut IntermediateAggregationResults,
) -> crate::Result<()> {
let field_type = self.column_type;
let name = agg_with_accessor.buckets.keys[self.accessor_idx].to_string();
let sub_agg = &agg_with_accessor.buckets.values[self.accessor_idx].sub_aggregation;
let name = agg_with_accessor.aggs.keys[self.accessor_idx].to_string();
let sub_agg = &agg_with_accessor.aggs.values[self.accessor_idx].sub_aggregation;
let buckets: FxHashMap<SerializedKey, IntermediateRangeBucketEntry> = self
.buckets
@@ -200,12 +204,9 @@ impl SegmentAggregationCollector for SegmentRangeCollector {
column_type: Some(self.column_type),
});
let buckets = Some(VecWithNames::from_entries(vec![(name, bucket)]));
results.push(name, IntermediateAggregationResult::Bucket(bucket))?;
Ok(IntermediateAggregationResults {
metrics: None,
buckets,
})
Ok(())
}
#[inline]
@@ -223,7 +224,7 @@ impl SegmentAggregationCollector for SegmentRangeCollector {
docs: &[crate::DocId],
agg_with_accessor: &mut AggregationsWithAccessor,
) -> crate::Result<()> {
let bucket_agg_accessor = &mut agg_with_accessor.buckets.values[self.accessor_idx];
let bucket_agg_accessor = &mut agg_with_accessor.aggs.values[self.accessor_idx];
bucket_agg_accessor
.column_block_accessor
@@ -245,7 +246,7 @@ impl SegmentAggregationCollector for SegmentRangeCollector {
fn flush(&mut self, agg_with_accessor: &mut AggregationsWithAccessor) -> crate::Result<()> {
let sub_aggregation_accessor =
&mut agg_with_accessor.buckets.values[self.accessor_idx].sub_aggregation;
&mut agg_with_accessor.aggs.values[self.accessor_idx].sub_aggregation;
for bucket in self.buckets.iter_mut() {
if let Some(sub_agg) = bucket.bucket.sub_aggregation.as_mut() {
@@ -260,8 +261,8 @@ impl SegmentAggregationCollector for SegmentRangeCollector {
impl SegmentRangeCollector {
pub(crate) fn from_req_and_validate(
req: &RangeAggregation,
sub_aggregation: &AggregationsWithAccessor,
limits: &AggregationLimits,
sub_aggregation: &mut AggregationsWithAccessor,
limits: &ResourceLimitGuard,
field_type: ColumnType,
accessor_idx: usize,
) -> crate::Result<Self> {
@@ -307,8 +308,7 @@ impl SegmentRangeCollector {
limits.add_memory_consumed(
buckets.len() as u64 * std::mem::size_of::<SegmentRangeAndBucketEntry>() as u64,
);
limits.validate_memory_consumption()?;
)?;
Ok(SegmentRangeCollector {
buckets,
@@ -445,14 +445,12 @@ mod tests {
use serde_json::Value;
use super::*;
use crate::aggregation::agg_req::{
Aggregation, Aggregations, BucketAggregation, BucketAggregationType, MetricAggregation,
};
use crate::aggregation::metric::AverageAggregation;
use crate::aggregation::agg_req::Aggregations;
use crate::aggregation::tests::{
exec_request, exec_request_with_query, get_test_index_2_segments,
get_test_index_with_num_docs,
};
use crate::aggregation::AggregationLimits;
pub fn get_collector_from_ranges(
ranges: Vec<RangeAggregationRange>,
@@ -466,8 +464,8 @@ mod tests {
SegmentRangeCollector::from_req_and_validate(
&req,
&Default::default(),
&Default::default(),
&mut Default::default(),
&AggregationLimits::default().new_guard(),
field_type,
0,
)
@@ -478,22 +476,18 @@ mod tests {
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()],
..Default::default()
}),
sub_aggregation: Default::default(),
}
.into(),
),
)]
.into_iter()
.collect();
let agg_req: Aggregations = serde_json::from_value(json!({
"range": {
"range": {
"field": "fraction_f64",
"ranges": [
{"from": 0.0, "to": 0.1},
{"from": 0.1, "to": 0.2},
]
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
@@ -513,31 +507,25 @@ mod tests {
fn range_fraction_test_with_sub_agg() -> crate::Result<()> {
let index = get_test_index_with_num_docs(false, 100)?;
let sub_agg_req: Aggregations = vec![(
"score_f64".to_string(),
Aggregation::Metric(MetricAggregation::Average(
AverageAggregation::from_field_name("score_f64".to_string()),
)),
)]
.into_iter()
.collect();
let sub_agg_req: Aggregations = serde_json::from_value(json!({
"avg": { "avg": { "field": "score_f64", } }
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()],
..Default::default()
}),
sub_aggregation: sub_agg_req,
}
.into(),
),
)]
.into_iter()
.collect();
}))
.unwrap();
let agg_req: Aggregations = serde_json::from_value(json!({
"range": {
"range": {
"field": "fraction_f64",
"ranges": [
{"from": 0.0, "to": 0.1},
{"from": 0.1, "to": 0.2},
]
},
"aggs": sub_agg_req
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
@@ -557,22 +545,19 @@ mod tests {
fn range_keyed_buckets_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()],
keyed: true,
}),
sub_aggregation: Default::default(),
}
.into(),
),
)]
.into_iter()
.collect();
let agg_req: Aggregations = serde_json::from_value(json!({
"range": {
"range": {
"field": "fraction_f64",
"ranges": [
{"from": 0.0, "to": 0.1},
{"from": 0.1, "to": 0.2},
],
"keyed": true
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
@@ -597,33 +582,19 @@ mod tests {
fn range_custom_key_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![
RangeAggregationRange {
key: Some("custom-key-0-to-0.1".to_string()),
from: Some(0f64),
to: Some(0.1f64),
},
RangeAggregationRange {
key: None,
from: Some(0.1f64),
to: Some(0.2f64),
},
],
keyed: false,
}),
sub_aggregation: Default::default(),
}
.into(),
),
)]
.into_iter()
.collect();
let agg_req: Aggregations = serde_json::from_value(json!({
"range": {
"range": {
"field": "fraction_f64",
"ranges": [
{"key": "custom-key-0-to-0.1", "from": 0.0, "to": 0.1},
{"from": 0.1, "to": 0.2},
],
"keyed": false
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
@@ -657,33 +628,19 @@ mod tests {
fn range_date_test_with_opt(merge_segments: bool) -> crate::Result<()> {
let index = get_test_index_2_segments(merge_segments)?;
let agg_req: Aggregations = vec![(
"date_ranges".to_string(),
Aggregation::Bucket(
BucketAggregation {
bucket_agg: BucketAggregationType::Range(RangeAggregation {
field: "date".to_string(),
ranges: vec![
RangeAggregationRange {
key: None,
from: None,
to: Some(1546300800000000.0f64),
},
RangeAggregationRange {
key: None,
from: Some(1546300800000000.0f64),
to: Some(1546387200000000.0f64),
},
],
keyed: false,
}),
sub_aggregation: Default::default(),
}
.into(),
),
)]
.into_iter()
.collect();
let agg_req: Aggregations = serde_json::from_value(json!({
"date_ranges": {
"range": {
"field": "date",
"ranges": [
{"to": 1546300800000000000i64},
{"from": 1546300800000000000i64, "to": 1546387200000000000i64},
],
"keyed": false
},
}
}))
.unwrap();
let agg_res = exec_request(agg_req, &index)?;
@@ -722,26 +679,18 @@ mod tests {
fn range_custom_key_keyed_buckets_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![RangeAggregationRange {
key: Some("custom-key-0-to-0.1".to_string()),
from: Some(0f64),
to: Some(0.1f64),
}],
keyed: true,
}),
sub_aggregation: Default::default(),
}
.into(),
),
)]
.into_iter()
.collect();
let agg_req: Aggregations = serde_json::from_value(json!({
"range": {
"range": {
"field": "fraction_f64",
"ranges": [
{"key": "custom-key-0-to-0.1", "from": 0.0, "to": 0.1},
],
"keyed": true
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;

1471
src/aggregation/bucket/term_agg.rs
View File

File diff suppressed because it is too large Load Diff

476
src/aggregation/bucket/term_missing_agg.rs Normal file
View File

@@ -0,0 +1,476 @@
use rustc_hash::FxHashMap;
use crate::aggregation::agg_req_with_accessor::AggregationsWithAccessor;
use crate::aggregation::intermediate_agg_result::{
IntermediateAggregationResult, IntermediateAggregationResults, IntermediateBucketResult,
IntermediateKey, IntermediateTermBucketEntry, IntermediateTermBucketResult,
};
use crate::aggregation::segment_agg_result::{
build_segment_agg_collector, SegmentAggregationCollector,
};
/// The specialized missing term aggregation.
#[derive(Default, Debug, Clone)]
pub struct TermMissingAgg {
missing_count: u32,
accessor_idx: usize,
sub_agg: Option<Box<dyn SegmentAggregationCollector>>,
}
impl TermMissingAgg {
pub(crate) fn new(
accessor_idx: usize,
sub_aggregations: &mut AggregationsWithAccessor,
) -> crate::Result<Self> {
let has_sub_aggregations = !sub_aggregations.is_empty();
let sub_agg = if has_sub_aggregations {
let sub_aggregation = build_segment_agg_collector(sub_aggregations)?;
Some(sub_aggregation)
} else {
None
};
Ok(Self {
accessor_idx,
sub_agg,
..Default::default()
})
}
}
impl SegmentAggregationCollector for TermMissingAgg {
fn add_intermediate_aggregation_result(
self: Box<Self>,
agg_with_accessor: &AggregationsWithAccessor,
results: &mut IntermediateAggregationResults,
) -> crate::Result<()> {
let name = agg_with_accessor.aggs.keys[self.accessor_idx].to_string();
let agg_with_accessor = &agg_with_accessor.aggs.values[self.accessor_idx];
let term_agg = agg_with_accessor
.agg
.agg
.as_term()
.expect("TermMissingAgg collector must be term agg req");
let missing = term_agg
.missing
.as_ref()
.expect("TermMissingAgg collector, but no missing found in agg req")
.clone();
let mut entries: FxHashMap<IntermediateKey, IntermediateTermBucketEntry> =
Default::default();
let mut missing_entry = IntermediateTermBucketEntry {
doc_count: self.missing_count,
sub_aggregation: Default::default(),
};
if let Some(sub_agg) = self.sub_agg {
let mut res = IntermediateAggregationResults::default();
sub_agg.add_intermediate_aggregation_result(
&agg_with_accessor.sub_aggregation,
&mut res,
)?;
missing_entry.sub_aggregation = res;
}
entries.insert(missing.into(), missing_entry);
let bucket = IntermediateBucketResult::Terms(IntermediateTermBucketResult {
entries,
sum_other_doc_count: 0,
doc_count_error_upper_bound: 0,
});
results.push(name, IntermediateAggregationResult::Bucket(bucket))?;
Ok(())
}
fn collect(
&mut self,
doc: crate::DocId,
agg_with_accessor: &mut AggregationsWithAccessor,
) -> crate::Result<()> {
let agg = &mut agg_with_accessor.aggs.values[self.accessor_idx];
let has_value = agg.accessors.iter().any(|acc| acc.index.has_value(doc));
if !has_value {
self.missing_count += 1;
if let Some(sub_agg) = self.sub_agg.as_mut() {
sub_agg.collect(doc, &mut agg.sub_aggregation)?;
}
}
Ok(())
}
fn collect_block(
&mut self,
docs: &[crate::DocId],
agg_with_accessor: &mut AggregationsWithAccessor,
) -> crate::Result<()> {
for doc in docs {
self.collect(*doc, agg_with_accessor)?;
}
Ok(())
}
}
#[cfg(test)]
mod tests {
use crate::aggregation::agg_req::Aggregations;
use crate::aggregation::tests::exec_request_with_query;
use crate::schema::{Schema, FAST};
use crate::Index;
#[test]
fn terms_aggregation_missing_mixed_type_mult_seg_sub_agg() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let json = schema_builder.add_json_field("json", FAST);
let score = schema_builder.add_f64_field("score", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
// => Segment with all values numeric
index_writer
.add_document(doc!(score => 1.0, json => json!({"mixed_type": 10.0})))
.unwrap();
index_writer.add_document(doc!(score => 5.0))?;
// index_writer.commit().unwrap();
//// => Segment with all values text
index_writer
.add_document(doc!(score => 1.0, json => json!({"mixed_type": "blue"})))
.unwrap();
index_writer.add_document(doc!(score => 5.0))?;
// index_writer.commit().unwrap();
// => Segment with mixed values
index_writer.add_document(doc!(json => json!({"mixed_type": "red"})))?;
index_writer.add_document(doc!(json => json!({"mixed_type": -20.5})))?;
index_writer.add_document(doc!(json => json!({"mixed_type": true})))?;
index_writer.add_document(doc!(score => 5.0))?;
index_writer.commit().unwrap();
let agg_req: Aggregations = serde_json::from_value(json!({
"replace_null": {
"terms": {
"field": "json.mixed_type",
"missing": "NULL"
},
"aggs": {
"sum_score": {
"sum": {
"field": "score"
}
}
}
},
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// text field
assert_eq!(res["replace_null"]["buckets"][0]["key"], "NULL");
assert_eq!(res["replace_null"]["buckets"][0]["doc_count"], 3);
assert_eq!(
res["replace_null"]["buckets"][0]["sum_score"]["value"],
15.0
);
assert_eq!(res["replace_null"]["sum_other_doc_count"], 0);
assert_eq!(res["replace_null"]["doc_count_error_upper_bound"], 0);
Ok(())
}
#[test]
fn terms_aggregation_missing_mixed_type_sub_agg_reg1() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let json = schema_builder.add_json_field("json", FAST);
let score = schema_builder.add_f64_field("score", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
// => Segment with all values numeric
index_writer.add_document(doc!(score => 1.0, json => json!({"mixed_type": 10.0})))?;
index_writer.add_document(doc!(score => 5.0))?;
index_writer.add_document(doc!(score => 5.0))?;
index_writer.commit().unwrap();
let agg_req: Aggregations = serde_json::from_value(json!({
"replace_null": {
"terms": {
"field": "json.mixed_type",
"missing": "NULL"
},
"aggs": {
"sum_score": {
"sum": {
"field": "score"
}
}
}
},
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// text field
assert_eq!(res["replace_null"]["buckets"][0]["key"], "NULL");
assert_eq!(res["replace_null"]["buckets"][0]["doc_count"], 2);
assert_eq!(
res["replace_null"]["buckets"][0]["sum_score"]["value"],
10.0
);
assert_eq!(res["replace_null"]["sum_other_doc_count"], 0);
assert_eq!(res["replace_null"]["doc_count_error_upper_bound"], 0);
Ok(())
}
#[test]
fn terms_aggregation_missing_mult_seg_empty() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let score = schema_builder.add_f64_field("score", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
index_writer.add_document(doc!(score => 5.0))?;
index_writer.commit().unwrap();
index_writer.add_document(doc!(score => 5.0))?;
index_writer.commit().unwrap();
index_writer.add_document(doc!(score => 5.0))?;
index_writer.commit().unwrap();
let agg_req: Aggregations = serde_json::from_value(json!({
"replace_null": {
"terms": {
"field": "json.mixed_type",
"missing": "NULL"
},
"aggs": {
"sum_score": {
"sum": {
"field": "score"
}
}
}
},
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// text field
assert_eq!(res["replace_null"]["buckets"][0]["key"], "NULL");
assert_eq!(res["replace_null"]["buckets"][0]["doc_count"], 3);
assert_eq!(
res["replace_null"]["buckets"][0]["sum_score"]["value"],
15.0
);
assert_eq!(res["replace_null"]["sum_other_doc_count"], 0);
assert_eq!(res["replace_null"]["doc_count_error_upper_bound"], 0);
Ok(())
}
#[test]
fn terms_aggregation_missing_single_seg_empty() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let score = schema_builder.add_f64_field("score", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
index_writer.add_document(doc!(score => 5.0))?;
index_writer.add_document(doc!(score => 5.0))?;
index_writer.add_document(doc!(score => 5.0))?;
index_writer.commit().unwrap();
let agg_req: Aggregations = serde_json::from_value(json!({
"replace_null": {
"terms": {
"field": "json.mixed_type",
"missing": "NULL"
},
"aggs": {
"sum_score": {
"sum": {
"field": "score"
}
}
}
},
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// text field
assert_eq!(res["replace_null"]["buckets"][0]["key"], "NULL");
assert_eq!(res["replace_null"]["buckets"][0]["doc_count"], 3);
assert_eq!(
res["replace_null"]["buckets"][0]["sum_score"]["value"],
15.0
);
assert_eq!(res["replace_null"]["sum_other_doc_count"], 0);
assert_eq!(res["replace_null"]["doc_count_error_upper_bound"], 0);
Ok(())
}
#[test]
fn terms_aggregation_missing_mixed_type_mult_seg() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let json = schema_builder.add_json_field("json", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
// => Segment with all values numeric
index_writer
.add_document(doc!(json => json!({"mixed_type": 10.0})))
.unwrap();
index_writer.add_document(doc!())?;
index_writer.commit().unwrap();
//// => Segment with all values text
index_writer
.add_document(doc!(json => json!({"mixed_type": "blue"})))
.unwrap();
index_writer.add_document(doc!())?;
index_writer.commit().unwrap();
// => Segment with mixed values
index_writer
.add_document(doc!(json => json!({"mixed_type": "red"})))
.unwrap();
index_writer
.add_document(doc!(json => json!({"mixed_type": -20.5})))
.unwrap();
index_writer
.add_document(doc!(json => json!({"mixed_type": true})))
.unwrap();
index_writer.add_document(doc!())?;
index_writer.commit().unwrap();
let agg_req: Aggregations = serde_json::from_value(json!({
"replace_null": {
"terms": {
"field": "json.mixed_type",
"missing": "NULL"
},
},
"replace_num": {
"terms": {
"field": "json.mixed_type",
"missing": 1337
},
},
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// text field
assert_eq!(res["replace_null"]["buckets"][0]["key"], "NULL");
assert_eq!(res["replace_null"]["buckets"][0]["doc_count"], 3);
assert_eq!(res["replace_num"]["buckets"][0]["key"], 1337.0);
assert_eq!(res["replace_num"]["buckets"][0]["doc_count"], 3);
assert_eq!(res["replace_null"]["sum_other_doc_count"], 0);
assert_eq!(res["replace_null"]["doc_count_error_upper_bound"], 0);
Ok(())
}
#[test]
fn terms_aggregation_missing_str_on_numeric_field() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let json = schema_builder.add_json_field("json", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
// => Segment with all values numeric
index_writer
.add_document(doc!(json => json!({"mixed_type": 10.0})))
.unwrap();
index_writer.add_document(doc!())?;
index_writer.add_document(doc!())?;
index_writer
.add_document(doc!(json => json!({"mixed_type": -20.5})))
.unwrap();
index_writer.add_document(doc!())?;
index_writer.commit().unwrap();
let agg_req: Aggregations = serde_json::from_value(json!({
"replace_null": {
"terms": {
"field": "json.mixed_type",
"missing": "NULL"
},
},
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// text field
assert_eq!(res["replace_null"]["buckets"][0]["key"], "NULL");
assert_eq!(res["replace_null"]["buckets"][0]["doc_count"], 3);
assert_eq!(res["replace_null"]["sum_other_doc_count"], 0);
assert_eq!(res["replace_null"]["doc_count_error_upper_bound"], 0);
Ok(())
}
#[test]
fn terms_aggregation_missing_mixed_type_one_seg() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let json = schema_builder.add_json_field("json", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
// => Segment with all values numeric
index_writer
.add_document(doc!(json => json!({"mixed_type": 10.0})))
.unwrap();
index_writer.add_document(doc!())?;
//// => Segment with all values text
index_writer
.add_document(doc!(json => json!({"mixed_type": "blue"})))
.unwrap();
index_writer.add_document(doc!())?;
// => Segment with mixed values
index_writer
.add_document(doc!(json => json!({"mixed_type": "red"})))
.unwrap();
index_writer
.add_document(doc!(json => json!({"mixed_type": -20.5})))
.unwrap();
index_writer
.add_document(doc!(json => json!({"mixed_type": true})))
.unwrap();
index_writer.add_document(doc!())?;
index_writer.commit().unwrap();
let agg_req: Aggregations = serde_json::from_value(json!({
"replace_null": {
"terms": {
"field": "json.mixed_type",
"missing": "NULL"
},
},
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// text field
assert_eq!(res["replace_null"]["buckets"][0]["key"], "NULL");
assert_eq!(res["replace_null"]["buckets"][0]["doc_count"], 3);
assert_eq!(res["replace_null"]["sum_other_doc_count"], 0);
assert_eq!(res["replace_null"]["doc_count_error_upper_bound"], 0);
Ok(())
}
}

7
src/aggregation/buf_collector.rs
View File

@@ -35,11 +35,12 @@ impl BufAggregationCollector {
impl SegmentAggregationCollector for BufAggregationCollector {
#[inline]
fn into_intermediate_aggregations_result(
fn add_intermediate_aggregation_result(
self: Box<Self>,
agg_with_accessor: &AggregationsWithAccessor,
) -> crate::Result<IntermediateAggregationResults> {
Box::new(self.collector).into_intermediate_aggregations_result(agg_with_accessor)
results: &mut IntermediateAggregationResults,
) -> crate::Result<()> {
Box::new(self.collector).add_intermediate_aggregation_result(agg_with_accessor, results)
}
#[inline]

20
src/aggregation/collector.rs
View File

@@ -6,7 +6,7 @@ use super::intermediate_agg_result::IntermediateAggregationResults;
use super::segment_agg_result::{
build_segment_agg_collector, AggregationLimits, SegmentAggregationCollector,
};
use crate::aggregation::agg_req_with_accessor::get_aggs_with_accessor_and_validate;
use crate::aggregation::agg_req_with_accessor::get_aggs_with_segment_accessor_and_validate;
use crate::collector::{Collector, SegmentCollector};
use crate::{DocId, SegmentReader, TantivyError};
@@ -104,7 +104,7 @@ impl Collector for AggregationCollector {
segment_fruits: Vec<<Self::Child as SegmentCollector>::Fruit>,
) -> crate::Result<Self::Fruit> {
let res = merge_fruits(segment_fruits)?;
res.into_final_bucket_result(self.agg.clone(), &self.limits)
res.into_final_result(self.agg.clone(), &self.limits)
}
}
@@ -114,7 +114,7 @@ fn merge_fruits(
if let Some(fruit) = segment_fruits.pop() {
let mut fruit = fruit?;
for next_fruit in segment_fruits {
fruit.merge_fruits(next_fruit?);
fruit.merge_fruits(next_fruit?)?;
}
Ok(fruit)
} else {
@@ -137,9 +137,10 @@ impl AggregationSegmentCollector {
reader: &SegmentReader,
limits: &AggregationLimits,
) -> crate::Result<Self> {
let aggs_with_accessor = get_aggs_with_accessor_and_validate(agg, reader, limits)?;
let mut aggs_with_accessor =
get_aggs_with_segment_accessor_and_validate(agg, reader, limits)?;
let result =
BufAggregationCollector::new(build_segment_agg_collector(&aggs_with_accessor)?);
BufAggregationCollector::new(build_segment_agg_collector(&mut aggs_with_accessor)?);
Ok(AggregationSegmentCollector {
aggs_with_accessor,
agg_collector: result,
@@ -184,6 +185,13 @@ impl SegmentCollector for AggregationSegmentCollector {
return Err(err);
}
self.agg_collector.flush(&mut self.aggs_with_accessor)?;
Box::new(self.agg_collector).into_intermediate_aggregations_result(&self.aggs_with_accessor)
let mut sub_aggregation_res = IntermediateAggregationResults::default();
Box::new(self.agg_collector).add_intermediate_aggregation_result(
&self.aggs_with_accessor,
&mut sub_aggregation_res,
)?;
Ok(sub_aggregation_res)
}
}

12
src/aggregation/date.rs
View File

@@ -4,13 +4,11 @@ use time::OffsetDateTime;
use crate::TantivyError;
pub(crate) fn format_date(val: i64) -> crate::Result<String> {
let datetime =
OffsetDateTime::from_unix_timestamp_nanos(1_000 * (val as i128)).map_err(|err| {
TantivyError::InvalidArgument(format!(
"Could not convert {:?} to OffsetDateTime, err {:?}",
val, err
))
})?;
let datetime = OffsetDateTime::from_unix_timestamp_nanos(val as i128).map_err(|err| {
TantivyError::InvalidArgument(format!(
"Could not convert {val:?} to OffsetDateTime, err {err:?}"
))
})?;
let key_as_string = datetime
.format(&Rfc3339)
.map_err(|_err| TantivyError::InvalidArgument("Could not serialize date".to_string()))?;

6
src/aggregation/error.rs
View File

@@ -5,6 +5,12 @@ use super::bucket::DateHistogramParseError;
/// Error that may occur when opening a directory
#[derive(Debug, Clone, PartialEq, Eq, Error)]
pub enum AggregationError {
/// InternalError Aggregation Request
#[error("InternalError: {0:?}")]
InternalError(String),
/// Invalid Aggregation Request
#[error("InvalidRequest: {0:?}")]
InvalidRequest(String),
/// Date histogram parse error
#[error("Date histogram parse error: {0:?}")]
DateHistogramParseError(#[from] DateHistogramParseError),

591
src/aggregation/intermediate_agg_result.rs
View File

@@ -3,50 +3,101 @@
//! indices.
use std::cmp::Ordering;
use std::collections::hash_map::Entry;
use std::hash::Hash;
use columnar::ColumnType;
use itertools::Itertools;
use rustc_hash::FxHashMap;
use serde::ser::SerializeSeq;
use serde::{Deserialize, Deserializer, Serialize, Serializer};
use serde::{Deserialize, Serialize};
use super::agg_req::{
Aggregations, AggregationsInternal, BucketAggregationInternal, BucketAggregationType,
MetricAggregation, RangeAggregation,
};
use super::agg_result::{AggregationResult, BucketResult, RangeBucketEntry};
use super::agg_req::{Aggregation, AggregationVariants, Aggregations};
use super::agg_result::{AggregationResult, BucketResult, MetricResult, RangeBucketEntry};
use super::bucket::{
cut_off_buckets, get_agg_name_and_property, intermediate_histogram_buckets_to_final_buckets,
GetDocCount, Order, OrderTarget, SegmentHistogramBucketEntry, TermsAggregation,
GetDocCount, Order, OrderTarget, RangeAggregation, TermsAggregation,
};
use super::metric::{
IntermediateAverage, IntermediateCount, IntermediateMax, IntermediateMin, IntermediateStats,
IntermediateSum,
IntermediateSum, PercentilesCollector,
};
use super::segment_agg_result::AggregationLimits;
use super::{format_date, AggregationError, Key, SerializedKey, VecWithNames};
use super::{format_date, AggregationError, Key, SerializedKey};
use crate::aggregation::agg_result::{AggregationResults, BucketEntries, BucketEntry};
use crate::aggregation::bucket::TermsAggregationInternal;
use crate::TantivyError;
/// Contains the intermediate aggregation result, which is optimized to be merged with other
/// intermediate results.
///
/// Notice: This struct should not be de/serialized via JSON format.
#[derive(Default, Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct IntermediateAggregationResults {
#[serde(skip_serializing_if = "Option::is_none")]
pub(crate) metrics: Option<VecWithNames<IntermediateMetricResult>>,
#[serde(skip_serializing_if = "Option::is_none")]
pub(crate) buckets: Option<VecWithNames<IntermediateBucketResult>>,
pub(crate) aggs_res: FxHashMap<String, IntermediateAggregationResult>,
}
#[derive(Clone, Debug, Serialize, Deserialize, PartialOrd, PartialEq)]
/// The key to identify a bucket.
/// This might seem redundant with `Key`, but the point is to have a different
/// Serialize implementation.
pub enum IntermediateKey {
/// String key
Str(String),
/// `f64` key
F64(f64),
}
impl From<Key> for IntermediateKey {
fn from(value: Key) -> Self {
match value {
Key::Str(s) => Self::Str(s),
Key::F64(f) => Self::F64(f),
}
}
}
impl From<IntermediateKey> for Key {
fn from(value: IntermediateKey) -> Self {
match value {
IntermediateKey::Str(s) => Self::Str(s),
IntermediateKey::F64(f) => Self::F64(f),
}
}
}
impl Eq for IntermediateKey {}
impl std::hash::Hash for IntermediateKey {
fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
core::mem::discriminant(self).hash(state);
match self {
IntermediateKey::Str(text) => text.hash(state),
IntermediateKey::F64(val) => val.to_bits().hash(state),
}
}
}
impl IntermediateAggregationResults {
/// Add a result
pub fn push(&mut self, key: String, value: IntermediateAggregationResult) -> crate::Result<()> {
let entry = self.aggs_res.entry(key);
match entry {
Entry::Occupied(mut e) => {
// In case of term aggregation over different types, we need to merge the results.
e.get_mut().merge_fruits(value)?;
}
Entry::Vacant(e) => {
e.insert(value);
}
}
Ok(())
}
/// Convert intermediate result and its aggregation request to the final result.
pub fn into_final_bucket_result(
pub fn into_final_result(
self,
req: Aggregations,
limits: &AggregationLimits,
) -> crate::Result<AggregationResults> {
let res = self.into_final_bucket_result_internal(&(req.into()), limits)?;
let res = self.into_final_result_internal(&req, limits)?;
let bucket_count = res.get_bucket_count() as u32;
if bucket_count > limits.get_bucket_limit() {
return Err(TantivyError::AggregationError(
@@ -60,152 +111,97 @@ impl IntermediateAggregationResults {
}
/// Convert intermediate result and its aggregation request to the final result.
///
/// Internal function, AggregationsInternal is used instead Aggregations, which is optimized
/// for internal processing, by splitting metric and buckets into separate groups.
pub(crate) fn into_final_bucket_result_internal(
pub(crate) fn into_final_result_internal(
self,
req: &AggregationsInternal,
req: &Aggregations,
limits: &AggregationLimits,
) -> crate::Result<AggregationResults> {
// Important assumption:
// When the tree contains buckets/metric, we expect it to have all buckets/metrics from the
// request
let mut results: FxHashMap<String, AggregationResult> = FxHashMap::default();
if let Some(buckets) = self.buckets {
convert_and_add_final_buckets_to_result(&mut results, buckets, &req.buckets, limits)?
} else {
// When there are no buckets, we create empty buckets, so that the serialized json
// format is constant
add_empty_final_buckets_to_result(&mut results, &req.buckets, limits)?
};
if let Some(metrics) = self.metrics {
convert_and_add_final_metrics_to_result(&mut results, metrics);
} else {
// When there are no metrics, we create empty metric results, so that the serialized
// json format is constant
add_empty_final_metrics_to_result(&mut results, &req.metrics)?;
for (key, agg_res) in self.aggs_res.into_iter() {
let req = req.get(key.as_str()).unwrap_or_else(|| {
panic!(
"Could not find key {:?} in request keys {:?}. This probably means that \
add_intermediate_aggregation_result passed the wrong agg object.",
key,
req.keys().collect::<Vec<_>>()
)
});
results.insert(key, agg_res.into_final_result(req, limits)?);
}
// Handle empty results
if results.len() != req.len() {
for (key, req) in req.iter() {
if !results.contains_key(key) {
let empty_res = empty_from_req(req);
results.insert(key.to_string(), empty_res.into_final_result(req, limits)?);
}
}
}
Ok(AggregationResults(results))
}
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))
};
pub(crate) fn empty_from_req(req: &Aggregations) -> Self {
let mut aggs_res: FxHashMap<String, IntermediateAggregationResult> = FxHashMap::default();
for (key, req) in req.iter() {
let empty_res = empty_from_req(req);
aggs_res.insert(key.to_string(), empty_res);
}
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 }
Self { aggs_res }
}
/// Merge another 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);
}
pub fn merge_fruits(&mut self, other: IntermediateAggregationResults) -> crate::Result<()> {
for (left, right) in self.aggs_res.values_mut().zip(other.aggs_res.into_values()) {
left.merge_fruits(right)?;
}
Ok(())
}
}
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);
}
pub(crate) fn empty_from_req(req: &Aggregation) -> IntermediateAggregationResult {
use AggregationVariants::*;
match req.agg {
Terms(_) => IntermediateAggregationResult::Bucket(IntermediateBucketResult::Terms(
Default::default(),
)),
Range(_) => IntermediateAggregationResult::Bucket(IntermediateBucketResult::Range(
Default::default(),
)),
Histogram(_) | DateHistogram(_) => {
IntermediateAggregationResult::Bucket(IntermediateBucketResult::Histogram {
buckets: Vec::new(),
column_type: None,
})
}
Average(_) => IntermediateAggregationResult::Metric(IntermediateMetricResult::Average(
IntermediateAverage::default(),
)),
Count(_) => IntermediateAggregationResult::Metric(IntermediateMetricResult::Count(
IntermediateCount::default(),
)),
Max(_) => IntermediateAggregationResult::Metric(IntermediateMetricResult::Max(
IntermediateMax::default(),
)),
Min(_) => IntermediateAggregationResult::Metric(IntermediateMetricResult::Min(
IntermediateMin::default(),
)),
Stats(_) => IntermediateAggregationResult::Metric(IntermediateMetricResult::Stats(
IntermediateStats::default(),
)),
Sum(_) => IntermediateAggregationResult::Metric(IntermediateMetricResult::Sum(
IntermediateSum::default(),
)),
Percentiles(_) => IntermediateAggregationResult::Metric(
IntermediateMetricResult::Percentiles(PercentilesCollector::default()),
),
}
}
fn convert_and_add_final_metrics_to_result(
results: &mut FxHashMap<String, AggregationResult>,
metrics: VecWithNames<IntermediateMetricResult>,
) {
results.extend(
metrics
.into_iter()
.map(|(key, metric)| (key, AggregationResult::MetricResult(metric.into()))),
);
}
fn add_empty_final_metrics_to_result(
results: &mut FxHashMap<String, AggregationResult>,
req_metrics: &VecWithNames<MetricAggregation>,
) -> crate::Result<()> {
results.extend(req_metrics.iter().map(|(key, req)| {
let empty_bucket = IntermediateMetricResult::empty_from_req(req);
(
key.to_string(),
AggregationResult::MetricResult(empty_bucket.into()),
)
}));
Ok(())
}
fn add_empty_final_buckets_to_result(
results: &mut FxHashMap<String, AggregationResult>,
req_buckets: &VecWithNames<BucketAggregationInternal>,
limits: &AggregationLimits,
) -> crate::Result<()> {
let requested_buckets = req_buckets.iter();
for (key, req) in requested_buckets {
let empty_bucket =
AggregationResult::BucketResult(BucketResult::empty_from_req(req, limits)?);
results.insert(key.to_string(), empty_bucket);
}
Ok(())
}
fn convert_and_add_final_buckets_to_result(
results: &mut FxHashMap<String, AggregationResult>,
buckets: VecWithNames<IntermediateBucketResult>,
req_buckets: &VecWithNames<BucketAggregationInternal>,
limits: &AggregationLimits,
) -> crate::Result<()> {
assert_eq!(buckets.len(), req_buckets.len());
let buckets_with_request = buckets.into_iter().zip(req_buckets.values());
for ((key, bucket), req) in buckets_with_request {
let result = AggregationResult::BucketResult(bucket.into_final_bucket_result(req, limits)?);
results.insert(key, result);
}
Ok(())
}
/// An aggregation is either a bucket or a metric.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub enum IntermediateAggregationResult {
@@ -215,9 +211,42 @@ pub enum IntermediateAggregationResult {
Metric(IntermediateMetricResult),
}
impl IntermediateAggregationResult {
pub(crate) fn into_final_result(
self,
req: &Aggregation,
limits: &AggregationLimits,
) -> crate::Result<AggregationResult> {
let res = match self {
IntermediateAggregationResult::Bucket(bucket) => {
AggregationResult::BucketResult(bucket.into_final_bucket_result(req, limits)?)
}
IntermediateAggregationResult::Metric(metric) => {
AggregationResult::MetricResult(metric.into_final_metric_result(req))
}
};
Ok(res)
}
fn merge_fruits(&mut self, other: IntermediateAggregationResult) -> crate::Result<()> {
match (self, other) {
(
IntermediateAggregationResult::Bucket(b1),
IntermediateAggregationResult::Bucket(b2),
) => b1.merge_fruits(b2),
(
IntermediateAggregationResult::Metric(m1),
IntermediateAggregationResult::Metric(m2),
) => m1.merge_fruits(m2),
_ => panic!("aggregation result type mismatch (mixed metric and buckets)"),
}
}
}
/// Holds the intermediate data for metric results
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub enum IntermediateMetricResult {
/// Intermediate average result.
Percentiles(PercentilesCollector),
/// Intermediate average result.
Average(IntermediateAverage),
/// Intermediate count result.
@@ -233,23 +262,34 @@ pub enum IntermediateMetricResult {
}
impl IntermediateMetricResult {
pub(crate) fn empty_from_req(req: &MetricAggregation) -> Self {
match req {
MetricAggregation::Average(_) => {
IntermediateMetricResult::Average(IntermediateAverage::default())
fn into_final_metric_result(self, req: &Aggregation) -> MetricResult {
match self {
IntermediateMetricResult::Average(intermediate_avg) => {
MetricResult::Average(intermediate_avg.finalize().into())
}
MetricAggregation::Count(_) => {
IntermediateMetricResult::Count(IntermediateCount::default())
IntermediateMetricResult::Count(intermediate_count) => {
MetricResult::Count(intermediate_count.finalize().into())
}
MetricAggregation::Max(_) => IntermediateMetricResult::Max(IntermediateMax::default()),
MetricAggregation::Min(_) => IntermediateMetricResult::Min(IntermediateMin::default()),
MetricAggregation::Stats(_) => {
IntermediateMetricResult::Stats(IntermediateStats::default())
IntermediateMetricResult::Max(intermediate_max) => {
MetricResult::Max(intermediate_max.finalize().into())
}
MetricAggregation::Sum(_) => IntermediateMetricResult::Sum(IntermediateSum::default()),
IntermediateMetricResult::Min(intermediate_min) => {
MetricResult::Min(intermediate_min.finalize().into())
}
IntermediateMetricResult::Stats(intermediate_stats) => {
MetricResult::Stats(intermediate_stats.finalize())
}
IntermediateMetricResult::Sum(intermediate_sum) => {
MetricResult::Sum(intermediate_sum.finalize().into())
}
IntermediateMetricResult::Percentiles(percentiles) => MetricResult::Percentiles(
percentiles
.into_final_result(req.agg.as_percentile().expect("unexpected metric type")),
),
}
}
fn merge_fruits(&mut self, other: IntermediateMetricResult) {
fn merge_fruits(&mut self, other: IntermediateMetricResult) -> crate::Result<()> {
match (self, other) {
(
IntermediateMetricResult::Average(avg_left),
@@ -278,10 +318,18 @@ impl IntermediateMetricResult {
(IntermediateMetricResult::Sum(sum_left), IntermediateMetricResult::Sum(sum_right)) => {
sum_left.merge_fruits(sum_right);
}
(
IntermediateMetricResult::Percentiles(left),
IntermediateMetricResult::Percentiles(right),
) => {
left.merge_fruits(right)?;
}
_ => {
panic!("incompatible fruit types in tree");
panic!("incompatible fruit types in tree or missing merge_fruits handler");
}
}
Ok(())
}
}
@@ -307,7 +355,7 @@ pub enum IntermediateBucketResult {
impl IntermediateBucketResult {
pub(crate) fn into_final_bucket_result(
self,
req: &BucketAggregationInternal,
req: &Aggregation,
limits: &AggregationLimits,
) -> crate::Result<BucketResult> {
match self {
@@ -317,8 +365,9 @@ impl IntermediateBucketResult {
.into_values()
.map(|bucket| {
bucket.into_final_bucket_entry(
&req.sub_aggregation,
req.as_range()
req.sub_aggregation(),
req.agg
.as_range()
.expect("unexpected aggregation, expected histogram aggregation"),
range_res.column_type,
limits,
@@ -333,6 +382,7 @@ impl IntermediateBucketResult {
});
let is_keyed = req
.agg
.as_range()
.expect("unexpected aggregation, expected range aggregation")
.keyed;
@@ -353,13 +403,14 @@ impl IntermediateBucketResult {
buckets,
} => {
let histogram_req = &req
.agg
.as_histogram()?
.expect("unexpected aggregation, expected histogram aggregation");
let buckets = intermediate_histogram_buckets_to_final_buckets(
buckets,
column_type,
histogram_req,
&req.sub_aggregation,
req.sub_aggregation(),
limits,
)?;
@@ -376,33 +427,22 @@ impl IntermediateBucketResult {
Ok(BucketResult::Histogram { buckets })
}
IntermediateBucketResult::Terms(terms) => terms.into_final_result(
req.as_term()
req.agg
.as_term()
.expect("unexpected aggregation, expected term aggregation"),
&req.sub_aggregation,
req.sub_aggregation(),
limits,
),
}
}
pub(crate) fn empty_from_req(req: &BucketAggregationType) -> Self {
match req {
BucketAggregationType::Terms(_) => IntermediateBucketResult::Terms(Default::default()),
BucketAggregationType::Range(_) => IntermediateBucketResult::Range(Default::default()),
BucketAggregationType::Histogram(_) | BucketAggregationType::DateHistogram(_) => {
IntermediateBucketResult::Histogram {
buckets: vec![],
column_type: None,
}
}
}
}
fn merge_fruits(&mut self, other: IntermediateBucketResult) {
fn merge_fruits(&mut self, other: IntermediateBucketResult) -> crate::Result<()> {
match (self, other) {
(
IntermediateBucketResult::Terms(term_res_left),
IntermediateBucketResult::Terms(term_res_right),
) => {
merge_key_maps(&mut term_res_left.entries, term_res_right.entries);
merge_maps(&mut term_res_left.entries, term_res_right.entries)?;
term_res_left.sum_other_doc_count += term_res_right.sum_other_doc_count;
term_res_left.doc_count_error_upper_bound +=
term_res_right.doc_count_error_upper_bound;
@@ -412,7 +452,7 @@ impl IntermediateBucketResult {
IntermediateBucketResult::Range(range_res_left),
IntermediateBucketResult::Range(range_res_right),
) => {
merge_serialized_key_maps(&mut range_res_left.buckets, range_res_right.buckets);
merge_maps(&mut range_res_left.buckets, range_res_right.buckets)?;
}
(
IntermediateBucketResult::Histogram {
@@ -424,22 +464,23 @@ impl IntermediateBucketResult {
..
},
) => {
let buckets = buckets_left
.drain(..)
.merge_join_by(buckets_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();
let buckets: Result<Vec<IntermediateHistogramBucketEntry>, TantivyError> =
buckets_left
.drain(..)
.merge_join_by(buckets_right, |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)?;
Ok(left)
}
itertools::EitherOrBoth::Left(left) => Ok(left),
itertools::EitherOrBoth::Right(right) => Ok(right),
})
.collect::<Result<_, _>>();
*buckets_left = buckets;
*buckets_left = buckets?;
}
(IntermediateBucketResult::Range(_), _) => {
panic!("try merge on different types")
@@ -451,6 +492,7 @@ impl IntermediateBucketResult {
panic!("try merge on different types")
}
}
Ok(())
}
}
@@ -464,59 +506,31 @@ pub struct IntermediateRangeBucketResult {
#[derive(Default, Clone, Debug, PartialEq, Serialize, Deserialize)]
/// Term aggregation including error counts
pub struct IntermediateTermBucketResult {
#[serde(
serialize_with = "serialize_entries",
deserialize_with = "deserialize_entries"
)]
pub(crate) entries: FxHashMap<Key, IntermediateTermBucketEntry>,
pub(crate) entries: FxHashMap<IntermediateKey, IntermediateTermBucketEntry>,
pub(crate) sum_other_doc_count: u64,
pub(crate) doc_count_error_upper_bound: u64,
}
// Serialize into a Vec to circument the JSON limitation, where keys can't be numbers
fn serialize_entries<S>(
entries: &FxHashMap<Key, IntermediateTermBucketEntry>,
serializer: S,
) -> Result<S::Ok, S::Error>
where
S: Serializer,
{
let mut seq = serializer.serialize_seq(Some(entries.len()))?;
for (k, v) in entries {
seq.serialize_element(&(k, v))?;
}
seq.end()
}
fn deserialize_entries<'de, D>(
deserializer: D,
) -> Result<FxHashMap<Key, IntermediateTermBucketEntry>, D::Error>
where D: Deserializer<'de> {
let vec_entries: Vec<(Key, IntermediateTermBucketEntry)> =
Deserialize::deserialize(deserializer)?;
Ok(vec_entries.into_iter().collect())
}
impl IntermediateTermBucketResult {
pub(crate) fn into_final_result(
self,
req: &TermsAggregation,
sub_aggregation_req: &AggregationsInternal,
sub_aggregation_req: &Aggregations,
limits: &AggregationLimits,
) -> crate::Result<BucketResult> {
let req = TermsAggregationInternal::from_req(req);
let mut buckets: Vec<BucketEntry> = self
.entries
.into_iter()
.filter(|bucket| bucket.1.doc_count >= req.min_doc_count)
.filter(|bucket| bucket.1.doc_count as u64 >= req.min_doc_count)
.map(|(key, entry)| {
Ok(BucketEntry {
key_as_string: None,
key,
doc_count: entry.doc_count,
key: key.into(),
doc_count: entry.doc_count as u64,
sub_aggregation: entry
.sub_aggregation
.into_final_bucket_result_internal(sub_aggregation_req, limits)?,
.into_final_result_internal(sub_aggregation_req, limits)?,
})
})
.collect::<crate::Result<_>>()?;
@@ -587,37 +601,23 @@ impl IntermediateTermBucketResult {
}
trait MergeFruits {
fn merge_fruits(&mut self, other: Self);
fn merge_fruits(&mut self, other: Self) -> crate::Result<()>;
}
fn merge_serialized_key_maps<V: MergeFruits + Clone>(
entries_left: &mut FxHashMap<SerializedKey, V>,
mut entries_right: FxHashMap<SerializedKey, V>,
) {
fn merge_maps<V: MergeFruits + Clone, T: Eq + PartialEq + Hash>(
entries_left: &mut FxHashMap<T, V>,
mut entries_right: FxHashMap<T, V>,
) -> crate::Result<()> {
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);
}
}
fn merge_key_maps<V: MergeFruits + Clone>(
entries_left: &mut FxHashMap<Key, V>,
mut entries_right: FxHashMap<Key, 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);
entry_left.merge_fruits(entry_right)?;
}
}
for (key, res) in entries_right.into_iter() {
entries_left.entry(key).or_insert(res);
}
Ok(())
}
/// This is the histogram entry for a bucket, which contains a key, count, and optionally
@@ -635,7 +635,7 @@ pub struct IntermediateHistogramBucketEntry {
impl IntermediateHistogramBucketEntry {
pub(crate) fn into_final_bucket_entry(
self,
req: &AggregationsInternal,
req: &Aggregations,
limits: &AggregationLimits,
) -> crate::Result<BucketEntry> {
Ok(BucketEntry {
@@ -644,53 +644,41 @@ impl IntermediateHistogramBucketEntry {
doc_count: self.doc_count,
sub_aggregation: self
.sub_aggregation
.into_final_bucket_result_internal(req, limits)?,
.into_final_result_internal(req, limits)?,
})
}
}
impl From<SegmentHistogramBucketEntry> for IntermediateHistogramBucketEntry {
fn from(entry: SegmentHistogramBucketEntry) -> Self {
IntermediateHistogramBucketEntry {
key: entry.key,
doc_count: entry.doc_count,
sub_aggregation: Default::default(),
}
}
}
/// 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 unique key the bucket is identified with.
pub key: IntermediateKey,
/// The number of documents in the bucket.
pub doc_count: 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 IntermediateRangeBucketEntry {
pub(crate) fn into_final_bucket_entry(
self,
req: &AggregationsInternal,
req: &Aggregations,
_range_req: &RangeAggregation,
column_type: Option<ColumnType>,
limits: &AggregationLimits,
) -> crate::Result<RangeBucketEntry> {
let mut range_bucket_entry = RangeBucketEntry {
key: self.key,
key: self.key.into(),
doc_count: self.doc_count,
sub_aggregation: self
.sub_aggregation
.into_final_bucket_result_internal(req, limits)?,
.into_final_result_internal(req, limits)?,
to: self.to,
from: self.from,
to_as_string: None,
@@ -719,29 +707,32 @@ impl IntermediateRangeBucketEntry {
#[derive(Clone, Default, Debug, PartialEq, Serialize, Deserialize)]
pub struct IntermediateTermBucketEntry {
/// The number of documents in the bucket.
pub doc_count: u64,
pub doc_count: u32,
/// The sub_aggregation in this bucket.
pub sub_aggregation: IntermediateAggregationResults,
}
impl MergeFruits for IntermediateTermBucketEntry {
fn merge_fruits(&mut self, other: IntermediateTermBucketEntry) {
fn merge_fruits(&mut self, other: IntermediateTermBucketEntry) -> crate::Result<()> {
self.doc_count += other.doc_count;
self.sub_aggregation.merge_fruits(other.sub_aggregation);
self.sub_aggregation.merge_fruits(other.sub_aggregation)?;
Ok(())
}
}
impl MergeFruits for IntermediateRangeBucketEntry {
fn merge_fruits(&mut self, other: IntermediateRangeBucketEntry) {
fn merge_fruits(&mut self, other: IntermediateRangeBucketEntry) -> crate::Result<()> {
self.doc_count += other.doc_count;
self.sub_aggregation.merge_fruits(other.sub_aggregation);
self.sub_aggregation.merge_fruits(other.sub_aggregation)?;
Ok(())
}
}
impl MergeFruits for IntermediateHistogramBucketEntry {
fn merge_fruits(&mut self, other: IntermediateHistogramBucketEntry) {
fn merge_fruits(&mut self, other: IntermediateHistogramBucketEntry) -> crate::Result<()> {
self.doc_count += other.doc_count;
self.sub_aggregation.merge_fruits(other.sub_aggregation);
self.sub_aggregation.merge_fruits(other.sub_aggregation)?;
Ok(())
}
}
@@ -760,7 +751,7 @@ mod tests {
buckets.insert(
key.to_string(),
IntermediateRangeBucketEntry {
key: Key::Str(key.to_string()),
key: IntermediateKey::Str(key.to_string()),
doc_count: *doc_count,
sub_aggregation: Default::default(),
from: None,
@@ -770,14 +761,15 @@ mod tests {
}
map.insert(
"my_agg_level2".to_string(),
IntermediateBucketResult::Range(IntermediateRangeBucketResult {
buckets,
column_type: None,
}),
IntermediateAggregationResult::Bucket(IntermediateBucketResult::Range(
IntermediateRangeBucketResult {
buckets,
column_type: None,
},
)),
);
IntermediateAggregationResults {
buckets: Some(VecWithNames::from_entries(map.into_iter().collect())),
metrics: Default::default(),
aggs_res: map.into_iter().collect(),
}
}
@@ -790,7 +782,7 @@ mod tests {
buckets.insert(
key.to_string(),
IntermediateRangeBucketEntry {
key: Key::Str(key.to_string()),
key: IntermediateKey::Str(key.to_string()),
doc_count: *doc_count,
from: None,
to: None,
@@ -803,14 +795,15 @@ mod tests {
}
map.insert(
"my_agg_level1".to_string(),
IntermediateBucketResult::Range(IntermediateRangeBucketResult {
buckets,
column_type: None,
}),
IntermediateAggregationResult::Bucket(IntermediateBucketResult::Range(
IntermediateRangeBucketResult {
buckets,
column_type: None,
},
)),
);
IntermediateAggregationResults {
buckets: Some(VecWithNames::from_entries(map.into_iter().collect())),
metrics: Default::default(),
aggs_res: map.into_iter().collect(),
}
}
@@ -825,7 +818,7 @@ mod tests {
("blue".to_string(), 25, "1900".to_string(), 50),
]);
tree_left.merge_fruits(tree_right);
tree_left.merge_fruits(tree_right).unwrap();
let tree_expected = get_intermediat_tree_with_ranges(&[
("red".to_string(), 110, "1900".to_string(), 55),
@@ -846,7 +839,7 @@ mod tests {
("green".to_string(), 25, "1900".to_string(), 50),
]);
tree_left.merge_fruits(tree_right);
tree_left.merge_fruits(tree_right).unwrap();
let tree_expected = get_intermediat_tree_with_ranges(&[
("red".to_string(), 110, "1900".to_string(), 55),
@@ -866,30 +859,10 @@ mod tests {
let orig = tree_left.clone();
tree_left.merge_fruits(IntermediateAggregationResults::default());
tree_left
.merge_fruits(IntermediateAggregationResults::default())
.unwrap();
assert_eq!(tree_left, orig);
}
#[test]
fn test_term_bucket_json_roundtrip() {
let term_buckets = IntermediateTermBucketResult {
entries: vec![(
Key::F64(5.0),
IntermediateTermBucketEntry {
doc_count: 10,
sub_aggregation: Default::default(),
},
)]
.into_iter()
.collect(),
sum_other_doc_count: 0,
doc_count_error_upper_bound: 0,
};
let term_buckets_round: IntermediateTermBucketResult =
serde_json::from_str(&serde_json::to_string(&term_buckets).unwrap()).unwrap();
assert_eq!(term_buckets, term_buckets_round);
}
}

11
src/aggregation/metric/average.rs
View File

@@ -20,12 +20,21 @@ use super::{IntermediateStats, SegmentStatsCollector};
pub struct AverageAggregation {
/// The field name to compute the average on.
pub field: String,
/// The missing parameter defines how documents that are missing a value should be treated.
/// By default they will be ignored but it is also possible to treat them as if they had a
/// value. Examples in JSON format:
/// { "field": "my_numbers", "missing": "10.0" }
#[serde(default)]
pub missing: Option<f64>,
}
impl AverageAggregation {
/// Creates a new [`AverageAggregation`] instance from a field name.
pub fn from_field_name(field_name: String) -> Self {
Self { field: field_name }
Self {
field: field_name,
missing: None,
}
}
/// Returns the field name the aggregation is computed on.
pub fn field_name(&self) -> &str {

15
src/aggregation/metric/count.rs
View File

@@ -18,14 +18,23 @@ use super::{IntermediateStats, SegmentStatsCollector};
/// ```
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct CountAggregation {
/// The field name to compute the minimum on.
/// The field name to compute the count on.
pub field: String,
/// The missing parameter defines how documents that are missing a value should be treated.
/// By default they will be ignored but it is also possible to treat them as if they had a
/// value. Examples in JSON format:
/// { "field": "my_numbers", "missing": "10.0" }
#[serde(default)]
pub missing: Option<f64>,
}
impl CountAggregation {
/// Creates a new [`CountAggregation`] instance from a field name.
pub fn from_field_name(field_name: String) -> Self {
Self { field: field_name }
Self {
field: field_name,
missing: None,
}
}
/// Returns the field name the aggregation is computed on.
pub fn field_name(&self) -> &str {
@@ -51,7 +60,7 @@ impl IntermediateCount {
pub fn merge_fruits(&mut self, other: IntermediateCount) {
self.stats.merge_fruits(other.stats);
}
/// Computes the final minimum value.
/// Computes the final count value.
pub fn finalize(&self) -> Option<f64> {
Some(self.stats.finalize().count as f64)
}

63
src/aggregation/metric/max.rs
View File

@@ -20,12 +20,21 @@ use super::{IntermediateStats, SegmentStatsCollector};
pub struct MaxAggregation {
/// The field name to compute the maximum on.
pub field: String,
/// The missing parameter defines how documents that are missing a value should be treated.
/// By default they will be ignored but it is also possible to treat them as if they had a
/// value. Examples in JSON format:
/// { "field": "my_numbers", "missing": "10.0" }
#[serde(default)]
pub missing: Option<f64>,
}
impl MaxAggregation {
/// Creates a new [`MaxAggregation`] instance from a field name.
pub fn from_field_name(field_name: String) -> Self {
Self { field: field_name }
Self {
field: field_name,
missing: None,
}
}
/// Returns the field name the aggregation is computed on.
pub fn field_name(&self) -> &str {
@@ -56,3 +65,55 @@ impl IntermediateMax {
self.stats.finalize().max
}
}
#[cfg(test)]
mod tests {
use crate::aggregation::agg_req::Aggregations;
use crate::aggregation::tests::exec_request_with_query;
use crate::schema::{Schema, FAST};
use crate::Index;
#[test]
fn test_max_agg_with_missing() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let json = schema_builder.add_json_field("json", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
// => Segment with empty json
index_writer.add_document(doc!()).unwrap();
index_writer.commit().unwrap();
// => Segment with json, but no field partially_empty
index_writer
.add_document(doc!(json => json!({"different_field": "blue"})))
.unwrap();
index_writer.commit().unwrap();
//// => Segment with field partially_empty
index_writer
.add_document(doc!(json => json!({"partially_empty": 10.0})))
.unwrap();
index_writer.add_document(doc!())?;
index_writer.commit().unwrap();
let agg_req: Aggregations = serde_json::from_value(json!({
"my_stats": {
"max": {
"field": "json.partially_empty",
"missing": 100.0,
}
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
assert_eq!(
res["my_stats"],
json!({
"value": 100.0,
})
);
Ok(())
}
}

11
src/aggregation/metric/min.rs
View File

@@ -20,12 +20,21 @@ use super::{IntermediateStats, SegmentStatsCollector};
pub struct MinAggregation {
/// The field name to compute the minimum on.
pub field: String,
/// The missing parameter defines how documents that are missing a value should be treated.
/// By default they will be ignored but it is also possible to treat them as if they had a
/// value. Examples in JSON format:
/// { "field": "my_numbers", "missing": "10.0" }
#[serde(default)]
pub missing: Option<f64>,
}
impl MinAggregation {
/// Creates a new [`MinAggregation`] instance from a field name.
pub fn from_field_name(field_name: String) -> Self {
Self { field: field_name }
Self {
field: field_name,
missing: None,
}
}
/// Returns the field name the aggregation is computed on.
pub fn field_name(&self) -> &str {

48
src/aggregation/metric/mod.rs
View File

@@ -1,17 +1,34 @@
//! Module for all metric aggregations.
//!
//! The aggregations in this family compute metrics, see [super::agg_req::MetricAggregation] for
//! details.
//! 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.
//!
//! ## Supported Metric Aggregations
//! - [Average](AverageAggregation)
//! - [Stats](StatsAggregation)
//! - [Min](MinAggregation)
//! - [Max](MaxAggregation)
//! - [Sum](SumAggregation)
//! - [Count](CountAggregation)
//! - [Percentiles](PercentilesAggregationReq)
mod average;
mod count;
mod max;
mod min;
mod percentiles;
mod stats;
mod sum;
pub use average::*;
pub use count::*;
pub use max::*;
pub use min::*;
pub use percentiles::*;
use rustc_hash::FxHashMap;
use serde::{Deserialize, Serialize};
pub use stats::*;
pub use sum::*;
@@ -37,6 +54,33 @@ impl From<Option<f64>> for SingleMetricResult {
}
}
/// This is the wrapper of percentile entries, which can be vector or hashmap
/// depending on if it's keyed or not.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
#[serde(untagged)]
pub enum PercentileValues {
/// Vector format percentile entries
Vec(Vec<PercentileValuesVecEntry>),
/// HashMap format percentile entries. Key is the serialized percentile
HashMap(FxHashMap<String, f64>),
}
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
/// The entry when requesting percentiles with keyed: false
pub struct PercentileValuesVecEntry {
key: f64,
value: f64,
}
/// 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 PercentilesMetricResult {
/// The result of the percentile metric.
pub values: PercentileValues,
}
#[cfg(test)]
mod tests {
use crate::aggregation::agg_req::Aggregations;

690
src/aggregation/metric/percentiles.rs Normal file
View File

@@ -0,0 +1,690 @@
use std::fmt::Debug;
use columnar::ColumnType;
use serde::{Deserialize, Serialize};
use super::*;
use crate::aggregation::agg_req_with_accessor::{
AggregationWithAccessor, AggregationsWithAccessor,
};
use crate::aggregation::intermediate_agg_result::{
IntermediateAggregationResult, IntermediateAggregationResults, IntermediateMetricResult,
};
use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
use crate::aggregation::{f64_from_fastfield_u64, f64_to_fastfield_u64, AggregationError};
use crate::{DocId, TantivyError};
/// # Percentiles
///
/// The percentiles aggregation is a useful tool for understanding the distribution
/// of a data set. It calculates the values below which a given percentage of the
/// data falls. For instance, the 95th percentile indicates the value below which
/// 95% of the data points can be found.
///
/// This aggregation can be particularly interesting for analyzing website or service response
/// times. For example, if the 95th percentile website load time is significantly higher than the
/// median, this indicates that a small percentage of users are experiencing much slower load times
/// than the majority.
///
/// To use the percentiles aggregation, you'll need to provide a field to
/// aggregate on. In the case of website load times, this would typically be a
/// field containing the duration of time it takes for the site to load.
///
/// The following example demonstrates a request for the percentiles of the "load_time"
/// field:
///
/// ```JSON
/// {
/// "percentiles": {
/// "field": "load_time"
/// }
/// }
/// ```
///
/// This request will return an object containing the default percentiles (1, 5,
/// 25, 50 (median), 75, 95, and 99). You can also customize the percentiles you want to
/// calculate by providing an array of values in the "percents" parameter:
///
/// ```JSON
/// {
/// "percentiles": {
/// "field": "load_time",
/// "percents": [10, 20, 30, 40, 50, 60, 70, 80, 90]
/// }
/// }
/// ```
///
/// In this example, the aggregation will return the 10th, 20th, 30th, 40th, 50th,
/// 60th, 70th, 80th, and 90th percentiles of the "load_time" field.
///
/// Analyzing the percentiles of website load times can help you understand the
/// user experience and identify areas for optimization. For example, if the 95th
/// percentile load time is significantly higher than the median, this indicates
/// that a small percentage of users are experiencing much slower load times than
/// the majority.
///
/// # Estimating Percentiles
///
/// While percentiles provide valuable insights into the distribution of data, it's
/// important to understand that they are often estimates. This is because
/// calculating exact percentiles for large data sets can be computationally
/// expensive and time-consuming. As a result, many percentile aggregation
/// algorithms use approximation techniques to provide faster results.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct PercentilesAggregationReq {
/// The field name to compute the percentiles on.
pub field: String,
/// The percentiles to compute.
/// Defaults to [1.0, 5.0, 25.0, 50.0, 75.0, 95.0, 99.0]
pub percents: Option<Vec<f64>>,
/// Whether to return the percentiles as a hash map
#[serde(default = "default_as_true")]
pub keyed: bool,
/// The missing parameter defines how documents that are missing a value should be treated.
/// By default they will be ignored but it is also possible to treat them as if they had a
/// value. Examples in JSON format:
/// { "field": "my_numbers", "missing": "10.0" }
#[serde(skip_serializing_if = "Option::is_none", default)]
pub missing: Option<f64>,
}
fn default_percentiles() -> &'static [f64] {
&[1.0, 5.0, 25.0, 50.0, 75.0, 95.0, 99.0]
}
fn default_as_true() -> bool {
true
}
impl PercentilesAggregationReq {
/// Creates a new [`PercentilesAggregationReq`] instance from a field name.
pub fn from_field_name(field_name: String) -> Self {
PercentilesAggregationReq {
field: field_name,
percents: None,
keyed: default_as_true(),
missing: None,
}
}
/// Returns the field name the aggregation is computed on.
pub fn field_name(&self) -> &str {
&self.field
}
fn validate(&self) -> crate::Result<()> {
if let Some(percents) = self.percents.as_ref() {
let all_in_range = percents
.iter()
.cloned()
.all(|percent| (0.0..=100.0).contains(&percent));
if !all_in_range {
return Err(TantivyError::AggregationError(
AggregationError::InvalidRequest(
"All percentiles have to be between 0.0 and 100.0".to_string(),
),
));
}
}
Ok(())
}
}
#[derive(Clone, Debug, PartialEq)]
pub(crate) struct SegmentPercentilesCollector {
field_type: ColumnType,
pub(crate) percentiles: PercentilesCollector,
pub(crate) accessor_idx: usize,
val_cache: Vec<u64>,
missing: Option<u64>,
}
#[derive(Clone, Serialize, Deserialize)]
/// The percentiles collector used during segment collection and for merging results.
pub struct PercentilesCollector {
sketch: sketches_ddsketch::DDSketch,
}
impl Default for PercentilesCollector {
fn default() -> Self {
Self::new()
}
}
impl Debug for PercentilesCollector {
fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
f.debug_struct("IntermediatePercentiles")
.field("sketch_len", &self.sketch.length())
.finish()
}
}
impl PartialEq for PercentilesCollector {
fn eq(&self, _other: &Self) -> bool {
false
}
}
fn format_percentil(percentil: f64) -> String {
let mut out = percentil.to_string();
// Slightly silly way to format trailing decimals
if !out.contains('.') {
out.push_str(".0");
}
out
}
impl PercentilesCollector {
/// Convert result into final result. This will query the quantils from the underlying quantil
/// collector.
pub fn into_final_result(self, req: &PercentilesAggregationReq) -> PercentilesMetricResult {
let percentiles: &[f64] = req
.percents
.as_ref()
.map(|el| el.as_ref())
.unwrap_or(default_percentiles());
let iter_quantile_and_values = percentiles.iter().cloned().map(|percentile| {
(
percentile,
self.sketch
.quantile(percentile / 100.0)
.expect(
"quantil out of range. This error should have been caught during \
validation phase",
)
.unwrap_or(f64::NAN),
)
});
let values = if req.keyed {
PercentileValues::HashMap(
iter_quantile_and_values
.map(|(val, quantil)| (format_percentil(val), quantil))
.collect(),
)
} else {
PercentileValues::Vec(
iter_quantile_and_values
.map(|(key, value)| PercentileValuesVecEntry { key, value })
.collect(),
)
};
PercentilesMetricResult { values }
}
fn new() -> Self {
let ddsketch_config = sketches_ddsketch::Config::defaults();
let sketch = sketches_ddsketch::DDSketch::new(ddsketch_config);
Self { sketch }
}
fn collect(&mut self, val: f64) {
self.sketch.add(val);
}
pub(crate) fn merge_fruits(&mut self, right: PercentilesCollector) -> crate::Result<()> {
self.sketch.merge(&right.sketch).map_err(|err| {
TantivyError::AggregationError(AggregationError::InternalError(format!(
"Error while merging percentiles {err:?}"
)))
})?;
Ok(())
}
}
impl SegmentPercentilesCollector {
pub fn from_req_and_validate(
req: &PercentilesAggregationReq,
field_type: ColumnType,
accessor_idx: usize,
) -> crate::Result<Self> {
req.validate()?;
let missing = req
.missing
.and_then(|val| f64_to_fastfield_u64(val, &field_type));
Ok(Self {
field_type,
percentiles: PercentilesCollector::new(),
accessor_idx,
val_cache: Default::default(),
missing,
})
}
#[inline]
pub(crate) fn collect_block_with_field(
&mut self,
docs: &[DocId],
agg_accessor: &mut AggregationWithAccessor,
) {
if let Some(missing) = self.missing.as_ref() {
agg_accessor.column_block_accessor.fetch_block_with_missing(
docs,
&agg_accessor.accessor,
*missing,
);
} else {
agg_accessor
.column_block_accessor
.fetch_block(docs, &agg_accessor.accessor);
}
for val in agg_accessor.column_block_accessor.iter_vals() {
let val1 = f64_from_fastfield_u64(val, &self.field_type);
self.percentiles.collect(val1);
}
}
}
impl SegmentAggregationCollector for SegmentPercentilesCollector {
#[inline]
fn add_intermediate_aggregation_result(
self: Box<Self>,
agg_with_accessor: &AggregationsWithAccessor,
results: &mut IntermediateAggregationResults,
) -> crate::Result<()> {
let name = agg_with_accessor.aggs.keys[self.accessor_idx].to_string();
let intermediate_metric_result = IntermediateMetricResult::Percentiles(self.percentiles);
results.push(
name,
IntermediateAggregationResult::Metric(intermediate_metric_result),
)?;
Ok(())
}
#[inline]
fn collect(
&mut self,
doc: crate::DocId,
agg_with_accessor: &mut AggregationsWithAccessor,
) -> crate::Result<()> {
let field = &agg_with_accessor.aggs.values[self.accessor_idx].accessor;
if let Some(missing) = self.missing {
let mut has_val = false;
for val in field.values_for_doc(doc) {
let val1 = f64_from_fastfield_u64(val, &self.field_type);
self.percentiles.collect(val1);
has_val = true;
}
if !has_val {
self.percentiles
.collect(f64_from_fastfield_u64(missing, &self.field_type));
}
} else {
for val in field.values_for_doc(doc) {
let val1 = f64_from_fastfield_u64(val, &self.field_type);
self.percentiles.collect(val1);
}
}
Ok(())
}
#[inline]
fn collect_block(
&mut self,
docs: &[crate::DocId],
agg_with_accessor: &mut AggregationsWithAccessor,
) -> crate::Result<()> {
let field = &mut agg_with_accessor.aggs.values[self.accessor_idx];
self.collect_block_with_field(docs, field);
Ok(())
}
}
#[cfg(test)]
mod tests {
use itertools::Itertools;
use more_asserts::{assert_ge, assert_le};
use rand::rngs::StdRng;
use rand::SeedableRng;
use serde_json::Value;
use crate::aggregation::agg_req::Aggregations;
use crate::aggregation::agg_result::AggregationResults;
use crate::aggregation::tests::{
exec_request_with_query, get_test_index_from_values, get_test_index_from_values_and_terms,
};
use crate::aggregation::AggregationCollector;
use crate::query::AllQuery;
use crate::schema::{Schema, FAST};
use crate::Index;
#[test]
fn test_aggregation_percentiles_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 = serde_json::from_value(json!({
"percentiles": {
"percentiles": {
"field": "score",
}
},
}))
.unwrap();
let collector = AggregationCollector::from_aggs(agg_req_1, Default::default());
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["percentiles"]["values"],
json!({
"1.0": Value::Null,
"5.0": Value::Null,
"25.0": Value::Null,
"50.0": Value::Null,
"75.0": Value::Null,
"95.0": Value::Null,
"99.0": Value::Null,
})
);
Ok(())
}
#[test]
fn test_aggregation_percentile_simple() -> crate::Result<()> {
let values = vec![10.0];
let index = get_test_index_from_values(false, &values)?;
let agg_req_1: Aggregations = serde_json::from_value(json!({
"percentiles": {
"percentiles": {
"field": "score",
}
},
}))
.unwrap();
let collector = AggregationCollector::from_aggs(agg_req_1, Default::default());
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)?)?;
let percents = vec!["1.0", "5.0", "25.0", "50.0", "75.0", "95.0", "99.0"];
let range = 9.9..10.1;
for percent in percents {
let val = res["percentiles"]["values"][percent].as_f64().unwrap();
assert!(range.contains(&val));
}
Ok(())
}
#[test]
fn test_aggregation_percentile_parameters() -> crate::Result<()> {
let values = vec![10.0];
let index = get_test_index_from_values(false, &values)?;
let agg_req_str = r#"
{
"mypercentiles": {
"percentiles": {
"field": "score",
"percents": [ 95, 99, 99.9 ]
}
}
} "#;
let agg_req_1: Aggregations = serde_json::from_str(agg_req_str).unwrap();
let collector = AggregationCollector::from_aggs(agg_req_1, Default::default());
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)?)?;
let percents = vec!["95.0", "99.0", "99.9"];
let expected_range = 9.9..10.1;
for percent in percents {
let val = res["mypercentiles"]["values"][percent].as_f64().unwrap();
assert!(expected_range.contains(&val));
}
// Keyed false
//
let agg_req_str = r#"
{
"mypercentiles": {
"percentiles": {
"field": "score",
"percents": [ 95, 99, 99.9 ],
"keyed": false
}
}
} "#;
let agg_req_1: Aggregations = serde_json::from_str(agg_req_str).unwrap();
let collector = AggregationCollector::from_aggs(agg_req_1, Default::default());
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)?)?;
let vals = &res["mypercentiles"]["values"];
assert_eq!(vals[0]["key"].as_f64().unwrap(), 95.0);
assert_eq!(vals[1]["key"].as_f64().unwrap(), 99.0);
assert_eq!(vals[2]["key"].as_f64().unwrap(), 99.9);
assert_eq!(vals[3]["key"], serde_json::Value::Null);
assert!(expected_range.contains(&vals[0]["value"].as_f64().unwrap()));
assert!(expected_range.contains(&vals[1]["value"].as_f64().unwrap()));
assert!(expected_range.contains(&vals[2]["value"].as_f64().unwrap()));
Ok(())
}
#[test]
fn test_aggregation_percentiles_single_seg() -> crate::Result<()> {
test_aggregation_percentiles(true)
}
#[test]
fn test_aggregation_percentiles_multi_seg() -> crate::Result<()> {
test_aggregation_percentiles(false)
}
fn test_aggregation_percentiles(merge_segments: bool) -> crate::Result<()> {
use rand_distr::Distribution;
let num_values_in_segment = [100, 30_000, 8000];
let lg_norm = rand_distr::LogNormal::new(2.996f64, 0.979f64).unwrap();
let mut rng = StdRng::from_seed([1u8; 32]);
let segment_data = |i| {
(0..num_values_in_segment[i])
.map(|_| lg_norm.sample(&mut rng))
.collect_vec()
};
let values = (0..=2).map(segment_data).collect_vec();
let mut all_values = values
.iter()
.flat_map(|el| el.iter().cloned())
.collect_vec();
all_values.sort_unstable_by(|a, b| a.total_cmp(b));
fn get_exact_quantil(q: f64, all_values: &[f64]) -> f64 {
let q = q / 100.0;
assert!((0f64..=1f64).contains(&q));
let index = (all_values.len() as f64 * q).ceil() as usize;
let index = index.min(all_values.len() - 1);
all_values[index]
}
let segment_and_values = values
.into_iter()
.map(|segment_data| {
segment_data
.into_iter()
.map(|val| (val, val.to_string()))
.collect_vec()
})
.collect_vec();
let index =
get_test_index_from_values_and_terms(merge_segments, &segment_and_values).unwrap();
let reader = index.reader()?;
let agg_req_str = r#"
{
"mypercentiles": {
"percentiles": {
"field": "score_f64",
"percents": [ 95, 99, 99.9 ]
}
}
} "#;
let agg_req_1: Aggregations = serde_json::from_str(agg_req_str).unwrap();
let collector = AggregationCollector::from_aggs(agg_req_1, Default::default());
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)?)?;
let vals = &res["mypercentiles"]["values"];
let check_quantil = |exact_quantil: f64, val: f64| {
let lower = exact_quantil - exact_quantil * 0.02;
let upper = exact_quantil + exact_quantil * 0.02;
assert_le!(val, upper);
assert_ge!(val, lower);
};
let val = vals["95.0"].as_f64().unwrap();
let exact_quantil = get_exact_quantil(95.0, &all_values);
check_quantil(exact_quantil, val);
let val = vals["99.0"].as_f64().unwrap();
let exact_quantil = get_exact_quantil(99.0, &all_values);
check_quantil(exact_quantil, val);
let val = vals["99.9"].as_f64().unwrap();
let exact_quantil = get_exact_quantil(99.9, &all_values);
check_quantil(exact_quantil, val);
Ok(())
}
#[test]
fn test_percentiles_missing_sub_agg() -> crate::Result<()> {
// This test verifies the `collect` method (in contrast to `collect_block`), which is
// called when the sub-aggregations are flushed.
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("texts", FAST);
let score_field_f64 = schema_builder.add_f64_field("score", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{
let mut index_writer = index.writer_for_tests()?;
// writing the segment
index_writer.add_document(doc!(
score_field_f64 => 10.0f64,
text_field => "a"
))?;
index_writer.add_document(doc!(
score_field_f64 => 10.0f64,
text_field => "a"
))?;
index_writer.add_document(doc!(text_field => "a"))?;
index_writer.commit()?;
}
let agg_req: Aggregations = {
serde_json::from_value(json!({
"range_with_stats": {
"terms": {
"field": "texts"
},
"aggs": {
"percentiles": {
"percentiles": {
"field": "score",
"missing": 5.0
}
}
}
}
}))
.unwrap()
};
let res = exec_request_with_query(agg_req, &index, None)?;
assert_eq!(res["range_with_stats"]["buckets"][0]["doc_count"], 3);
assert_eq!(
res["range_with_stats"]["buckets"][0]["percentiles"]["values"]["1.0"],
5.0028295751107414
);
assert_eq!(
res["range_with_stats"]["buckets"][0]["percentiles"]["values"]["99.0"],
10.07469668951144
);
Ok(())
}
#[test]
fn test_percentiles_missing() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("texts", FAST);
let score_field_f64 = schema_builder.add_f64_field("score", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{
let mut index_writer = index.writer_for_tests()?;
// writing the segment
index_writer.add_document(doc!(
score_field_f64 => 10.0f64,
text_field => "a"
))?;
index_writer.add_document(doc!(
score_field_f64 => 10.0f64,
text_field => "a"
))?;
index_writer.add_document(doc!(text_field => "a"))?;
index_writer.commit()?;
}
let agg_req: Aggregations = {
serde_json::from_value(json!({
"percentiles": {
"percentiles": {
"field": "score",
"missing": 5.0
}
}
}))
.unwrap()
};
let res = exec_request_with_query(agg_req, &index, None)?;
assert_eq!(res["percentiles"]["values"]["1.0"], 5.0028295751107414);
assert_eq!(res["percentiles"]["values"]["99.0"], 10.07469668951144);
Ok(())
}
}

367
src/aggregation/metric/stats.rs
View File

@@ -3,13 +3,13 @@ use serde::{Deserialize, Serialize};
use super::*;
use crate::aggregation::agg_req_with_accessor::{
AggregationsWithAccessor, MetricAggregationWithAccessor,
AggregationWithAccessor, AggregationsWithAccessor,
};
use crate::aggregation::intermediate_agg_result::{
IntermediateAggregationResults, IntermediateMetricResult,
IntermediateAggregationResult, IntermediateAggregationResults, IntermediateMetricResult,
};
use crate::aggregation::segment_agg_result::SegmentAggregationCollector;
use crate::aggregation::{f64_from_fastfield_u64, VecWithNames};
use crate::aggregation::{f64_from_fastfield_u64, f64_to_fastfield_u64};
use crate::{DocId, TantivyError};
/// A multi-value metric aggregation that computes a collection of statistics on numeric values that
@@ -29,12 +29,21 @@ use crate::{DocId, TantivyError};
pub struct StatsAggregation {
/// The field name to compute the stats on.
pub field: String,
/// The missing parameter defines how documents that are missing a value should be treated.
/// By default they will be ignored but it is also possible to treat them as if they had a
/// value. Examples in JSON format:
/// { "field": "my_numbers", "missing": "10.0" }
#[serde(default)]
pub missing: Option<f64>,
}
impl StatsAggregation {
/// Creates a new [`StatsAggregation`] instance from a field name.
pub fn from_field_name(field_name: String) -> Self {
StatsAggregation { field: field_name }
StatsAggregation {
field: field_name,
missing: None,
}
}
/// Returns the field name the aggregation is computed on.
pub fn field_name(&self) -> &str {
@@ -66,8 +75,7 @@ impl Stats {
"max" => Ok(self.max),
"avg" => Ok(self.avg),
_ => Err(TantivyError::InvalidArgument(format!(
"Unknown property {} on stats metric aggregation",
agg_property
"Unknown property {agg_property} on stats metric aggregation"
))),
}
}
@@ -154,6 +162,7 @@ pub(crate) enum SegmentStatsType {
#[derive(Clone, Debug, PartialEq)]
pub(crate) struct SegmentStatsCollector {
missing: Option<u64>,
field_type: ColumnType,
pub(crate) collecting_for: SegmentStatsType,
pub(crate) stats: IntermediateStats,
@@ -166,12 +175,15 @@ impl SegmentStatsCollector {
field_type: ColumnType,
collecting_for: SegmentStatsType,
accessor_idx: usize,
missing: Option<f64>,
) -> Self {
let missing = missing.and_then(|val| f64_to_fastfield_u64(val, &field_type));
Self {
field_type,
collecting_for,
stats: IntermediateStats::default(),
accessor_idx,
missing,
val_cache: Default::default(),
}
}
@@ -179,12 +191,19 @@ impl SegmentStatsCollector {
pub(crate) fn collect_block_with_field(
&mut self,
docs: &[DocId],
agg_accessor: &mut MetricAggregationWithAccessor,
agg_accessor: &mut AggregationWithAccessor,
) {
agg_accessor
.column_block_accessor
.fetch_block(docs, &agg_accessor.accessor);
if let Some(missing) = self.missing.as_ref() {
agg_accessor.column_block_accessor.fetch_block_with_missing(
docs,
&agg_accessor.accessor,
*missing,
);
} else {
agg_accessor
.column_block_accessor
.fetch_block(docs, &agg_accessor.accessor);
}
for val in agg_accessor.column_block_accessor.iter_vals() {
let val1 = f64_from_fastfield_u64(val, &self.field_type);
self.stats.collect(val1);
@@ -194,11 +213,12 @@ impl SegmentStatsCollector {
impl SegmentAggregationCollector for SegmentStatsCollector {
#[inline]
fn into_intermediate_aggregations_result(
fn add_intermediate_aggregation_result(
self: Box<Self>,
agg_with_accessor: &AggregationsWithAccessor,
) -> crate::Result<IntermediateAggregationResults> {
let name = agg_with_accessor.metrics.keys[self.accessor_idx].to_string();
results: &mut IntermediateAggregationResults,
) -> crate::Result<()> {
let name = agg_with_accessor.aggs.keys[self.accessor_idx].to_string();
let intermediate_metric_result = match self.collecting_for {
SegmentStatsType::Average => {
@@ -219,15 +239,12 @@ impl SegmentAggregationCollector for SegmentStatsCollector {
}
};
let metrics = Some(VecWithNames::from_entries(vec![(
results.push(
name,
intermediate_metric_result,
)]));
IntermediateAggregationResult::Metric(intermediate_metric_result),
)?;
Ok(IntermediateAggregationResults {
metrics,
buckets: None,
})
Ok(())
}
#[inline]
@@ -236,11 +253,23 @@ impl SegmentAggregationCollector for SegmentStatsCollector {
doc: crate::DocId,
agg_with_accessor: &mut AggregationsWithAccessor,
) -> crate::Result<()> {
let field = &agg_with_accessor.metrics.values[self.accessor_idx].accessor;
for val in field.values_for_doc(doc) {
let val1 = f64_from_fastfield_u64(val, &self.field_type);
self.stats.collect(val1);
let field = &agg_with_accessor.aggs.values[self.accessor_idx].accessor;
if let Some(missing) = self.missing {
let mut has_val = false;
for val in field.values_for_doc(doc) {
let val1 = f64_from_fastfield_u64(val, &self.field_type);
self.stats.collect(val1);
has_val = true;
}
if !has_val {
self.stats
.collect(f64_from_fastfield_u64(missing, &self.field_type));
}
} else {
for val in field.values_for_doc(doc) {
let val1 = f64_from_fastfield_u64(val, &self.field_type);
self.stats.collect(val1);
}
}
Ok(())
@@ -252,7 +281,7 @@ impl SegmentAggregationCollector for SegmentStatsCollector {
docs: &[crate::DocId],
agg_with_accessor: &mut AggregationsWithAccessor,
) -> crate::Result<()> {
let field = &mut agg_with_accessor.metrics.values[self.accessor_idx];
let field = &mut agg_with_accessor.aggs.values[self.accessor_idx];
self.collect_block_with_field(docs, field);
Ok(())
}
@@ -261,21 +290,17 @@ impl SegmentAggregationCollector for SegmentStatsCollector {
#[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_req::{Aggregation, Aggregations};
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::tests::{
exec_request_with_query, 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;
use crate::schema::{IndexRecordOption, Schema, FAST};
use crate::{Index, Term};
#[test]
fn test_aggregation_stats_empty_index() -> crate::Result<()> {
@@ -284,14 +309,14 @@ mod tests {
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 agg_req_1: Aggregations = serde_json::from_value(json!({
"stats": {
"stats": {
"field": "score",
},
}
}))
.unwrap();
let collector = AggregationCollector::from_aggs(agg_req_1, Default::default());
@@ -316,19 +341,18 @@ mod tests {
#[test]
fn test_aggregation_stats_simple() -> crate::Result<()> {
// test index without segments
let values = vec![10.0];
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 agg_req_1: Aggregations = serde_json::from_value(json!({
"stats": {
"stats": {
"field": "score",
},
}
}))
.unwrap();
let collector = AggregationCollector::from_aggs(agg_req_1, Default::default());
@@ -363,52 +387,42 @@ mod tests {
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(),
],
..Default::default()
}),
sub_aggregation: iter::once((
"stats".to_string(),
Aggregation::Metric(MetricAggregation::Stats(
StatsAggregation::from_field_name("score".to_string()),
)),
))
.collect(),
let range_agg: Aggregation = {
serde_json::from_value(json!({
"range": {
"field": "score",
"ranges": [ { "from": 3.0f64, "to": 7.0f64 }, { "from": 7.0f64, "to": 19.0f64 }, { "from": 19.0f64, "to": 20.0f64 } ]
},
"aggs": {
"stats": {
"stats": {
"field": "score"
}
}
.into(),
),
),
]
.into_iter()
.collect();
}
}))
.unwrap()
};
let agg_req_1: Aggregations = serde_json::from_value(json!({
"stats_i64": {
"stats": {
"field": "score_i64",
},
},
"stats_f64": {
"stats": {
"field": "score_f64",
},
},
"stats": {
"stats": {
"field": "score",
},
},
"range": range_agg
}))
.unwrap();
let collector = AggregationCollector::from_aggs(agg_req_1, Default::default());
@@ -473,4 +487,159 @@ mod tests {
Ok(())
}
#[test]
fn test_stats_json() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let json = schema_builder.add_json_field("json", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
// => Segment with empty json
index_writer.add_document(doc!()).unwrap();
index_writer.commit().unwrap();
// => Segment with json, but no field partially_empty
index_writer
.add_document(doc!(json => json!({"different_field": "blue"})))
.unwrap();
index_writer.commit().unwrap();
//// => Segment with field partially_empty
index_writer
.add_document(doc!(json => json!({"partially_empty": 10.0})))
.unwrap();
index_writer.add_document(doc!())?;
index_writer.commit().unwrap();
let agg_req: Aggregations = serde_json::from_value(json!({
"my_stats": {
"stats": {
"field": "json.partially_empty"
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
assert_eq!(
res["my_stats"],
json!({
"avg": 10.0,
"count": 1,
"max": 10.0,
"min": 10.0,
"sum": 10.0
})
);
Ok(())
}
#[test]
fn test_stats_json_missing() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let json = schema_builder.add_json_field("json", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_for_tests().unwrap();
// => Segment with empty json
index_writer.add_document(doc!()).unwrap();
index_writer.commit().unwrap();
// => Segment with json, but no field partially_empty
index_writer
.add_document(doc!(json => json!({"different_field": "blue"})))
.unwrap();
index_writer.commit().unwrap();
//// => Segment with field partially_empty
index_writer
.add_document(doc!(json => json!({"partially_empty": 10.0})))
.unwrap();
index_writer.add_document(doc!())?;
index_writer.commit().unwrap();
let agg_req: Aggregations = serde_json::from_value(json!({
"my_stats": {
"stats": {
"field": "json.partially_empty",
"missing": 0.0
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
assert_eq!(
res["my_stats"],
json!({
"avg": 2.5,
"count": 4,
"max": 10.0,
"min": 0.0,
"sum": 10.0
})
);
Ok(())
}
#[test]
fn test_stats_json_missing_sub_agg() -> crate::Result<()> {
// This test verifies the `collect` method (in contrast to `collect_block`), which is
// called when the sub-aggregations are flushed.
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("texts", FAST);
let score_field_f64 = schema_builder.add_f64_field("score", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{
let mut index_writer = index.writer_for_tests()?;
// writing the segment
index_writer.add_document(doc!(
score_field_f64 => 10.0f64,
text_field => "a"
))?;
index_writer.add_document(doc!(text_field => "a"))?;
index_writer.commit()?;
}
let agg_req: Aggregations = {
serde_json::from_value(json!({
"range_with_stats": {
"terms": {
"field": "texts"
},
"aggs": {
"my_stats": {
"stats": {
"field": "score",
"missing": 0.0
}
}
}
}
}))
.unwrap()
};
let res = exec_request_with_query(agg_req, &index, None)?;
assert_eq!(
res["range_with_stats"]["buckets"][0]["my_stats"]["count"],
2
);
assert_eq!(
res["range_with_stats"]["buckets"][0]["my_stats"]["min"],
0.0
);
assert_eq!(
res["range_with_stats"]["buckets"][0]["my_stats"]["avg"],
5.0
);
Ok(())
}
}

11
src/aggregation/metric/sum.rs
View File

@@ -20,12 +20,21 @@ use super::{IntermediateStats, SegmentStatsCollector};
pub struct SumAggregation {
/// The field name to compute the minimum on.
pub field: String,
/// The missing parameter defines how documents that are missing a value should be treated.
/// By default they will be ignored but it is also possible to treat them as if they had a
/// value. Examples in JSON format:
/// { "field": "my_numbers", "missing": "10.0" }
#[serde(default)]
pub missing: Option<f64>,
}
impl SumAggregation {
/// Creates a new [`SumAggregation`] instance from a field name.
pub fn from_field_name(field_name: String) -> Self {
Self { field: field_name }
Self {
field: field_name,
missing: None,
}
}
/// Returns the field name the aggregation is computed on.
pub fn field_name(&self) -> &str {

118
src/aggregation/mod.rs
View File

@@ -24,6 +24,9 @@
//! ## JSON Format
//! Aggregations request and result structures de/serialize into elasticsearch compatible JSON.
//!
//! Notice: Intermediate aggregation results should not be de/serialized via JSON format.
//! Postcard is a good choice.
//!
//! ```verbatim
//! let agg_req: Aggregations = serde_json::from_str(json_request_string).unwrap();
//! let collector = AggregationCollector::from_aggs(agg_req, None);
@@ -35,6 +38,7 @@
//! ## Supported Aggregations
//! - [Bucket](bucket)
//! - [Histogram](bucket::HistogramAggregation)
//! - [DateHistogram](bucket::DateHistogramAggregationReq)
//! - [Range](bucket::RangeAggregation)
//! - [Terms](bucket::TermsAggregation)
//! - [Metric](metric)
@@ -44,39 +48,12 @@
//! - [Max](metric::MaxAggregation)
//! - [Sum](metric::SumAggregation)
//! - [Count](metric::CountAggregation)
//! - [Percentiles](metric::PercentilesAggregationReq)
//!
//! # Example
//! Compute the average metric, by building [`agg_req::Aggregations`], which is built from an
//! `(String, agg_req::Aggregation)` iterator.
//!
//! ```
//! use tantivy::aggregation::agg_req::{Aggregations, Aggregation, MetricAggregation};
//! use tantivy::aggregation::AggregationCollector;
//! use tantivy::aggregation::metric::AverageAggregation;
//! use tantivy::query::AllQuery;
//! use tantivy::aggregation::agg_result::AggregationResults;
//! use tantivy::IndexReader;
//!
//! # #[allow(dead_code)]
//! fn aggregate_on_index(reader: &IndexReader) {
//! let agg_req: Aggregations = vec![
//! (
//! "average".to_string(),
//! Aggregation::Metric(MetricAggregation::Average(
//! AverageAggregation::from_field_name("score".to_string()),
//! )),
//! ),
//! ]
//! .into_iter()
//! .collect();
//!
//! let collector = AggregationCollector::from_aggs(agg_req, Default::default());
//!
//! let searcher = reader.searcher();
//! let agg_res: AggregationResults = searcher.search(&AllQuery, &collector).unwrap();
//! }
//! ```
//! # Example JSON
//! Requests are compatible with the elasticsearch JSON request format.
//!
//! ```
@@ -116,32 +93,24 @@
//! aggregation and then calculate the average on each bucket.
//! ```
//! use tantivy::aggregation::agg_req::*;
//! use tantivy::aggregation::metric::AverageAggregation;
//! use tantivy::aggregation::bucket::RangeAggregation;
//! let sub_agg_req_1: Aggregations = vec![(
//! "average_in_range".to_string(),
//! Aggregation::Metric(MetricAggregation::Average(
//! AverageAggregation::from_field_name("score".to_string()),
//! )),
//! )]
//! .into_iter()
//! .collect();
//! use serde_json::json;
//!
//! let agg_req_1: Aggregations = vec![
//! (
//! "range".to_string(),
//! Aggregation::Bucket(Box::new(BucketAggregation {
//! bucket_agg: BucketAggregationType::Range(RangeAggregation{
//! field: "score".to_string(),
//! ranges: vec![(3f64..7f64).into(), (7f64..20f64).into()],
//! keyed: false,
//! }),
//! sub_aggregation: sub_agg_req_1.clone(),
//! })),
//! ),
//! ]
//! .into_iter()
//! .collect();
//! let agg_req_1: Aggregations = serde_json::from_value(json!({
//! "rangef64": {
//! "range": {
//! "field": "score",
//! "ranges": [
//! { "from": 3, "to": 7000 },
//! { "from": 7000, "to": 20000 },
//! { "from": 50000, "to": 60000 }
//! ]
//! },
//! "aggs": {
//! "average_in_range": { "avg": { "field": "score" } }
//! }
//! },
//! }))
//! .unwrap();
//! ```
//!
//! # Distributed Aggregation
@@ -153,7 +122,7 @@
//! [`merge_fruits`](intermediate_agg_result::IntermediateAggregationResults::merge_fruits) method
//! to merge multiple results. The merged result can then be converted into
//! [`AggregationResults`](agg_result::AggregationResults) via the
//! [`into_final_bucket_result`](intermediate_agg_result::IntermediateAggregationResults::into_final_bucket_result) method.
//! [`into_final_result`](intermediate_agg_result::IntermediateAggregationResults::into_final_result) method.
mod agg_limits;
pub mod agg_req;
@@ -174,6 +143,8 @@ use std::fmt::Display;
#[cfg(test)]
mod agg_tests;
mod agg_bench;
pub use agg_limits::AggregationLimits;
pub use collector::{
AggregationCollector, AggregationSegmentCollector, DistributedAggregationCollector,
@@ -186,13 +157,22 @@ use itertools::Itertools;
use serde::{Deserialize, Serialize};
/// Represents an associative array `(key => values)` in a very efficient manner.
#[derive(Clone, PartialEq, Serialize, Deserialize)]
pub(crate) struct VecWithNames<T: Clone> {
#[derive(PartialEq, Serialize, Deserialize)]
pub(crate) struct VecWithNames<T> {
pub(crate) values: Vec<T>,
keys: Vec<String>,
}
impl<T: Clone> Default for VecWithNames<T> {
impl<T: Clone> Clone for VecWithNames<T> {
fn clone(&self) -> Self {
Self {
values: self.values.clone(),
keys: self.keys.clone(),
}
}
}
impl<T> Default for VecWithNames<T> {
fn default() -> Self {
Self {
values: Default::default(),
@@ -201,24 +181,19 @@ impl<T: Clone> Default for VecWithNames<T> {
}
}
impl<T: Clone + std::fmt::Debug> std::fmt::Debug for VecWithNames<T> {
impl<T: std::fmt::Debug> std::fmt::Debug for VecWithNames<T> {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
f.debug_map().entries(self.iter()).finish()
}
}
impl<T: Clone> From<HashMap<String, T>> for VecWithNames<T> {
impl<T> From<HashMap<String, T>> for VecWithNames<T> {
fn from(map: HashMap<String, T>) -> Self {
VecWithNames::from_entries(map.into_iter().collect_vec())
}
}
impl<T: Clone> VecWithNames<T> {
fn extend(&mut self, entries: VecWithNames<T>) {
self.keys.extend(entries.keys);
self.values.extend(entries.values);
}
impl<T> VecWithNames<T> {
fn from_entries(mut entries: Vec<(String, T)>) -> Self {
// Sort to ensure order of elements match across multiple instances
entries.sort_by(|left, right| left.0.cmp(&right.0));
@@ -233,21 +208,12 @@ impl<T: Clone> VecWithNames<T> {
keys: data_names,
}
}
fn into_iter(self) -> impl Iterator<Item = (String, T)> {
self.keys.into_iter().zip(self.values.into_iter())
}
fn iter(&self) -> impl Iterator<Item = (&str, &T)> + '_ {
self.keys().zip(self.values.iter())
}
fn keys(&self) -> impl Iterator<Item = &str> + '_ {
self.keys.iter().map(|key| key.as_str())
}
fn into_values(self) -> impl Iterator<Item = T> {
self.values.into_iter()
}
fn values(&self) -> impl Iterator<Item = &T> + '_ {
self.values.iter()
}
fn values_mut(&mut self) -> impl Iterator<Item = &mut T> + '_ {
self.values.iter_mut()
}
@@ -316,7 +282,7 @@ pub(crate) fn f64_from_fastfield_u64(val: u64, field_type: &ColumnType) -> f64 {
ColumnType::I64 | ColumnType::DateTime => i64::from_u64(val) as f64,
ColumnType::F64 => f64::from_u64(val),
_ => {
panic!("unexpected type {:?}. This should not happen", field_type)
panic!("unexpected type {field_type:?}. This should not happen")
}
}
}
@@ -461,7 +427,7 @@ mod tests {
let index = Index::create_in_ram(schema_builder.build());
{
// let mut index_writer = index.writer_for_tests()?;
let mut index_writer = index.writer_with_num_threads(1, 30_000_000)?;
let mut index_writer = index.writer_with_num_threads(1, 20_000_000)?;
index_writer.set_merge_policy(Box::new(NoMergePolicy));
for values in segment_and_values {
for (i, term) in values {

271
src/aggregation/segment_agg_result.rs
View File

@@ -6,24 +6,23 @@
use std::fmt::Debug;
pub(crate) use super::agg_limits::AggregationLimits;
use super::agg_req::MetricAggregation;
use super::agg_req_with_accessor::{
AggregationsWithAccessor, BucketAggregationWithAccessor, MetricAggregationWithAccessor,
};
use super::agg_req::AggregationVariants;
use super::agg_req_with_accessor::{AggregationWithAccessor, AggregationsWithAccessor};
use super::bucket::{SegmentHistogramCollector, SegmentRangeCollector, SegmentTermCollector};
use super::intermediate_agg_result::IntermediateAggregationResults;
use super::metric::{
AverageAggregation, CountAggregation, MaxAggregation, MinAggregation, SegmentStatsCollector,
SegmentStatsType, StatsAggregation, SumAggregation,
AverageAggregation, CountAggregation, MaxAggregation, MinAggregation,
SegmentPercentilesCollector, SegmentStatsCollector, SegmentStatsType, StatsAggregation,
SumAggregation,
};
use super::VecWithNames;
use crate::aggregation::agg_req::BucketAggregationType;
use crate::aggregation::bucket::TermMissingAgg;
pub(crate) trait SegmentAggregationCollector: CollectorClone + Debug {
fn into_intermediate_aggregations_result(
fn add_intermediate_aggregation_result(
self: Box<Self>,
agg_with_accessor: &AggregationsWithAccessor,
) -> crate::Result<IntermediateAggregationResults>;
results: &mut IntermediateAggregationResults,
) -> crate::Result<()>;
fn collect(
&mut self,
@@ -63,92 +62,104 @@ impl Clone for Box<dyn SegmentAggregationCollector> {
}
pub(crate) fn build_segment_agg_collector(
req: &AggregationsWithAccessor,
req: &mut AggregationsWithAccessor,
) -> crate::Result<Box<dyn SegmentAggregationCollector>> {
// Single metric special case
if req.buckets.is_empty() && req.metrics.len() == 1 {
let req = &req.metrics.values[0];
// Single collector special case
if req.aggs.len() == 1 {
let req = &mut req.aggs.values[0];
let accessor_idx = 0;
return build_metric_segment_agg_collector(req, accessor_idx);
}
// Single bucket special case
if req.metrics.is_empty() && req.buckets.len() == 1 {
let req = &req.buckets.values[0];
let accessor_idx = 0;
return build_bucket_segment_agg_collector(req, accessor_idx);
return build_single_agg_segment_collector(req, accessor_idx);
}
let agg = GenericSegmentAggregationResultsCollector::from_req_and_validate(req)?;
Ok(Box::new(agg))
}
pub(crate) fn build_metric_segment_agg_collector(
req: &MetricAggregationWithAccessor,
pub(crate) fn build_single_agg_segment_collector(
req: &mut AggregationWithAccessor,
accessor_idx: usize,
) -> crate::Result<Box<dyn SegmentAggregationCollector>> {
let stats_collector = match &req.metric {
MetricAggregation::Average(AverageAggregation { .. }) => {
SegmentStatsCollector::from_req(req.field_type, SegmentStatsType::Average, accessor_idx)
use AggregationVariants::*;
match &req.agg.agg {
Terms(terms_req) => {
if req.accessors.is_empty() {
Ok(Box::new(SegmentTermCollector::from_req_and_validate(
terms_req,
&mut req.sub_aggregation,
req.field_type,
accessor_idx,
)?))
} else {
Ok(Box::new(TermMissingAgg::new(
accessor_idx,
&mut req.sub_aggregation,
)?))
}
}
MetricAggregation::Count(CountAggregation { .. }) => {
SegmentStatsCollector::from_req(req.field_type, SegmentStatsType::Count, accessor_idx)
Range(range_req) => Ok(Box::new(SegmentRangeCollector::from_req_and_validate(
range_req,
&mut req.sub_aggregation,
&req.limits,
req.field_type,
accessor_idx,
)?)),
Histogram(histogram) => Ok(Box::new(SegmentHistogramCollector::from_req_and_validate(
histogram.clone(),
&mut req.sub_aggregation,
req.field_type,
accessor_idx,
)?)),
DateHistogram(histogram) => Ok(Box::new(SegmentHistogramCollector::from_req_and_validate(
histogram.to_histogram_req()?,
&mut req.sub_aggregation,
req.field_type,
accessor_idx,
)?)),
Average(AverageAggregation { missing, .. }) => {
Ok(Box::new(SegmentStatsCollector::from_req(
req.field_type,
SegmentStatsType::Average,
accessor_idx,
*missing,
)))
}
MetricAggregation::Max(MaxAggregation { .. }) => {
SegmentStatsCollector::from_req(req.field_type, SegmentStatsType::Max, accessor_idx)
}
MetricAggregation::Min(MinAggregation { .. }) => {
SegmentStatsCollector::from_req(req.field_type, SegmentStatsType::Min, accessor_idx)
}
MetricAggregation::Stats(StatsAggregation { .. }) => {
SegmentStatsCollector::from_req(req.field_type, SegmentStatsType::Stats, accessor_idx)
}
MetricAggregation::Sum(SumAggregation { .. }) => {
SegmentStatsCollector::from_req(req.field_type, SegmentStatsType::Sum, accessor_idx)
}
};
Ok(Box::new(stats_collector))
}
pub(crate) fn build_bucket_segment_agg_collector(
req: &BucketAggregationWithAccessor,
accessor_idx: usize,
) -> crate::Result<Box<dyn SegmentAggregationCollector>> {
match &req.bucket_agg {
BucketAggregationType::Terms(terms_req) => {
Ok(Box::new(SegmentTermCollector::from_req_and_validate(
terms_req,
&req.sub_aggregation,
Count(CountAggregation { missing, .. }) => Ok(Box::new(SegmentStatsCollector::from_req(
req.field_type,
SegmentStatsType::Count,
accessor_idx,
*missing,
))),
Max(MaxAggregation { missing, .. }) => Ok(Box::new(SegmentStatsCollector::from_req(
req.field_type,
SegmentStatsType::Max,
accessor_idx,
*missing,
))),
Min(MinAggregation { missing, .. }) => Ok(Box::new(SegmentStatsCollector::from_req(
req.field_type,
SegmentStatsType::Min,
accessor_idx,
*missing,
))),
Stats(StatsAggregation { missing, .. }) => Ok(Box::new(SegmentStatsCollector::from_req(
req.field_type,
SegmentStatsType::Stats,
accessor_idx,
*missing,
))),
Sum(SumAggregation { missing, .. }) => Ok(Box::new(SegmentStatsCollector::from_req(
req.field_type,
SegmentStatsType::Sum,
accessor_idx,
*missing,
))),
Percentiles(percentiles_req) => Ok(Box::new(
SegmentPercentilesCollector::from_req_and_validate(
percentiles_req,
req.field_type,
accessor_idx,
)?))
}
BucketAggregationType::Range(range_req) => {
Ok(Box::new(SegmentRangeCollector::from_req_and_validate(
range_req,
&req.sub_aggregation,
&req.limits,
req.field_type,
accessor_idx,
)?))
}
BucketAggregationType::Histogram(histogram) => {
Ok(Box::new(SegmentHistogramCollector::from_req_and_validate(
histogram,
&req.sub_aggregation,
req.field_type,
accessor_idx,
)?))
}
BucketAggregationType::DateHistogram(histogram) => {
Ok(Box::new(SegmentHistogramCollector::from_req_and_validate(
&histogram.to_histogram_req()?,
&req.sub_aggregation,
req.field_type,
accessor_idx,
)?))
}
)?,
)),
}
}
@@ -157,50 +168,28 @@ pub(crate) fn build_bucket_segment_agg_collector(
/// can handle arbitrary complexity of sub-aggregations. Ideally we never have to pick this one
/// and can provide specialized versions instead, that remove some of its overhead.
pub(crate) struct GenericSegmentAggregationResultsCollector {
pub(crate) metrics: Option<Vec<Box<dyn SegmentAggregationCollector>>>,
pub(crate) buckets: Option<Vec<Box<dyn SegmentAggregationCollector>>>,
pub(crate) aggs: Vec<Box<dyn SegmentAggregationCollector>>,
}
impl Debug for GenericSegmentAggregationResultsCollector {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
f.debug_struct("SegmentAggregationResultsCollector")
.field("metrics", &self.metrics)
.field("buckets", &self.buckets)
.field("aggs", &self.aggs)
.finish()
}
}
impl SegmentAggregationCollector for GenericSegmentAggregationResultsCollector {
fn into_intermediate_aggregations_result(
fn add_intermediate_aggregation_result(
self: Box<Self>,
agg_with_accessor: &AggregationsWithAccessor,
) -> crate::Result<IntermediateAggregationResults> {
let buckets = if let Some(buckets) = self.buckets {
let mut intermeditate_buckets = VecWithNames::default();
for bucket in buckets {
// TODO too many allocations?
let res = bucket.into_intermediate_aggregations_result(agg_with_accessor)?;
// unwrap is fine since we only have buckets here
intermeditate_buckets.extend(res.buckets.unwrap());
}
Some(intermeditate_buckets)
} else {
None
};
let metrics = if let Some(metrics) = self.metrics {
let mut intermeditate_metrics = VecWithNames::default();
for metric in metrics {
// TODO too many allocations?
let res = metric.into_intermediate_aggregations_result(agg_with_accessor)?;
// unwrap is fine since we only have metrics here
intermeditate_metrics.extend(res.metrics.unwrap());
}
Some(intermeditate_metrics)
} else {
None
};
results: &mut IntermediateAggregationResults,
) -> crate::Result<()> {
for agg in self.aggs {
agg.add_intermediate_aggregation_result(agg_with_accessor, results)?;
}
Ok(IntermediateAggregationResults { metrics, buckets })
Ok(())
}
fn collect(
@@ -218,66 +207,30 @@ impl SegmentAggregationCollector for GenericSegmentAggregationResultsCollector {
docs: &[crate::DocId],
agg_with_accessor: &mut AggregationsWithAccessor,
) -> crate::Result<()> {
if let Some(metrics) = self.metrics.as_mut() {
for collector in metrics {
collector.collect_block(docs, agg_with_accessor)?;
}
}
if let Some(buckets) = self.buckets.as_mut() {
for collector in buckets {
collector.collect_block(docs, agg_with_accessor)?;
}
for collector in &mut self.aggs {
collector.collect_block(docs, agg_with_accessor)?;
}
Ok(())
}
fn flush(&mut self, agg_with_accessor: &mut AggregationsWithAccessor) -> crate::Result<()> {
if let Some(metrics) = &mut self.metrics {
for collector in metrics {
collector.flush(agg_with_accessor)?;
}
}
if let Some(buckets) = &mut self.buckets {
for collector in buckets {
collector.flush(agg_with_accessor)?;
}
for collector in &mut self.aggs {
collector.flush(agg_with_accessor)?;
}
Ok(())
}
}
impl GenericSegmentAggregationResultsCollector {
pub(crate) fn from_req_and_validate(req: &AggregationsWithAccessor) -> crate::Result<Self> {
let buckets = req
.buckets
.iter()
pub(crate) fn from_req_and_validate(req: &mut AggregationsWithAccessor) -> crate::Result<Self> {
let aggs = req
.aggs
.values_mut()
.enumerate()
.map(|(accessor_idx, (_key, req))| {
build_bucket_segment_agg_collector(req, accessor_idx)
})
.collect::<crate::Result<Vec<Box<dyn SegmentAggregationCollector>>>>()?;
let metrics = req
.metrics
.iter()
.enumerate()
.map(|(accessor_idx, (_key, req))| {
build_metric_segment_agg_collector(req, accessor_idx)
})
.map(|(accessor_idx, req)| build_single_agg_segment_collector(req, accessor_idx))
.collect::<crate::Result<Vec<Box<dyn SegmentAggregationCollector>>>>()?;
let metrics = if metrics.is_empty() {
None
} else {
Some(metrics)
};
let buckets = if buckets.is_empty() {
None
} else {
Some(buckets)
};
Ok(GenericSegmentAggregationResultsCollector { metrics, buckets })
Ok(GenericSegmentAggregationResultsCollector { aggs })
}
}

2
src/collector/count_collector.rs
View File

@@ -16,7 +16,7 @@ use crate::{DocId, Score, SegmentOrdinal, SegmentReader};
/// let schema = schema_builder.build();
/// let index = Index::create_in_ram(schema);
///
/// let mut index_writer = index.writer(3_000_000).unwrap();
/// let mut index_writer = index.writer(15_000_000).unwrap();
/// index_writer.add_document(doc!(title => "The Name of the Wind")).unwrap();
/// index_writer.add_document(doc!(title => "The Diary of Muadib")).unwrap();
/// index_writer.add_document(doc!(title => "A Dairy Cow")).unwrap();

37
src/collector/facet_collector.rs
View File

@@ -89,7 +89,7 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
/// let schema = schema_builder.build();
/// let index = Index::create_in_ram(schema);
/// {
/// let mut index_writer = index.writer(3_000_000)?;
/// let mut index_writer = index.writer(15_000_000)?;
/// // a document can be associated with any number of facets
/// index_writer.add_document(doc!(
/// title => "The Name of the Wind",
@@ -161,6 +161,21 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
/// ]);
/// }
///
/// {
/// let mut facet_collector = FacetCollector::for_field("facet");
/// facet_collector.add_facet("/");
/// let facet_counts = searcher.search(&AllQuery, &facet_collector)?;
///
/// // This lists all of the facet counts
/// let facets: Vec<(&Facet, u64)> = facet_counts
/// .get("/")
/// .collect();
/// assert_eq!(facets, vec![
/// (&Facet::from("/category"), 4),
/// (&Facet::from("/lang"), 4)
/// ]);
/// }
///
/// Ok(())
/// }
/// # assert!(example().is_ok());
@@ -285,6 +300,9 @@ fn is_child_facet(parent_facet: &[u8], possible_child_facet: &[u8]) -> bool {
if !possible_child_facet.starts_with(parent_facet) {
return false;
}
if parent_facet.is_empty() {
return true;
}
possible_child_facet.get(parent_facet.len()).copied() == Some(0u8)
}
@@ -414,8 +432,8 @@ impl FacetCounts {
pub fn get<T>(&self, facet_from: T) -> FacetChildIterator<'_>
where Facet: From<T> {
let facet = Facet::from(facet_from);
let left_bound = Bound::Excluded(facet.clone());
let right_bound = if facet.is_root() {
let lower_bound = Bound::Excluded(facet.clone());
let upper_bound = if facet.is_root() {
Bound::Unbounded
} else {
let mut facet_after_bytes: String = facet.encoded_str().to_owned();
@@ -424,7 +442,7 @@ impl FacetCounts {
Bound::Excluded(facet_after)
};
let underlying: btree_map::Range<'_, _, _> =
self.facet_counts.range((left_bound, right_bound));
self.facet_counts.range((lower_bound, upper_bound));
FacetChildIterator { underlying }
}
@@ -789,6 +807,15 @@ mod tests {
);
Ok(())
}
#[test]
fn is_child_facet() {
assert!(super::is_child_facet(&b"foo"[..], &b"foo\0bar"[..]));
assert!(super::is_child_facet(&b""[..], &b"foo\0bar"[..]));
assert!(super::is_child_facet(&b""[..], &b"foo"[..]));
assert!(!super::is_child_facet(&b"foo\0bar"[..], &b"foo"[..]));
assert!(!super::is_child_facet(&b"foo"[..], &b"foobar\0baz"[..]));
}
}
#[cfg(all(test, feature = "unstable"))]
@@ -812,7 +839,7 @@ mod bench {
let mut docs = vec![];
for val in 0..50 {
let facet = Facet::from(&format!("/facet_{}", val));
let facet = Facet::from(&format!("/facet_{val}"));
for _ in 0..val * val {
docs.push(doc!(facet_field=>facet.clone()));
}

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