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:0.20.2
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:experimental-do-not-merge
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

155 Commits
0.20.2 ... experiment

Author SHA1 Message Date
Paul Masurel
4e64801ec9 Experimental do not merge 2024-04-19 18:32:03 +09:00
Paul Masurel
34cb0c48bf Owned value relying on Vec instead of BTreeMap 2024-04-19 15:08:50 +09:00
PSeitz
74940e9345 clippy (#2349) 
* fix clippy

* fix clippy

* fix duplicate imports
 2024-04-09 07:54:44 +02:00
PSeitz
1e9fc51535 update ahash (#2344) 2024-04-09 06:35:39 +02:00
PSeitz
92c32979d2 fix postcard compatibility for top_hits, add postcard test (#2346) 
* fix postcard compatibility for top_hits, add postcard test

* fix top_hits naming, delay data fetch

closes #2347

* fix import
 2024-04-09 06:17:25 +02:00
PSeitz
b644d78a32 fix null byte handling in JSON paths (#2345) 
* fix null byte handling in JSON paths

closes https://github.com/quickwit-oss/tantivy/issues/2193
closes https://github.com/quickwit-oss/tantivy/issues/2340

* avoid repeated term truncation

* fix test

* Apply suggestions from code review

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

* add comment

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2024-04-05 09:53:35 +02:00
PSeitz
4e79e11007 add collect_block to BoxableSegmentCollector (#2331) 2024-03-21 09:10:25 +01:00
PSeitz
67ebba3c3c expose collect_block buffer size (#2326) 
* expose buffer of collect_block

* flip shard_size segment_size
 2024-03-15 08:02:08 +01:00
PSeitz
7ce950f141 add method to fetch block of first vals in columnar (#2330) 
* add method to fetch block of first vals in columnar

add method to fetch block of first vals in columnar (this is way faster
than single calls for full columns)
add benchmark
fix import warnings

```
test bench_get_block_first_on_full_column                  ... bench:          56 ns/iter (+/- 26)
test bench_get_block_first_on_full_column_single_calls     ... bench:         311 ns/iter (+/- 6)
test bench_get_block_first_on_multi_column                 ... bench:         378 ns/iter (+/- 15)
test bench_get_block_first_on_multi_column_single_calls    ... bench:         546 ns/iter (+/- 13)
test bench_get_block_first_on_optional_column              ... bench:         291 ns/iter (+/- 6)
test bench_get_block_first_on_optional_column_single_calls ... bench:         362 ns/iter (+/- 8)
```

* use remainder
 2024-03-15 08:01:47 +01:00
dependabot[bot]
0cffe5fb09 Update base64 requirement from 0.21.0 to 0.22.0 (#2324) 
Updates the requirements on [base64](https://github.com/marshallpierce/rust-base64) to permit the latest version.
- [Changelog](https://github.com/marshallpierce/rust-base64/blob/master/RELEASE-NOTES.md)
- [Commits](https://github.com/marshallpierce/rust-base64/compare/v0.21.0...v0.22.0)

---
updated-dependencies:
- dependency-name: base64
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2024-03-15 15:50:34 +09:00
PSeitz
b0e65560a1 handle ip adresses in term aggregation (#2319) 
* handle ip adresses in term aggregation

Stores IpAdresses during the segment term aggregation via u64 representation
and convert to u128(IpV6Adress) via downcast when converting to intermediate results.

Enable Downcasting on `ColumnValues`
Expose u64 variant for u128 encoded data via `open_u64_lenient` method.
Remove lifetime in VecColumn, to avoid 'static lifetime requirement coming
from downcast trait.

* rename method
 2024-03-14 09:41:18 +01:00
PSeitz
ec37295b2f add fast path for full columns in fetch_block (#2328) 
Spotted in `range_date_histogram` query in quickwit benchmark:
5% of time copying docs around, which is not needed in the full index case

remove Column to ColumnIndex deref
 2024-03-14 04:07:11 +01:00
trinity-1686a
f6b0cc1aab allow some mixing of occur and bool in strict query parser (#2323) 
* allow some mixing of occur and bool in strict query parser

* allow all mixing of binary and occur in strict parser
 2024-03-07 15:17:48 +01:00
PSeitz
7e41d31c6e agg: support to deserialize f64 from string (#2311) 
* agg: support to deserialize f64 from string

* remove visit_string

* disallow NaN
 2024-03-05 05:49:41 +01:00
Adam Reichold
40aa4abfe5 Make FacetCounts defaultable and cloneable. (#2322) 2024-03-05 04:11:11 +01:00
dependabot[bot]
2650317622 Update fs4 requirement from 0.7.0 to 0.8.0 (#2321) 
Updates the requirements on [fs4](https://github.com/al8n/fs4-rs) to permit the latest version.
- [Release notes](https://github.com/al8n/fs4-rs/releases)
- [Commits](https://github.com/al8n/fs4-rs/commits)

---
updated-dependencies:
- dependency-name: fs4
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2024-02-27 03:38:04 +01:00
Paul Masurel
6739357314 Removing split_size and adding split_size and shard_size as segmnet_size (#2320) 
aliases.
 2024-02-26 11:35:22 +01:00
PSeitz
d57622d54b support bool type in term aggregation (#2318) 
* support bool type in term aggregation

* add Bool to Intermediate Key
 2024-02-20 03:22:22 +01:00
PSeitz
f745dbc054 fix Clone for TopNComputer, add top_hits bench (#2315) 
* fix Clone for TopNComputer, add top_hits bench

add top_hits agg bench

test aggregation::agg_bench::bench::bench_aggregation_terms_many_with_sub_agg                                            ... bench: 123,475,175 ns/iter (+/- 30,608,889)
test aggregation::agg_bench::bench::bench_aggregation_terms_many_with_sub_agg_multi                                      ... bench: 194,170,414 ns/iter (+/- 36,495,516)
test aggregation::agg_bench::bench::bench_aggregation_terms_many_with_sub_agg_opt                                        ... bench: 179,742,809 ns/iter (+/- 29,976,507)
test aggregation::agg_bench::bench::bench_aggregation_terms_many_with_sub_agg_sparse                                     ... bench:  27,592,534 ns/iter (+/- 2,672,370)
test aggregation::agg_bench::bench::bench_aggregation_terms_many_with_top_hits_agg                                       ... bench: 552,851,227 ns/iter (+/- 71,975,886)
test aggregation::agg_bench::bench::bench_aggregation_terms_many_with_top_hits_agg_multi                                 ... bench: 558,616,384 ns/iter (+/- 100,890,124)
test aggregation::agg_bench::bench::bench_aggregation_terms_many_with_top_hits_agg_opt                                   ... bench: 554,031,368 ns/iter (+/- 165,452,650)
test aggregation::agg_bench::bench::bench_aggregation_terms_many_with_top_hits_agg_sparse                                ... bench:  46,435,919 ns/iter (+/- 13,681,935)

* add comment
 2024-02-20 03:22:00 +01:00
PSeitz
79b041f81f clippy (#2314) 2024-02-13 05:56:31 +01:00
PSeitz
0e16ed9ef7 Fix serde for TopNComputer (#2313) 
* Fix serde for TopNComputer

The top hits aggregation changed the TopNComputer to be serializable,
but capacity needs to be carried over, as it contains logic which is
checked against when pushing elements (capacity == 0 is not allowed).

* use serde from deser

* remove pub, clippy
 2024-02-07 12:52:06 +01:00
mochi
88a3275dbb add shared search executor (#2312) 2024-02-05 09:33:00 +01:00
PSeitz
1223a87eb2 add fuzz test for hashmap (#2310) 2024-01-31 10:30:21 +01:00
PSeitz
48630ceec9 move into new index module (#2259) 
move core modules to index module
 2024-01-31 10:30:04 +01:00
Adam Reichold
72002e8a89 Make test builds Clippy clean. (#2277) 2024-01-31 02:47:06 +01:00
trinity-1686a
3c9297dd64 report if posting list was actually loaded when warming it up (#2309) 2024-01-29 15:23:16 +01:00
Tushar
0e04ec3136 feat(aggregators/metric): Add a top_hits aggregator (#2198) 
* feat(aggregators/metric): Implement a top_hits aggregator

* fix: Expose get_fields

* fix: Serializer for top_hits request

Also removes extraneous the extraneous third-party
serialization helper.

* chore: Avert panick on parsing invalid top_hits query

* refactor: Allow multiple field names from aggregations

* perf: Replace binary heap with TopNComputer

* fix: Avoid comparator inversion by ComparableDoc

* fix: Rank missing field values lower than present values

* refactor: Make KeyOrder a struct

* feat: Rough attempt at docvalue_fields

* feat: Complete stab at docvalue_fields

- Rename "SearchResult*" => "Retrieval*"
- Revert Vec => HashMap for aggregation accessors.
- Split accessors for core aggregation and field retrieval.
- Resolve globbed field names in docvalue_fields retrieval.
- Handle strings/bytes and other column types with DynamicColumn

* test(unit): Add tests for top_hits aggregator

* fix: docfield_value field globbing

* test(unit): Include dynamic fields

* fix: Value -> OwnedValue

* fix: Use OwnedValue's native Null variant

* chore: Improve readability of test asserts

* chore: Remove DocAddress from top_hits result

* docs: Update aggregator doc

* revert: accidental doc test

* chore: enable time macros only for tests

* chore: Apply suggestions from review

* chore: Apply suggestions from review

* fix: Retrieve all values for fields

* test(unit): Update for multi-value retrieval

* chore: Assert term existence

* feat: Include all columns for a column name

Since a (name, type) constitutes a unique column.

* fix: Resolve json fields

Introduces a translation step to bridge the difference between
ColumnarReaders null `\0` separated json field keys to the common
`.` separated used by SegmentReader. Although, this should probably
be the default behavior for ColumnarReader's public API perhaps.

* chore: Address review on mutability

* chore: s/segment_id/segment_ordinal instances of SegmentOrdinal

* chore: Revert erroneous grammar change
 2024-01-26 16:46:41 +01:00
Paul Masurel
9b7f3a55cf Bumped census version 2024-01-26 19:32:02 +09:00
PSeitz
1dacdb6c85 add histogram agg test on empty index (#2306) 2024-01-23 16:27:34 +01:00
François Massot
30483310ca Minor improvement of README.md (#2305) 
* Update README.md

* Remove useless paragraph

* Wording.
 2024-01-19 17:46:48 +09:00
Tushar
e1d18b5114 chore: Expose TopDocs::order_by_u64_field again (#2282) 2024-01-18 05:58:24 +01:00
trinity-1686a
108f30ba23 allow newline where we allow space in query parser (#2302) 
fix regression from the new parser
 2024-01-17 14:38:35 +01:00
PSeitz
5943ee46bd Truncate keys to u16::MAX in term hashmap (#2299) 
Truncate keys to u16::MAX, instead e.g. storing 0 bytes for keys with length u16::MAX + 1

The term hashmap has a hidden API contract to only accept terms with lenght up u16::MAX.
 2024-01-11 10:19:12 +01:00
PSeitz
f95a76293f add memory arena test (#2298) 
* add memory arena test

* add assert

* Update stacker/src/memory_arena.rs

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

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2024-01-11 07:18:48 +01:00
Paul Masurel
014328e378 Fix bug that can cause get_docids_for_value_range to panic. (#2295) 
* Fix bug that can cause `get_docids_for_value_range` to panic.

When `selected_docid_range.end == num_rows`, we would get a panic
as we try to access a non-existing blockmeta.

This PR accepts calls to rank with any value.
For any value above num_rows we simply return non_null_rows.

Fixes #2293

* add tests, merge variables

---------

Co-authored-by: Pascal Seitz <pascal.seitz@gmail.com>
 2024-01-09 14:52:20 +01:00
Adam Reichold
53f2fe1fbe Forward regex parser errors to enable understandin their reason. (#2288) 2023-12-22 11:01:10 +01:00
PSeitz
9c75942aaf fix merge panic for JSON fields (#2284) 
Root cause was the positions buffer had residue positions from the
previous term, when the terms were alternating between having and not
having positions in JSON (terms have positions, but not numerics).

Fixes #2283
 2023-12-21 11:05:34 +01:00
PSeitz
bff7c58497 improve indexing benchmark (#2275) 2023-12-11 09:04:42 +01:00
trinity-1686a
9ebc5ed053 use fst for sstable index (#2268) 
* read path for new fst based index

* implement BlockAddrStoreWriter

* extract slop/derivation computation

* use better linear approximator and allow negative correction to approximator

* document format and reorder some fields

* optimize single block sstable size

* plug backward compat
 2023-12-04 15:13:15 +01:00
PSeitz
0b56c88e69 Revert "Preparing for 0.21.2 release." (#2258) 
* Revert "Preparing for 0.21.2 release. (#2256)"

This reverts commit 9caab45136.

* bump version to 0.21.1

* set version to 0.22.0-dev
 2023-12-01 13:46:12 +01:00
PSeitz
24841f0b2a update bitpacker dep (#2269) 2023-12-01 13:45:52 +01:00
PSeitz
1a9fc10be9 add fields_metadata to SegmentReader, add columnar docs (#2222) 
* add fields_metadata to SegmentReader, add columnar docs

* use schema to resolve field, add test

* normalize paths

* merge for FieldsMetadata, add fields_metadata on Index

* Update src/core/segment_reader.rs

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

* merge code paths

* add Hash

* move function oustide

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-11-22 12:29:53 +01:00
PSeitz
07573a7f19 update fst (#2267) 
update fst to 0.5 (deduplicates regex-syntax in the dep tree)
deps cleanup
 2023-11-21 16:06:57 +01:00
BlackHoleFox
daad2dc151 Take string references instead of owned values building Facet paths (#2265) 2023-11-20 09:40:44 +01:00
PSeitz
054f49dc31 support escaped dot, add agg test (#2250) 
add agg test for nested JSON
allow escaping of dot
 2023-11-20 03:00:57 +01:00
PSeitz
47009ed2d3 remove unused deps (#2264) 
found with cargo machete
remove pprof (doesn't work)
 2023-11-20 02:59:59 +01:00
PSeitz
0aae31d7d7 reduce number of allocations (#2257) 
* reduce number of allocations

Explanation makes up around 50% of all allocations (numbers not perf).
It's created during serialization but not called.

- Make Explanation optional in BM25
- Avoid allocations when using Explanation

* use Cow
 2023-11-16 13:47:36 +01:00
Paul Masurel
9caab45136 Preparing for 0.21.2 release. (#2256) 2023-11-15 10:43:36 +09:00
Chris Tam
6d9a7b7eb0 Derive Debug for SchemaBuilder (#2254) 2023-11-15 01:03:44 +01:00
dependabot[bot]
7a2c5804b1 Update itertools requirement from 0.11.0 to 0.12.0 (#2255) 
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/v0.11.0...v0.12.0)

---
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>
 2023-11-15 01:03:08 +01:00
François Massot
5319977171 Merge pull request #2253 from quickwit-oss/issue/2251-bug-merge-json-object-with-number 
Fix bug occuring when merging JSON object indexed with positions.
 2023-11-14 17:28:29 +01:00
trinity-1686a
828632e8c4 rustfmt 2023-11-14 15:05:16 +01:00
Paul Masurel
6b59ec6fd5 Fix bug occuring when merging JSON object indexed with positions. 
In JSON Object field the presence of term frequencies depend on the
field.
Typically, a string with postiions indexed will have positions
while numbers won't.

The presence or absence of term freqs for a given term is unfortunately
encoded in a very passive way.

It is given by the presence of extra information in the skip info, or
the lack of term freqs after decoding vint blocks.

Before, after writing a segment, we would encode the segment correctly
(without any term freq for number in json object field).
However during merge, we would get the default term freq=1 value.
(this is default in the absence of encoded term freqs)

The merger would then proceed and attempt to decode 1 position when
there are in fact none.

This PR requires to explictly tell the posting serialize whether
term frequencies should be serialized for each new term.

Closes #2251
 2023-11-14 22:41:48 +09:00
PSeitz
b60d862150 docid deltas while indexing (#2249) 
* docid deltas while indexing

storing deltas is especially helpful for repetitive data like logs.
In those cases, recording a doc on a term costed 4 bytes instead of 1
byte now.

HDFS Indexing 1.1GB Total memory consumption:
Before:  760 MB
Now:     590 MB

* use scan for delta decoding
 2023-11-13 05:14:27 +01:00
PSeitz
4837c7811a add missing inlines (#2245) 2023-11-10 08:00:42 +01:00
PSeitz
5a2397d57e add sstable ord_to_term benchmark (#2242) 2023-11-10 07:27:48 +01:00
PSeitz
927b4432c9 Perf: use term hashmap in fastfield (#2243) 
* add shared arena hashmap

* bench fastfield indexing

* use shared arena hashmap in columnar

lower minimum resize in hashtable

* clippy

* add comments
 2023-11-09 13:44:02 +01:00
trinity-1686a
7a0064db1f bump index version (#2237) 
* bump index version

and add constant for lowest supported version

* use range instead of handcoded bounds
 2023-11-06 19:02:37 +01:00
PSeitz
2e7327205d fix coverage run (#2232) 
coverage run uses the compare_hash_only feature which is not compativle
with the test_hashmap_size test
 2023-11-06 11:18:38 +00:00
Paul Masurel
7bc5bf78e2 Fixing functional tests. (#2239) 2023-11-05 18:18:39 +09:00
giovannicuccu
ef603c8c7e rename ReloadPolicy onCommit to onCommitWithDelay (#2235) 
* rename ReloadPolicy onCommit to onCommitWithDelay

* fix format issues

---------

Co-authored-by: Giovanni Cuccu <gcuccu@imolainformatica.it>
 2023-11-03 12:22:10 +01:00
PSeitz
28dd6b6546 collect json paths in indexing (#2231) 
* collect json paths in indexing

* remove unsafe iter_mut_keys
 2023-11-01 11:25:17 +01:00
trinity-1686a
1dda2bb537 handle * inside term in query parser (#2228) 2023-10-27 08:57:02 +02:00
PSeitz
bf6544cf28 fix mmap::Advice reexport (#2230) 2023-10-27 14:09:25 +09:00
PSeitz
ccecf946f7 tantivy 0.21.1 (#2227) 2023-10-27 05:01:44 +02:00
PSeitz
19a859d6fd term hashmap remove copy in is_empty, unused unordered_id (#2229) 2023-10-27 05:01:32 +02:00
PSeitz
83af14caa4 Fix range query (#2226) 
Fix range query end check in advance
Rename vars to reduce ambiguity
add tests

Fixes #2225
 2023-10-25 09:17:31 +02:00
PSeitz
4feeb2323d fix clippy (#2223) 2023-10-24 10:05:22 +02:00
PSeitz
07bf66a197 json path writer (#2224) 
* refactor logic to JsonPathWriter

* use in encode_column_name

* add inlines

* move unsafe block
 2023-10-24 09:45:50 +02:00
trinity-1686a
0d4589219b encode some part of posting list as -1 instead of direct values (#2185) 
* add support for delta-1 encoding posting list

* encode term frequency minus one

* don't emit tf for json integer terms

* make skipreader not pub(crate) mutable
 2023-10-20 16:58:26 +02:00
PSeitz
c2b0469180 improve docs, rework exports (#2220) 
* rework exports

move snippet and advice
make indexer pub, remove indexer reexports

* add deprecation warning

* add architecture overview
 2023-10-18 09:22:24 +02:00
PSeitz
7e1980b218 run coverage only after merge (#2212) 
* run coverage only after merge

coverage is a quite slow step in CI. It can be run only after merging

* Apply suggestions from code review

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

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2023-10-18 07:19:36 +02:00
PSeitz
ecb9a89a9f add compat mode for JSON (#2219) 2023-10-17 10:00:55 +02:00
PSeitz
5e06e504e6 split into ReferenceValueLeaf (#2217) 2023-10-16 16:31:30 +02:00
PSeitz
182f58cea6 remove Document: DocumentDeserialize dependency (#2211) 
* remove Document: DocumentDeserialize dependency

The dependency requires users to implement an API they may not use.

* remove unnecessary Document bounds
 2023-10-13 07:59:54 +02:00
dependabot[bot]
337ffadefd Update lru requirement from 0.11.0 to 0.12.0 (#2208) 
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/0.11.0...0.12.0)

---
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>
 2023-10-12 12:09:56 +02:00
dependabot[bot]
22aa4daf19 Update zstd requirement from 0.12 to 0.13 (#2214) 
Updates the requirements on [zstd](https://github.com/gyscos/zstd-rs) to permit the latest version.
- [Release notes](https://github.com/gyscos/zstd-rs/releases)
- [Commits](https://github.com/gyscos/zstd-rs/compare/v0.12.0...v0.13.0)

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

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2023-10-12 04:24:44 +02:00
PSeitz
493f9b2f2a Read list of JSON fields encoded in dictionary (#2184) 
* Read list of JSON fields encoded in dictionary

add method to get list of fields on InvertedIndexReader

* add field type
 2023-10-09 12:06:22 +02:00
PSeitz
e246e5765d replace ReferenceValue with Self in Value (#2210) 2023-10-06 08:22:15 +02:00
PSeitz
6097235eff fix numeric order, refactor Document (#2209) 
fix numeric order to prefer i64
rename and move Document stuff
 2023-10-05 16:39:56 +02:00
PSeitz
b700c42246 add AsRef, expose object and array iter on Value (#2207) 
add AsRef
expose object and array iter
add to_json on Document
 2023-10-05 03:55:35 +02:00
PSeitz
5b1bf1a993 replace Field with field name (#2196) 2023-10-04 06:21:40 +02:00
PSeitz
041d4fced7 move to_named_doc to Document trait (#2205) 2023-10-04 06:03:07 +02:00
dependabot[bot]
166fc15239 Update memmap2 requirement from 0.7.1 to 0.9.0 (#2204) 
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/v0.7.1...v0.9.0)

---
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>
 2023-10-04 05:00:46 +02:00
PSeitz
514a6e7fef fix bench compile, fix Document reexport (#2203) 2023-10-03 17:28:36 +02:00
dependabot[bot]
82d9127191 Update fs4 requirement from 0.6.3 to 0.7.0 (#2199) 
Updates the requirements on [fs4](https://github.com/al8n/fs4-rs) to permit the latest version.
- [Release notes](https://github.com/al8n/fs4-rs/releases)
- [Commits](https://github.com/al8n/fs4-rs/commits/0.7.0)

---
updated-dependencies:
- dependency-name: fs4
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2023-10-03 04:43:09 +02:00
PSeitz
03a1f40767 rename DocValue to Value (#2197) 
rename DocValue to Value to avoid confusion with lucene DocValues
rename Value to OwnedValue
 2023-10-02 17:03:00 +02:00
Harrison Burt
1c7c6fd591 POC: Tantivy documents as a trait (#2071) 
* fix windows build (#1)

* Fix windows build

* Add doc traits

* Add field value iter

* Add value and serialization

* Adjust order

* Fix bug

* Correct type

* Fix generic bugs

* Reformat code

* Add generic to index writer which I forgot about

* Fix missing generics on single segment writer

* Add missing type export

* Add default methods for convenience

* Cleanup

* Fix more-like-this query to use standard types

* Update API and fix tests

* Add doc traits

* Add field value iter

* Add value and serialization

* Adjust order

* Fix bug

* Correct type

* Rebase main and fix conflicts

* Reformat code

* Merge upstream

* Fix missing generics on single segment writer

* Add missing type export

* Add default methods for convenience

* Cleanup

* Fix more-like-this query to use standard types

* Update API and fix tests

* Add tokenizer improvements from previous commits

* Add tokenizer improvements from previous commits

* Reformat

* Fix unit tests

* Fix unit tests

* Use enum in changes

* Stage changes

* Add new deserializer logic

* Add serializer integration

* Add document deserializer

* Implement new (de)serialization api for existing types

* Fix bugs and type errors

* Add helper implementations

* Fix errors

* Reformat code

* Add unit tests and some code organisation for serialization

* Add unit tests to deserializer

* Add some small docs

* Add support for deserializing serde values

* Reformat

* Fix typo

* Fix typo

* Change repr of facet

* Remove unused trait methods

* Add child value type

* Resolve comments

* Fix build

* Fix more build errors

* Fix more build errors

* Fix the tests I missed

* Fix examples

* fix numerical order, serialize PreTok Str

* fix coverage

* rename Document to TantivyDocument, rename DocumentAccess to Document

add Binary prefix to binary de/serialization

* fix coverage

---------

Co-authored-by: Pascal Seitz <pascal.seitz@gmail.com>
 2023-10-02 10:01:16 +02:00
PSeitz
b525f653c0 replace BinaryHeap for TopN (#2186) 
* replace BinaryHeap for TopN

replace BinaryHeap for TopN with variant that selects the median with QuickSort,
which runs in O(n) time.

add merge_fruits fast path

* call truncate unconditionally, extend test

* remove special early exit

* add TODO, fmt

* truncate top n instead median, return vec

* simplify code
 2023-09-27 09:25:30 +02:00
ethever.eth
90586bc1e2 chore: remove unused Seek impl for Writers (#2187) (#2189) 
Co-authored-by: famouscat <onismaa@gmail.com>
 2023-09-26 17:03:28 +09:00
PSeitz
832f1633de handle exclusive out of bounds ranges on fastfield range queries (#2174) 
closes https://github.com/quickwit-oss/quickwit/issues/3790
 2023-09-26 08:00:40 +02:00
PSeitz
38db53c465 make column_index pub (#2181) 2023-09-22 08:06:45 +02:00
PSeitz
34920d31f5 Fix DateHistogram bucket gap (#2183) 
* Fix DateHistogram bucket gap

Fixes a computation issue of the number of buckets needed in the
DateHistogram.

This is due to a missing normalization from request values (ms) to fast field
values (ns), when converting an intermediate result to the final result.
This results in a wrong computation by a factor 1_000_000.
The Histogram normalizes values to nanoseconds, to make the user input like
extended_bounds (ms precision) and the values from the fast field (ns precision for date type) compatible.
This normalization happens only for date type fields, as other field types don't have precision settings.
The normalization does not happen due a missing `column_type`, which is not
correctly passed after merging an empty aggregation (which does not have a `column_type` set), with a regular aggregation.

Another related issue is an empty aggregation, which will not have
`column_type` set, will not convert the result to human readable format.

This PR fixes the issue by:
- Limit the allowed field types of DateHistogram to DateType
- Instead of passing the column_type, which is only available on the segment level, we flag the aggregation as `is_date_agg`.
- Fix the merge logic

Add a flag to to normalization only once. This is not an issue
currently, but it could become easily one.

closes https://github.com/quickwit-oss/quickwit/issues/3837

* use older nightly for time crate (breaks build)
 2023-09-21 10:41:35 +02:00
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
286 changed files with 17233 additions and 5125 deletions
Expand all files Collapse all files

8
.github/workflows/coverage.yml vendored
View File

@@ -3,8 +3,6 @@ name: Coverage
on:
push:
branches: [main]
pull_request:
branches: [main]
# Ensures that we cancel running jobs for the same PR / same workflow.
concurrency:
@@ -15,13 +13,13 @@ 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
run: rustup toolchain install nightly-2023-09-10 --profile minimal --component llvm-tools-preview
- uses: Swatinem/rust-cache@v2
- uses: taiki-e/install-action@cargo-llvm-cov
- name: Generate code coverage
run: cargo +nightly llvm-cov --all-features --workspace --doctests --lcov --output-path lcov.info
run: cargo +nightly-2023-09-10 llvm-cov --all-features --workspace --doctests --lcov --output-path lcov.info
- name: Upload coverage to Codecov
uses: codecov/codecov-action@v3
continue-on-error: true

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

@@ -19,7 +19,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v3
- uses: actions/checkout@v4
- name: Install stable
uses: actions-rs/toolchain@v1
with:

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

@@ -20,7 +20,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v3
- uses: actions/checkout@v4
- name: Install nightly
uses: actions-rs/toolchain@v1
@@ -39,6 +39,13 @@ jobs:
- name: Check Formatting
run: cargo +nightly fmt --all -- --check
- name: Check Stable Compilation
run: cargo build --all-features
- name: Check Bench Compilation
run: cargo +nightly bench --no-run --profile=dev --all-features
- uses: actions-rs/clippy-check@v1
with:
@@ -53,14 +60,14 @@ 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

46
CHANGELOG.md
View File

@@ -1,5 +1,47 @@
Tantivy 0.21.1
================================
#### Bugfixes
- Range queries on fast fields with less values on that field than documents had an invalid end condition, leading to missing results. [#2226](https://github.com/quickwit-oss/tantivy/issues/2226)(@appaquet @PSeitz)
- Increase the minimum memory budget from 3MB to 15MB to avoid single doc segments (API fix). [#2176](https://github.com/quickwit-oss/tantivy/issues/2176)(@PSeitz)
Tantivy 0.20 [Unreleased]
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)
@@ -38,12 +80,14 @@ Tantivy 0.20 [Unreleased]
- 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)

66
Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy"
version = "0.20.2"
version = "0.22.0-dev"
authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT"
categories = ["database-implementations", "data-structures"]
@@ -11,60 +11,57 @@ repository = "https://github.com/quickwit-oss/tantivy"
readme = "README.md"
keywords = ["search", "information", "retrieval"]
edition = "2021"
rust-version = "1.62"
rust-version = "1.63"
exclude = ["benches/*.json", "benches/*.txt"]
[dependencies]
oneshot = "0.1.5"
base64 = "0.21.0"
base64 = "0.22.0"
byteorder = "1.4.3"
crc32fast = "1.3.2"
once_cell = "1.10.0"
regex = { version = "1.5.5", default-features = false, features = ["std", "unicode"] }
aho-corasick = "1.0"
tantivy-fst = "0.4.0"
memmap2 = { version = "0.6.0", optional = true }
lz4_flex = { version = "0.10", default-features = false, features = ["checked-decode"], optional = true }
brotli = { version = "3.3.4", optional = true }
zstd = { version = "0.12", optional = true, default-features = false }
snap = { version = "1.0.5", optional = true }
tantivy-fst = "0.5"
memmap2 = { version = "0.9.0", optional = true }
lz4_flex = { version = "0.11", default-features = false, optional = true }
zstd = { version = "0.13", optional = true, default-features = false }
tempfile = { version = "3.3.0", optional = true }
log = "0.4.16"
serde = { version = "1.0.136", features = ["derive"] }
serde_json = "1.0.79"
num_cpus = "1.13.1"
fs4 = { version = "0.6.3", optional = true }
fs4 = { version = "0.8.0", optional = true }
levenshtein_automata = "0.2.1"
uuid = { version = "1.0.0", features = ["v4", "serde"] }
crossbeam-channel = "0.5.4"
rust-stemmers = "1.2.0"
downcast-rs = "1.2.0"
bitpacking = { version = "0.8.4", default-features = false, features = ["bitpacker4x"] }
census = "0.4.0"
bitpacking = { version = "0.9.2", default-features = false, features = ["bitpacker4x"] }
census = "0.4.2"
rustc-hash = "1.1.0"
thiserror = "1.0.30"
htmlescape = "0.3.1"
fail = "0.5.0"
murmurhash32 = "0.3.0"
fail = { version = "0.5.0", optional = true }
time = { version = "0.3.10", features = ["serde-well-known"] }
smallvec = "1.8.0"
rayon = "1.5.2"
lru = "0.10.0"
lru = "0.12.0"
fastdivide = "0.4.0"
itertools = "0.10.3"
itertools = "0.12.0"
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= "0.20.0", path="./query-grammar", package = "tantivy-query-grammar" }
tantivy-bitpacker = { version= "0.4", 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 }
fnv = "1.0.7"
[target.'cfg(windows)'.dependencies]
winapi = "0.3.9"
@@ -75,14 +72,18 @@ maplit = "1.0.2"
matches = "0.1.9"
pretty_assertions = "1.2.1"
proptest = "1.0.0"
criterion = "0.5"
test-log = "0.2.10"
env_logger = "0.10.0"
pprof = { git = "https://github.com/PSeitz/pprof-rs/", rev = "53af24b", features = ["flamegraph", "criterion"] } # temp fork that works with criterion 0.5
futures = "0.3.21"
paste = "1.0.11"
more-asserts = "0.3.1"
rand_distr = "0.4.3"
time = { version = "0.3.10", features = ["serde-well-known", "macros"] }
postcard = { version = "1.0.4", features = [
"use-std",
], default-features = false }
[target.'cfg(not(windows))'.dev-dependencies]
criterion = { version = "0.5", default-features = false }
[dev-dependencies.fail]
version = "0.5.0"
@@ -107,16 +108,19 @@ 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", "futures-util"]
# Compares only the hash of a string when indexing data.
# Increases indexing speed, but may lead to extremely rare missing terms, when there's a hash collision.
# Uses 64bit ahash.
compare_hash_only = ["stacker/compare_hash_only"]
[workspace]
members = ["query-grammar", "bitpacker", "common", "ownedbytes", "stacker", "sstable", "tokenizer-api", "columnar"]
@@ -130,7 +134,7 @@ 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"

35
README.md
View File

@@ -5,19 +5,18 @@
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
[![Crates.io](https://img.shields.io/crates/v/tantivy.svg)](https://crates.io/crates/tantivy)
![Tantivy](https://tantivy-search.github.io/logo/tantivy-logo.png)
<img src="https://tantivy-search.github.io/logo/tantivy-logo.png" alt="Tantivy, the fastest full-text search engine library written in Rust" height="250">
**Tantivy** is a **full-text search engine library** written in Rust.
## Fast full-text search engine library written in Rust
It is closer to [Apache Lucene](https://lucene.apache.org/) than to [Elasticsearch](https://www.elastic.co/products/elasticsearch) or [Apache Solr](https://lucene.apache.org/solr/) in the sense it is not
an off-the-shelf search engine server, but rather a crate that can be used
to build such a search engine.
**If you are looking for an alternative to Elasticsearch or Apache Solr, check out [Quickwit](https://github.com/quickwit-oss/quickwit), our distributed search engine built on top of Tantivy.**
Tantivy is closer to [Apache Lucene](https://lucene.apache.org/) than to [Elasticsearch](https://www.elastic.co/products/elasticsearch) or [Apache Solr](https://lucene.apache.org/solr/) in the sense it is not
an off-the-shelf search engine server, but rather a crate that can be used to build such a search engine.
Tantivy is, in fact, strongly inspired by Lucene's design.
If you are looking for an alternative to Elasticsearch or Apache Solr, check out [Quickwit](https://github.com/quickwit-oss/quickwit), our search engine built on top of Tantivy.
# Benchmark
## Benchmark
The following [benchmark](https://tantivy-search.github.io/bench/) breakdowns
performance for different types of queries/collections.
@@ -28,7 +27,7 @@ Your mileage WILL vary depending on the nature of queries and their load.
Details about the benchmark can be found at this [repository](https://github.com/quickwit-oss/search-benchmark-game).
# Features
## Features
- Full-text search
- Configurable tokenizer (stemming available for 17 Latin languages) with third party support for Chinese ([tantivy-jieba](https://crates.io/crates/tantivy-jieba) and [cang-jie](https://crates.io/crates/cang-jie)), Japanese ([lindera](https://github.com/lindera-morphology/lindera-tantivy), [Vaporetto](https://crates.io/crates/vaporetto_tantivy), and [tantivy-tokenizer-tiny-segmenter](https://crates.io/crates/tantivy-tokenizer-tiny-segmenter)) and Korean ([lindera](https://github.com/lindera-morphology/lindera-tantivy) + [lindera-ko-dic-builder](https://github.com/lindera-morphology/lindera-ko-dic-builder))
@@ -44,7 +43,7 @@ Details about the benchmark can be found at this [repository](https://github.com
- Single valued and multivalued u64, i64, and f64 fast fields (equivalent of doc values in Lucene)
- `&[u8]` fast fields
- Text, i64, u64, f64, dates, ip, bool, and hierarchical facet fields
- Compressed document store (LZ4, Zstd, None, Brotli, Snap)
- Compressed document store (LZ4, Zstd, None)
- Range queries
- Faceted search
- Configurable indexing (optional term frequency and position indexing)
@@ -54,11 +53,11 @@ Details about the benchmark can be found at this [repository](https://github.com
- Searcher Warmer API
- Cheesy logo with a horse
## Non-features
### Non-features
Distributed search is out of the scope of Tantivy, but if you are looking for this feature, check out [Quickwit](https://github.com/quickwit-oss/quickwit/).
# Getting started
## Getting started
Tantivy works on stable Rust and supports Linux, macOS, and Windows.
@@ -68,7 +67,7 @@ index documents, and search via the CLI or a small server with a REST API.
It walks you through getting a Wikipedia search engine up and running in a few minutes.
- [Reference doc for the last released version](https://docs.rs/tantivy/)
# How can I support this project?
## How can I support this project?
There are many ways to support this project.
@@ -79,16 +78,16 @@ There are many ways to support this project.
- Contribute code (you can join [our Discord server](https://discord.gg/MT27AG5EVE))
- Talk about Tantivy around you
# Contributing code
## Contributing code
We use the GitHub Pull Request workflow: reference a GitHub ticket and/or include a comprehensive commit message when opening a PR.
Feel free to update CHANGELOG.md with your contribution.
## Tokenizer
### Tokenizer
When implementing a tokenizer for tantivy depend on the `tantivy-tokenizer-api` crate.
## Clone and build locally
### Clone and build locally
Tantivy compiles on stable Rust.
To check out and run tests, you can simply run:
@@ -99,7 +98,7 @@ cd tantivy
cargo test
```
# Companies Using Tantivy
## Companies Using Tantivy
<p align="left">
<img align="center" src="doc/assets/images/etsy.png" alt="Etsy" height="25" width="auto" />&nbsp;
@@ -111,7 +110,7 @@ cargo test
<img align="center" src="doc/assets/images/element-dark-theme.png#gh-dark-mode-only" alt="Element.io" height="25" width="auto" />
</p>
# FAQ
## FAQ
### Can I use Tantivy in other languages?

25
benches/analyzer.rs
View File

@@ -1,5 +1,7 @@
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");
@@ -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);

262
benches/index-bench.rs
View File

@@ -1,14 +1,99 @@
use criterion::{criterion_group, criterion_main, Criterion, Throughput};
use pprof::criterion::{Output, PProfProfiler};
use tantivy::schema::{FAST, INDEXED, STORED, STRING, TEXT};
use tantivy::Index;
use criterion::{criterion_group, criterion_main, BatchSize, Bencher, Criterion, Throughput};
use tantivy::schema::{TantivyDocument, FAST, INDEXED, STORED, STRING, TEXT};
use tantivy::{tokenizer, Index, IndexWriter};
const HDFS_LOGS: &str = include_str!("hdfs.json");
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()
fn benchmark(
b: &mut Bencher,
input: &str,
schema: tantivy::schema::Schema,
commit: bool,
parse_json: bool,
is_dynamic: bool,
) {
if is_dynamic {
benchmark_dynamic_json(b, input, schema, commit, parse_json)
} else {
_benchmark(b, input, schema, commit, parse_json, |schema, doc_json| {
TantivyDocument::parse_json(&schema, doc_json).unwrap()
})
}
}
fn get_index(schema: tantivy::schema::Schema) -> Index {
let mut index = Index::create_in_ram(schema.clone());
let ff_tokenizer_manager = tokenizer::TokenizerManager::default();
ff_tokenizer_manager.register(
"raw",
tokenizer::TextAnalyzer::builder(tokenizer::RawTokenizer::default())
.filter(tokenizer::RemoveLongFilter::limit(255))
.build(),
);
index.set_fast_field_tokenizers(ff_tokenizer_manager.clone());
index
}
fn _benchmark(
b: &mut Bencher,
input: &str,
schema: tantivy::schema::Schema,
commit: bool,
include_json_parsing: bool,
create_doc: impl Fn(&tantivy::schema::Schema, &str) -> TantivyDocument,
) {
if include_json_parsing {
let lines: Vec<&str> = input.trim().split('\n').collect();
b.iter(|| {
let index = get_index(schema.clone());
let mut index_writer: IndexWriter =
index.writer_with_num_threads(1, 100_000_000).unwrap();
for doc_json in &lines {
let doc = create_doc(&schema, doc_json);
index_writer.add_document(doc).unwrap();
}
if commit {
index_writer.commit().unwrap();
}
})
} else {
let docs: Vec<_> = input
.trim()
.split('\n')
.map(|doc_json| create_doc(&schema, doc_json))
.collect();
b.iter_batched(
|| docs.clone(),
|docs| {
let index = get_index(schema.clone());
let mut index_writer: IndexWriter =
index.writer_with_num_threads(1, 100_000_000).unwrap();
for doc in docs {
index_writer.add_document(doc).unwrap();
}
if commit {
index_writer.commit().unwrap();
}
},
BatchSize::SmallInput,
)
}
}
fn benchmark_dynamic_json(
b: &mut Bencher,
input: &str,
schema: tantivy::schema::Schema,
commit: bool,
parse_json: bool,
) {
let json_field = schema.get_field("json").unwrap();
_benchmark(b, input, schema, commit, parse_json, |_schema, doc_json| {
let json_val: serde_json::Map<String, serde_json::Value> =
serde_json::from_str(doc_json).unwrap();
tantivy::doc!(json_field=>json_val)
})
}
pub fn hdfs_index_benchmark(c: &mut Criterion) {
@@ -19,7 +104,14 @@ pub fn hdfs_index_benchmark(c: &mut Criterion) {
schema_builder.add_text_field("severity", STRING);
schema_builder.build()
};
let schema_with_store = {
let schema_only_fast = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_u64_field("timestamp", FAST);
schema_builder.add_text_field("body", FAST);
schema_builder.add_text_field("severity", FAST);
schema_builder.build()
};
let _schema_with_store = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_u64_field("timestamp", INDEXED | STORED);
schema_builder.add_text_field("body", TEXT | STORED);
@@ -28,74 +120,39 @@ pub fn hdfs_index_benchmark(c: &mut Criterion) {
};
let dynamic_schema = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_json_field("json", TEXT);
schema_builder.add_json_field("json", TEXT | FAST);
schema_builder.build()
};
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 doc_json in &lines {
let doc = schema.parse_document(doc_json).unwrap();
index_writer.add_document(doc).unwrap();
let benches = [
("only-indexed-".to_string(), schema, false),
//("stored-".to_string(), _schema_with_store, false),
("only-fast-".to_string(), schema_only_fast, false),
("dynamic-".to_string(), dynamic_schema, true),
];
for (prefix, schema, is_dynamic) in benches {
for commit in [false, true] {
let suffix = if commit { "with-commit" } else { "no-commit" };
for parse_json in [false] {
// for parse_json in [false, true] {
let suffix = if parse_json {
format!("{}-with-json-parsing", suffix)
} else {
format!("{}", suffix)
};
let bench_name = format!("{}{}", prefix, suffix);
group.bench_function(bench_name, |b| {
benchmark(b, HDFS_LOGS, schema.clone(), commit, parse_json, is_dynamic)
});
}
})
});
group.bench_function("index-hdfs-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 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 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 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 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 gh_index_benchmark(c: &mut Criterion) {
@@ -104,38 +161,24 @@ pub fn gh_index_benchmark(c: &mut Criterion) {
schema_builder.add_json_field("json", TEXT | FAST);
schema_builder.build()
};
let dynamic_schema_fast = {
let mut schema_builder = tantivy::schema::SchemaBuilder::new();
schema_builder.add_json_field("json", FAST);
schema_builder.build()
};
let mut group = c.benchmark_group("index-gh");
group.throughput(Throughput::Bytes(GH_LOGS.len() as u64));
group.bench_function("index-gh-no-commit", |b| {
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 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();
}
})
benchmark_dynamic_json(b, GH_LOGS, dynamic_schema.clone(), false, false)
});
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 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-gh-fast", |b| {
benchmark_dynamic_json(b, GH_LOGS, dynamic_schema_fast.clone(), false, false)
});
group.bench_function("index-gh-fast-with-commit", |b| {
benchmark_dynamic_json(b, GH_LOGS, dynamic_schema_fast.clone(), true, false)
});
}
@@ -150,33 +193,10 @@ pub fn wiki_index_benchmark(c: &mut Criterion) {
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();
}
})
benchmark_dynamic_json(b, WIKI, dynamic_schema.clone(), false, false)
});
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();
})
benchmark_dynamic_json(b, WIKI, dynamic_schema.clone(), true, false)
});
}
@@ -187,12 +207,12 @@ criterion_group! {
}
criterion_group! {
name = gh_benches;
config = Criterion::default().with_profiler(PProfProfiler::new(100, Output::Flamegraph(None)));
config = Criterion::default();
targets = gh_index_benchmark
}
criterion_group! {
name = wiki_benches;
config = Criterion::default().with_profiler(PProfProfiler::new(100, Output::Flamegraph(None)));
config = Criterion::default();
targets = wiki_index_benchmark
}
criterion_main!(benches, gh_benches, wiki_benches);

4
bitpacker/Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy-bitpacker"
version = "0.4.0"
version = "0.5.0"
edition = "2021"
authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT"
@@ -15,7 +15,7 @@ homepage = "https://github.com/quickwit-oss/tantivy"
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
[dependencies]
bitpacking = {version="0.8", default-features=false, features = ["bitpacker1x"]}
bitpacking = { version = "0.9.2", default-features = false, features = ["bitpacker1x"] }
[dev-dependencies]
rand = "0.8"

3
bitpacker/src/bitpacker.rs
View File

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

4
bitpacker/src/blocked_bitpacker.rs
View File

@@ -64,10 +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/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;

1
cliff.toml
View File

@@ -32,6 +32,7 @@ postprocessors = [
{ 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
]

16
columnar/Cargo.toml
View File

@@ -1,23 +1,23 @@
[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"
desciption = "column oriented storage for tantivy"
description = "column oriented storage for tantivy"
categories = ["database-implementations", "data-structures", "compression"]
[dependencies]
itertools = "0.10.5"
fnv = "1.0.7"
itertools = "0.12.0"
fastdivide = "0.4.0"
stacker = { version= "0.1", path = "../stacker", package="tantivy-stacker"}
sstable = { version= "0.1", path = "../sstable", package = "tantivy-sstable" }
common = { version= "0.5", path = "../common", package = "tantivy-common" }
tantivy-bitpacker = { version= "0.4", 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"
downcast-rs = "1.2.0"
[dev-dependencies]
proptest = "1"

155
columnar/benches/bench_first_vals.rs Normal file
View File

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

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

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

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

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

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

148
columnar/src/block_accessor.rs
View File

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

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)
}

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

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

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

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

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

@@ -168,8 +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]);
}
@@ -200,8 +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]);
}

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

@@ -140,7 +140,7 @@ mod tests {
#[test]
fn test_merge_column_index_optional_shuffle() {
let optional_index: ColumnIndex = OptionalIndex::for_test(2, &[0]).into();
let column_indexes = vec![optional_index, ColumnIndex::Full];
let column_indexes = [optional_index, ColumnIndex::Full];
let row_addrs = vec![
RowAddr {
segment_ord: 0u32,
@@ -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]);

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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))
}
}

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

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

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

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

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

@@ -63,7 +63,6 @@ impl ColumnValues for BitpackedReader {
fn get_val(&self, doc: u32) -> u64 {
self.stats.min_value + self.stats.gcd.get() * self.bit_unpacker.get(doc, &self.data)
}
#[inline]
fn min_value(&self) -> u64 {
self.stats.min_value
@@ -83,7 +82,8 @@ 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;

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

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

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

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

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

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

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

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

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

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

4
columnar/src/columnar/merge/merge_mapping.rs
View File

@@ -52,8 +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),

186
columnar/src/columnar/merge/mod.rs
View File

@@ -2,7 +2,7 @@ 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;
@@ -18,7 +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.
@@ -28,14 +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 {
@@ -83,9 +86,20 @@ pub fn merge_columnar(
.iter()
.map(|reader| reader.num_rows())
.collect::<Vec<u32>>();
let columns_to_merge =
group_columns_for_merge(columnar_readers, required_columns, &merge_row_order)?;
for ((column_name, column_type), columns) in columns_to_merge {
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.start_serialize_column(column_name.as_bytes(), column_type);
merge_column(
@@ -97,6 +111,7 @@ pub fn merge_columnar(
)?;
column_serializer.finalize()?;
}
serializer.finalize(merge_row_order.num_rows())?;
Ok(())
}
@@ -210,40 +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.
@@ -265,11 +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
@@ -293,7 +345,7 @@ fn merged_numerical_columns_type<'a>(
fn is_empty_after_merge(
merge_row_order: &MergeRowOrder,
column: &DynamicColumn,
columnar_id: usize,
columnar_ord: usize,
) -> bool {
if column.num_values() == 0u32 {
// It was empty before the merge.
@@ -305,7 +357,7 @@ fn is_empty_after_merge(
false
}
MergeRowOrder::Shuffled(shuffled) => {
if let Some(alive_bitset) = &shuffled.alive_bitsets[columnar_id] {
if let Some(alive_bitset) = &shuffled.alive_bitsets[columnar_ord] {
let column_index = column.column_index();
match column_index {
ColumnIndex::Empty { .. } => true,
@@ -348,56 +400,34 @@ fn is_empty_after_merge(
}
}
#[allow(clippy::type_complexity)]
fn group_columns_for_merge(
columnar_readers: &[&ColumnarReader],
required_columns: &[(String, ColumnType)],
merge_row_order: &MergeRowOrder,
) -> 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();
/// 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()?;
// 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.
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()?;
if is_empty_after_merge(merge_row_order, &column, columnar_id) {
continue;
}
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(

116
columnar/src/columnar/merge/tests.rs
View File

@@ -1,5 +1,3 @@
use itertools::Itertools;
use super::*;
use crate::{Cardinality, ColumnarWriter, HasAssociatedColumnType, RowId};
@@ -27,22 +25,10 @@ fn test_column_coercion_to_u64() {
let columnar2 = make_columnar("numbers", &[u64::MAX]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
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 columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(columnars, &[], &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]
@@ -51,24 +37,24 @@ fn test_column_coercion_to_i64() {
let columnar2 = make_columnar("numbers", &[2u64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
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 merge_order = StackMergeOrder::stack(&[&columnar1]).into();
let group_error = group_columns_for_merge(
&[&columnar1],
&[("numbers".to_string(), ColumnType::I64)],
&merge_order,
)
.unwrap_err();
assert_eq!(group_error.kind(), io::ErrorKind::InvalidInput);
}
//#[test]
// fn test_impossible_coercion_returns_an_error() {
// let columnar1 = make_columnar("numbers", &[u64::MAX]);
// let 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() {
@@ -76,7 +62,7 @@ fn test_group_columns_with_required_column() {
let columnar2 = make_columnar("numbers", &[2u64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(
&[&columnar1, &columnar2],
&[("numbers".to_string(), ColumnType::U64)],
@@ -84,7 +70,7 @@ fn test_group_columns_with_required_column() {
)
.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]
@@ -93,17 +79,17 @@ fn test_group_columns_required_column_with_no_existing_columns() {
let columnar2 = make_columnar("numbers", &[2u64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(
columnars,
&[("required_col".to_string(), ColumnType::Str)],
&merge_order,
)
.unwrap();
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());
@@ -115,7 +101,7 @@ fn test_group_columns_required_column_is_above_all_columns_have_the_same_type_ru
let columnar2 = make_columnar("numbers", &[2i64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(
columnars,
&[("numbers".to_string(), ColumnType::U64)],
@@ -123,7 +109,7 @@ fn test_group_columns_required_column_is_above_all_columns_have_the_same_type_ru
)
.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]
@@ -132,21 +118,23 @@ fn test_missing_column() {
let columnar2 = make_columnar("numbers2", &[2u64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
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());
}
@@ -244,7 +232,9 @@ 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);
@@ -270,7 +260,9 @@ 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| {
@@ -317,7 +309,9 @@ 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();
@@ -371,7 +365,9 @@ 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();
@@ -423,7 +419,9 @@ 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![]);
@@ -433,7 +431,9 @@ 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> {

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

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

50
columnar/src/columnar/writer/mod.rs
View File

@@ -13,9 +13,7 @@ pub(crate) use serializer::ColumnarSerializer;
use stacker::{Addr, ArenaHashMap, MemoryArena};
use crate::column_index::SerializableColumnIndex;
use crate::column_values::{
ColumnValues, MonotonicallyMappableToU128, MonotonicallyMappableToU64, VecColumn,
};
use crate::column_values::{MonotonicallyMappableToU128, MonotonicallyMappableToU64};
use crate::columnar::column_type::ColumnType;
use crate::columnar::writer::column_writers::{
ColumnWriter, NumericalColumnWriter, StrOrBytesColumnWriter,
@@ -79,7 +77,6 @@ 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 +84,11 @@ 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 +100,15 @@ 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();
@@ -328,7 +336,7 @@ impl ColumnarWriter {
let mut columns: Vec<(&[u8], ColumnType, Addr)> = self
.numerical_field_hash_map
.iter()
.map(|(column_name, addr, _)| {
.map(|(column_name, addr)| {
let numerical_column_writer: NumericalColumnWriter =
self.numerical_field_hash_map.read(addr);
let column_type = numerical_column_writer.numerical_type().into();
@@ -338,27 +346,27 @@ impl ColumnarWriter {
columns.extend(
self.bytes_field_hash_map
.iter()
.map(|(term, addr, _)| (term, ColumnType::Bytes, addr)),
.map(|(term, addr)| (term, ColumnType::Bytes, addr)),
);
columns.extend(
self.str_field_hash_map
.iter()
.map(|(column_name, addr, _)| (column_name, ColumnType::Str, addr)),
.map(|(column_name, addr)| (column_name, ColumnType::Str, addr)),
);
columns.extend(
self.bool_field_hash_map
.iter()
.map(|(column_name, addr, _)| (column_name, ColumnType::Bool, addr)),
.map(|(column_name, addr)| (column_name, ColumnType::Bool, addr)),
);
columns.extend(
self.ip_addr_field_hash_map
.iter()
.map(|(column_name, addr, _)| (column_name, ColumnType::IpAddr, addr)),
.map(|(column_name, addr)| (column_name, ColumnType::IpAddr, addr)),
);
columns.extend(
self.datetime_field_hash_map
.iter()
.map(|(column_name, addr, _)| (column_name, ColumnType::DateTime, addr)),
.map(|(column_name, addr)| (column_name, ColumnType::DateTime, addr)),
);
columns.sort_unstable_by_key(|(column_name, col_type, _)| (*column_name, *col_type));
@@ -427,6 +435,7 @@ impl ColumnarWriter {
&mut symbol_byte_buffer,
),
buffers,
&self.arena,
&mut column_serializer,
)?;
column_serializer.finalize()?;
@@ -480,6 +489,7 @@ impl ColumnarWriter {
// Serialize [Dictionary, Column, dictionary num bytes U32::LE]
// Column: [Column Index, Column Values, column index num bytes U32::LE]
#[allow(clippy::too_many_arguments)]
fn serialize_bytes_or_str_column(
cardinality: Cardinality,
num_docs: RowId,
@@ -487,6 +497,7 @@ fn serialize_bytes_or_str_column(
dictionary_builder: &DictionaryBuilder,
operation_it: impl Iterator<Item = ColumnOperation<UnorderedId>>,
buffers: &mut SpareBuffers,
arena: &MemoryArena,
wrt: impl io::Write,
) -> io::Result<()> {
let SpareBuffers {
@@ -495,7 +506,8 @@ fn serialize_bytes_or_str_column(
..
} = buffers;
let mut counting_writer = CountingWriter::wrap(wrt);
let term_id_mapping: TermIdMapping = dictionary_builder.serialize(&mut counting_writer)?;
let term_id_mapping: TermIdMapping =
dictionary_builder.serialize(arena, &mut counting_writer)?;
let dictionary_num_bytes: u32 = counting_writer.written_bytes() as u32;
let mut wrt = counting_writer.finish();
let operation_iterator = operation_it.map(|symbol: ColumnOperation<UnorderedId>| {
@@ -631,10 +643,7 @@ fn send_to_serialize_column_mappable_to_u128<
value_index_builders: &mut PreallocatedIndexBuilders,
values: &mut Vec<T>,
mut wrt: impl io::Write,
) -> io::Result<()>
where
for<'a> VecColumn<'a, T>: ColumnValues<T>,
{
) -> io::Result<()> {
values.clear();
// TODO: split index and values
let serializable_column_index = match cardinality {
@@ -687,10 +696,7 @@ fn send_to_serialize_column_mappable_to_u64(
value_index_builders: &mut PreallocatedIndexBuilders,
values: &mut Vec<u64>,
mut wrt: impl io::Write,
) -> io::Result<()>
where
for<'a> VecColumn<'a, u64>: ColumnValues<u64>,
{
) -> io::Result<()> {
values.clear();
let serializable_column_index = match cardinality {
Cardinality::Full => {

10
columnar/src/columnar/writer/serializer.rs
View File

@@ -18,7 +18,12 @@ pub struct ColumnarSerializer<W: io::Write> {
/// code.
fn prepare_key(key: &[u8], column_type: ColumnType, buffer: &mut Vec<u8>) {
buffer.clear();
buffer.extend_from_slice(key);
// Convert 0 bytes to '0' string, as 0 bytes are reserved for the end of the path.
if key.contains(&0u8) {
buffer.extend(key.iter().map(|&b| if b == 0 { b'0' } else { b }));
} else {
buffer.extend_from_slice(key);
}
buffer.push(0u8);
buffer.push(column_type.to_code());
}
@@ -96,14 +101,13 @@ impl<'a, W: io::Write> io::Write for ColumnSerializer<'a, W> {
#[cfg(test)]
mod tests {
use super::*;
use crate::columnar::column_type::ColumnType;
#[test]
fn test_prepare_key_bytes() {
let mut buffer: Vec<u8> = b"somegarbage".to_vec();
prepare_key(b"root\0child", ColumnType::Str, &mut buffer);
assert_eq!(buffer.len(), 12);
assert_eq!(&buffer[..10], b"root\0child");
assert_eq!(&buffer[..10], b"root0child");
assert_eq!(buffer[10], 0u8);
assert_eq!(buffer[11], ColumnType::Str.to_code());
}

49
columnar/src/dictionary.rs
View File

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

23
columnar/src/dynamic_column.rs
View File

@@ -8,7 +8,7 @@ use common::{ByteCount, DateTime, HasLen, OwnedBytes};
use crate::column::{BytesColumn, Column, StrColumn};
use crate::column_values::{monotonic_map_column, StrictlyMonotonicFn};
use crate::columnar::ColumnType;
use crate::{Cardinality, ColumnIndex, NumericalType};
use crate::{Cardinality, ColumnIndex, ColumnValues, NumericalType};
#[derive(Clone)]
pub enum DynamicColumn {
@@ -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 +247,12 @@ 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, ip, or datetime.
///
/// Notice that for IpAddr, the fastfield reader will return the u64 representation of the
/// IpAddr.
/// In order to convert to u128 back cast to `CompactSpaceU64Accessor` and call
/// `compact_to_u128`.
///
/// If not, the fastfield reader will returns the u64-value associated with the original
/// FastValue.
@@ -258,9 +263,15 @@ 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::IpAddr => {
let column = crate::column::open_column_u128_as_compact_u64(column_bytes)?;
Ok(Some(column))
}
ColumnType::Bool
| ColumnType::I64
| ColumnType::U64
| ColumnType::F64
| ColumnType::DateTime => {
let column = crate::column::open_column_u64::<u64>(column_bytes)?;
Ok(Some(column))
}

24
columnar/src/lib.rs
View File

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

62
columnar/src/tests.rs
View File

@@ -26,7 +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(), 87);
assert_eq!(cols[0].num_bytes(), 73);
}
#[test]
@@ -40,7 +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(), 87);
assert_eq!(cols[0].num_bytes(), 73);
}
#[test]
@@ -57,7 +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,]);
}
@@ -79,7 +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
@@ -101,7 +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,
@@ -134,7 +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));
@@ -198,7 +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);
@@ -230,7 +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();
@@ -318,9 +330,9 @@ fn bytes_strategy() -> impl Strategy<Value = &'static [u8]> {
// A random column value
fn column_value_strategy() -> impl Strategy<Value = ColumnValue> {
prop_oneof![
10 => string_strategy().prop_map(|s| ColumnValue::Str(s)),
1 => bytes_strategy().prop_map(|b| ColumnValue::Bytes(b)),
40 => num_strategy().prop_map(|n| ColumnValue::Numerical(n)),
10 => string_strategy().prop_map(ColumnValue::Str),
1 => bytes_strategy().prop_map(ColumnValue::Bytes),
40 => num_strategy().prop_map(ColumnValue::Numerical),
1 => (1u16..3u16).prop_map(|ip_addr_byte| ColumnValue::IpAddr(Ipv6Addr::new(
127,
0,
@@ -331,7 +343,7 @@ fn column_value_strategy() -> impl Strategy<Value = ColumnValue> {
0,
ip_addr_byte
))),
1 => any::<bool>().prop_map(|b| ColumnValue::Bool(b)),
1 => any::<bool>().prop_map(ColumnValue::Bool),
1 => (0_679_723_993i64..1_679_723_995i64)
.prop_map(|val| { ColumnValue::DateTime(DateTime::from_timestamp_secs(val)) })
]
@@ -407,8 +419,8 @@ fn build_columnar_with_mapping(
columnar_writer
.serialize(num_docs, old_to_new_row_ids_opt, &mut buffer)
.unwrap();
let columnar_reader = ColumnarReader::open(buffer).unwrap();
columnar_reader
ColumnarReader::open(buffer).unwrap()
}
fn build_columnar(docs: &[Vec<(&'static str, ColumnValue)>]) -> ColumnarReader {
@@ -533,28 +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);
}
}
@@ -726,7 +746,7 @@ proptest! {
let stack_merge_order = StackMergeOrder::stack(&columnar_readers_arr[..]).into();
crate::merge_columnar(&columnar_readers_arr[..], &[], stack_merge_order, &mut output).unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
let concat_rows: Vec<Vec<(&'static str, ColumnValue)>> = columnar_docs.iter().cloned().flatten().collect();
let concat_rows: Vec<Vec<(&'static str, ColumnValue)>> = columnar_docs.iter().flatten().cloned().collect();
let expected_merged_columnar = build_columnar(&concat_rows[..]);
assert_columnar_eq_strict(&merged_columnar, &expected_merged_columnar);
}
@@ -752,7 +772,7 @@ fn test_columnar_merging_empty_columnar() {
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
let concat_rows: Vec<Vec<(&'static str, ColumnValue)>> =
columnar_docs.iter().cloned().flatten().collect();
columnar_docs.iter().flatten().cloned().collect();
let expected_merged_columnar = build_columnar(&concat_rows[..]);
assert_columnar_eq_strict(&merged_columnar, &expected_merged_columnar);
}
@@ -789,7 +809,7 @@ fn test_columnar_merging_number_columns() {
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
let concat_rows: Vec<Vec<(&'static str, ColumnValue)>> =
columnar_docs.iter().cloned().flatten().collect();
columnar_docs.iter().flatten().cloned().collect();
let expected_merged_columnar = build_columnar(&concat_rows[..]);
assert_columnar_eq_strict(&merged_columnar, &expected_merged_columnar);
}

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"] }

1
common/src/bitset.rs
View File

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

17
common/src/datetime.rs
View File

@@ -1,11 +1,14 @@
#![allow(deprecated)]
use std::fmt;
use std::io::{Read, Write};
use serde::{Deserialize, Serialize};
use time::format_description::well_known::Rfc3339;
use time::{OffsetDateTime, PrimitiveDateTime, UtcOffset};
use crate::BinarySerializable;
/// Precision with which datetimes are truncated when stored in fast fields. This setting is only
/// relevant for fast fields. In the docstore, datetimes are always saved with nanosecond precision.
#[derive(
@@ -37,7 +40,7 @@ pub type DatePrecision = DateTimePrecision;
/// All constructors and conversions are provided as explicit
/// functions and not by implementing any `From`/`Into` traits
/// to prevent unintended usage.
#[derive(Clone, Default, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
#[derive(Clone, Default, Copy, PartialEq, Eq, PartialOrd, Ord, Hash, Serialize, Deserialize)]
pub struct DateTime {
// Timestamp in nanoseconds.
pub(crate) timestamp_nanos: i64,
@@ -164,3 +167,15 @@ impl fmt::Debug for DateTime {
f.write_str(&utc_rfc3339)
}
}
impl BinarySerializable for DateTime {
fn serialize<W: Write + ?Sized>(&self, writer: &mut W) -> std::io::Result<()> {
let timestamp_micros = self.into_timestamp_micros();
<i64 as BinarySerializable>::serialize(&timestamp_micros, writer)
}
fn deserialize<R: Read>(reader: &mut R) -> std::io::Result<Self> {
let timestamp_micros = <i64 as BinarySerializable>::deserialize(reader)?;
Ok(Self::from_timestamp_micros(timestamp_micros))
}
}

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 {

112
common/src/json_path_writer.rs Normal file
View File

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

3
common/src/lib.rs
View File

@@ -9,6 +9,7 @@ mod byte_count;
mod datetime;
pub mod file_slice;
mod group_by;
mod json_path_writer;
mod serialize;
mod vint;
mod writer;
@@ -18,6 +19,7 @@ pub use byte_count::ByteCount;
pub use datetime::DatePrecision;
pub use datetime::{DateTime, DateTimePrecision};
pub use group_by::GroupByIteratorExtended;
pub use json_path_writer::JsonPathWriter;
pub use ownedbytes::{OwnedBytes, StableDeref};
pub use serialize::{BinarySerializable, DeserializeFrom, FixedSize};
pub use vint::{
@@ -116,6 +118,7 @@ pub fn u64_to_f64(val: u64) -> f64 {
///
/// This function assumes that the needle is rarely contained in the bytes string
/// and offers a fast path if the needle is not present.
#[inline]
pub fn replace_in_place(needle: u8, replacement: u8, bytes: &mut [u8]) {
if !bytes.contains(&needle) {
return;

41
common/src/serialize.rs
View File

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

6
examples/aggregation.rs
View File

@@ -12,7 +12,7 @@ use tantivy::aggregation::agg_result::AggregationResults;
use tantivy::aggregation::AggregationCollector;
use tantivy::query::AllQuery;
use tantivy::schema::{self, IndexRecordOption, Schema, TextFieldIndexing, FAST};
use tantivy::Index;
use tantivy::{Index, IndexWriter, TantivyDocument};
fn main() -> tantivy::Result<()> {
// # Create Schema
@@ -132,10 +132,10 @@ fn main() -> tantivy::Result<()> {
let stream = Deserializer::from_str(data).into_iter::<Value>();
let mut index_writer = index.writer(50_000_000)?;
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let mut num_indexed = 0;
for value in stream {
let doc = schema.parse_document(&serde_json::to_string(&value.unwrap())?)?;
let doc = TantivyDocument::parse_json(&schema, &serde_json::to_string(&value.unwrap())?)?;
index_writer.add_document(doc)?;
num_indexed += 1;
if num_indexed > 4 {

26
examples/basic_search.rs
View File

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

4
examples/custom_collector.rs
View File

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

10
examples/custom_tokenizer.rs
View File

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

18
examples/date_time_field.rs
View File

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

12
examples/deleting_updating_documents.rs
View File

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

4
examples/faceted_search.rs
View File

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

9
examples/faceted_search_with_tweaked_score.rs
View File

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

10
examples/fuzzy_search.rs
View File

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

4
examples/index_with_json.rs
View File

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

4
examples/integer_range_search.rs
View File

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

13
examples/ip_field.rs
View File

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

4
examples/iterating_docs_and_positions.rs
View File

@@ -10,7 +10,7 @@
// ---
// Importing tantivy...
use tantivy::schema::*;
use tantivy::{doc, DocSet, Index, Postings, TERMINATED};
use tantivy::{doc, DocSet, Index, IndexWriter, Postings, TERMINATED};
fn main() -> tantivy::Result<()> {
// We first create a schema for the sake of the
@@ -24,7 +24,7 @@ fn main() -> tantivy::Result<()> {
let index = Index::create_in_ram(schema);
let mut index_writer = index.writer_with_num_threads(1, 50_000_000)?;
let mut index_writer: IndexWriter = index.writer_with_num_threads(1, 50_000_000)?;
index_writer.add_document(doc!(title => "The Old Man and the Sea"))?;
index_writer.add_document(doc!(title => "Of Mice and Men"))?;
index_writer.add_document(doc!(title => "The modern Promotheus"))?;

10
examples/json_field.rs
View File

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

14
examples/phrase_prefix_search.rs
View File

@@ -1,7 +1,7 @@
use tantivy::collector::TopDocs;
use tantivy::query::QueryParser;
use tantivy::schema::*;
use tantivy::{doc, Index, ReloadPolicy, Result};
use tantivy::{doc, Index, IndexWriter, ReloadPolicy, Result};
use tempfile::TempDir;
fn main() -> Result<()> {
@@ -17,7 +17,7 @@ fn main() -> Result<()> {
let index = Index::create_in_dir(&index_path, schema)?;
let mut index_writer = index.writer(50_000_000)?;
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
index_writer.add_document(doc!(
title => "The Old Man and the Sea",
@@ -51,7 +51,7 @@ fn main() -> Result<()> {
let reader = index
.reader_builder()
.reload_policy(ReloadPolicy::OnCommit)
.reload_policy(ReloadPolicy::OnCommitWithDelay)
.try_into()?;
let searcher = reader.searcher();
@@ -67,8 +67,12 @@ fn main() -> Result<()> {
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();
let doc = searcher.doc::<TantivyDocument>(doc_address)?;
let title = doc
.get_first(title)
.and_then(|v| v.as_str())
.unwrap()
.to_owned();
Ok(title)
})
.collect::<Result<Vec<_>>>()?;

12
examples/pre_tokenized_text.rs
View File

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

12
examples/snippet.rs
View File

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

8
examples/stop_words.rs
View File

@@ -15,7 +15,7 @@ use tantivy::collector::TopDocs;
use tantivy::query::QueryParser;
use tantivy::schema::*;
use tantivy::tokenizer::*;
use tantivy::{doc, Index};
use tantivy::{doc, Index, IndexWriter};
fn main() -> tantivy::Result<()> {
// this example assumes you understand the content in `basic_search`
@@ -60,7 +60,7 @@ fn main() -> tantivy::Result<()> {
index.tokenizers().register("stoppy", tokenizer);
let mut index_writer = index.writer(50_000_000)?;
let mut index_writer: IndexWriter = index.writer(50_000_000)?;
let title = schema.get_field("title").unwrap();
let body = schema.get_field("body").unwrap();
@@ -105,9 +105,9 @@ fn main() -> tantivy::Result<()> {
let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
for (score, doc_address) in top_docs {
let retrieved_doc = searcher.doc(doc_address)?;
let retrieved_doc: TantivyDocument = searcher.doc(doc_address)?;
println!("\n==\nDocument score {score}:");
println!("{}", schema.to_json(&retrieved_doc));
println!("{}", retrieved_doc.to_json(&schema));
}
Ok(())

75
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,
doc, DocAddress, DocId, Index, IndexWriter, 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: IndexWriter = 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"

108
ownedbytes/src/lib.rs
View File

@@ -1,7 +1,6 @@
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 +25,8 @@ 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 +56,12 @@ 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 +73,14 @@ 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)
@@ -99,55 +105,45 @@ 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());
let byte = self.as_slice()[0];
self.advance(1);
byte
self.advance(1)[0]
}
/// 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)
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 {
assert!(self.len() > 3);
u32::from_le_bytes(self.read_n())
}
let quad: [u8; 4] = self.as_slice()[..4].try_into().unwrap();
self.advance(4);
u32::from_le_bytes(quad)
/// Reads an `u64` encoded as little-endian from the `OwnedBytes` and advance by 8 bytes.
#[inline]
pub fn read_u64(&mut self) -> u64 {
u64::from_le_bytes(self.read_n())
}
}
@@ -201,32 +197,33 @@ 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]
@@ -242,13 +239,6 @@ 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};

6
query-grammar/Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy-query-grammar"
version = "0.20.0"
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::multispace0)(input)
.map(|(left, (spaces, errors))| (left, (spaces.expect("multispace0 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::multispace1)(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))
}

13
query-grammar/src/lib.rs
View File

@@ -1,19 +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;
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)
}

1622
query-grammar/src/query_grammar.rs
View File

File diff suppressed because it is too large Load Diff

48
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,
@@ -16,6 +16,34 @@ 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 {
@@ -28,6 +56,7 @@ impl Debug for UserInputLeaf {
ref upper,
} => {
if let Some(ref field) = field {
// TODO properly escape field (in case of \")
write!(formatter, "\"{field}\":")?;
}
lower.display_lower(formatter)?;
@@ -37,6 +66,7 @@ impl Debug for UserInputLeaf {
}
UserInputLeaf::Set { field, elements } => {
if let Some(ref field) = field {
// TODO properly escape field (in case of \")
write!(formatter, "\"{field}\": ")?;
}
write!(formatter, "IN [")?;
@@ -44,11 +74,15 @@ impl Debug for UserInputLeaf {
if i != 0 {
write!(formatter, " ")?;
}
// TODO properly escape element
write!(formatter, "\"{text}\"")?;
}
write!(formatter, "]")
}
UserInputLeaf::All => write!(formatter, "*"),
UserInputLeaf::Exists { field } => {
write!(formatter, "\"{field}\":*")
}
}
}
}
@@ -60,7 +94,7 @@ pub enum Delimiter {
None,
}
#[derive(PartialEq)]
#[derive(PartialEq, Clone)]
pub struct UserInputLiteral {
pub field_name: Option<String>,
pub phrase: String,
@@ -72,16 +106,20 @@ pub struct UserInputLiteral {
impl fmt::Debug for UserInputLiteral {
fn fmt(&self, formatter: &mut fmt::Formatter) -> Result<(), fmt::Error> {
if let Some(ref field) = self.field_name {
// 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)?;
}
}
@@ -94,7 +132,7 @@ impl fmt::Debug for UserInputLiteral {
}
}
#[derive(PartialEq)]
#[derive(PartialEq, Debug, Clone)]
pub enum UserInputBound {
Inclusive(String),
Exclusive(String),
@@ -104,6 +142,7 @@ pub enum UserInputBound {
impl UserInputBound {
fn display_lower(&self, formatter: &mut fmt::Formatter) -> Result<(), fmt::Error> {
match *self {
// TODO properly escape word if required
UserInputBound::Inclusive(ref word) => write!(formatter, "[\"{word}\""),
UserInputBound::Exclusive(ref word) => write!(formatter, "{{\"{word}\""),
UserInputBound::Unbounded => write!(formatter, "{{\"*\""),
@@ -112,6 +151,7 @@ impl UserInputBound {
fn display_upper(&self, formatter: &mut fmt::Formatter) -> Result<(), fmt::Error> {
match *self {
// TODO properly escape word if required
UserInputBound::Inclusive(ref word) => write!(formatter, "\"{word}\"]"),
UserInputBound::Exclusive(ref word) => write!(formatter, "\"{word}\"}}"),
UserInputBound::Unbounded => write!(formatter, "\"*\"}}"),
@@ -127,6 +167,7 @@ impl UserInputBound {
}
}
#[derive(PartialEq, Clone)]
pub enum UserInputAst {
Clause(Vec<(Option<Occur>, UserInputAst)>),
Leaf(Box<UserInputLeaf>),
@@ -196,6 +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, "(")?;

39
src/aggregation/agg_bench.rs
View File

@@ -48,7 +48,7 @@ mod bench {
let score_field_f64 = schema_builder.add_f64_field("score_f64", score_fieldtype.clone());
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 few_terms_data = ["INFO", "ERROR", "WARN", "DEBUG"];
let lg_norm = rand_distr::LogNormal::new(2.996f64, 0.979f64).unwrap();
@@ -85,7 +85,7 @@ mod bench {
if cardinality == Cardinality::Sparse {
doc_with_value /= 20;
}
let val_max = 1_000_000.0;
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) {
@@ -290,6 +290,41 @@ mod bench {
});
}
bench_all_cardinalities!(bench_aggregation_terms_many_with_top_hits_agg);
fn bench_aggregation_terms_many_with_top_hits_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": "text_many_terms" },
"aggs": {
"top_hits": { "top_hits":
{
"sort": [
{ "score": "desc" }
],
"size": 2,
"doc_value_fields": ["score_f64"]
}
}
}
},
}))
.unwrap();
let collector = get_collector(agg_req);
let searcher = reader.searcher();
searcher.search(&AllQuery, &collector).unwrap()
});
}
bench_all_cardinalities!(bench_aggregation_terms_many_with_sub_agg);
fn bench_aggregation_terms_many_with_sub_agg_card(b: &mut Bencher, cardinality: Cardinality) {

143
src/aggregation/agg_limits.rs
View File

@@ -73,9 +73,9 @@ impl AggregationLimits {
/// 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.
// The counter which is shared between the aggregations for one request.
memory_consumption: Arc::clone(&self.memory_consumption),
/// The memory_limit in bytes
// The memory_limit in bytes
memory_limit: self.memory_limit,
allocated_with_the_guard: 0,
}
@@ -134,3 +134,142 @@ impl Drop for ResourceLimitGuard {
.fetch_sub(self.allocated_with_the_guard, Ordering::Relaxed);
}
}
#[cfg(test)]
mod tests {
use crate::aggregation::tests::exec_request_with_query;
// https://github.com/quickwit-oss/quickwit/issues/3837
#[test]
fn test_agg_limits_with_empty_merge() {
use crate::aggregation::agg_req::Aggregations;
use crate::aggregation::bucket::tests::get_test_index_from_docs;
let docs = vec![
vec![r#"{ "date": "2015-01-02T00:00:00Z", "text": "bbb", "text2": "bbb" }"#],
vec![r#"{ "text": "aaa", "text2": "bbb" }"#],
];
let index = get_test_index_from_docs(false, &docs).unwrap();
{
let elasticsearch_compatible_json = json!(
{
"1": {
"terms": {"field": "text2", "min_doc_count": 0},
"aggs": {
"2":{
"date_histogram": {
"field": "date",
"fixed_interval": "1d",
"extended_bounds": {
"min": "2015-01-01T00:00:00Z",
"max": "2015-01-10T00:00:00Z"
}
}
}
}
}
}
);
let agg_req: Aggregations = serde_json::from_str(
&serde_json::to_string(&elasticsearch_compatible_json).unwrap(),
)
.unwrap();
let res = exec_request_with_query(agg_req, &index, Some(("text", "bbb"))).unwrap();
let expected_res = json!({
"1": {
"buckets": [
{
"2": {
"buckets": [
{ "doc_count": 0, "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": 0, "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" },
{ "doc_count": 0, "key": 1420675200000.0, "key_as_string": "2015-01-08T00:00:00Z" },
{ "doc_count": 0, "key": 1420761600000.0, "key_as_string": "2015-01-09T00:00:00Z" },
{ "doc_count": 0, "key": 1420848000000.0, "key_as_string": "2015-01-10T00:00:00Z" }
]
},
"doc_count": 1,
"key": "bbb"
}
],
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0
}
});
assert_eq!(res, expected_res);
}
}
// https://github.com/quickwit-oss/quickwit/issues/3837
#[test]
fn test_agg_limits_with_empty_data() {
use crate::aggregation::agg_req::Aggregations;
use crate::aggregation::bucket::tests::get_test_index_from_docs;
let docs = vec![vec![r#"{ "text": "aaa", "text2": "bbb" }"#]];
let index = get_test_index_from_docs(false, &docs).unwrap();
{
// Empty result since there is no doc with dates
let elasticsearch_compatible_json = json!(
{
"1": {
"terms": {"field": "text2", "min_doc_count": 0},
"aggs": {
"2":{
"date_histogram": {
"field": "date",
"fixed_interval": "1d",
"extended_bounds": {
"min": "2015-01-01T00:00:00Z",
"max": "2015-01-10T00:00:00Z"
}
}
}
}
}
}
);
let agg_req: Aggregations = serde_json::from_str(
&serde_json::to_string(&elasticsearch_compatible_json).unwrap(),
)
.unwrap();
let res = exec_request_with_query(agg_req, &index, Some(("text", "bbb"))).unwrap();
let expected_res = json!({
"1": {
"buckets": [
{
"2": {
"buckets": [
{ "doc_count": 0, "key": 1420070400000.0, "key_as_string": "2015-01-01T00:00:00Z" },
{ "doc_count": 0, "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": 0, "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" },
{ "doc_count": 0, "key": 1420675200000.0, "key_as_string": "2015-01-08T00:00:00Z" },
{ "doc_count": 0, "key": 1420761600000.0, "key_as_string": "2015-01-09T00:00:00Z" },
{ "doc_count": 0, "key": 1420848000000.0, "key_as_string": "2015-01-10T00:00:00Z" }
]
},
"doc_count": 0,
"key": "bbb"
}
],
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0
}
});
assert_eq!(res, expected_res);
}
}
}

71
src/aggregation/agg_req.rs
View File

@@ -35,7 +35,7 @@ use super::bucket::{
};
use super::metric::{
AverageAggregation, CountAggregation, MaxAggregation, MinAggregation,
PercentilesAggregationReq, StatsAggregation, SumAggregation,
PercentilesAggregationReq, StatsAggregation, SumAggregation, TopHitsAggregation,
};
/// The top-level aggregation request structure, which contains [`Aggregation`] and their user
@@ -44,29 +44,61 @@ use super::metric::{
/// The key is the user defined name of the aggregation.
pub type Aggregations = HashMap<String, Aggregation>;
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
/// 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,
/// The sub_aggregations, only valid for bucket type aggregations. 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,
}
/// 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,
})
}
}
impl Aggregation {
pub(crate) fn sub_aggregation(&self) -> &Aggregations {
&self.sub_aggregation
}
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(
self.agg
.get_fast_field_names()
.iter()
.map(|s| s.to_string()),
);
fast_field_names.extend(get_fast_field_names(&self.sub_aggregation));
}
}
@@ -120,22 +152,27 @@ pub enum AggregationVariants {
/// Computes the sum of the extracted values.
#[serde(rename = "percentiles")]
Percentiles(PercentilesAggregationReq),
/// Finds the top k values matching some order
#[serde(rename = "top_hits")]
TopHits(TopHitsAggregation),
}
impl AggregationVariants {
fn get_fast_field_name(&self) -> &str {
/// Returns the name of the fields used by the aggregation.
pub fn get_fast_field_names(&self) -> Vec<&str> {
match self {
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(),
AggregationVariants::Terms(terms) => vec![terms.field.as_str()],
AggregationVariants::Range(range) => vec![range.field.as_str()],
AggregationVariants::Histogram(histogram) => vec![histogram.field.as_str()],
AggregationVariants::DateHistogram(histogram) => vec![histogram.field.as_str()],
AggregationVariants::Average(avg) => vec![avg.field_name()],
AggregationVariants::Count(count) => vec![count.field_name()],
AggregationVariants::Max(max) => vec![max.field_name()],
AggregationVariants::Min(min) => vec![min.field_name()],
AggregationVariants::Stats(stats) => vec![stats.field_name()],
AggregationVariants::Sum(sum) => vec![sum.field_name()],
AggregationVariants::Percentiles(per) => vec![per.field_name()],
AggregationVariants::TopHits(top_hits) => top_hits.field_names(),
}
}

368
src/aggregation/agg_req_with_accessor.rs
View File

@@ -1,6 +1,9 @@
//! This will enhance the request tree with access to the fastfield and metadata.
use columnar::{Column, ColumnBlockAccessor, ColumnType, StrColumn};
use std::collections::HashMap;
use std::io;
use columnar::{Column, ColumnBlockAccessor, ColumnType, DynamicColumn, StrColumn};
use super::agg_limits::ResourceLimitGuard;
use super::agg_req::{Aggregation, AggregationVariants, Aggregations};
@@ -13,7 +16,8 @@ use super::metric::{
};
use super::segment_agg_result::AggregationLimits;
use super::VecWithNames;
use crate::SegmentReader;
use crate::aggregation::{f64_to_fastfield_u64, Key};
use crate::{SegmentOrdinal, SegmentReader};
#[derive(Default)]
pub(crate) struct AggregationsWithAccessor {
@@ -31,100 +35,313 @@ impl AggregationsWithAccessor {
}
pub struct AggregationWithAccessor {
pub(crate) segment_ordinal: SegmentOrdinal,
/// In general there can be buckets without fast field access, e.g. buckets that are created
/// 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,
/// In case there are multiple types of fast fields, e.g. string and numeric.
/// Only used for term aggregations currently.
pub(crate) accessor2: Option<(Column<u64>, ColumnType)>,
pub(crate) sub_aggregation: AggregationsWithAccessor,
pub(crate) limits: ResourceLimitGuard,
pub(crate) column_block_accessor: ColumnBlockAccessor<u64>,
/// Used for missing term aggregation, which checks all columns for existence.
/// And also for `top_hits` aggregation, which may sort on multiple fields.
/// 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.
// NOTE: we can make all other aggregations use this instead of the `accessor` and `field_type`
// (making them obsolete) But will it have a performance impact?
pub(crate) accessors: Vec<(Column<u64>, ColumnType)>,
/// Map field names to all associated column accessors.
/// This field is used for `docvalue_fields`, which is currently only supported for `top_hits`.
pub(crate) value_accessors: HashMap<String, Vec<DynamicColumn>>,
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,
segment_ordinal: SegmentOrdinal,
limits: AggregationLimits,
) -> crate::Result<AggregationWithAccessor> {
let mut str_dict_column = None;
let mut accessor2 = None;
) -> crate::Result<Vec<AggregationWithAccessor>> {
let mut agg = agg.clone();
let add_agg_with_accessor = |agg: &Aggregation,
accessor: Column<u64>,
column_type: ColumnType,
aggs: &mut Vec<AggregationWithAccessor>|
-> crate::Result<()> {
let res = AggregationWithAccessor {
segment_ordinal,
accessor,
accessors: Default::default(),
value_accessors: Default::default(),
field_type: column_type,
sub_aggregation: get_aggs_with_segment_accessor_and_validate(
sub_aggregation,
reader,
segment_ordinal,
&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 add_agg_with_accessors = |agg: &Aggregation,
accessors: Vec<(Column<u64>, ColumnType)>,
aggs: &mut Vec<AggregationWithAccessor>,
value_accessors: HashMap<String, Vec<DynamicColumn>>|
-> crate::Result<()> {
let (accessor, field_type) = accessors.first().expect("at least one accessor");
let res = AggregationWithAccessor {
segment_ordinal,
// TODO: We should do away with the `accessor` field altogether
accessor: accessor.clone(),
value_accessors,
field_type: *field_type,
accessors,
sub_aggregation: get_aggs_with_segment_accessor_and_validate(
sub_aggregation,
reader,
segment_ordinal,
&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::*;
let (accessor, field_type) = match &agg.agg {
match agg.agg {
Range(RangeAggregation {
field: field_name, ..
}) => get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?,
Histogram(HistogramAggregation {
field: field_name, ..
}) => get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?,
DateHistogram(DateHistogramAggregationReq {
field: field_name, ..
}) => get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?,
Terms(TermsAggregation {
field: field_name, ..
field: ref field_name,
..
}) => {
str_dict_column = reader.fast_fields().str(field_name)?;
let (accessor, column_type) =
get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?;
add_agg_with_accessor(&agg, accessor, column_type, &mut res)?;
}
Histogram(HistogramAggregation {
field: ref field_name,
..
}) => {
let (accessor, column_type) =
get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?;
add_agg_with_accessor(&agg, accessor, column_type, &mut res)?;
}
DateHistogram(DateHistogramAggregationReq {
field: ref field_name,
..
}) => {
let (accessor, column_type) =
// Only DateTime is supported for DateHistogram
get_ff_reader(reader, field_name, Some(&[ColumnType::DateTime]))?;
add_agg_with_accessor(&agg, accessor, column_type, &mut res)?;
}
Terms(TermsAggregation {
field: ref field_name,
ref 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::Bool,
ColumnType::IpAddr,
// ColumnType::Bytes Unsupported
// ColumnType::Bool Unsupported
// ColumnType::IpAddr Unsupported
// ColumnType::DateTime Unsupported
];
let mut columns =
get_all_ff_reader_or_empty(reader, field_name, Some(&allowed_column_types))?;
let first = columns.pop().unwrap();
accessor2 = columns.pop();
first
}
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, field_type) =
get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?;
(accessor, field_type)
// 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 = column_and_types
.iter()
.map(|c_t| (c_t.0.clone(), c_t.1))
.collect();
add_agg_with_accessors(&agg, accessors, &mut res, Default::default())?;
}
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_names()[0])?
} else {
None
};
let agg = AggregationWithAccessor {
segment_ordinal,
missing_value_for_accessor,
accessor,
accessors: Default::default(),
value_accessors: Default::default(),
field_type: column_type,
sub_aggregation: get_aggs_with_segment_accessor_and_validate(
sub_aggregation,
reader,
segment_ordinal,
&limits,
)?,
agg: agg.clone(),
str_dict_column: str_dict_column.clone(),
limits: limits.new_guard(),
column_block_accessor: Default::default(),
};
res.push(agg);
}
}
Percentiles(percentiles) => {
let (accessor, field_type) = get_ff_reader(
Average(AverageAggregation {
field: ref field_name,
..
})
| Count(CountAggregation {
field: ref field_name,
..
})
| Max(MaxAggregation {
field: ref field_name,
..
})
| Min(MinAggregation {
field: ref field_name,
..
})
| Stats(StatsAggregation {
field: ref field_name,
..
})
| Sum(SumAggregation {
field: ref field_name,
..
}) => {
let (accessor, column_type) =
get_ff_reader(reader, field_name, Some(get_numeric_or_date_column_types()))?;
add_agg_with_accessor(&agg, accessor, column_type, &mut res)?;
}
Percentiles(ref percentiles) => {
let (accessor, column_type) = get_ff_reader(
reader,
percentiles.field_name(),
Some(get_numeric_or_date_column_types()),
)?;
(accessor, field_type)
add_agg_with_accessor(&agg, accessor, column_type, &mut res)?;
}
TopHits(ref mut top_hits) => {
top_hits.validate_and_resolve_field_names(reader.fast_fields().columnar())?;
let accessors: Vec<(Column<u64>, ColumnType)> = top_hits
.field_names()
.iter()
.map(|field| {
get_ff_reader(reader, field, Some(get_numeric_or_date_column_types()))
})
.collect::<crate::Result<_>>()?;
let value_accessors = top_hits
.value_field_names()
.iter()
.map(|field_name| {
Ok((
field_name.to_string(),
get_dynamic_columns(reader, field_name)?,
))
})
.collect::<crate::Result<_>>()?;
add_agg_with_accessors(&agg, accessors, &mut res, value_accessors)?;
}
};
let sub_aggregation = sub_aggregation.clone();
Ok(AggregationWithAccessor {
accessor,
accessor2,
field_type,
sub_aggregation: get_aggs_with_segment_accessor_and_validate(
&sub_aggregation,
reader,
&limits,
)?,
agg: agg.clone(),
str_dict_column,
limits: limits.new_guard(),
column_block_accessor: Default::default(),
})
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] {
&[
ColumnType::F64,
@@ -137,19 +354,21 @@ fn get_numeric_or_date_column_types() -> &'static [ColumnType] {
pub(crate) fn get_aggs_with_segment_accessor_and_validate(
aggs: &Aggregations,
reader: &SegmentReader,
segment_ordinal: SegmentOrdinal,
limits: &AggregationLimits,
) -> crate::Result<AggregationsWithAccessor> {
let mut aggss = Vec::new();
for (key, agg) in aggs.iter() {
aggss.push((
key.to_string(),
AggregationWithAccessor::try_from_agg(
agg,
agg.sub_aggregation(),
reader,
limits.clone(),
)?,
));
let aggs = AggregationWithAccessor::try_from_agg(
agg,
agg.sub_aggregation(),
reader,
segment_ordinal,
limits.clone(),
)?;
for agg in aggs {
aggss.push((key.to_string(), agg));
}
}
Ok(AggregationsWithAccessor::from_data(
VecWithNames::from_entries(aggss),
@@ -174,6 +393,19 @@ fn get_ff_reader(
Ok(ff_field_with_type)
}
fn get_dynamic_columns(
reader: &SegmentReader,
field_name: &str,
) -> crate::Result<Vec<columnar::DynamicColumn>> {
let ff_fields = reader.fast_fields().dynamic_column_handles(field_name)?;
let cols = ff_fields
.iter()
.map(|h| h.open())
.collect::<io::Result<_>>()?;
assert!(!ff_fields.is_empty(), "field {} not found", field_name);
Ok(cols)
}
/// Get all fast field reader or empty as default.
///
/// Is guaranteed to return at least one column.
@@ -181,15 +413,13 @@ 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()),
ColumnType::U64,
));
ff_field_with_type.push((Column::build_empty_column(reader.num_docs()), fallback_type));
}
Ok(ff_field_with_type)
}

9
src/aggregation/agg_result.rs
View File

@@ -8,7 +8,7 @@ use rustc_hash::FxHashMap;
use serde::{Deserialize, Serialize};
use super::bucket::GetDocCount;
use super::metric::{PercentilesMetricResult, SingleMetricResult, Stats};
use super::metric::{PercentilesMetricResult, SingleMetricResult, Stats, TopHitsMetricResult};
use super::{AggregationError, Key};
use crate::TantivyError;
@@ -90,8 +90,10 @@ pub enum MetricResult {
Stats(Stats),
/// Sum metric result.
Sum(SingleMetricResult),
/// Sum metric result.
/// Percentiles metric result.
Percentiles(PercentilesMetricResult),
/// Top hits metric result
TopHits(TopHitsMetricResult),
}
impl MetricResult {
@@ -106,6 +108,9 @@ impl MetricResult {
MetricResult::Percentiles(_) => Err(TantivyError::AggregationError(
AggregationError::InvalidRequest("percentiles can't be used to order".to_string()),
)),
MetricResult::TopHits(_) => Err(TantivyError::AggregationError(
AggregationError::InvalidRequest("top_hits can't be used to order".to_string()),
)),
}
}
}

138
src/aggregation/agg_tests.rs
View File

@@ -4,12 +4,13 @@ use crate::aggregation::agg_req::{Aggregation, Aggregations};
use crate::aggregation::agg_result::AggregationResults;
use crate::aggregation::buf_collector::DOC_BLOCK_SIZE;
use crate::aggregation::collector::AggregationCollector;
use crate::aggregation::intermediate_agg_result::IntermediateAggregationResults;
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;
use crate::query::{AllQuery, TermQuery};
use crate::schema::{IndexRecordOption, Schema, FAST};
use crate::{Index, Term};
use crate::{Index, IndexWriter, Term};
fn get_avg_req(field_name: &str) -> Aggregation {
serde_json::from_value(json!({
@@ -66,6 +67,22 @@ fn test_aggregation_flushing(
}
}
},
"top_hits_test":{
"terms": {
"field": "string_id"
},
"aggs": {
"bucketsL2": {
"top_hits": {
"size": 2,
"sort": [
{ "score": "asc" }
],
"docvalue_fields": ["score"]
}
}
}
},
"histogram_test":{
"histogram": {
"field": "score",
@@ -108,6 +125,16 @@ fn test_aggregation_flushing(
let searcher = reader.searcher();
let intermediate_agg_result = searcher.search(&AllQuery, &collector).unwrap();
// Test postcard roundtrip serialization
let intermediate_agg_result_bytes = postcard::to_allocvec(&intermediate_agg_result).expect(
"Postcard Serialization failed, flatten etc. is not supported in the intermediate \
result",
);
let intermediate_agg_result: IntermediateAggregationResults =
postcard::from_bytes(&intermediate_agg_result_bytes)
.expect("Post deserialization failed");
intermediate_agg_result
.into_final_result(agg_req, &Default::default())
.unwrap()
@@ -558,10 +585,10 @@ fn test_aggregation_invalid_requests() -> crate::Result<()> {
assert_eq!(agg_req_1.is_err(), true);
// TODO: This should list valid values
assert_eq!(
agg_req_1.unwrap_err().to_string(),
"no variant of enum AggregationVariants found in flattened data"
);
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();
@@ -586,7 +613,10 @@ fn test_aggregation_on_json_object() {
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();
let mut index_writer: IndexWriter = index.writer_for_tests().unwrap();
index_writer
.add_document(doc!(json => json!({"color": "red"})))
.unwrap();
index_writer
.add_document(doc!(json => json!({"color": "red"})))
.unwrap();
@@ -614,12 +644,74 @@ fn test_aggregation_on_json_object() {
&serde_json::json!({
"jsonagg": {
"buckets": [
{"doc_count": 2, "key": "red"},
{"doc_count": 1, "key": "blue"},
],
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0
}
})
);
}
#[test]
fn test_aggregation_on_nested_json_object() {
let mut schema_builder = Schema::builder();
let json = schema_builder.add_json_field("json.blub", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
let mut index_writer: IndexWriter = index.writer_for_tests().unwrap();
index_writer
.add_document(doc!(json => json!({"color.dot": "red", "color": {"nested":"red"} })))
.unwrap();
index_writer
.add_document(doc!(json => json!({"color.dot": "blue", "color": {"nested":"blue"} })))
.unwrap();
index_writer
.add_document(doc!(json => json!({"color.dot": "blue", "color": {"nested":"blue"} })))
.unwrap();
index_writer.commit().unwrap();
let reader = index.reader().unwrap();
let searcher = reader.searcher();
let agg: Aggregations = serde_json::from_value(json!({
"jsonagg1": {
"terms": {
"field": "json\\.blub.color\\.dot",
}
},
"jsonagg2": {
"terms": {
"field": "json\\.blub.color.nested",
}
}
}))
.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();
assert_eq!(
&aggregation_res_json,
&serde_json::json!({
"jsonagg1": {
"buckets": [
{"doc_count": 2, "key": "blue"},
{"doc_count": 1, "key": "red"}
],
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0
},
"jsonagg2": {
"buckets": [
{"doc_count": 2, "key": "blue"},
{"doc_count": 1, "key": "red"}
],
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0
}
})
);
}
@@ -630,7 +722,7 @@ fn test_aggregation_on_json_object_empty_columns() {
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();
let mut index_writer: IndexWriter = index.writer_for_tests().unwrap();
// => Empty column when accessing color
index_writer
.add_document(doc!(json => json!({"price": 10.0})))
@@ -748,32 +840,41 @@ fn test_aggregation_on_json_object_mixed_types() {
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();
let mut index_writer: IndexWriter = index.writer_for_tests().unwrap();
// => Segment with all values numeric
index_writer
.add_document(doc!(json => json!({"mixed_type": 10.0})))
.add_document(doc!(json => json!({"mixed_type": 10.0, "mixed_price": 10.0})))
.unwrap();
index_writer.commit().unwrap();
// => Segment with all values text
index_writer
.add_document(doc!(json => json!({"mixed_type": "blue"})))
.add_document(doc!(json => json!({"mixed_type": "blue", "mixed_price": 5.0})))
.unwrap();
index_writer
.add_document(doc!(json => json!({"mixed_type": "blue", "mixed_price": 5.0})))
.unwrap();
index_writer
.add_document(doc!(json => json!({"mixed_type": "blue", "mixed_price": 5.0})))
.unwrap();
index_writer.commit().unwrap();
// => Segment with all boolen
index_writer
.add_document(doc!(json => json!({"mixed_type": true})))
.add_document(doc!(json => json!({"mixed_type": true, "mixed_price": "no_price"})))
.unwrap();
index_writer.commit().unwrap();
// => Segment with mixed values
index_writer
.add_document(doc!(json => json!({"mixed_type": "red"})))
.add_document(doc!(json => json!({"mixed_type": "red", "mixed_price": 1.0})))
.unwrap();
index_writer
.add_document(doc!(json => json!({"mixed_type": -20.5})))
.add_document(doc!(json => json!({"mixed_type": "red", "mixed_price": 1.0})))
.unwrap();
index_writer
.add_document(doc!(json => json!({"mixed_type": true})))
.add_document(doc!(json => json!({"mixed_type": -20.5, "mixed_price": -20.5})))
.unwrap();
index_writer
.add_document(doc!(json => json!({"mixed_type": true, "mixed_price": "no_price"})))
.unwrap();
index_writer.commit().unwrap();
@@ -787,7 +888,7 @@ fn test_aggregation_on_json_object_mixed_types() {
"order": { "min_price": "desc" }
},
"aggs": {
"min_price": { "min": { "field": "json.mixed_type" } }
"min_price": { "min": { "field": "json.mixed_price" } }
}
},
"rangeagg": {
@@ -811,6 +912,7 @@ fn test_aggregation_on_json_object_mixed_types() {
let aggregation_results = searcher.search(&AllQuery, &aggregation_collector).unwrap();
let aggregation_res_json = serde_json::to_value(aggregation_results).unwrap();
use pretty_assertions::assert_eq;
assert_eq!(
&aggregation_res_json,
&serde_json::json!({
@@ -825,10 +927,10 @@ fn test_aggregation_on_json_object_mixed_types() {
"termagg": {
"buckets": [
{ "doc_count": 1, "key": 10.0, "min_price": { "value": 10.0 } },
{ "doc_count": 3, "key": "blue", "min_price": { "value": 5.0 } },
{ "doc_count": 2, "key": "red", "min_price": { "value": 1.0 } },
{ "doc_count": 1, "key": -20.5, "min_price": { "value": -20.5 } },
// TODO bool is also not yet handled in aggregation
{ "doc_count": 1, "key": "blue", "min_price": { "value": null } },
{ "doc_count": 1, "key": "red", "min_price": { "value": null } },
{ "doc_count": 2, "key": 1.0, "key_as_string": "true", "min_price": { "value": null } },
],
"sum_other_doc_count": 0
}

35
src/aggregation/bucket/histogram/date_histogram.rs
View File

@@ -1,7 +1,7 @@
use serde::{Deserialize, Serialize};
use super::{HistogramAggregation, HistogramBounds};
use crate::aggregation::AggregationError;
use crate::aggregation::*;
/// DateHistogramAggregation is similar to `HistogramAggregation`, but it can only be used with date
/// type.
@@ -132,6 +132,7 @@ impl DateHistogramAggregationReq {
hard_bounds: self.hard_bounds,
extended_bounds: self.extended_bounds,
keyed: self.keyed,
is_normalized_to_ns: false,
})
}
@@ -243,15 +244,15 @@ fn parse_into_milliseconds(input: &str) -> Result<i64, AggregationError> {
}
#[cfg(test)]
mod tests {
pub mod tests {
use pretty_assertions::assert_eq;
use super::*;
use crate::aggregation::agg_req::Aggregations;
use crate::aggregation::tests::exec_request;
use crate::indexer::NoMergePolicy;
use crate::schema::{Schema, FAST};
use crate::Index;
use crate::schema::{Schema, FAST, STRING};
use crate::{Index, IndexWriter, TantivyDocument};
#[test]
fn test_parse_into_millisecs() {
@@ -306,7 +307,9 @@ mod tests {
) -> crate::Result<Index> {
let mut schema_builder = Schema::builder();
schema_builder.add_date_field("date", FAST);
schema_builder.add_text_field("text", FAST);
schema_builder.add_json_field("mixed", FAST);
schema_builder.add_text_field("text", FAST | STRING);
schema_builder.add_text_field("text2", FAST | STRING);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema.clone());
{
@@ -314,7 +317,7 @@ mod tests {
index_writer.set_merge_policy(Box::new(NoMergePolicy));
for values in segment_and_docs {
for doc_str in values {
let doc = schema.parse_document(doc_str)?;
let doc = TantivyDocument::parse_json(&schema, doc_str)?;
index_writer.add_document(doc)?;
}
// writing the segment
@@ -326,7 +329,7 @@ mod tests {
.searchable_segment_ids()
.expect("Searchable segments failed.");
if segment_ids.len() > 1 {
let mut index_writer = index.writer_for_tests()?;
let mut index_writer: IndexWriter = index.writer_for_tests()?;
index_writer.merge(&segment_ids).wait()?;
index_writer.wait_merging_threads()?;
}
@@ -349,8 +352,10 @@ mod tests {
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": "2015-01-01T11:11:30Z", "text": "bbb" }"#],
vec![r#"{ "date": "2015-01-02T00:00:00Z", "text": "bbb" }"#],
vec![r#"{ "date": "2015-01-06T00:00:00Z", "text": "ccc" }"#],
vec![r#"{ "date": "2015-01-06T00:00:00Z", "text": "ccc" }"#],
];
let index = get_test_index_from_docs(merge_segments, &docs).unwrap();
@@ -379,7 +384,7 @@ mod tests {
{
"key_as_string" : "2015-01-01T00:00:00Z",
"key" : 1420070400000.0,
"doc_count" : 4
"doc_count" : 6
}
]
}
@@ -417,15 +422,15 @@ mod tests {
{
"key_as_string" : "2015-01-01T00:00:00Z",
"key" : 1420070400000.0,
"doc_count" : 4,
"doc_count" : 6,
"texts": {
"buckets": [
{
"doc_count": 2,
"doc_count": 3,
"key": "bbb"
},
{
"doc_count": 1,
"doc_count": 2,
"key": "ccc"
},
{
@@ -464,7 +469,7 @@ mod tests {
"sales_over_time": {
"buckets": [
{
"doc_count": 2,
"doc_count": 3,
"key": 1420070400000.0,
"key_as_string": "2015-01-01T00:00:00Z"
},
@@ -489,7 +494,7 @@ mod tests {
"key_as_string": "2015-01-05T00:00:00Z"
},
{
"doc_count": 1,
"doc_count": 2,
"key": 1420502400000.0,
"key_as_string": "2015-01-06T00:00:00Z"
}
@@ -530,7 +535,7 @@ mod tests {
"key_as_string": "2014-12-31T00:00:00Z"
},
{
"doc_count": 2,
"doc_count": 3,
"key": 1420070400000.0,
"key_as_string": "2015-01-01T00:00:00Z"
},
@@ -555,7 +560,7 @@ mod tests {
"key_as_string": "2015-01-05T00:00:00Z"
},
{
"doc_count": 1,
"doc_count": 2,
"key": 1420502400000.0,
"key_as_string": "2015-01-06T00:00:00Z"
},

85
src/aggregation/bucket/histogram/histogram.rs
View File

@@ -1,8 +1,5 @@
use std::cmp::Ordering;
use std::fmt::Display;
use columnar::ColumnType;
use itertools::Itertools;
use rustc_hash::FxHashMap;
use serde::{Deserialize, Serialize};
use tantivy_bitpacker::minmax;
@@ -18,9 +15,9 @@ use crate::aggregation::intermediate_agg_result::{
IntermediateHistogramBucketEntry,
};
use crate::aggregation::segment_agg_result::{
build_segment_agg_collector, AggregationLimits, SegmentAggregationCollector,
build_segment_agg_collector, SegmentAggregationCollector,
};
use crate::aggregation::{f64_from_fastfield_u64, format_date};
use crate::aggregation::*;
use crate::TantivyError;
/// Histogram is a bucket aggregation, where buckets are created dynamically for given `interval`.
@@ -73,6 +70,7 @@ pub struct HistogramAggregation {
pub field: String,
/// The interval to chunk your data range. Each bucket spans a value range of [0..interval).
/// Must be a positive value.
#[serde(deserialize_with = "deserialize_f64")]
pub interval: f64,
/// Intervals implicitly defines an absolute grid of buckets `[interval * k, interval * (k +
/// 1))`.
@@ -85,6 +83,7 @@ pub struct HistogramAggregation {
/// fall into the buckets with the key 0 and 10.
/// With offset 5 and interval 10, they would both fall into the bucket with they key 5 and the
/// range [5..15)
#[serde(default, deserialize_with = "deserialize_option_f64")]
pub offset: Option<f64>,
/// The minimum number of documents in a bucket to be returned. Defaults to 0.
pub min_doc_count: Option<u64>,
@@ -122,11 +121,14 @@ pub struct HistogramAggregation {
/// Whether to return the buckets as a hash map
#[serde(default)]
pub keyed: bool,
/// Whether the values are normalized to ns for date time values. Defaults to false.
#[serde(default)]
pub is_normalized_to_ns: bool,
}
impl HistogramAggregation {
pub(crate) fn normalize(&mut self, column_type: ColumnType) {
if column_type.is_date_time() {
pub(crate) fn normalize_date_time(&mut self) {
if !self.is_normalized_to_ns {
// values are provided in ms, but the fastfield is in nano seconds
self.interval *= 1_000_000.0;
self.offset = self.offset.map(|off| off * 1_000_000.0);
@@ -138,6 +140,7 @@ impl HistogramAggregation {
min: bounds.min * 1_000_000.0,
max: bounds.max * 1_000_000.0,
});
self.is_normalized_to_ns = true;
}
}
@@ -304,7 +307,10 @@ impl SegmentAggregationCollector for SegmentHistogramCollector {
.column_block_accessor
.fetch_block(docs, &bucket_agg_accessor.accessor);
for (doc, val) in bucket_agg_accessor.column_block_accessor.iter_docid_vals() {
for (doc, val) in bucket_agg_accessor
.column_block_accessor
.iter_docid_vals(docs, &bucket_agg_accessor.accessor)
{
let val = self.f64_from_fastfield_u64(val);
let bucket_pos = get_bucket_pos(val);
@@ -351,6 +357,7 @@ impl SegmentHistogramCollector {
let buckets_mem = self.buckets.memory_consumption();
self_mem + sub_aggs_mem + buckets_mem
}
/// Converts the collector result into a intermediate bucket result.
pub fn into_intermediate_bucket_result(
self,
agg_with_accessor: &AggregationWithAccessor,
@@ -369,7 +376,7 @@ impl SegmentHistogramCollector {
Ok(IntermediateBucketResult::Histogram {
buckets,
column_type: Some(self.column_type),
is_date_agg: self.column_type == ColumnType::DateTime,
})
}
@@ -380,7 +387,9 @@ impl SegmentHistogramCollector {
accessor_idx: usize,
) -> crate::Result<Self> {
req.validate()?;
req.normalize(field_type);
if field_type == ColumnType::DateTime {
req.normalize_date_time();
}
let sub_aggregation_blueprint = if sub_aggregation.is_empty() {
None
@@ -438,6 +447,7 @@ fn intermediate_buckets_to_final_buckets_fill_gaps(
// memory check upfront
let (_, first_bucket_num, last_bucket_num) =
generate_bucket_pos_with_opt_minmax(histogram_req, min_max);
// It's based on user input, so we need to account for overflows
let added_buckets = ((last_bucket_num.saturating_sub(first_bucket_num)).max(0) as u64)
.saturating_sub(buckets.len() as u64);
@@ -453,15 +463,12 @@ fn intermediate_buckets_to_final_buckets_fill_gaps(
let final_buckets: Vec<BucketEntry> = buckets
.into_iter()
.merge_join_by(
fill_gaps_buckets.into_iter(),
|existing_bucket, fill_gaps_bucket| {
existing_bucket
.key
.partial_cmp(fill_gaps_bucket)
.unwrap_or(Ordering::Equal)
},
)
.merge_join_by(fill_gaps_buckets, |existing_bucket, fill_gaps_bucket| {
existing_bucket
.key
.partial_cmp(fill_gaps_bucket)
.unwrap_or(Ordering::Equal)
})
.map(|either| match either {
// Ignore the generated bucket
itertools::EitherOrBoth::Both(existing, _) => existing,
@@ -484,7 +491,7 @@ fn intermediate_buckets_to_final_buckets_fill_gaps(
// Convert to BucketEntry
pub(crate) fn intermediate_histogram_buckets_to_final_buckets(
buckets: Vec<IntermediateHistogramBucketEntry>,
column_type: Option<ColumnType>,
is_date_agg: bool,
histogram_req: &HistogramAggregation,
sub_aggregation: &Aggregations,
limits: &AggregationLimits,
@@ -493,8 +500,8 @@ pub(crate) fn intermediate_histogram_buckets_to_final_buckets(
// The request used in the the call to final is not yet be normalized.
// Normalization is changing the precision from milliseconds to nanoseconds.
let mut histogram_req = histogram_req.clone();
if let Some(column_type) = column_type {
histogram_req.normalize(column_type);
if is_date_agg {
histogram_req.normalize_date_time();
}
let mut buckets = if histogram_req.min_doc_count() == 0 {
// With min_doc_count != 0, we may need to add buckets, so that there are no
@@ -518,7 +525,7 @@ pub(crate) fn intermediate_histogram_buckets_to_final_buckets(
// If we have a date type on the histogram buckets, we add the `key_as_string` field as rfc339
// and normalize from nanoseconds to milliseconds
if column_type == Some(ColumnType::DateTime) {
if is_date_agg {
for bucket in buckets.iter_mut() {
if let crate::aggregation::Key::F64(ref mut val) = bucket.key {
let key_as_string = format_date(*val as i64)?;
@@ -590,11 +597,12 @@ mod tests {
use serde_json::Value;
use super::*;
use crate::aggregation::agg_req::Aggregations;
use crate::aggregation::agg_result::AggregationResults;
use crate::aggregation::tests::{
exec_request, exec_request_with_query, exec_request_with_query_and_memory_limit,
get_test_index_2_segments, get_test_index_from_values, get_test_index_with_num_docs,
};
use crate::query::AllQuery;
#[test]
fn histogram_test_crooked_values() -> crate::Result<()> {
@@ -1346,6 +1354,35 @@ mod tests {
})
);
Ok(())
}
#[test]
fn test_aggregation_histogram_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!({
"myhisto": {
"histogram": {
"field": "score",
"interval": 10.0
},
}
}))
.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)?)?;
// Make sure the result structure is correct
assert_eq!(res["myhisto"]["buckets"].as_array().unwrap().len(), 0);
Ok(())
}
}

10
src/aggregation/bucket/mod.rs
View File

@@ -15,19 +15,25 @@
//! 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)]

28
src/aggregation/bucket/range.rs
View File

@@ -1,7 +1,6 @@
use std::fmt::Debug;
use std::ops::Range;
use columnar::{ColumnType, MonotonicallyMappableToU64};
use rustc_hash::FxHashMap;
use serde::{Deserialize, Serialize};
@@ -14,9 +13,7 @@ use crate::aggregation::intermediate_agg_result::{
use crate::aggregation::segment_agg_result::{
build_segment_agg_collector, SegmentAggregationCollector,
};
use crate::aggregation::{
f64_from_fastfield_u64, f64_to_fastfield_u64, format_date, Key, SerializedKey,
};
use crate::aggregation::*;
use crate::TantivyError;
/// Provide user-defined buckets to aggregate on.
@@ -72,11 +69,19 @@ pub struct RangeAggregationRange {
pub key: Option<String>,
/// The from range value, which is inclusive in the range.
/// `None` equals to an open ended interval.
#[serde(skip_serializing_if = "Option::is_none", default)]
#[serde(
skip_serializing_if = "Option::is_none",
default,
deserialize_with = "deserialize_option_f64"
)]
pub from: Option<f64>,
/// The to range value, which is not inclusive in the range.
/// `None` equals to an open ended interval.
#[serde(skip_serializing_if = "Option::is_none", default)]
#[serde(
skip_serializing_if = "Option::is_none",
default,
deserialize_with = "deserialize_option_f64"
)]
pub to: Option<f64>,
}
@@ -230,7 +235,10 @@ impl SegmentAggregationCollector for SegmentRangeCollector {
.column_block_accessor
.fetch_block(docs, &bucket_agg_accessor.accessor);
for (doc, val) in bucket_agg_accessor.column_block_accessor.iter_docid_vals() {
for (doc, val) in bucket_agg_accessor
.column_block_accessor
.iter_docid_vals(docs, &bucket_agg_accessor.accessor)
{
let bucket_pos = self.get_bucket_pos(val);
let bucket = &mut self.buckets[bucket_pos];
@@ -262,7 +270,7 @@ impl SegmentRangeCollector {
pub(crate) fn from_req_and_validate(
req: &RangeAggregation,
sub_aggregation: &mut AggregationsWithAccessor,
limits: &mut ResourceLimitGuard,
limits: &ResourceLimitGuard,
field_type: ColumnType,
accessor_idx: usize,
) -> crate::Result<Self> {
@@ -441,7 +449,6 @@ pub(crate) fn range_to_key(range: &Range<u64>, field_type: &ColumnType) -> crate
#[cfg(test)]
mod tests {
use columnar::MonotonicallyMappableToU64;
use serde_json::Value;
use super::*;
@@ -450,7 +457,6 @@ mod 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>,
@@ -465,7 +471,7 @@ mod tests {
SegmentRangeCollector::from_req_and_validate(
&req,
&mut Default::default(),
&mut AggregationLimits::default().new_guard(),
&AggregationLimits::default().new_guard(),
field_type,
0,
)

747
src/aggregation/bucket/term_agg.rs
View File

@@ -1,6 +1,10 @@
use std::fmt::Debug;
use std::net::Ipv6Addr;
use columnar::ColumnType;
use columnar::column_values::CompactSpaceU64Accessor;
use columnar::{
BytesColumn, ColumnType, MonotonicallyMappableToU128, MonotonicallyMappableToU64, StrColumn,
};
use rustc_hash::FxHashMap;
use serde::{Deserialize, Serialize};
@@ -9,7 +13,6 @@ use crate::aggregation::agg_limits::MemoryConsumption;
use crate::aggregation::agg_req_with_accessor::{
AggregationWithAccessor, AggregationsWithAccessor,
};
use crate::aggregation::f64_from_fastfield_u64;
use crate::aggregation::intermediate_agg_result::{
IntermediateAggregationResult, IntermediateAggregationResults, IntermediateBucketResult,
IntermediateKey, IntermediateTermBucketEntry, IntermediateTermBucketResult,
@@ -17,6 +20,7 @@ use crate::aggregation::intermediate_agg_result::{
use crate::aggregation::segment_agg_result::{
build_segment_agg_collector, SegmentAggregationCollector,
};
use crate::aggregation::{f64_from_fastfield_u64, format_date, Key};
use crate::error::DataCorruption;
use crate::TantivyError;
@@ -99,23 +103,14 @@ pub struct TermsAggregation {
#[serde(skip_serializing_if = "Option::is_none", default)]
pub size: Option<u32>,
/// Unused by tantivy.
///
/// Since tantivy doesn't know shards, this parameter is merely there to be used by consumers
/// of tantivy. shard_size is the number of terms returned by each shard.
/// The default value in elasticsearch is size * 1.5 + 10.
///
/// Should never be smaller than size.
#[serde(skip_serializing_if = "Option::is_none", default)]
#[serde(alias = "shard_size")]
pub split_size: Option<u32>,
/// The get more accurate results, we fetch more than `size` from each segment.
/// To get more accurate results, we fetch more than `size` from each segment.
///
/// Increasing this value is will increase the cost for more accuracy.
///
/// Defaults to 10 * size.
#[serde(skip_serializing_if = "Option::is_none", default)]
#[serde(alias = "shard_size")]
#[serde(alias = "split_size")]
pub segment_size: Option<u32>,
/// If you set the `show_term_doc_count_error` parameter to true, the terms aggregation will
@@ -146,6 +141,28 @@ pub struct TermsAggregation {
/// { "average_price": "asc" }
#[serde(skip_serializing_if = "Option::is_none", default)]
pub order: Option<CustomOrder>,
/// 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:
/// { "missing": "NO_DATA" }
///
/// # Internal
///
/// Internally, `missing` requires some specialized handling in some scenarios.
///
/// Simple Case:
/// In the simplest case, we can just put the missing value in the termmap use that. In case of
/// text we put a special u64::MAX and replace it at the end with the actual missing value,
/// when loading the text.
/// Special Case 1:
/// If we have multiple columns on one field, we need to have a union on the indices on both
/// columns, to find docids without a value. That requires a special missing aggreggation.
/// Special Case 2: if the key is of type text and the column is numerical, we also need to use
/// the special missing aggregation, since there is no mechanism in the numerical column to
/// add text.
#[serde(skip_serializing_if = "Option::is_none", default)]
pub missing: Option<Key>,
}
/// Same as TermsAggregation, but with populated defaults.
@@ -176,6 +193,7 @@ pub(crate) struct TermsAggregationInternal {
pub min_doc_count: u64,
pub order: CustomOrder,
pub missing: Option<Key>,
}
impl TermsAggregationInternal {
@@ -195,6 +213,7 @@ impl TermsAggregationInternal {
.unwrap_or_else(|| order == CustomOrder::default()),
min_doc_count: req.min_doc_count.unwrap_or(1),
order,
missing: req.missing.clone(),
}
}
}
@@ -224,110 +243,6 @@ impl TermBuckets {
}
}
/// The composite collector is used, when we have different types under one field, to support a term
/// aggregation on both.
#[derive(Clone, Debug)]
pub struct SegmentTermCollectorComposite {
term_agg1: SegmentTermCollector, // field type 1, e.g. strings
term_agg2: SegmentTermCollector, // field type 2, e.g. u64
accessor_idx: usize,
}
impl SegmentAggregationCollector for SegmentTermCollectorComposite {
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 bucket = self
.term_agg1
.into_intermediate_bucket_result(agg_with_accessor)?;
results.push(
name.to_string(),
IntermediateAggregationResult::Bucket(bucket),
)?;
let bucket = self
.term_agg2
.into_intermediate_bucket_result(agg_with_accessor)?;
results.push(name, IntermediateAggregationResult::Bucket(bucket))?;
Ok(())
}
#[inline]
fn collect(
&mut self,
doc: crate::DocId,
agg_with_accessor: &mut AggregationsWithAccessor,
) -> crate::Result<()> {
self.term_agg1.collect_block(&[doc], agg_with_accessor)?;
self.swap_accessor(&mut agg_with_accessor.aggs.values[self.accessor_idx]);
self.term_agg2.collect_block(&[doc], agg_with_accessor)?;
self.swap_accessor(&mut agg_with_accessor.aggs.values[self.accessor_idx]);
Ok(())
}
#[inline]
fn collect_block(
&mut self,
docs: &[crate::DocId],
agg_with_accessor: &mut AggregationsWithAccessor,
) -> crate::Result<()> {
self.term_agg1.collect_block(docs, agg_with_accessor)?;
self.swap_accessor(&mut agg_with_accessor.aggs.values[self.accessor_idx]);
self.term_agg2.collect_block(docs, agg_with_accessor)?;
self.swap_accessor(&mut agg_with_accessor.aggs.values[self.accessor_idx]);
Ok(())
}
fn flush(&mut self, agg_with_accessor: &mut AggregationsWithAccessor) -> crate::Result<()> {
self.term_agg1.flush(agg_with_accessor)?;
self.swap_accessor(&mut agg_with_accessor.aggs.values[self.accessor_idx]);
self.term_agg2.flush(agg_with_accessor)?;
self.swap_accessor(&mut agg_with_accessor.aggs.values[self.accessor_idx]);
Ok(())
}
}
impl SegmentTermCollectorComposite {
/// Swaps the accessor and field type with the second accessor and field type.
/// This way we can use the same code for both aggregations.
fn swap_accessor(&self, aggregations: &mut AggregationWithAccessor) {
if let Some(accessor) = aggregations.accessor2.as_mut() {
std::mem::swap(&mut accessor.0, &mut aggregations.accessor);
std::mem::swap(&mut accessor.1, &mut aggregations.field_type);
}
}
pub(crate) fn from_req_and_validate(
req: &TermsAggregation,
sub_aggregations: &mut AggregationsWithAccessor,
field_type: ColumnType,
field_type2: ColumnType,
accessor_idx: usize,
) -> crate::Result<Self> {
Ok(Self {
term_agg1: SegmentTermCollector::from_req_and_validate(
req,
sub_aggregations,
field_type,
accessor_idx,
)?,
term_agg2: SegmentTermCollector::from_req_and_validate(
req,
sub_aggregations,
field_type2,
accessor_idx,
)?,
accessor_idx,
})
}
}
/// The collector puts values from the fast field into the correct buckets and does a conversion to
/// the correct datatype.
#[derive(Clone, Debug)]
@@ -336,7 +251,7 @@ pub struct SegmentTermCollector {
term_buckets: TermBuckets,
req: TermsAggregationInternal,
blueprint: Option<Box<dyn SegmentAggregationCollector>>,
field_type: ColumnType,
column_type: ColumnType,
accessor_idx: usize,
}
@@ -379,16 +294,26 @@ impl SegmentAggregationCollector for SegmentTermCollector {
let mem_pre = self.get_memory_consumption();
bucket_agg_accessor
.column_block_accessor
.fetch_block(docs, &bucket_agg_accessor.accessor);
if let Some(missing) = bucket_agg_accessor.missing_value_for_accessor {
bucket_agg_accessor
.column_block_accessor
.fetch_block_with_missing(docs, &bucket_agg_accessor.accessor, missing);
} else {
bucket_agg_accessor
.column_block_accessor
.fetch_block(docs, &bucket_agg_accessor.accessor);
}
for term_id in bucket_agg_accessor.column_block_accessor.iter_vals() {
let entry = self.term_buckets.entries.entry(term_id).or_default();
*entry += 1;
}
// has subagg
if let Some(blueprint) = self.blueprint.as_ref() {
for (doc, term_id) in bucket_agg_accessor.column_block_accessor.iter_docid_vals() {
for (doc, term_id) in bucket_agg_accessor
.column_block_accessor
.iter_docid_vals(docs, &bucket_agg_accessor.accessor)
{
let sub_aggregations = self
.term_buckets
.sub_aggs
@@ -428,7 +353,7 @@ impl SegmentTermCollector {
field_type: ColumnType,
accessor_idx: usize,
) -> crate::Result<Self> {
if field_type == ColumnType::Bytes || field_type == ColumnType::Bool {
if field_type == ColumnType::Bytes {
return Err(TantivyError::InvalidArgument(format!(
"terms aggregation is not supported for column type {:?}",
field_type
@@ -462,7 +387,7 @@ impl SegmentTermCollector {
req: TermsAggregationInternal::from_req(req),
term_buckets,
blueprint,
field_type,
column_type: field_type,
accessor_idx,
})
}
@@ -539,23 +464,46 @@ impl SegmentTermCollector {
Ok(intermediate_entry)
};
if self.field_type == ColumnType::Str {
if self.column_type == ColumnType::Str {
let term_dict = agg_with_accessor
.str_dict_column
.as_ref()
.expect("internal error: term dictionary not found for term aggregation");
.cloned()
.unwrap_or_else(|| {
StrColumn::wrap(BytesColumn::empty(agg_with_accessor.accessor.num_docs()))
});
let mut buffer = String::new();
for (term_id, doc_count) in entries {
if !term_dict.ord_to_str(term_id, &mut buffer)? {
return Err(TantivyError::InternalError(format!(
"Couldn't find term_id {term_id} in dict"
)));
}
let intermediate_entry = into_intermediate_bucket_entry(term_id, doc_count)?;
dict.insert(IntermediateKey::Str(buffer.to_string()), intermediate_entry);
// Special case for missing key
if term_id == u64::MAX {
let missing_key = self
.req
.missing
.as_ref()
.expect("Found placeholder term_id but `missing` is None");
match missing_key {
Key::Str(missing) => {
buffer.clear();
buffer.push_str(missing);
dict.insert(
IntermediateKey::Str(buffer.to_string()),
intermediate_entry,
);
}
Key::F64(val) => {
buffer.push_str(&val.to_string());
dict.insert(IntermediateKey::F64(*val), intermediate_entry);
}
}
} else {
if !term_dict.ord_to_str(term_id, &mut buffer)? {
return Err(TantivyError::InternalError(format!(
"Couldn't find term_id {term_id} in dict"
)));
}
dict.insert(IntermediateKey::Str(buffer.to_string()), intermediate_entry);
}
}
if self.req.min_doc_count == 0 {
// TODO: Handle rev streaming for descending sorting by keys
@@ -581,21 +529,55 @@ impl SegmentTermCollector {
});
}
}
} else if self.column_type == ColumnType::DateTime {
for (val, doc_count) in entries {
let intermediate_entry = into_intermediate_bucket_entry(val, doc_count)?;
let val = i64::from_u64(val);
let date = format_date(val)?;
dict.insert(IntermediateKey::Str(date), intermediate_entry);
}
} else if self.column_type == ColumnType::Bool {
for (val, doc_count) in entries {
let intermediate_entry = into_intermediate_bucket_entry(val, doc_count)?;
let val = bool::from_u64(val);
dict.insert(IntermediateKey::Bool(val), intermediate_entry);
}
} else if self.column_type == ColumnType::IpAddr {
let compact_space_accessor = agg_with_accessor
.accessor
.values
.clone()
.downcast_arc::<CompactSpaceU64Accessor>()
.map_err(|_| {
TantivyError::AggregationError(
crate::aggregation::AggregationError::InternalError(
"Type mismatch: Could not downcast to CompactSpaceU64Accessor"
.to_string(),
),
)
})?;
for (val, doc_count) in entries {
let intermediate_entry = into_intermediate_bucket_entry(val, doc_count)?;
let val: u128 = compact_space_accessor.compact_to_u128(val as u32);
let val = Ipv6Addr::from_u128(val);
dict.insert(IntermediateKey::IpAddr(val), intermediate_entry);
}
} else {
for (val, doc_count) in entries {
let intermediate_entry = into_intermediate_bucket_entry(val, doc_count)?;
let val = f64_from_fastfield_u64(val, &self.field_type);
let val = f64_from_fastfield_u64(val, &self.column_type);
dict.insert(IntermediateKey::F64(val), intermediate_entry);
}
};
Ok(IntermediateBucketResult::Terms(
IntermediateTermBucketResult {
Ok(IntermediateBucketResult::Terms {
buckets: IntermediateTermBucketResult {
entries: dict,
sum_other_doc_count,
doc_count_error_upper_bound: term_doc_count_before_cutoff,
},
))
})
}
}
@@ -633,6 +615,12 @@ pub(crate) fn cut_off_buckets<T: GetDocCount + Debug>(
#[cfg(test)]
mod tests {
use std::net::IpAddr;
use std::str::FromStr;
use common::DateTime;
use time::{Date, Month};
use crate::aggregation::agg_req::Aggregations;
use crate::aggregation::tests::{
exec_request, exec_request_with_query, exec_request_with_query_and_memory_limit,
@@ -640,8 +628,8 @@ mod tests {
};
use crate::aggregation::AggregationLimits;
use crate::indexer::NoMergePolicy;
use crate::schema::{Schema, FAST, STRING};
use crate::Index;
use crate::schema::{IntoIpv6Addr, Schema, FAST, STRING};
use crate::{Index, IndexWriter};
#[test]
fn terms_aggregation_test_single_segment() -> crate::Result<()> {
@@ -1222,9 +1210,9 @@ mod tests {
assert_eq!(res["my_texts"]["buckets"][0]["key"], "terma");
assert_eq!(res["my_texts"]["buckets"][0]["doc_count"], 4);
assert_eq!(res["my_texts"]["buckets"][1]["key"], "termc");
assert_eq!(res["my_texts"]["buckets"][1]["key"], "termb");
assert_eq!(res["my_texts"]["buckets"][1]["doc_count"], 0);
assert_eq!(res["my_texts"]["buckets"][2]["key"], "termb");
assert_eq!(res["my_texts"]["buckets"][2]["key"], "termc");
assert_eq!(res["my_texts"]["buckets"][2]["doc_count"], 0);
assert_eq!(res["my_texts"]["sum_other_doc_count"], 0);
assert_eq!(res["my_texts"]["doc_count_error_upper_bound"], 0);
@@ -1321,6 +1309,7 @@ mod tests {
];
let index = get_test_index_from_terms(false, &terms_per_segment)?;
assert_eq!(index.searchable_segments().unwrap().len(), 2);
let agg_req: Aggregations = serde_json::from_value(json!({
"my_texts": {
@@ -1404,7 +1393,7 @@ mod tests {
#[test]
fn terms_aggregation_different_tokenizer_on_ff_test() -> crate::Result<()> {
let terms = vec!["Hello Hello", "Hallo Hallo"];
let terms = vec!["Hello Hello", "Hallo Hallo", "Hallo Hallo"];
let index = get_test_index_from_terms(true, &[terms])?;
@@ -1422,7 +1411,7 @@ mod tests {
println!("{}", serde_json::to_string_pretty(&res).unwrap());
assert_eq!(res["my_texts"]["buckets"][0]["key"], "Hallo Hallo");
assert_eq!(res["my_texts"]["buckets"][0]["doc_count"], 1);
assert_eq!(res["my_texts"]["buckets"][0]["doc_count"], 2);
assert_eq!(res["my_texts"]["buckets"][1]["key"], "Hello Hello");
assert_eq!(res["my_texts"]["buckets"][1]["doc_count"], 1);
@@ -1506,6 +1495,47 @@ mod tests {
Ok(())
}
#[test]
fn terms_empty_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: IndexWriter = 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": "blue"})))
.unwrap();
index_writer.add_document(doc!())?;
index_writer.commit().unwrap();
let agg_req: Aggregations = serde_json::from_value(json!({
"my_texts": {
"terms": {
"field": "json.partially_empty"
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
assert_eq!(res["my_texts"]["buckets"][0]["key"], "blue");
assert_eq!(res["my_texts"]["buckets"][0]["doc_count"], 1);
assert_eq!(res["my_texts"]["buckets"][1], serde_json::Value::Null);
assert_eq!(res["my_texts"]["sum_other_doc_count"], 0);
assert_eq!(res["my_texts"]["doc_count_error_upper_bound"], 0);
Ok(())
}
#[test]
fn terms_aggregation_bytes() -> crate::Result<()> {
@@ -1543,4 +1573,429 @@ mod tests {
Ok(())
}
#[test]
fn terms_aggregation_missing_multi_value() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("text", FAST);
let id_field = schema_builder.add_u64_field("id", FAST);
let index = Index::create_in_ram(schema_builder.build());
{
let mut index_writer = index.writer_with_num_threads(1, 20_000_000)?;
index_writer.set_merge_policy(Box::new(NoMergePolicy));
index_writer.add_document(doc!(
text_field => "Hello Hello",
text_field => "Hello Hello",
id_field => 1u64,
id_field => 1u64,
))?;
// Missing
index_writer.add_document(doc!())?;
index_writer.add_document(doc!(
text_field => "Hello Hello",
))?;
index_writer.add_document(doc!(
text_field => "Hello Hello",
))?;
index_writer.commit()?;
// Empty segment special case
index_writer.add_document(doc!())?;
index_writer.commit()?;
// Full segment special case
index_writer.add_document(doc!(
text_field => "Hello Hello",
id_field => 1u64,
))?;
index_writer.commit()?;
}
let agg_req: Aggregations = serde_json::from_value(json!({
"my_texts": {
"terms": {
"field": "text",
"missing": "Empty"
},
},
"my_texts2": {
"terms": {
"field": "text",
"missing": 1337
},
},
"my_ids": {
"terms": {
"field": "id",
"missing": 1337
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// text field
assert_eq!(res["my_texts"]["buckets"][0]["key"], "Hello Hello");
assert_eq!(res["my_texts"]["buckets"][0]["doc_count"], 5);
assert_eq!(res["my_texts"]["buckets"][1]["key"], "Empty");
assert_eq!(res["my_texts"]["buckets"][1]["doc_count"], 2);
assert_eq!(
res["my_texts"]["buckets"][2]["key"],
serde_json::Value::Null
);
// text field with numner as missing fallback
assert_eq!(res["my_texts2"]["buckets"][0]["key"], "Hello Hello");
assert_eq!(res["my_texts2"]["buckets"][0]["doc_count"], 5);
assert_eq!(res["my_texts2"]["buckets"][1]["key"], 1337.0);
assert_eq!(res["my_texts2"]["buckets"][1]["doc_count"], 2);
assert_eq!(
res["my_texts2"]["buckets"][2]["key"],
serde_json::Value::Null
);
assert_eq!(res["my_texts"]["sum_other_doc_count"], 0);
assert_eq!(res["my_texts"]["doc_count_error_upper_bound"], 0);
// id field
assert_eq!(res["my_ids"]["buckets"][0]["key"], 1337.0);
assert_eq!(res["my_ids"]["buckets"][0]["doc_count"], 4);
assert_eq!(res["my_ids"]["buckets"][1]["key"], 1.0);
assert_eq!(res["my_ids"]["buckets"][1]["doc_count"], 3);
assert_eq!(res["my_ids"]["buckets"][2]["key"], serde_json::Value::Null);
Ok(())
}
#[test]
fn terms_aggregation_missing_simple_id() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let id_field = schema_builder.add_u64_field("id", FAST);
let index = Index::create_in_ram(schema_builder.build());
{
let mut index_writer = index.writer_with_num_threads(1, 20_000_000)?;
index_writer.set_merge_policy(Box::new(NoMergePolicy));
index_writer.add_document(doc!(
id_field => 1u64,
))?;
// Missing
index_writer.add_document(doc!())?;
index_writer.add_document(doc!())?;
index_writer.commit()?;
}
let agg_req: Aggregations = serde_json::from_value(json!({
"my_ids": {
"terms": {
"field": "id",
"missing": 1337
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// id field
assert_eq!(res["my_ids"]["buckets"][0]["key"], 1337.0);
assert_eq!(res["my_ids"]["buckets"][0]["doc_count"], 2);
assert_eq!(res["my_ids"]["buckets"][1]["key"], 1.0);
assert_eq!(res["my_ids"]["buckets"][1]["doc_count"], 1);
assert_eq!(res["my_ids"]["buckets"][2]["key"], serde_json::Value::Null);
Ok(())
}
#[test]
fn terms_aggregation_missing1() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let text_field = schema_builder.add_text_field("text", FAST);
let id_field = schema_builder.add_u64_field("id", FAST);
let index = Index::create_in_ram(schema_builder.build());
{
let mut index_writer = index.writer_with_num_threads(1, 20_000_000)?;
index_writer.set_merge_policy(Box::new(NoMergePolicy));
index_writer.add_document(doc!(
text_field => "Hello Hello",
id_field => 1u64,
))?;
// Missing
index_writer.add_document(doc!())?;
index_writer.add_document(doc!(
text_field => "Hello Hello",
))?;
index_writer.add_document(doc!(
text_field => "Hello Hello",
))?;
index_writer.commit()?;
// Empty segment special case
index_writer.add_document(doc!())?;
index_writer.commit()?;
// Full segment special case
index_writer.add_document(doc!(
text_field => "Hello Hello",
id_field => 1u64,
))?;
index_writer.commit()?;
}
let agg_req: Aggregations = serde_json::from_value(json!({
"my_texts": {
"terms": {
"field": "text",
"missing": "Empty"
},
},
"my_texts2": {
"terms": {
"field": "text",
"missing": 1337
},
},
"my_ids": {
"terms": {
"field": "id",
"missing": 1337
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// text field
assert_eq!(res["my_texts"]["buckets"][0]["key"], "Hello Hello");
assert_eq!(res["my_texts"]["buckets"][0]["doc_count"], 4);
assert_eq!(res["my_texts"]["buckets"][1]["key"], "Empty");
assert_eq!(res["my_texts"]["buckets"][1]["doc_count"], 2);
assert_eq!(
res["my_texts"]["buckets"][2]["key"],
serde_json::Value::Null
);
// text field with numner as missing fallback
assert_eq!(res["my_texts2"]["buckets"][0]["key"], "Hello Hello");
assert_eq!(res["my_texts2"]["buckets"][0]["doc_count"], 4);
assert_eq!(res["my_texts2"]["buckets"][1]["key"], 1337.0);
assert_eq!(res["my_texts2"]["buckets"][1]["doc_count"], 2);
assert_eq!(
res["my_texts2"]["buckets"][2]["key"],
serde_json::Value::Null
);
assert_eq!(res["my_texts"]["sum_other_doc_count"], 0);
assert_eq!(res["my_texts"]["doc_count_error_upper_bound"], 0);
// id field
assert_eq!(res["my_ids"]["buckets"][0]["key"], 1337.0);
assert_eq!(res["my_ids"]["buckets"][0]["doc_count"], 4);
assert_eq!(res["my_ids"]["buckets"][1]["key"], 1.0);
assert_eq!(res["my_ids"]["buckets"][1]["doc_count"], 2);
assert_eq!(res["my_ids"]["buckets"][2]["key"], serde_json::Value::Null);
Ok(())
}
#[test]
fn terms_aggregation_missing_empty() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
schema_builder.add_text_field("text", FAST);
schema_builder.add_u64_field("id", FAST);
let index = Index::create_in_ram(schema_builder.build());
{
let mut index_writer = index.writer_with_num_threads(1, 20_000_000)?;
index_writer.set_merge_policy(Box::new(NoMergePolicy));
// Empty segment special case
index_writer.add_document(doc!())?;
index_writer.commit()?;
}
let agg_req: Aggregations = serde_json::from_value(json!({
"my_texts": {
"terms": {
"field": "text",
"missing": "Empty"
},
},
"my_texts2": {
"terms": {
"field": "text",
"missing": 1337
},
},
"my_ids": {
"terms": {
"field": "id",
"missing": 1337
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// text field
assert_eq!(res["my_texts"]["buckets"][0]["key"], "Empty");
assert_eq!(res["my_texts"]["buckets"][0]["doc_count"], 1);
assert_eq!(
res["my_texts"]["buckets"][1]["key"],
serde_json::Value::Null
);
// text field with number as missing fallback
assert_eq!(res["my_texts2"]["buckets"][0]["key"], 1337.0);
assert_eq!(res["my_texts2"]["buckets"][0]["doc_count"], 1);
assert_eq!(
res["my_texts2"]["buckets"][1]["key"],
serde_json::Value::Null
);
assert_eq!(res["my_texts"]["sum_other_doc_count"], 0);
assert_eq!(res["my_texts"]["doc_count_error_upper_bound"], 0);
// id field
assert_eq!(res["my_ids"]["buckets"][0]["key"], 1337.0);
assert_eq!(res["my_ids"]["buckets"][0]["doc_count"], 1);
assert_eq!(res["my_ids"]["buckets"][1]["key"], serde_json::Value::Null);
Ok(())
}
#[test]
fn terms_aggregation_date() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let date_field = schema_builder.add_date_field("date_field", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{
let mut writer = index.writer_with_num_threads(1, 15_000_000)?;
writer.add_document(doc!(date_field=>DateTime::from_primitive(Date::from_calendar_date(1982, Month::September, 17)?.with_hms(0, 0, 0)?)))?;
writer.add_document(doc!(date_field=>DateTime::from_primitive(Date::from_calendar_date(1982, Month::September, 17)?.with_hms(0, 0, 0)?)))?;
writer.add_document(doc!(date_field=>DateTime::from_primitive(Date::from_calendar_date(1983, Month::September, 27)?.with_hms(0, 0, 0)?)))?;
writer.commit()?;
}
let agg_req: Aggregations = serde_json::from_value(json!({
"my_date": {
"terms": {
"field": "date_field"
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// date_field field
assert_eq!(res["my_date"]["buckets"][0]["key"], "1982-09-17T00:00:00Z");
assert_eq!(res["my_date"]["buckets"][0]["doc_count"], 2);
assert_eq!(res["my_date"]["buckets"][1]["key"], "1983-09-27T00:00:00Z");
assert_eq!(res["my_date"]["buckets"][1]["doc_count"], 1);
assert_eq!(res["my_date"]["buckets"][2]["key"], serde_json::Value::Null);
Ok(())
}
#[test]
fn terms_aggregation_date_missing() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let date_field = schema_builder.add_date_field("date_field", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{
let mut writer = index.writer_with_num_threads(1, 15_000_000)?;
writer.add_document(doc!(date_field=>DateTime::from_primitive(Date::from_calendar_date(1982, Month::September, 17)?.with_hms(0, 0, 0)?)))?;
writer.add_document(doc!(date_field=>DateTime::from_primitive(Date::from_calendar_date(1982, Month::September, 17)?.with_hms(0, 0, 0)?)))?;
writer.add_document(doc!(date_field=>DateTime::from_primitive(Date::from_calendar_date(1983, Month::September, 27)?.with_hms(0, 0, 0)?)))?;
writer.add_document(doc!())?;
writer.commit()?;
}
let agg_req: Aggregations = serde_json::from_value(json!({
"my_date": {
"terms": {
"field": "date_field",
"missing": "1982-09-17T00:00:00Z"
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// date_field field
assert_eq!(res["my_date"]["buckets"][0]["key"], "1982-09-17T00:00:00Z");
assert_eq!(res["my_date"]["buckets"][0]["doc_count"], 3);
assert_eq!(res["my_date"]["buckets"][1]["key"], "1983-09-27T00:00:00Z");
assert_eq!(res["my_date"]["buckets"][1]["doc_count"], 1);
assert_eq!(res["my_date"]["buckets"][2]["key"], serde_json::Value::Null);
Ok(())
}
#[test]
fn terms_aggregation_bool() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let field = schema_builder.add_bool_field("bool_field", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{
let mut writer = index.writer_with_num_threads(1, 15_000_000)?;
writer.add_document(doc!(field=>true))?;
writer.add_document(doc!(field=>false))?;
writer.add_document(doc!(field=>true))?;
writer.commit()?;
}
let agg_req: Aggregations = serde_json::from_value(json!({
"my_bool": {
"terms": {
"field": "bool_field"
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
assert_eq!(res["my_bool"]["buckets"][0]["key"], 1.0);
assert_eq!(res["my_bool"]["buckets"][0]["key_as_string"], "true");
assert_eq!(res["my_bool"]["buckets"][0]["doc_count"], 2);
assert_eq!(res["my_bool"]["buckets"][1]["key"], 0.0);
assert_eq!(res["my_bool"]["buckets"][1]["key_as_string"], "false");
assert_eq!(res["my_bool"]["buckets"][1]["doc_count"], 1);
assert_eq!(res["my_bool"]["buckets"][2]["key"], serde_json::Value::Null);
Ok(())
}
#[test]
fn terms_aggregation_ip_addr() -> crate::Result<()> {
let mut schema_builder = Schema::builder();
let field = schema_builder.add_ip_addr_field("ip_field", FAST);
let schema = schema_builder.build();
let index = Index::create_in_ram(schema);
{
let mut writer = index.writer_with_num_threads(1, 15_000_000)?;
// IpV6 loopback
writer.add_document(doc!(field=>IpAddr::from_str("::1").unwrap().into_ipv6_addr()))?;
writer.add_document(doc!(field=>IpAddr::from_str("::1").unwrap().into_ipv6_addr()))?;
// IpV4
writer.add_document(
doc!(field=>IpAddr::from_str("127.0.0.1").unwrap().into_ipv6_addr()),
)?;
writer.commit()?;
}
let agg_req: Aggregations = serde_json::from_value(json!({
"my_bool": {
"terms": {
"field": "ip_field"
},
}
}))
.unwrap();
let res = exec_request_with_query(agg_req, &index, None)?;
// print as json
// println!("{}", serde_json::to_string_pretty(&res).unwrap());
assert_eq!(res["my_bool"]["buckets"][0]["key"], "::1");
assert_eq!(res["my_bool"]["buckets"][0]["doc_count"], 2);
assert_eq!(res["my_bool"]["buckets"][1]["key"], "127.0.0.1");
assert_eq!(res["my_bool"]["buckets"][1]["doc_count"], 1);
assert_eq!(res["my_bool"]["buckets"][2]["key"], serde_json::Value::Null);
Ok(())
}
}

481
src/aggregation/bucket/term_missing_agg.rs Normal file
View File

@@ -0,0 +1,481 @@
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 {
buckets: 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, IndexWriter};
#[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: IndexWriter = 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: IndexWriter = 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: IndexWriter = 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: IndexWriter = 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: IndexWriter = 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: IndexWriter = 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: IndexWriter = 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(())
}
}

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