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

Compare commits

base: rust: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:ownedbytes-v0.9.0
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

332 Commits
0.20.2 ... ownedbytes

Author SHA1 Message Date
Pascal Seitz
c6e77d27c6 chore: Release 2025-04-09 16:58:45 +08:00
Pascal Seitz
db6587ed9b chore: Release 2025-04-09 16:57:04 +08:00
Paul Masurel
3fa90e70e2 Merge pull request #2618 from quickwit-oss/release_tantivy 
fix tantivy-query-grammar version
 2025-04-09 09:54:09 +02:00
Pascal Seitz
6ab4102253 fix tantivy-query-grammar version 2025-04-09 14:35:23 +08:00
PSeitz
11c6329ca5 temp unbump version (#2501) 
temp unbump to 0.22 for easier release with `cargo release`
 2025-04-09 08:09:41 +02:00
PSeitz
ab8bb93928 update changelog (#2617) 2025-04-09 03:31:30 +02:00
PSeitz
2b668bd2bf readability improvement on executor (#2615) 2025-04-08 18:28:49 +02:00
Paul Masurel
97a7137ef8 Merge pull request #2606 from katlim-br/add_serde_serialize 
Add serde json serialize to UserInputAst
 2025-04-03 15:57:03 +02:00
Kat Lim Ruiz
ffa7cdf397 agreed with Remi, about the final json structure, having "type" tag and using "clauses" is more accurate 2025-04-03 08:35:16 -05:00
Kat Lim Ruiz
caf1275e60 Merge pull request #1 from quickwit-oss/tagged-user-input-ast 
Tag UserInputAst
 2025-04-03 08:30:07 -05:00
Remi Dettai
fb12b7be28 Tag UserInputAst 2025-04-03 10:07:34 +02:00
Kat Lim Ruiz
6f77083493 create more complex unit test 2025-04-02 18:06:20 -05:00
Kat Lim Ruiz
cd7745da7a set Leaf untagged, leave clause and boost the same (with own property) 2025-04-02 17:52:18 -05:00
Kat Lim Ruiz
eb8304dee9 remove untitled file 2025-04-02 08:47:58 -05:00
Kat Lim Ruiz
e5638112a9 all json should be snake_case 2025-04-02 08:45:33 -05:00
Kat Lim Ruiz
81110152fb add unit test for unbounded 2025-04-01 18:08:04 -05:00
Kat Lim Ruiz
ae88a7ece5 add tag type and content value to UserInputBound 2025-04-01 18:06:40 -05:00
Kat Lim Ruiz
bdd5f80fd9 add clause unit test 2025-04-01 18:04:19 -05:00
Kat Lim Ruiz
3f62ef22e5 set tag=type only for Leaf 2025-04-01 17:52:36 -05:00
Kat Lim Ruiz
8102e19e48 set Error as serializable because is part of the possible outcomes (however, I think using this empty Error struct is not a good pattern) 2025-04-01 17:43:24 -05:00
Kat Lim Ruiz
175c853ea7 add serialization test for LenientError 2025-04-01 17:38:23 -05:00
Kat Lim Ruiz
c992cf3f37 Revert "set all enum to be snake_case when serializing" 
This reverts commit 83f6c2f265.
 2025-04-01 17:27:28 -05:00
Kat Lim Ruiz
83f6c2f265 set all enum to be snake_case when serializing 2025-04-01 17:13:04 -05:00
Kat Lim Ruiz
17bf8aa092 Merge branch 'quickwit-oss:main' into add_serde_serialize 2025-04-01 08:32:08 -05:00
trinity-1686a
6fc0e96ff8 Merge pull request #2610 from quickwit-oss/fix-compilation-stability 
Fix compilation stability
 2025-04-01 10:45:58 +02:00
Remi Dettai
06d2dcf469 Further fix type inference tests 2025-04-01 09:52:22 +02:00
Remi Dettai
b681ec9335 Fix compilation stability 2025-04-01 09:33:33 +02:00
Kat Lim Ruiz
da2ff5712a fix fmt nightly 2025-03-31 08:21:54 -05:00
Kat Lim Ruiz
18da402e27 cargo fmt 2025-03-30 22:10:38 -05:00
Kat Lim Ruiz
18ae3ffe94 uniformize root cargo.toml 2025-03-30 21:55:51 -05:00
Kat Lim Ruiz
0a37b7acaa update to latest serde and serde_json (and follow the pattern to use patch versions) 2025-03-30 11:35:58 -05:00
Kat Lim Ruiz
1a9fd885dd allow LenientError to be serializable too 2025-03-30 11:26:20 -05:00
Kat Lim Ruiz
3e660905a7 unit test parse_query_lenient 2025-03-30 11:22:22 -05:00
Kat Lim Ruiz
0c2b984cb4 add tests 2025-03-30 11:12:15 -05:00
Kat Lim Ruiz
a69b1c609c add error to be debuggable 2025-03-30 11:12:12 -05:00
Kat Lim Ruiz
8d4a6fcaba deserialize is not needed 2025-03-30 11:11:55 -05:00
Kat Lim Ruiz
feced4762f update root cargo.toml 2025-03-30 11:01:22 -05:00
Kat Lim Ruiz
0149317c5a set 0.23 2025-03-30 10:55:48 -05:00
Kat Lim Ruiz
3fcb6f9597 add unit tests 2025-03-30 10:41:43 -05:00
Kat Lim Ruiz
388fcd763b add serde, and allow UserInputAst to be json serialized/deserialized 2025-03-30 10:36:43 -05:00
trinity-1686a
e488f9e6a2 Merge pull request #2598 from quickwit-oss/1686a/agg-key-eq 
fix invalid impl of Eq on Key
 2025-03-14 15:24:31 +01:00
trinity Pointard
9426d5be7b fix agg Key PartialEq impl 2025-03-14 14:57:45 +01:00
PSeitz
d5d2d41264 merge column: small refactors (#2579) 
* merge column: small refactors

* make ord dependency more explicit

* add columnar merge crashtest proptest

* fix naming
 2025-03-07 18:52:34 +08:00
Paul Masurel
80f5f1ecd4 Merge pull request #2586 from quickwit-oss/issue/2577-get_batch_multiply_overflow 
follow up on the fix of multiply with overflow
 2025-03-05 11:17:12 +01:00
Paul Masurel
519e5d2ed1 clippy warnings 2025-03-05 11:15:06 +01:00
Paul Masurel
df2d52a84e follow up on the fix of multiply with overflow 2025-03-05 11:15:05 +01:00
Paul Masurel
371dba9414 Merge pull request #2591 from quickwit-oss/cargo-fmt 
Cargo fmt
 2025-03-05 11:08:06 +01:00
Paul Masurel
0afabad494 Cargo fmt 2025-03-05 11:07:46 +01:00
Remi Dettai
89b052cd42 Catch panics during merges (#2582) 
* Adding panic handler for the rayon merge thread pool

* Return panic message in error

---------

Co-authored-by: Paul Masurel <paul.masurel@datadoghq.com>
 2025-03-05 10:36:48 +01:00
SteveLauC
c48c649436 refactor: use std AtomicU64 and remove wrapper (#2585) 2025-02-24 03:56:15 +01:00
Paul Masurel
58c0739953 Merge pull request #2581 from quickwit-oss/merge_dict_column_repro 
use usize in bitpacker
 2025-02-21 10:53:07 +09:00
Pascal Seitz
e7daf69de9 use usize in bitpacker 
use usize in bitpacker to enable larger columns in the columnar store

Godbolt comparison with u32 vs u64 for get access: https://godbolt.org/z/cjf7nenYP

Add a mini-tool to inspect columnar files created by tantivy. (very basic functionality which can be extended later)
 2025-02-20 15:39:10 +01:00
trinity-1686a
f060e86bc6 Merge pull request #2578 from quickwit-oss/1686a/buildable-histo-agg 
make DateHistogramAggregationReq buildable
 2025-02-18 15:30:54 +01:00
trinity Pointard
0368162ef0 make DateHistogramAggregationReq buildable 2025-02-18 11:45:24 +01:00
trinity-1686a
e843c71015 Merge pull request #2568 from quickwit-oss/trinity/wildcard-query-parser 
allow term starting with wildcard in query parser
 2025-02-12 16:47:25 +01:00
trinity Pointard
5cea16ef9f improve handling of spcial char after exist query 2025-01-22 16:04:31 +01:00
dependabot[bot]
4aa8cd2470 Update downcast-rs requirement from 1.2.1 to 2.0.1 (#2566) 
Updates the requirements on [downcast-rs](https://github.com/marcianx/downcast-rs) to permit the latest version.
- [Changelog](https://github.com/marcianx/downcast-rs/blob/master/CHANGELOG.md)
- [Commits](https://github.com/marcianx/downcast-rs/compare/v1.2.1...v2.0.1)

---
updated-dependencies:
- dependency-name: downcast-rs
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2025-01-22 10:32:24 +01:00
trinity Pointard
4d4ee1b0ac allow term starting with wildcard in query parser 2025-01-15 10:27:48 +01:00
dependabot[bot]
43c89b4360 Update itertools requirement from 0.13.0 to 0.14.0 (#2563) 
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/v0.13.0...v0.14.0)

---
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>
 2025-01-08 17:11:46 +01:00
trinity-1686a
d281ca3e65 Merge pull request #2559 from quickwit-oss/trinity/sstable-partial-automaton 
allow warming partially an sstable for an automaton
 2025-01-08 16:35:35 +01:00
trinity Pointard
be17daf658 split iterator 2025-01-08 16:24:34 +01:00
trinity Pointard
6ca84a61fa make termdict always clone 2025-01-08 16:19:54 +01:00
trinity Pointard
037d12c9c9 fix deadlocking on automaton warmup 2025-01-06 11:58:58 +01:00
Remi Dettai
71cf19870b Exist queries match subpath fields (#2558) 
* Exist queries match subpath fields

* Make subpath check optional

* Add async subpath listing
 2025-01-06 10:17:39 +01:00
trinity Pointard
175a529c41 use executor for cpu-heavy sstable decompression for automaton 2025-01-03 19:14:07 +01:00
trinity Pointard
fe0c7c5408 change rangebound style 2025-01-02 11:56:05 +01:00
Harrison Burt
148594f0f9 Improve IndexWriter customisation via builder (#2562) 
* Improve `IndexWriter` customisation via builder

* Remove change noise from PR

* Correct documentation

* Resolve comments and add test
 2025-01-02 09:43:22 +01:00
dependabot[bot]
8edb439440 Update rustc-hash requirement from 1.1.0 to 2.1.0 (#2551) 
---
updated-dependencies:
- dependency-name: rustc-hash
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2024-12-26 10:25:05 +01:00
trinity Pointard
dfff5f3bcb rename merge_holes_under => merge_holes_under_bytes 2024-12-23 16:17:44 +01:00
trinity-1686a
ebf4d84553 add comment about cpu-intensive operation in async context 2024-12-20 12:23:49 +01:00
trinity-1686a
42efc7f7c8 clippy 2024-12-20 11:00:11 +01:00
trinity-1686a
192395c311 attempt at simplifying can_block_match_automaton 2024-12-20 10:25:38 +01:00
trinity-1686a
a1447cc9c2 remove breaking change in sstable public api 2024-12-19 17:30:05 +01:00
trinity-1686a
c39d91f827 Merge pull request #2547 from quickwit-oss/trinity/count-str 
add support for counting non integer in aggregation
 2024-12-17 15:27:30 +01:00
trinity Pointard
32b6e9711b add tests 2024-12-13 16:06:24 +01:00
trinity-1686a
24c5dc2398 allow warming up automaton 2024-12-10 13:32:12 +01:00
trinity-1686a
9e2ddec4b3 merge adjacent block when building delta for automaton 2024-12-10 13:32:12 +01:00
trinity-1686a
1f6a8e74bb support iterating over partially loaded sstable 2024-12-10 13:32:12 +01:00
trinity-1686a
7e901f523b get iter for blocks of sstable matching automaton 2024-12-10 13:32:12 +01:00
trinity-1686a
3c30a41c14 add helper to figure if block can match automaton 2024-12-10 13:32:12 +01:00
dependabot[bot]
0f99d4f420 Update measure_time requirement from 0.8.2 to 0.9.0 (#2557) 
---
updated-dependencies:
- dependency-name: measure_time
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2024-12-09 21:39:01 +01:00
Pierre Barre
6e02c5cb25 Make NUM_MERGE_THREADS configurable (#2535) 
* Make `NUM_MERGE_THREADS` configurable

* Remove unused import

* Reword comment src/index/index.rs

Co-authored-by: PSeitz <PSeitz@users.noreply.github.com>

---------

Co-authored-by: PSeitz <PSeitz@users.noreply.github.com>
 2024-12-09 16:53:11 +08:00
PSeitz
876a579e5d queryparser: add field respecification test (#2550) 2024-12-02 14:17:12 +01:00
PSeitz
4c52499622 clippy (#2549) 2024-11-29 16:08:21 +08:00
trinity-1686a
0bac391291 add support for counting non integer in aggregation 2024-11-28 19:52:47 +01:00
PSeitz
52d4e81e70 update CHANGELOG (#2546) 2024-11-27 20:49:35 +08:00
dependabot[bot]
c71ea7b2ef Update thiserror requirement from 1.0.30 to 2.0.1 (#2542) 
Updates the requirements on [thiserror](https://github.com/dtolnay/thiserror) to permit the latest version.
- [Release notes](https://github.com/dtolnay/thiserror/releases)
- [Commits](https://github.com/dtolnay/thiserror/compare/1.0.30...2.0.1)

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

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2024-11-09 08:08:34 +08:00
Paul Masurel
c35a782747 Updating rustc-hash and clippy fixes (#2532) 
* Updating rustc-hash and clippy fixes

* fix terms_aggregation_min_doc_count_special_case

---------

Co-authored-by: Pascal Seitz <pascal.seitz@gmail.com>
 2024-11-01 13:46:26 +08:00
dependabot[bot]
c66af2c0a9 Update binggan requirement from 0.12.0 to 0.14.0 (#2530) 
* Update binggan requirement from 0.12.0 to 0.14.0

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

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

* fix build

---------

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Pascal Seitz <pascal.seitz@gmail.com>
 2024-10-24 09:41:35 +08:00
Joan Antoni RE
f9ac055847 Fix some links in architecture docs (#2528) 2024-10-23 21:06:54 +09:00
PSeitz
21d057059e clippy (#2527) 
* clippy

* clippy

* clippy

* clippy

* convert allow to expect and remove unused

* cargo fmt

* cleanup

* export sample

* clippy
 2024-10-22 09:26:54 +08:00
PSeitz
dca508b4ca remove read_postings_no_deletes (#2526) 
closes #2525
 2024-10-22 09:52:43 +09:00
PSeitz
aebae9965d add RegexPhraseQuery (#2516) 
* add RegexPhraseQuery

RegexPhraseQuery supports phrase queries with regex. It supports regex
and wildcards. E.g. a query with wildcards:
"b* b* wolf" matches "big bad wolf"
Slop is supported as well:
"b* wolf"~2 matches "big bad wolf"

Regex queries may match a lot of terms where we still need to
keep track which term hit to load the positions.
The phrase query algorithm groups terms by their frequency
together in the union to prefilter groups early.

This PR comes with some new datastructures:

SimpleUnion - A union docset for a list of docsets. It doesn't do any
caching and is therefore well suited for datasets with lots of skipping.
(phrase search, but intersections in general)

LoadedPostings - Like SegmentPostings, but all docs and positions are loaded in
memory. SegmentPostings uses 1840 bytes per instance with its caches,
which is equivalent to 460 docids.
LoadedPostings is used for terms which have less than 100 docs.
LoadedPostings is only used to reduce memory consumption.

BitSetPostingUnion - Creates a `Posting` that uses the bitset for docid
hits and the docsets for positions. The BitSet is the precalculated
union of the docsets
In the RegexPhraseQuery there is a size limit of 512 docsets per PreAggregatedUnion,
before creating a new one.

Renamed Union to BufferedUnionScorer
Added proptests to test different union types.

* cleanup

* use Box instead of Vec

* use RefCell instead of term_freq(&mut)

* remove wildcard mode

* move RefCell to outer

* clippy
 2024-10-21 18:29:17 +08:00
Marvin
e7e3e3f44c make casing in docs more consistent (#2524) 
* make casing in docs more consistent

* more

* lowercase tantivy
 2024-10-21 17:59:41 +09:00
PSeitz
2f2db16ec1 store DateTime as nanoseconds in doc store (#2486) 
* store DateTime as nanoseconds in doc store

The doc store DateTime was truncated to microseconds previously. This
removes this truncation, while still keeping backwards compatibility.

This is done by adding the trait `ConfigurableBinarySerializable`, which
works like `BinarySerializable`, but with a config that allows de/serialize
as different date time precision currently.

bump version format to 7.
add compat test to check the date time truncation.

* remove configurable binary serialize, add enum for doc store version

* test doc store version ord
 2024-10-18 10:50:20 +08:00
Paul Masurel
d152e29687 Fixed citation (#2523) 2024-10-17 10:19:50 +09:00
Paul Masurel
285bcc25c9 Added citation.cff (#2522) 2024-10-17 09:43:35 +09:00
PSeitz
7b65ad922d use binggan for stacker bench (#2492) 
* use binggan for stacker bench

```
alice (num terms: 174693)
hashmap                    Memory: 1.3 MB     Avg: 367.19 MiB/s (-1.34%)    Median: 368.10 MiB/s (-1.34%)    [378.75 MiB/s .. 352.81 MiB/s]
hasmap with postings       Memory: 2.4 MB     Avg: 237.29 MiB/s (-2.19%)    Median: 240.22 MiB/s (-1.61%)    [248.26 MiB/s .. 210.66 MiB/s]
fxhashmap ref postings     Memory: 2.9 MB     Avg: 171.94 MiB/s (-3.22%)    Median: 174.13 MiB/s (-2.69%)    [185.94 MiB/s .. 152.43 MiB/s]
fxhasmap owned postings    Memory: 3.5 MB     Avg: 96.993 MiB/s (-4.20%)    Median: 97.410 MiB/s (-4.48%)    [102.78 MiB/s .. 82.745 MiB/s]
numbers unique 100k
hashmap                 Memory: 5.2 MB     Avg: 334.17 MiB/s (-3.06%)    Median: 352.61 MiB/s (+0.77%)    [362.60 MiB/s .. 213.03 MiB/s]
hasmap with postings    Memory: 6.3 MB     Avg: 316.96 MiB/s (-0.02%)    Median: 325.16 MiB/s (-0.04%)    [338.36 MiB/s .. 218.60 MiB/s]
zipfs numbers 100k
hashmap                 Memory: 1.3 MB     Avg: 1.2342 GiB/s (+2.87%)    Median: 1.2677 GiB/s (+4.66%)    [1.3130 GiB/s .. 915.93 MiB/s]
hasmap with postings    Memory: 2.4 MB     Avg: 485.16 MiB/s (+2.68%)    Median: 494.70 MiB/s (+4.42%)    [505.31 MiB/s .. 413.14 MiB/s]
numbers unique 1mio
hashmap                 Memory: 35.7 MB     Avg: 169.68 MiB/s (-1.08%)    Median: 166.80 MiB/s (-3.87%)    [201.33 MiB/s .. 154.26 MiB/s]
hasmap with postings    Memory: 39.8 MB     Avg: 149.49 MiB/s (-3.07%)    Median: 150.85 MiB/s (-1.45%)    [160.76 MiB/s .. 130.94 MiB/s]
zipfs numbers 1mio
hashmap                 Memory: 1.3 MB     Avg: 1.2185 GiB/s (-2.33%)     Median: 1.2291 GiB/s (-2.33%)     [1.2905 GiB/s .. 1.0742 GiB/s]
hasmap with postings    Memory: 5.5 MB     Avg: 358.43 MiB/s (-11.63%)    Median: 356.95 MiB/s (-12.85%)    [444.94 MiB/s .. 302.46 MiB/s]
numbers unique 2mio
hashmap                 Memory: 70.3 MB     Avg: 163.65 MiB/s (+8.37%)    Median: 162.83 MiB/s (+8.80%)    [190.20 MiB/s .. 144.70 MiB/s]
hasmap with postings    Memory: 78.6 MB     Avg: 148.00 MiB/s (+7.75%)    Median: 151.53 MiB/s (+9.11%)    [166.92 MiB/s .. 120.09 MiB/s]
zipfs numbers 2mio
hashmap                 Memory: 1.3 MB     Avg: 1.2535 GiB/s (+2.59%)    Median: 1.2654 GiB/s (+0.36%)    [1.2938 GiB/s .. 1.0592 GiB/s]
hasmap with postings    Memory: 9.7 MB     Avg: 377.96 MiB/s (-4.94%)    Median: 381.82 MiB/s (-3.67%)    [426.14 MiB/s .. 335.66 MiB/s]
numbers unique 5mio
hashmap                 Memory: 277.9 MB     Avg: 121.30 MiB/s (+2.00%)    Median: 121.99 MiB/s (+2.99%)    [132.51 MiB/s .. 110.32 MiB/s]
hasmap with postings    Memory: 295.7 MB     Avg: 114.23 MiB/s (+2.13%)    Median: 115.26 MiB/s (+2.94%)    [124.08 MiB/s .. 103.38 MiB/s]
zipfs numbers 5mio
hashmap                 Memory: 1.3 MB      Avg: 1.2326 GiB/s (+0.63%)    Median: 1.2400 GiB/s (+0.71%)    [1.2755 GiB/s .. 1.0923 GiB/s]
hasmap with postings    Memory: 25.4 MB     Avg: 360.49 MiB/s (+1.07%)    Median: 363.44 MiB/s (+1.27%)    [404.88 MiB/s .. 300.38 MiB/s]
```

* rename bench

* update binggan

* rename to HASHMAP_CAPACITY
 2024-10-16 11:41:33 +08:00
dependabot[bot]
99be20cedd Update binggan requirement from 0.10.0 to 0.12.0 (#2519) 
* Update binggan requirement from 0.10.0 to 0.12.0

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

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

* fix build

---------

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Pascal Seitz <pascal.seitz@gmail.com>
 2024-10-16 11:36:04 +08:00
Bruce Mitchener
5f026901b8 Update MSRV to 1.75 (#2515) 
This is required by the `fs4` dependency. There are other
things that need something later than 1.66.

Both quickwit and the Python binding already require something
newer.
 2024-10-16 10:32:16 +08:00
baishen
6dfa2df06f fix OwnedBytes debug panic (#2512) 2024-10-16 10:31:40 +08:00
Bruce Mitchener
c17e513377 Reduce typo count. (#2510) 2024-10-10 09:55:37 +08:00
PSeitz
2f5a269c70 update packages (#2500) 
fixes some warnings
 2024-09-25 17:46:18 +08:00
PSeitz
50532260e3 update changelog (#2496) 2024-09-25 10:28:53 +08:00
Tri
8bd6eb06e6 feat: make SegmentMeta.with_max_doc public (#2499) 
* chore: add container

* feat: make max doc editable externally

* chore: expose another method

* chore: remove comments

* remove unused devcontainer

* chore: manually match nightly format

* chore: change weird formating

* revert format change

* fix: format with nightly
 2024-09-23 12:39:36 +08:00
PSeitz
55b0b52457 Fix AggregationLimits (#2495) 
* change AggregationLimits behavior

This fixes an issue encountered with the current behaviour of
AggregationLimits.
Previously we had AggregationLimits and RessourceLimitGuard, which both
track the memory, but only RessourceLimitGuard released memory when
dropped, while AggregationLimits did not.

This PR changes AggregationLimits to be a guard itself and removes the
RessourceLimitGuard.

* rename AggregationLimits to AggregationLimitsGuard
 2024-09-17 14:25:47 +08:00
dependabot[bot]
56fc56c5b9 Update binggan requirement from 0.8.0 to 0.10.0 (#2493) 
* Update binggan requirement from 0.8.0 to 0.10.0

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

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

* update PR

---------

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Pascal Seitz <pascal.seitz@gmail.com>
 2024-09-10 14:26:06 +08:00
trinity-1686a
85395d942a fix clippy lints from 1.80-1.81 (#2488) 
* fix some clippy lints

* fix clippy::doc_lazy_continuation

* fix some lints for 1.82
 2024-09-05 14:33:05 +02:00
PSeitz
a206c3ccd3 add compat tests (#2485) 2024-09-04 18:26:57 +08:00
Chaya
dc5d31c116 grammar and misspellings (#2483) 
* grammar

* grammar

* misspelling
 2024-09-04 12:45:31 +08:00
gezihuzi
95a4ddea3e Fix: Improve collapse_overlapped_ranges function (#2474) 
* Fix: Improve collapse_overlapped_ranges function

- Refactor into separate sort_and_deduplicate_ranges and merge_overlapping_ranges functions
- Enhance sorting to consider both start and end of ranges
- Optimize merging logic to handle adjacent ranges
- Add comprehensive examples in function documentation
- Ensure proper handling of duplicate and unsorted input ranges
- Improve overall efficiency and readability of range collapsing algorithm

* move debug_assert

---------

Co-authored-by: PSeitz <PSeitz@users.noreply.github.com>
 2024-09-04 12:39:13 +08:00
trinity-1686a
ab5125d3dc remove unused trait bounds and outdated doc comment (#2478) 2024-09-03 16:31:51 +02:00
trinity-1686a
9f81d59ecd make find_field_with_default return json fields without path (#2476) 
* make find_field_with_default return json fields without path

* add tests for find_field_with_default
 2024-08-19 15:25:29 +02:00
PSeitz
c71ec8086d add FastFieldRangeQuery, rename (#2477) 
* add FastFieldRangeQuery, rename

* remove Query impl
 2024-08-19 09:02:00 +02:00
PSeitz
27be6aed91 lift clauses in LogicalAst (#2449) 
(a OR b) OR (c OR d) can be simplified to (a OR b OR c OR d)
(a AND b) AND (c AND d) can be simplified to (a AND b AND c AND d)

This directly affects how queries are executed

remove unused SumWithCoordsCombiner
the number of fields is unused and private
 2024-08-14 19:21:26 +02:00
PSeitz
3d1c4b313a support ff range queries on json fields (#2456) 
* support ff range queries on json fields

* fix term date truncation

* use inverted index range query for phrase prefix queries

* rename to InvertedIndexRangeQuery

* fix column filter, add mixed column test
 2024-08-02 00:06:50 +08:00
PSeitz
0d4e319965 add Key::I64 and Key::U64 variants in aggregation (#2468) 
* add Key::I64 and Key::U64 variants in aggregation

Currently all `Key` numerical values are returned as f64. This causes problems in some
cases with the precision and the way f64 is serialized.

This PR adds `Key::I64` and `Key::U64` variants and uses them in the term
aggregation.

* add clarification comment
 2024-07-31 20:29:32 +08:00
PSeitz
75dc3eb298 extend custom order deserialization (#2451) 
allow arrays
improve validation
closes https://github.com/quickwit-oss/tantivy/issues/2435
 2024-07-30 18:36:08 +08:00
PSeitz
3f6d225086 fix potential endless loop in merge (#2457) 
avoid single segments lists without deletes as merge candidates, as they will be moved
to a merge operation and filtered for merging in the next
consider_merge_options call. In rare cases this may end up in a endless
merge loop where only single segments where nothing is to be done are
merged.
 2024-07-30 16:37:20 +08:00
PSeitz
d8843c608c make FastFieldRangeWeight::new pub (#2460) 2024-07-29 10:39:27 +08:00
PSeitz
7ebcc15b17 add support for str fast field range query (#2453) 
* add support for str fast field range query

Add support for range queries on fast fields, by converting term bounds to
term ordinals bounds.

closes https://github.com/quickwit-oss/tantivy/issues/2023

* extend tests, rename

* update comment

* update comment
 2024-07-17 09:31:42 +08:00
PSeitz
1b4076691f refactor fast field query (#2452) 
As preparation of #2023 and #1709

* Use Term to pass parameters
* merge u64 and ip fast field range query

Side note: I did not rename range_query_u64_fastfield, because then git can't track the changes.
 2024-07-15 18:08:05 +08:00
Robert Caulk
eab660873a doc: fix typo in readme (#2450) 2024-07-09 15:12:22 +08:00
PSeitz
232f37126e fix coverage (#2448) 2024-07-05 12:04:18 +08:00
PSeitz
13e9885dfd faster term aggregation fetch terms (#2447) 
big impact for term aggregations with large `size` parameter (e.g. 1000)
add top 1000 term agg bench

full
terms_few                                      Memory: 27.3 KB (+79.09%)    Avg: 3.8058ms (+2.40%)      Median: 3.7192ms (+3.47%)       [3.6224ms .. 4.3721ms]
terms_many                                     Memory: 6.9 MB               Avg: 12.6102ms (-4.70%)     Median: 12.1389ms (-6.58%)      [10.2847ms .. 15.4857ms]
terms_many_top_1000                            Memory: 6.9 MB               Avg: 15.8216ms (-83.19%)    Median: 15.4899ms (-83.46%)     [13.4250ms .. 20.6897ms]
terms_many_order_by_term                       Memory: 6.9 MB               Avg: 14.7820ms (-3.95%)     Median: 14.2236ms (-4.28%)      [12.6669ms .. 21.0968ms]
terms_many_with_top_hits                       Memory: 58.2 MB              Avg: 551.6218ms (+7.18%)    Median: 549.8826ms (+11.01%)    [496.7371ms .. 592.1299ms]
terms_many_with_avg_sub_agg                    Memory: 27.8 MB              Avg: 197.7029ms (+2.66%)    Median: 190.1564ms (+0.64%)     [167.9226ms .. 245.6651ms]
terms_many_json_mixed_type_with_avg_sub_agg    Memory: 42.0 MB (+0.00%)     Avg: 242.0121ms (+0.92%)    Median: 237.7084ms (-2.85%)     [201.9959ms .. 302.2136ms]
terms_few_with_cardinality_agg                 Memory: 10.6 MB              Avg: 122.6036ms (+1.21%)    Median: 119.0033ms (+2.60%)     [109.2859ms .. 161.5858ms]
range_agg_with_term_agg_few                    Memory: 45.4 KB (+39.75%)    Avg: 24.5454ms (+2.14%)     Median: 24.2861ms (+2.44%)      [23.5109ms .. 27.8406ms]
range_agg_with_term_agg_many                   Memory: 6.9 MB               Avg: 56.8049ms (+3.01%)     Median: 50.9706ms (+1.52%)      [41.4517ms .. 90.3934ms]
dense
terms_few                                      Memory: 28.8 KB (+81.74%)    Avg: 8.9092ms (-2.24%)      Median: 8.7143ms (-1.31%)      [8.6148ms .. 10.3868ms]
terms_many                                     Memory: 6.9 MB (-0.00%)      Avg: 17.9604ms (-10.18%)    Median: 17.1552ms (-11.93%)    [14.8979ms .. 26.2779ms]
terms_many_top_1000                            Memory: 6.9 MB               Avg: 21.4963ms (-78.90%)    Median: 21.2924ms (-78.98%)    [18.2033ms .. 28.0087ms]
terms_many_order_by_term                       Memory: 6.9 MB               Avg: 20.4167ms (-9.13%)     Median: 19.5596ms (-11.37%)    [17.5153ms .. 29.5987ms]
terms_many_with_top_hits                       Memory: 58.2 MB              Avg: 518.4474ms (-6.41%)    Median: 514.9180ms (-9.44%)    [471.5550ms .. 579.0220ms]
terms_many_with_avg_sub_agg                    Memory: 27.8 MB              Avg: 263.6702ms (-2.78%)    Median: 260.8775ms (-2.55%)    [239.5754ms .. 304.6669ms]
terms_many_json_mixed_type_with_avg_sub_agg    Memory: 42.0 MB              Avg: 299.9791ms (-2.01%)    Median: 302.2180ms (-3.08%)    [239.2080ms .. 346.3649ms]
terms_few_with_cardinality_agg                 Memory: 10.6 MB              Avg: 136.3303ms (-3.12%)    Median: 132.3831ms (-2.88%)    [123.7564ms .. 164.7914ms]
range_agg_with_term_agg_few                    Memory: 47.1 KB (+37.81%)    Avg: 35.4538ms (+0.66%)     Median: 34.8754ms (-0.56%)     [34.2287ms .. 40.0884ms]
range_agg_with_term_agg_many                   Memory: 6.9 MB               Avg: 72.2269ms (-4.38%)     Median: 66.1174ms (-4.98%)     [55.5125ms .. 124.1622ms]
sparse
terms_few                                      Memory: 27.3 KB (+69.68%)    Avg: 19.6053ms (-1.15%)     Median: 19.4543ms (-0.38%)     [19.3056ms .. 24.0547ms]
terms_many                                     Memory: 1.8 MB               Avg: 21.2886ms (-6.28%)     Median: 21.1287ms (-6.65%)     [20.6640ms .. 24.6144ms]
terms_many_top_1000                            Memory: 2.6 MB               Avg: 23.4869ms (-85.53%)    Median: 23.3393ms (-85.61%)    [22.7789ms .. 25.0896ms]
terms_many_order_by_term                       Memory: 1.8 MB               Avg: 21.7437ms (-7.78%)     Median: 21.6272ms (-7.66%)     [21.0409ms .. 23.6517ms]
terms_many_with_top_hits                       Memory: 13.1 MB              Avg: 43.7926ms (-2.76%)     Median: 44.3602ms (+0.01%)     [37.8039ms .. 51.0451ms]
terms_many_with_avg_sub_agg                    Memory: 7.5 MB               Avg: 34.6307ms (+3.72%)     Median: 33.4522ms (+1.16%)     [32.4418ms .. 41.4196ms]
terms_many_json_mixed_type_with_avg_sub_agg    Memory: 7.4 MB               Avg: 46.4318ms (+1.16%)     Median: 46.4050ms (+2.03%)     [44.5986ms .. 48.5142ms]
terms_few_with_cardinality_agg                 Memory: 680.0 KB (-0.04%)    Avg: 35.4410ms (+2.05%)     Median: 35.1384ms (+1.19%)     [34.4402ms .. 39.1082ms]
range_agg_with_term_agg_few                    Memory: 45.7 KB (+39.44%)    Avg: 22.7760ms (+0.44%)     Median: 22.5152ms (-0.35%)     [22.3078ms .. 26.1567ms]
range_agg_with_term_agg_many                   Memory: 1.8 MB               Avg: 25.7696ms (-4.45%)     Median: 25.4009ms (-5.61%)     [24.7874ms .. 29.6434ms]
multivalue
terms_few                                      Memory: 244.4 KB            Avg: 15.1253ms (-2.85%)     Median: 15.0988ms (-0.54%)     [14.8790ms .. 15.8193ms]
terms_many                                     Memory: 6.9 MB (-0.00%)     Avg: 26.3019ms (-6.24%)     Median: 26.3662ms (-4.94%)     [21.3553ms .. 31.0564ms]
terms_many_top_1000                            Memory: 6.9 MB              Avg: 29.5212ms (-72.90%)    Median: 29.4257ms (-72.84%)    [24.2645ms .. 35.1607ms]
terms_many_order_by_term                       Memory: 6.9 MB              Avg: 28.6076ms (-4.93%)     Median: 28.1059ms (-6.64%)     [24.0845ms .. 34.1493ms]
terms_many_with_top_hits                       Memory: 58.3 MB             Avg: 570.1548ms (+1.52%)    Median: 572.7759ms (+0.53%)    [525.9567ms .. 617.0862ms]
terms_many_with_avg_sub_agg                    Memory: 27.8 MB             Avg: 305.5207ms (+0.24%)    Median: 296.0101ms (-0.22%)    [277.8579ms .. 373.5914ms]
terms_many_json_mixed_type_with_avg_sub_agg    Memory: 42.0 MB (-0.00%)    Avg: 324.7342ms (-2.51%)    Median: 319.0025ms (-2.58%)    [298.7122ms .. 368.6144ms]
terms_few_with_cardinality_agg                 Memory: 10.8 MB             Avg: 151.6126ms (-2.54%)    Median: 149.0616ms (-0.32%)    [136.5592ms .. 181.8942ms]
range_agg_with_term_agg_few                    Memory: 248.2 KB            Avg: 49.5225ms (+3.11%)     Median: 48.3994ms (+3.18%)     [46.4134ms .. 60.5989ms]
range_agg_with_term_agg_many                   Memory: 6.9 MB              Avg: 85.9824ms (-3.66%)     Median: 78.4266ms (-3.85%)     [64.1231ms .. 128.5279ms]
 2024-07-03 12:42:59 +08:00
PSeitz
56d79cb203 fix cardinality aggregation performance (#2446) 
* fix cardinality aggregation performance

fix cardinality performance by fetching multiple terms at once. This
avoids decompressing the same block and keeps the buffer state between
terms.

add cardinality aggregation benchmark

bump rust version to 1.66

Performance comparison to before (AllQuery)
```
full
cardinality_agg                   Memory: 3.5 MB (-0.00%)    Avg: 21.2256ms (-97.78%)    Median: 21.0042ms (-97.82%)    [20.4717ms .. 23.6206ms]
terms_few_with_cardinality_agg    Memory: 10.6 MB            Avg: 81.9293ms (-97.37%)    Median: 81.5526ms (-97.38%)    [79.7564ms .. 88.0374ms]
dense
cardinality_agg                   Memory: 3.6 MB (-0.00%)    Avg: 25.9372ms (-97.24%)    Median: 25.7744ms (-97.25%)    [24.7241ms .. 27.8793ms]
terms_few_with_cardinality_agg    Memory: 10.6 MB            Avg: 93.9897ms (-96.91%)    Median: 92.7821ms (-96.94%)    [90.3312ms .. 117.4076ms]
sparse
cardinality_agg                   Memory: 895.4 KB (-0.00%)    Avg: 22.5113ms (-95.01%)    Median: 22.5629ms (-94.99%)    [22.1628ms .. 22.9436ms]
terms_few_with_cardinality_agg    Memory: 680.2 KB             Avg: 26.4250ms (-94.85%)    Median: 26.4135ms (-94.86%)    [26.3210ms .. 26.6774ms]
```

* clippy

* assert for sorted ordinals
 2024-07-02 15:29:00 +08:00
Paul Masurel
0f4c2e27cf Fixes bug that causes out-of-order sstable key. (#2445) 
The previous way to address the problem was to replace \u{0000}
with 0 in different places.

This logic had several flaws:
Done on the serializer side (like it was for the columnar), there was
a collision problem.

If a document in the segment contained a json field with a \0 and
antoher doc contained the same json field but `0` then we were sending
the same field path twice to the serializer.

Another option would have been to normalizes all values on the writer
side.

This PR simplifies the logic and simply ignore json path containing a
\0, both in the columnar and the inverted index.

Closes #2442
 2024-07-01 15:40:07 +08:00
落叶乌龟
f9ae295507 feat(query): Make BooleanQuery supports minimum_number_should_match (#2405) 
* feat(query): Make `BooleanQuery` supports `minimum_number_should_match`. see issue #2398

In this commit, a novel scorer named DisjunctionScorer is introduced, which performs the union of inverted chains with the minimal required elements. BTW, it's implemented via a min-heap. Necessary modifications on `BooleanQuery` and `BooleanWeight` are performed as well.

* fixup! fix test

* fixup!: refactor code.

1. More meaningful names.
2. Add Cache for `Disjunction`'s scorers, and fix bug.
3. Optimize `BooleanWeight::complex_scorer`

Thanks
 Paul Masurel <paul@quickwit.io>

* squash!: come up with better variable naming.

* squash!: fix naming issues.

* squash!: fix typo.

* squash!: Remove CombinationMethod::FullIntersection
 2024-07-01 15:39:41 +08:00
Raphael Coeffic
d9db5302d9 feat: cardinality aggregation (#2337) 
* WiP: cardinality aggregation

* Collect unique entries first, then insert into HyperLogLog

* Handle `missing`

* Hybrid approach

* Review changes

- insert `missing` value at most once
- `term_id` -> `term_ord`
- iterate directly over entries without collecting first

* Use salted hasher to include column type

* fix: formatting

* More review fixes

* Add cardinality to test_aggregation_flushing

* Formatting
 2024-07-01 07:49:42 +08:00
Paul Masurel
e453848134 Recycling buffer in PrefixPhraseScorer (#2443) 2024-06-24 17:11:53 +09:00
PSeitz
59084143ef use optional index in multivalued index (#2439) 
* use optional index in multivalued index

For mostly empty multivalued indices there was a large overhead during
creation when iterating all docids. This is alleviated by placing an
optional index in the multivalued index to mark documents that have values.

There's some performance overhead when accessing values in a multivalued
index. The accessing cost is now optional index + multivalue index. The
sparse codec performs relatively bad with the binary_search when accessing
data. This is reflected in the benchmarks below.

This changes the format of columnar to v2, but code is added to handle the v1
formats.

```
     Running benches/bench_access.rs (/home/pascal/Development/tantivy/optional_multivalues/target/release/deps/bench_access-ea323c028db88db4)
multi sparse 1/13
access_values_for_doc        Avg: 42.8946ms (+241.80%)    Median: 42.8869ms (+244.10%)    [42.7484ms .. 43.1074ms]
access_first_vals            Avg: 42.8022ms (+421.93%)    Median: 42.7553ms (+439.84%)    [42.6794ms .. 43.7404ms]
multi 2x
access_values_for_doc        Avg: 31.1244ms (+24.17%)    Median: 30.8339ms (+23.46%)    [30.7192ms .. 33.6059ms]
access_first_vals            Avg: 24.3070ms (+70.92%)    Median: 24.0966ms (+70.18%)    [23.9328ms .. 26.4851ms]
sparse 1/13
access_values_for_doc        Avg: 42.2490ms (+0.61%)    Median: 42.2346ms (+2.28%)    [41.8988ms .. 43.7821ms]
access_first_vals            Avg: 43.6272ms (+0.23%)    Median: 43.6197ms (+1.78%)    [43.4920ms .. 43.9009ms]
dense 1/12
access_values_for_doc        Avg: 8.6184ms (+23.18%)    Median: 8.6126ms (+23.78%)    [8.5843ms .. 8.7527ms]
access_first_vals            Avg: 6.8112ms (+4.47%)     Median: 6.8002ms (+4.55%)     [6.7887ms .. 6.8991ms]
full
access_values_for_doc        Avg: 9.4073ms (-5.09%)    Median: 9.4023ms (-2.23%)    [9.3694ms .. 9.4568ms]
access_first_vals            Avg: 4.9531ms (+6.24%)    Median: 4.9502ms (+7.85%)    [4.9423ms .. 4.9718ms]
```

```
     Running benches/bench_merge.rs (/home/pascal/Development/tantivy/optional_multivalues/target/release/deps/bench_merge-475697dfceb3639f)
merge_multi 2x_and_multi 2x                          Avg: 20.2280ms (+34.33%)    Median: 20.1829ms (+35.33%)    [19.9933ms .. 20.8806ms]
merge_multi sparse 1/13_and_multi sparse 1/13        Avg: 0.8961ms (-78.04%)     Median: 0.8943ms (-77.61%)     [0.8899ms .. 0.9272ms]
merge_dense 1/12_and_dense 1/12                      Avg: 0.6619ms (-1.26%)      Median: 0.6616ms (+2.20%)      [0.6473ms .. 0.6837ms]
merge_sparse 1/13_and_sparse 1/13                    Avg: 0.5508ms (-0.85%)      Median: 0.5508ms (+2.80%)      [0.5420ms .. 0.5634ms]
merge_sparse 1/13_and_dense 1/12                     Avg: 0.6046ms (-4.64%)      Median: 0.6038ms (+2.80%)      [0.5939ms .. 0.6296ms]
merge_multi sparse 1/13_and_dense 1/12               Avg: 0.9111ms (-83.48%)     Median: 0.9063ms (-83.50%)     [0.9047ms .. 0.9663ms]
merge_multi sparse 1/13_and_sparse 1/13              Avg: 0.8451ms (-89.49%)     Median: 0.8428ms (-89.43%)     [0.8411ms .. 0.8563ms]
merge_multi 2x_and_dense 1/12                        Avg: 10.6624ms (-4.82%)     Median: 10.6568ms (-4.49%)     [10.5738ms .. 10.8353ms]
merge_multi 2x_and_sparse 1/13                       Avg: 10.6336ms (-22.95%)    Median: 10.5925ms (-22.33%)    [10.5149ms .. 11.5657ms]
```

* Update columnar/src/columnar/format_version.rs

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

* Update columnar/src/column_index/mod.rs

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

---------

Co-authored-by: Paul Masurel <paul@quickwit.io>
 2024-06-19 14:54:12 +08:00
PSeitz
511b027350 update columnar bench (#2438) 
* update columnar bench

* fix compile
 2024-06-14 10:42:35 +08:00
Philippe Noël
322f47eb47 Add ParadeDB to Companies List (#1) (#2437) 
* Add ParadeDB logo
 2024-06-14 09:12:58 +09:00
PSeitz
72f61ff89c remove index sorting (#2434) 
closes https://github.com/quickwit-oss/tantivy/issues/2352
 2024-06-13 15:51:53 +08:00
PSeitz
a141c3ec59 add columnar format compatibiliy tests (#2433) 
* add columnar format compatibiliy tests

* always try to write current format
 2024-06-13 15:04:52 +08:00
PSeitz
e90e7a25ae add access benchmark for columnar (#2432) 2024-06-12 14:29:15 +08:00
PSeitz
c3b92a5412 fix compiler warning, cleanup (#2393) 
fix compiler warning for missing feature flag
remove unused variables
cleanup unused methods
 2024-06-11 16:03:50 +08:00
PSeitz
2f55511064 extend indexwriter proptests (#2342) 
* index random values in proptest

* add proptest with multiple docs
 2024-06-11 16:02:57 +08:00
trinity-1686a
08b9fc0b31 fix de-escaping too much in query parser (#2427) 
* fix de-escaping too much in query parser
 2024-06-10 11:19:01 +02:00
PSeitz
714f363d43 add bench & test for columnar merging (#2428) 
* add merge columnar proptest

* add columnar merge benchmark
 2024-06-10 16:26:16 +08:00
PSeitz
93ff7365b0 reduce top hits aggregation memory consumption (#2426) 
move request structure out of top hits aggregation collector and use from the
passed structure instead

full
terms_many_with_top_hits    Memory: 58.2 MB (-43.64%)    Avg: 425.9680ms (-21.38%)    Median: 415.1097ms (-23.56%)    [395.5303ms .. 484.6325ms]
dense
terms_many_with_top_hits    Memory: 58.2 MB (-43.64%)    Avg: 440.0817ms (-19.68%)    Median: 432.2286ms (-21.10%)    [403.5632ms .. 497.7541ms]
sparse
terms_many_with_top_hits    Memory: 13.1 MB (-49.31%)    Avg: 33.3568ms (-32.19%)    Median: 33.0834ms (-31.86%)    [32.5126ms .. 35.7397ms]
multivalue
terms_many_with_top_hits    Memory: 58.2 MB (-43.64%)    Avg: 414.2340ms (-25.44%)    Median: 413.4144ms (-25.64%)    [403.9919ms .. 430.3170ms]
 2024-06-06 22:32:58 +08:00
Adam Reichold
8151925068 Panicking in spawned Rayon tasks will abort the process by default. (#2409) 2024-06-04 17:04:30 +09:00
dependabot[bot]
b960e40bc8 Update sketches-ddsketch requirement from 0.2.1 to 0.3.0 (#2423) 
Updates the requirements on [sketches-ddsketch](https://github.com/mheffner/rust-sketches-ddsketch) to permit the latest version.
- [Release notes](https://github.com/mheffner/rust-sketches-ddsketch/releases)
- [Commits](https://github.com/mheffner/rust-sketches-ddsketch/compare/v0.2.1...v0.3.0)

---
updated-dependencies:
- dependency-name: sketches-ddsketch
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2024-06-04 15:50:23 +08:00
giovannicuccu
1095c9b073 Issue 1787 extended stats (#2247) 
* first version of extended stats along with its tests

* using IntermediateExtendStats instead of IntermediateStats with all tests passing

* Created struct for request and response

* first test with extended_stats

* kahan summation and tests with approximate equality

* version ready for merge

* removed approx dependency

* refactor for using ExtendedStats only when needed

* interim version

* refined version with code formatted

* refactored a struct

* cosmetic refactor

* fix after merge

* fix format

* added extended_stat bench

* merge and new benchmark for extended stats

* split stat segment collectors

* wrapped intermediate extended stat with a box to limit memory usage

* Revert "wrapped intermediate extended stat with a box to limit memory usage"

This reverts commit 5b4aa9f393.

* some code reformat, commented kahan summation

* refactor after review

* refactor after code review

* fix after incorrectly restoring kahan summation

* modifications for code review + bug fix in merge_fruit

* refactor assert_nearly_equals macro

* update after code review

---------

Co-authored-by: Giovanni Cuccu <gcuccu@imolainformatica.it>
 2024-06-04 14:25:17 +08:00
PSeitz
c0686515a9 update one_shot (#2420) 2024-05-31 11:07:35 +08:00
trinity-1686a
455156f51c improve query parser (#2416) 
* support escape sequence in more place

and fix bug with singlequoted strings

* add query parser test for range query on default field
 2024-05-30 17:29:27 +02:00
Meng Zhang
4143d31865 chore: fix build as the rev is gone (#2417) 2024-05-29 09:49:16 +08:00
Hamir Mahal
0c634adbe1 style: simplify strings with string interpolation (#2412) 
* style: simplify strings with string interpolation

* fix: formatting
 2024-05-27 09:16:47 +02:00
PSeitz
2e3641c2ae return CompactDocValue instead of trait (#2410) 
The CompactDocValue is easier to handle than the trait in some cases like comparison
and conversion
 2024-05-27 07:33:50 +02:00
Paul Masurel
b806122c81 Fixing flaky test (#2407) 2024-05-22 10:10:55 +09:00
PSeitz
e1679f3fb9 compact doc (#2402) 
* compact doc

* add any value type

* pass references when building CompactDoc

* remove OwnedValue from API

* clippy

* clippy

* fail on large documents

* fmt

* cleanup

* cleanup

* implement Value for different types

fix serde_json date Value implementation

* fmt

* cleanup

* fmt

* cleanup

* store positions instead of pos+len

* remove nodes array

* remove mediumvec

* cleanup

* infallible serialize into vec

* remove positions indirection

* remove 24MB limitation in document

use u32 for Addr
Remove the 3 byte addressing limitation and use VInt instead

* cleanup

* extend test

* cleanup, add comments

* rename, remove pub
 2024-05-21 10:16:08 +02:00
dependabot[bot]
5a80420b10 --- (#2406) 
updated-dependencies:
- dependency-name: binggan
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2024-05-21 04:36:32 +02:00
dependabot[bot]
aa26ff5029 Update binggan requirement from 0.6.2 to 0.7.0 (#2401) 
---
updated-dependencies:
- dependency-name: binggan
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2024-05-17 02:53:25 +02:00
dependabot[bot]
e197b59258 Update itertools requirement from 0.12.0 to 0.13.0 (#2400) 
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/v0.12.0...v0.13.0)

---
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>
 2024-05-17 02:53:02 +02:00
PSeitz
5b7cca13e5 lower contention on AggregationLimits (#2394) 
PR https://github.com/quickwit-oss/quickwit/pull/4962 fixes an issue
where the AggregationLimits are not passed correctly. Since the
AggregationLimits are shared properly we run into contention issues.

This PR includes some straightforward improvement to reduce contention,
by only calling if the memory changed and avoiding the second read.

We probably need some sharding with multiple counters or local caching before updating the
global after some threshold.
 2024-05-15 12:25:40 +02:00
dependabot[bot]
a79590477e Update binggan requirement from 0.5.2 to 0.6.2 (#2399) 
---
updated-dependencies:
- dependency-name: binggan
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
 2024-05-15 05:40:37 +02:00
Paul Masurel
6181c1eb5e Small changes in the Executor API. (#2391) 
Warning, this change is mildly not backward compatible
so I bumped tantivy's version.
 2024-05-10 17:19:12 +09:00
Adam Reichold
1ee5f90761 Give allocation control to the caller instead of force a clone (#2389) 
Achieved by moving the boxes out of the temporary reference wrappers which are
cloneable themselves, i.e. if required the caller can clone them already or
consume them to reuse existing allocations.
 2024-05-09 16:01:13 +09:00
PSeitz
71f3b4e4e3 fix ReferenceValue API flaw (#2372) 
* fix ReferenceValue API flaw

Remove `Facet` and `TokenizedString` values from the `ReferenceValue` API,
as this requires the trait value to have them stored somewhere.

Since `TokenizedString` is quite niche, I just copy it into a Box,
instead of designing a reference API around it.

* fix comment link
 2024-05-09 06:14:42 +02:00
trinity-1686a
8cd7ddc535 run block decompression from executor (#2386) 
* run block decompression from executor

* add a wrapper with is_closed to oneshot channel

* add cancelation test to Executor::spawn_blocking
 2024-05-08 12:22:44 +02:00
Paul Masurel
2b76335a95 Removed usage of num_cpus (#2387) 
* Removed usage of num_cpus
* handling error
 2024-05-08 13:32:52 +09:00
PSeitz
c6b213d8f0 use bingang for agg benchmark (#2378) 
* use bingang for agg benchmark

use bingang for agg benchmark, which includes memory consumption

Output:
```
full
histogram                     Memory: 15.8 KB              Avg: 10.9322ms  (+5.44%)    Median: 10.8790ms  (+9.28%)     Min: 10.7470ms    Max: 11.3263ms
histogram_hard_bounds         Memory: 15.5 KB              Avg: 5.1939ms  (+6.61%)     Median: 5.1722ms  (+10.98%)     Min: 5.0432ms     Max: 5.3910ms
histogram_with_avg_sub_agg    Memory: 48.7 KB              Avg: 23.8165ms  (+4.57%)    Median: 23.7264ms  (+10.06%)    Min: 23.4995ms    Max: 24.8107ms
dense
histogram                     Memory: 17.3 KB              Avg: 15.6810ms  (-8.54%)    Median: 15.6174ms  (-8.89%)    Min: 15.4953ms    Max: 16.0702ms
histogram_hard_bounds         Memory: 15.4 KB              Avg: 10.0720ms  (-7.33%)    Median: 10.0572ms  (-7.06%)    Min: 9.8500ms     Max: 10.4819ms
histogram_with_avg_sub_agg    Memory: 50.1 KB              Avg: 33.0993ms  (-7.04%)    Median: 32.9499ms  (-6.86%)    Min: 32.8284ms    Max: 34.0529ms
sparse
histogram                     Memory: 16.3 KB              Avg: 19.2325ms  (-0.44%)    Median: 19.1211ms  (-1.26%)    Min: 19.0348ms    Max: 19.7902ms
histogram_hard_bounds         Memory: 16.1 KB              Avg: 18.5179ms  (-0.61%)    Median: 18.4552ms  (-0.90%)    Min: 18.3799ms    Max: 19.0535ms
histogram_with_avg_sub_agg    Memory: 34.7 KB              Avg: 21.2589ms  (-0.69%)    Median: 21.1867ms  (-1.05%)    Min: 21.0342ms    Max: 21.9900ms
```

* add more bench with term as sub agg
 2024-05-07 11:29:49 +02:00
PSeitz
eea70030bf cleanup top level exports (#2382) 
remove some top level exports
 2024-05-07 09:59:41 +02:00
PSeitz
92b5526310 allow more JSON values, fix i64 special case (#2383) 
This changes three things:
- Reuse positions_per_path hashmap instead of allocating one per
  indexed JSON value
- Try to cast u64 values to i64 to streamline with search behaviour
- Allow top level json values to be of any type, instead of limiting it
  to JSON objects. Remove special JSON object handling method.

TODO: We probably should also try to check f64 to i64 and u64 when
indexing, as values may get converted to f64 by the JSON parser
 2024-05-01 12:08:12 +02:00
PSeitz
99a59ad37e remove zero byte check (#2379) 
remove zero byte checks in columnar. zero bytes are converted during serialization now.
unify code paths
extend test for expected column names
 2024-04-26 06:03:28 +02:00
trinity-1686a
6a66a71cbb modify fastfield range query heuristic (#2375) 2024-04-25 10:06:11 +02:00
PSeitz
ff40764204 make convert_to_fast_value_and_append_to_json_term pub (#2370) 
* make convert_to_fast_value_and_append_to_json_term pub

* clippy
 2024-04-23 04:05:41 +02:00
PSeitz
047da20b5b add json path constructor to term (#2367) 2024-04-22 12:23:35 +02:00
PSeitz
1417eaf3a7 fix coverage (#2368) 2024-04-22 12:23:15 +02:00
PSeitz
4f8493d2de improve document docs (#2359) 2024-04-22 12:05:16 +02:00
Paul Masurel
8861366137 Owned value relying on Vec instead of BTreeMap (#2364) 
* Owned value relying on Vec instead of BTreeMap

* fmt

* fix build

* fix serialization

---------

Co-authored-by: Pascal Seitz <pascal.seitz@gmail.com>
 2024-04-22 09:38:05 +02:00
PSeitz
0e9fced336 remove JsonTermWriter (#2238) 
* remove JsonTermWriter

remove JsonTermWriter
remove path truncation logic, add assertion

* fix json_path_writer add sep logic
 2024-04-18 16:28:05 +02:00
PSeitz
b257b960b3 validate sort by field type (#2336) 
* validate sort by field type

* Update src/index/index.rs

Co-authored-by: Adam Reichold <adamreichold@users.noreply.github.com>

---------

Co-authored-by: Adam Reichold <adamreichold@users.noreply.github.com>
 2024-04-16 04:42:24 +02:00
Adam Reichold
4708171a32 Fix some of the things current Clippy complains about (#2363) 2024-04-16 04:27:06 +02:00
Adam Reichold
b493743f8d Fix trait bound of StoreReader::iter (#2360) 
* Fix trait bound of StoreReader::iter

Similar to `StoreReader::get`, `StoreReader::iter` should only require
`DocumentDeserialize` and not `Document`.

* Mark the iterator returned by SegmentReader::doc_ids_alive as Send so it can be used in impls of Stream/AsyncIterator.
 2024-04-15 15:50:02 +02:00
trinity-1686a
d2955a3fd2 extend field grouping (#2333) 
* extend field grouping
 2024-04-15 10:36:32 +02:00
PSeitz
17d5869ad6 update CHANGELOG, use github API in cliff (#2354) 
* update CHANGELOG, use github API in cliff

* reset version to 0.21.1, before release

* chore: Release

* remove unreleased from CHANGELOG
 2024-04-15 10:07:20 +02:00
PSeitz
dfa3aed32d check unsupported parameters top_hits (#2351) 
* check unsupported parameters top_hits

* move to function
 2024-04-10 08:20:52 +02:00
PSeitz
398817ce7b add index sorting deprecation warning (#2353) 
* add index sorting deprecation warning

* remove deprecated IntOptions and DatePrecision
 2024-04-10 08:09:09 +02:00
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
404 changed files with 30910 additions and 11803 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-2024-07-01 --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-2024-07-01 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

4
ARCHITECTURE.md
View File

@@ -46,7 +46,7 @@ The file of a segment has the format
```segment-id . ext```
The extension signals which data structure (or [`SegmentComponent`](src/core/segment_component.rs)) is stored in the file.
The extension signals which data structure (or [`SegmentComponent`](src/index/segment_component.rs)) is stored in the file.
A small `meta.json` file is in charge of keeping track of the list of segments, as well as the schema.
@@ -102,7 +102,7 @@ but users can extend tantivy with their own implementation.
Tantivy's document follows a very strict schema, decided before building any index.
The schema defines all of the fields that the indexes [`Document`](src/schema/document.rs) may and should contain, their types (`text`, `i64`, `u64`, `Date`, ...) as well as how it should be indexed / represented in tantivy.
The schema defines all of the fields that the indexes [`Document`](src/schema/document/mod.rs) may and should contain, their types (`text`, `i64`, `u64`, `Date`, ...) as well as how it should be indexed / represented in tantivy.
Depending on the type of the field, you can decide to

206
CHANGELOG.md
View File

@@ -1,5 +1,191 @@
Tantivy 0.23 - Unreleased
================================
Tantivy 0.23 will be backwards compatible with indices created with v0.22 and v0.21. The new minimum rust version will be 1.75.
Tantivy 0.20 [Unreleased]
#### Bugfixes
- fix potential endless loop in merge [#2457](https://github.com/quickwit-oss/tantivy/pull/2457)(@PSeitz)
- fix bug that causes out-of-order sstable key. [#2445](https://github.com/quickwit-oss/tantivy/pull/2445)(@fulmicoton)
- fix ReferenceValue API flaw [#2372](https://github.com/quickwit-oss/tantivy/pull/2372)(@PSeitz)
- fix `OwnedBytes` debug panic [#2512](https://github.com/quickwit-oss/tantivy/pull/2512)(@b41sh)
- catch panics during merges [#2582](https://github.com/quickwit-oss/tantivy/pull/2582)(@rdettai)
- switch from u32 to usize in bitpacker. This enables multivalued columns larger than 4GB, which crashed during merge before. [#2581](https://github.com/quickwit-oss/tantivy/pull/2581) [#2586](https://github.com/quickwit-oss/tantivy/pull/2586)(@fulmicoton-dd @PSeitz)
#### Breaking API Changes
- remove index sorting [#2434](https://github.com/quickwit-oss/tantivy/pull/2434)(@PSeitz)
#### Features/Improvements
- **Aggregation**
- Support for cardinality aggregation [#2337](https://github.com/quickwit-oss/tantivy/pull/2337) [#2446](https://github.com/quickwit-oss/tantivy/pull/2446) (@raphaelcoeffic @PSeitz)
- Support for extended stats aggregation [#2247](https://github.com/quickwit-oss/tantivy/pull/2247)(@giovannicuccu)
- Add Key::I64 and Key::U64 variants in aggregation to avoid f64 precision issues [#2468](https://github.com/quickwit-oss/tantivy/pull/2468)(@PSeitz)
- Faster term aggregation fetch terms [#2447](https://github.com/quickwit-oss/tantivy/pull/2447)(@PSeitz)
- Improve custom order deserialization [#2451](https://github.com/quickwit-oss/tantivy/pull/2451)(@PSeitz)
- Change AggregationLimits behavior [#2495](https://github.com/quickwit-oss/tantivy/pull/2495)(@PSeitz)
- lower contention on AggregationLimits [#2394](https://github.com/quickwit-oss/tantivy/pull/2394)(@PSeitz)
- fix postcard compatibility for top_hits, add postcard test [#2346](https://github.com/quickwit-oss/tantivy/pull/2346)(@PSeitz)
- reduce top hits memory consumption [#2426](https://github.com/quickwit-oss/tantivy/pull/2426)(@PSeitz)
- check unsupported parameters top_hits [#2351](https://github.com/quickwit-oss/tantivy/pull/2351)(@PSeitz)
- Change AggregationLimits to AggregationLimitsGuard [#2495](https://github.com/quickwit-oss/tantivy/pull/2495)(@PSeitz)
- add support for counting non integer in aggregation [#2547](https://github.com/quickwit-oss/tantivy/pull/2547)(@trinity-1686a)
- **Range Queries**
- Support fast field range queries on json fields [#2456](https://github.com/quickwit-oss/tantivy/pull/2456)(@PSeitz)
- Add support for str fast field range query [#2460](https://github.com/quickwit-oss/tantivy/pull/2460) [#2452](https://github.com/quickwit-oss/tantivy/pull/2452) [#2453](https://github.com/quickwit-oss/tantivy/pull/2453)(@PSeitz)
- modify fastfield range query heuristic [#2375](https://github.com/quickwit-oss/tantivy/pull/2375)(@trinity-1686a)
- add FastFieldRangeQuery for explicit range queries on fast field (for `RangeQuery` it is autodetected) [#2477](https://github.com/quickwit-oss/tantivy/pull/2477)(@PSeitz)
- add format backwards-compatibility tests [#2485](https://github.com/quickwit-oss/tantivy/pull/2485)(@PSeitz)
- add columnar format compatibility tests [#2433](https://github.com/quickwit-oss/tantivy/pull/2433)(@PSeitz)
- Improved snippet ranges algorithm [#2474](https://github.com/quickwit-oss/tantivy/pull/2474)(@gezihuzi)
- make find_field_with_default return json fields without path [#2476](https://github.com/quickwit-oss/tantivy/pull/2476)(@trinity-1686a)
- Make `BooleanQuery` support `minimum_number_should_match` [#2405](https://github.com/quickwit-oss/tantivy/pull/2405)(@LebranceBW)
- Make `NUM_MERGE_THREADS` configurable [#2535](https://github.com/quickwit-oss/tantivy/pull/2535)(@Barre)
- **RegexPhraseQuery**
`RegexPhraseQuery` supports phrase queries with regex. E.g. query "b.* b.* wolf" matches "big bad wolf". Slop is supported as well: "b.* wolf"~2 matches "big bad wolf" [#2516](https://github.com/quickwit-oss/tantivy/pull/2516)(@PSeitz)
- **Optional Index in Multivalue Columnar Index**
For mostly empty multivalued indices there was a large overhead during creation when iterating all docids (merge case).
This is alleviated by placing an optional index in the multivalued index to mark documents that have values.
This will slightly increase space and access time. [#2439](https://github.com/quickwit-oss/tantivy/pull/2439)(@PSeitz)
- **Store DateTime as nanoseconds in doc store** DateTime in the doc store was truncated to microseconds previously. This removes this truncation, while still keeping backwards compatibility. [#2486](https://github.com/quickwit-oss/tantivy/pull/2486)(@PSeitz)
- **Performace/Memory**
- lift clauses in LogicalAst for optimized ast during execution [#2449](https://github.com/quickwit-oss/tantivy/pull/2449)(@PSeitz)
- Use Vec instead of BTreeMap to back OwnedValue object [#2364](https://github.com/quickwit-oss/tantivy/pull/2364)(@fulmicoton)
- Replace TantivyDocument with CompactDoc. CompactDoc is much smaller and provides similar performance. [#2402](https://github.com/quickwit-oss/tantivy/pull/2402)(@PSeitz)
- Recycling buffer in PrefixPhraseScorer [#2443](https://github.com/quickwit-oss/tantivy/pull/2443)(@fulmicoton)
- **Json Type**
- JSON supports now all values on the root level. Previously an object was required. This enables support for flat mixed types. allow more JSON values, fix i64 special case [#2383](https://github.com/quickwit-oss/tantivy/pull/2383)(@PSeitz)
- add json path constructor to term [#2367](https://github.com/quickwit-oss/tantivy/pull/2367)(@PSeitz)
- **QueryParser**
- fix de-escaping too much in query parser [#2427](https://github.com/quickwit-oss/tantivy/pull/2427)(@trinity-1686a)
- improve query parser [#2416](https://github.com/quickwit-oss/tantivy/pull/2416)(@trinity-1686a)
- Support field grouping `title:(return AND "pink panther")` [#2333](https://github.com/quickwit-oss/tantivy/pull/2333)(@trinity-1686a)
- allow term starting with wildcard [#2568](https://github.com/quickwit-oss/tantivy/pull/2568)(@trinity-1686a)
- Exist queries match subpath fields [#2558](https://github.com/quickwit-oss/tantivy/pull/2558)(@rdettai)
- add access benchmark for columnar [#2432](https://github.com/quickwit-oss/tantivy/pull/2432)(@PSeitz)
- extend indexwriter proptests [#2342](https://github.com/quickwit-oss/tantivy/pull/2342)(@PSeitz)
- add bench & test for columnar merging [#2428](https://github.com/quickwit-oss/tantivy/pull/2428)(@PSeitz)
- Change in Executor API [#2391](https://github.com/quickwit-oss/tantivy/pull/2391)(@fulmicoton)
- Removed usage of num_cpus [#2387](https://github.com/quickwit-oss/tantivy/pull/2387)(@fulmicoton)
- use bingang for agg and stacker benchmark [#2378](https://github.com/quickwit-oss/tantivy/pull/2378)[#2492](https://github.com/quickwit-oss/tantivy/pull/2492)(@PSeitz)
- cleanup top level exports [#2382](https://github.com/quickwit-oss/tantivy/pull/2382)(@PSeitz)
- make convert_to_fast_value_and_append_to_json_term pub [#2370](https://github.com/quickwit-oss/tantivy/pull/2370)(@PSeitz)
- remove JsonTermWriter [#2238](https://github.com/quickwit-oss/tantivy/pull/2238)(@PSeitz)
- validate sort by field type [#2336](https://github.com/quickwit-oss/tantivy/pull/2336)(@PSeitz)
- Fix trait bound of StoreReader::iter [#2360](https://github.com/quickwit-oss/tantivy/pull/2360)(@adamreichold)
- remove read_postings_no_deletes [#2526](https://github.com/quickwit-oss/tantivy/pull/2526)(@PSeitz)
Tantivy 0.22
================================
Tantivy 0.22 will be able to read indices created with Tantivy 0.21.
#### Bugfixes
- Fix null byte handling in JSON paths (null bytes in json keys caused panic during indexing) [#2345](https://github.com/quickwit-oss/tantivy/pull/2345)(@PSeitz)
- Fix bug that can cause `get_docids_for_value_range` to panic. [#2295](https://github.com/quickwit-oss/tantivy/pull/2295)(@fulmicoton)
- Avoid 1 document indices by increase min memory to 15MB for indexing [#2176](https://github.com/quickwit-oss/tantivy/pull/2176)(@PSeitz)
- Fix merge panic for JSON fields [#2284](https://github.com/quickwit-oss/tantivy/pull/2284)(@PSeitz)
- Fix bug occurring when merging JSON object indexed with positions. [#2253](https://github.com/quickwit-oss/tantivy/pull/2253)(@fulmicoton)
- Fix empty DateHistogram gap bug [#2183](https://github.com/quickwit-oss/tantivy/pull/2183)(@PSeitz)
- Fix range query end check (fields with less than 1 value per doc are affected) [#2226](https://github.com/quickwit-oss/tantivy/pull/2226)(@PSeitz)
- Handle exclusive out of bounds ranges on fastfield range queries [#2174](https://github.com/quickwit-oss/tantivy/pull/2174)(@PSeitz)
#### Breaking API Changes
- rename ReloadPolicy onCommit to onCommitWithDelay [#2235](https://github.com/quickwit-oss/tantivy/pull/2235)(@giovannicuccu)
- Move exports from the root into modules [#2220](https://github.com/quickwit-oss/tantivy/pull/2220)(@PSeitz)
- Accept field name instead of `Field` in FilterCollector [#2196](https://github.com/quickwit-oss/tantivy/pull/2196)(@PSeitz)
- remove deprecated IntOptions and DateTime [#2353](https://github.com/quickwit-oss/tantivy/pull/2353)(@PSeitz)
#### Features/Improvements
- Tantivy documents as a trait: Index data directly without converting to tantivy types first [#2071](https://github.com/quickwit-oss/tantivy/pull/2071)(@ChillFish8)
- encode some part of posting list as -1 instead of direct values (smaller inverted indices) [#2185](https://github.com/quickwit-oss/tantivy/pull/2185)(@trinity-1686a)
- **Aggregation**
- Support to deserialize f64 from string [#2311](https://github.com/quickwit-oss/tantivy/pull/2311)(@PSeitz)
- Add a top_hits aggregator [#2198](https://github.com/quickwit-oss/tantivy/pull/2198)(@ditsuke)
- Support bool type in term aggregation [#2318](https://github.com/quickwit-oss/tantivy/pull/2318)(@PSeitz)
- Support ip addresses in term aggregation [#2319](https://github.com/quickwit-oss/tantivy/pull/2319)(@PSeitz)
- Support date type in term aggregation [#2172](https://github.com/quickwit-oss/tantivy/pull/2172)(@PSeitz)
- Support escaped dot when addressing field [#2250](https://github.com/quickwit-oss/tantivy/pull/2250)(@PSeitz)
- Add ExistsQuery to check documents that have a value [#2160](https://github.com/quickwit-oss/tantivy/pull/2160)(@imotov)
- Expose TopDocs::order_by_u64_field again [#2282](https://github.com/quickwit-oss/tantivy/pull/2282)(@ditsuke)
- **Memory/Performance**
- Faster TopN: replace BinaryHeap with TopNComputer [#2186](https://github.com/quickwit-oss/tantivy/pull/2186)(@PSeitz)
- reduce number of allocations during indexing [#2257](https://github.com/quickwit-oss/tantivy/pull/2257)(@PSeitz)
- Less Memory while indexing: docid deltas while indexing [#2249](https://github.com/quickwit-oss/tantivy/pull/2249)(@PSeitz)
- Faster indexing: use term hashmap in fastfield [#2243](https://github.com/quickwit-oss/tantivy/pull/2243)(@PSeitz)
- term hashmap remove copy in is_empty, unused unordered_id [#2229](https://github.com/quickwit-oss/tantivy/pull/2229)(@PSeitz)
- add method to fetch block of first values in columnar [#2330](https://github.com/quickwit-oss/tantivy/pull/2330)(@PSeitz)
- Faster aggregations: add fast path for full columns in fetch_block [#2328](https://github.com/quickwit-oss/tantivy/pull/2328)(@PSeitz)
- Faster sstable loading: use fst for sstable index [#2268](https://github.com/quickwit-oss/tantivy/pull/2268)(@trinity-1686a)
- **QueryParser**
- allow newline where we allow space in query parser [#2302](https://github.com/quickwit-oss/tantivy/pull/2302)(@trinity-1686a)
- allow some mixing of occur and bool in strict query parser [#2323](https://github.com/quickwit-oss/tantivy/pull/2323)(@trinity-1686a)
- handle * inside term in lenient query parser [#2228](https://github.com/quickwit-oss/tantivy/pull/2228)(@trinity-1686a)
- add support for exists query syntax in query parser [#2170](https://github.com/quickwit-oss/tantivy/pull/2170)(@trinity-1686a)
- Add shared search executor [#2312](https://github.com/quickwit-oss/tantivy/pull/2312)(@MochiXu)
- Truncate keys to u16::MAX in term hashmap [#2299](https://github.com/quickwit-oss/tantivy/pull/2299)(@PSeitz)
- report if a term matched when warming up posting list [#2309](https://github.com/quickwit-oss/tantivy/pull/2309)(@trinity-1686a)
- Support json fields in FuzzyTermQuery [#2173](https://github.com/quickwit-oss/tantivy/pull/2173)(@PingXia-at)
- Read list of fields encoded in term dictionary for JSON fields [#2184](https://github.com/quickwit-oss/tantivy/pull/2184)(@PSeitz)
- add collect_block to BoxableSegmentCollector [#2331](https://github.com/quickwit-oss/tantivy/pull/2331)(@PSeitz)
- expose collect_block buffer size [#2326](https://github.com/quickwit-oss/tantivy/pull/2326)(@PSeitz)
- Forward regex parser errors [#2288](https://github.com/quickwit-oss/tantivy/pull/2288)(@adamreichold)
- Make FacetCounts defaultable and cloneable. [#2322](https://github.com/quickwit-oss/tantivy/pull/2322)(@adamreichold)
- Derive Debug for SchemaBuilder [#2254](https://github.com/quickwit-oss/tantivy/pull/2254)(@GodTamIt)
- add missing inlines to tantivy options [#2245](https://github.com/quickwit-oss/tantivy/pull/2245)(@PSeitz)
Tantivy 0.21.1
================================
#### Bugfixes
- Range queries on fast fields with less values on that field than documents had an invalid end condition, leading to missing results. [#2226](https://github.com/quickwit-oss/tantivy/issues/2226)(@appaquet @PSeitz)
- Increase the minimum memory budget from 3MB to 15MB to avoid single doc segments (API fix). [#2176](https://github.com/quickwit-oss/tantivy/issues/2176)(@PSeitz)
Tantivy 0.21
================================
#### Bugfixes
- Fix track fast field memory consumption, which led to higher memory consumption than the budget allowed during indexing [#2148](https://github.com/quickwit-oss/tantivy/issues/2148)[#2147](https://github.com/quickwit-oss/tantivy/issues/2147)(@PSeitz)
- Fix a regression from 0.20 where sort index by date wasn't working anymore [#2124](https://github.com/quickwit-oss/tantivy/issues/2124)(@PSeitz)
- Fix getting the root facet on the `FacetCollector`. [#2086](https://github.com/quickwit-oss/tantivy/issues/2086)(@adamreichold)
- Align numerical type priority order of columnar and query. [#2088](https://github.com/quickwit-oss/tantivy/issues/2088)(@fmassot)
#### Breaking Changes
- Remove support for Brotli and Snappy compression [#2123](https://github.com/quickwit-oss/tantivy/issues/2123)(@adamreichold)
#### Features/Improvements
- Implement lenient query parser [#2129](https://github.com/quickwit-oss/tantivy/pull/2129)(@trinity-1686a)
- order_by_u64_field and order_by_fast_field allow sorting in ascending and descending order [#2111](https://github.com/quickwit-oss/tantivy/issues/2111)(@naveenann)
- Allow dynamic filters in text analyzer builder [#2110](https://github.com/quickwit-oss/tantivy/issues/2110)(@fulmicoton @fmassot)
- **Aggregation**
- Add missing parameter for term aggregation [#2149](https://github.com/quickwit-oss/tantivy/issues/2149)[#2103](https://github.com/quickwit-oss/tantivy/issues/2103)(@PSeitz)
- Add missing parameter for percentiles [#2157](https://github.com/quickwit-oss/tantivy/issues/2157)(@PSeitz)
- Add missing parameter for stats,min,max,count,sum,avg [#2151](https://github.com/quickwit-oss/tantivy/issues/2151)(@PSeitz)
- Improve aggregation deserialization error message [#2150](https://github.com/quickwit-oss/tantivy/issues/2150)(@PSeitz)
- Add validation for type Bytes to term_agg [#2077](https://github.com/quickwit-oss/tantivy/issues/2077)(@PSeitz)
- Alternative mixed field collection [#2135](https://github.com/quickwit-oss/tantivy/issues/2135)(@PSeitz)
- Add missing query_terms impl for TermSetQuery. [#2120](https://github.com/quickwit-oss/tantivy/issues/2120)(@adamreichold)
- Minor improvements to OwnedBytes [#2134](https://github.com/quickwit-oss/tantivy/issues/2134)(@adamreichold)
- Remove allocations in split compound words [#2080](https://github.com/quickwit-oss/tantivy/issues/2080)(@PSeitz)
- Ngram tokenizer now returns an error with invalid arguments [#2102](https://github.com/quickwit-oss/tantivy/issues/2102)(@fmassot)
- Make TextAnalyzerBuilder public [#2097](https://github.com/quickwit-oss/tantivy/issues/2097)(@adamreichold)
- Return an error when tokenizer is not found while indexing [#2093](https://github.com/quickwit-oss/tantivy/issues/2093)(@naveenann)
- Delayed column opening during merge [#2132](https://github.com/quickwit-oss/tantivy/issues/2132)(@PSeitz)
Tantivy 0.20.2
================================
- Align numerical type priority order on the search side. [#2088](https://github.com/quickwit-oss/tantivy/issues/2088) (@fmassot)
- Fix is_child_of function not considering the root facet. [#2086](https://github.com/quickwit-oss/tantivy/issues/2086) (@adamreichhold)
Tantivy 0.20.1
================================
- Fix building on windows with mmap [#2070](https://github.com/quickwit-oss/tantivy/issues/2070) (@ChillFish8)
Tantivy 0.20
================================
#### Bugfixes
- Fix phrase queries with slop (slop supports now transpositions, algorithm that carries slop so far for num terms > 2) [#2031](https://github.com/quickwit-oss/tantivy/issues/2031)[#2020](https://github.com/quickwit-oss/tantivy/issues/2020)(@PSeitz)
@@ -12,7 +198,7 @@ Tantivy 0.20 [Unreleased]
- Add PhrasePrefixQuery [#1842](https://github.com/quickwit-oss/tantivy/issues/1842) (@trinity-1686a)
- Add `coerce` option for text and numbers types (convert the value instead of returning an error during indexing) [#1904](https://github.com/quickwit-oss/tantivy/issues/1904) (@PSeitz)
- Add regex tokenizer [#1759](https://github.com/quickwit-oss/tantivy/issues/1759)(@mkleen)
- Move tokenizer API to seperate crate. Having a seperate crate with a stable API will allow us to use tokenizers with different tantivy versions. [#1767](https://github.com/quickwit-oss/tantivy/issues/1767) (@PSeitz)
- Move tokenizer API to separate crate. Having a separate crate with a stable API will allow us to use tokenizers with different tantivy versions. [#1767](https://github.com/quickwit-oss/tantivy/issues/1767) (@PSeitz)
- **Columnar crate**: New fast field handling (@fulmicoton @PSeitz) [#1806](https://github.com/quickwit-oss/tantivy/issues/1806)[#1809](https://github.com/quickwit-oss/tantivy/issues/1809)
- Support for fast fields with optional values. Previously tantivy supported only single-valued and multi-value fast fields. The encoding of optional fast fields is now very compact.
- Fast field Support for JSON (schemaless fast fields). Support multiple types on the same column. [#1876](https://github.com/quickwit-oss/tantivy/issues/1876) (@fulmicoton)
@@ -38,12 +224,14 @@ 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)
@@ -57,13 +245,13 @@ Tantivy 0.20 [Unreleased]
- Auto downgrade index record option, instead of vint error [#1857](https://github.com/quickwit-oss/tantivy/issues/1857) (@PSeitz)
- Enable range query on fast field for u64 compatible types [#1762](https://github.com/quickwit-oss/tantivy/issues/1762) (@PSeitz) [#1876]
- sstable
- Isolating sstable and stacker in independant crates. [#1718](https://github.com/quickwit-oss/tantivy/issues/1718) (@fulmicoton)
- Isolating sstable and stacker in independent crates. [#1718](https://github.com/quickwit-oss/tantivy/issues/1718) (@fulmicoton)
- New sstable format [#1943](https://github.com/quickwit-oss/tantivy/issues/1943)[#1953](https://github.com/quickwit-oss/tantivy/issues/1953) (@trinity-1686a)
- Use DeltaReader directly to implement Dictionnary::ord_to_term [#1928](https://github.com/quickwit-oss/tantivy/issues/1928) (@trinity-1686a)
- Use DeltaReader directly to implement Dictionnary::term_ord [#1925](https://github.com/quickwit-oss/tantivy/issues/1925) (@trinity-1686a)
- Add seperate tokenizer manager for fast fields [#2019](https://github.com/quickwit-oss/tantivy/issues/2019) (@PSeitz)
- Use DeltaReader directly to implement Dictionary::ord_to_term [#1928](https://github.com/quickwit-oss/tantivy/issues/1928) (@trinity-1686a)
- Use DeltaReader directly to implement Dictionary::term_ord [#1925](https://github.com/quickwit-oss/tantivy/issues/1925) (@trinity-1686a)
- Add separate tokenizer manager for fast fields [#2019](https://github.com/quickwit-oss/tantivy/issues/2019) (@PSeitz)
- Make construction of LevenshteinAutomatonBuilder for FuzzyTermQuery instances lazy. [#1756](https://github.com/quickwit-oss/tantivy/issues/1756) (@adamreichold)
- Added support for madvise when opening an mmaped Index [#2036](https://github.com/quickwit-oss/tantivy/issues/2036) (@fulmicoton)
- Added support for madvise when opening an mmapped Index [#2036](https://github.com/quickwit-oss/tantivy/issues/2036) (@fulmicoton)
- Rename `DatePrecision` to `DateTimePrecision` [#2051](https://github.com/quickwit-oss/tantivy/issues/2051) (@guilload)
- Query Parser
- Quotation mark can now be used for phrase queries. [#2050](https://github.com/quickwit-oss/tantivy/issues/2050) (@fulmicoton)
@@ -102,7 +290,7 @@ Tantivy 0.19
- Add support for phrase slop in query language [#1393](https://github.com/quickwit-oss/tantivy/pull/1393) (@saroh)
- Aggregation
- Add aggregation support for date type [#1693](https://github.com/quickwit-oss/tantivy/pull/1693)(@PSeitz)
- Add support for keyed parameter in range and histgram aggregations [#1424](https://github.com/quickwit-oss/tantivy/pull/1424) (@k-yomo)
- Add support for keyed parameter in range and histogram aggregations [#1424](https://github.com/quickwit-oss/tantivy/pull/1424) (@k-yomo)
- Add aggregation bucket limit [#1363](https://github.com/quickwit-oss/tantivy/pull/1363) (@PSeitz)
- Faster indexing
- [#1610](https://github.com/quickwit-oss/tantivy/pull/1610) (@PSeitz)
@@ -545,7 +733,7 @@ Tantivy 0.4.0
- Raise the limit of number of fields (previously 256 fields) (@fulmicoton)
- Removed u32 fields. They are replaced by u64 and i64 fields (#65) (@fulmicoton)
- Optimized skip in SegmentPostings (#130) (@lnicola)
- Replacing rustc_serialize by serde. Kudos to @KodrAus and @lnicola
- Replacing rustc_serialize by serde. Kudos to benchmark@KodrAus and @lnicola
- Using error-chain (@KodrAus)
- QueryParser: (@fulmicoton)
- Explicit error returned when searched for a term that is not indexed

10
CITATION.cff Normal file
View File

@@ -0,0 +1,10 @@
cff-version: 1.2.0
message: "If you use this software, please cite it as below."
authors:
- alias: Quickwit Inc.
website: "https://quickwit.io"
title: "tantivy"
version: 0.22.0
doi: 10.5281/zenodo.13942948
date-released: 2024-10-17
url: "https://github.com/quickwit-oss/tantivy"

117
Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy"
version = "0.20.2"
version = "0.24.0"
authors = ["Paul Masurel <paul.masurel@gmail.com>"]
license = "MIT"
categories = ["database-implementations", "data-structures"]
@@ -11,78 +11,87 @@ repository = "https://github.com/quickwit-oss/tantivy"
readme = "README.md"
keywords = ["search", "information", "retrieval"]
edition = "2021"
rust-version = "1.62"
rust-version = "1.75"
exclude = ["benches/*.json", "benches/*.txt"]
[dependencies]
oneshot = "0.1.5"
base64 = "0.21.0"
oneshot = "0.1.7"
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"] }
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 }
tempfile = { version = "3.3.0", 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 = "3.12.0", 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 }
serde = { version = "1.0.219", features = ["derive"] }
serde_json = "1.0.140"
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"
rustc-hash = "1.1.0"
thiserror = "1.0.30"
downcast-rs = "2.0.1"
bitpacking = { version = "0.9.2", default-features = false, features = [
"bitpacker4x",
] }
census = "0.4.2"
rustc-hash = "2.0.0"
thiserror = "2.0.1"
htmlescape = "0.3.1"
fail = "0.5.0"
murmurhash32 = "0.3.0"
time = { version = "0.3.10", features = ["serde-well-known"] }
fail = { version = "0.5.0", optional = true }
time = { version = "0.3.35", 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"
measure_time = "0.8.2"
async-trait = "0.1.53"
itertools = "0.14.0"
measure_time = "0.9.0"
arc-swap = "1.5.0"
bon = "3.3.1"
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" }
sketches-ddsketch = { version = "0.2.1", features = ["use_serde"] }
columnar = { version = "0.5", path = "./columnar", package = "tantivy-columnar" }
sstable = { version = "0.5", path = "./sstable", package = "tantivy-sstable", optional = true }
stacker = { version = "0.5", path = "./stacker", package = "tantivy-stacker" }
query-grammar = { version = "0.24.0", path = "./query-grammar", package = "tantivy-query-grammar" }
tantivy-bitpacker = { version = "0.8", path = "./bitpacker" }
common = { version = "0.9", path = "./common/", package = "tantivy-common" }
tokenizer-api = { version = "0.5", path = "./tokenizer-api", package = "tantivy-tokenizer-api" }
sketches-ddsketch = { version = "0.3.0", features = ["use_serde"] }
hyperloglogplus = { version = "0.4.1", features = ["const-loop"] }
futures-util = { version = "0.3.28", optional = true }
futures-channel = { version = "0.3.28", optional = true }
fnv = "1.0.7"
[target.'cfg(windows)'.dependencies]
winapi = "0.3.9"
[dev-dependencies]
binggan = "0.14.0"
rand = "0.8.5"
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"
@@ -107,18 +116,30 @@ 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"]
unstable = [] # useful for benches.
failpoints = ["fail", "fail/failpoints"]
unstable = [] # useful for benches.
quickwit = ["sstable", "futures-util"]
quickwit = ["sstable", "futures-util", "futures-channel"]
# 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"]
members = [
"query-grammar",
"bitpacker",
"common",
"ownedbytes",
"stacker",
"sstable",
"tokenizer-api",
"columnar",
]
# Following the "fail" crate best practises, we isolate
# tests that define specific behavior in fail check points
@@ -130,7 +151,7 @@ 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"
@@ -139,3 +160,7 @@ harness = false
[[bench]]
name = "index-bench"
harness = false
[[bench]]
name = "agg_bench"
harness = false

40
README.md
View File

@@ -5,21 +5,20 @@
[![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
The following [benchmark](https://tantivy-search.github.io/bench/) breaks down the
performance for different types of queries/collections.
Your mileage WILL vary depending on the nature of queries and their load.
@@ -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:
@@ -99,10 +98,11 @@ 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;
<img align="center" src="doc/assets/images/etsy.png" alt="Etsy" height="25" width="auto" /> &nbsp;
<img align="center" src="doc/assets/images/paradedb.png" alt="ParadeDB" height="25" width="auto" /> &nbsp;
<img align="center" src="doc/assets/images/Nuclia.png#gh-light-mode-only" alt="Nuclia" height="25" width="auto" /> &nbsp;
<img align="center" src="doc/assets/images/humanfirst.png#gh-light-mode-only" alt="Humanfirst.ai" height="30" width="auto" />
<img align="center" src="doc/assets/images/element.io.svg#gh-light-mode-only" alt="Element.io" height="25" width="auto" />
@@ -111,7 +111,7 @@ 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?

2
TODO.txt
View File

@@ -1,7 +1,7 @@
Make schema_builder API fluent.
fix doc serialization and prevent compression problems
u64 , etc. shoudl return Resutl<Option> now that we support optional missing a column is really not an error
u64 , etc. should return Result<Option> now that we support optional missing a column is really not an error
remove fastfield codecs
ditch the first_or_default trick. if it is still useful, improve its implementation.
rename FastFieldReaders::open to load

462
benches/agg_bench.rs Normal file
View File

@@ -0,0 +1,462 @@
use binggan::plugins::PeakMemAllocPlugin;
use binggan::{black_box, InputGroup, PeakMemAlloc, INSTRUMENTED_SYSTEM};
use rand::prelude::SliceRandom;
use rand::rngs::StdRng;
use rand::{Rng, SeedableRng};
use rand_distr::Distribution;
use serde_json::json;
use tantivy::aggregation::agg_req::Aggregations;
use tantivy::aggregation::AggregationCollector;
use tantivy::query::{AllQuery, TermQuery};
use tantivy::schema::{IndexRecordOption, Schema, TextFieldIndexing, FAST, STRING};
use tantivy::{doc, Index, Term};
#[global_allocator]
pub static GLOBAL: &PeakMemAlloc<std::alloc::System> = &INSTRUMENTED_SYSTEM;
/// Mini macro to register a function via its name
/// runner.register("average_u64", move |index| average_u64(index));
macro_rules! register {
($runner:expr, $func:ident) => {
$runner.register(stringify!($func), move |index| {
$func(index);
})
};
}
fn main() {
let inputs = vec![
("full", get_test_index_bench(Cardinality::Full).unwrap()),
(
"dense",
get_test_index_bench(Cardinality::OptionalDense).unwrap(),
),
(
"sparse",
get_test_index_bench(Cardinality::OptionalSparse).unwrap(),
),
(
"multivalue",
get_test_index_bench(Cardinality::Multivalued).unwrap(),
),
];
bench_agg(InputGroup::new_with_inputs(inputs));
}
fn bench_agg(mut group: InputGroup<Index>) {
group.add_plugin(PeakMemAllocPlugin::new(GLOBAL));
register!(group, average_u64);
register!(group, average_f64);
register!(group, average_f64_u64);
register!(group, stats_f64);
register!(group, extendedstats_f64);
register!(group, percentiles_f64);
register!(group, terms_few);
register!(group, terms_many);
register!(group, terms_many_top_1000);
register!(group, terms_many_order_by_term);
register!(group, terms_many_with_top_hits);
register!(group, terms_many_with_avg_sub_agg);
register!(group, terms_many_json_mixed_type_with_avg_sub_agg);
register!(group, cardinality_agg);
register!(group, terms_few_with_cardinality_agg);
register!(group, range_agg);
register!(group, range_agg_with_avg_sub_agg);
register!(group, range_agg_with_term_agg_few);
register!(group, range_agg_with_term_agg_many);
register!(group, histogram);
register!(group, histogram_hard_bounds);
register!(group, histogram_with_avg_sub_agg);
register!(group, avg_and_range_with_avg_sub_agg);
group.run();
}
fn exec_term_with_agg(index: &Index, agg_req: serde_json::Value) {
let agg_req: Aggregations = serde_json::from_value(agg_req).unwrap();
let reader = index.reader().unwrap();
let text_field = reader.searcher().schema().get_field("text").unwrap();
let term_query = TermQuery::new(
Term::from_field_text(text_field, "cool"),
IndexRecordOption::Basic,
);
let collector = get_collector(agg_req);
let searcher = reader.searcher();
black_box(searcher.search(&term_query, &collector).unwrap());
}
fn average_u64(index: &Index) {
let agg_req = json!({
"average": { "avg": { "field": "score", } }
});
exec_term_with_agg(index, agg_req)
}
fn average_f64(index: &Index) {
let agg_req = json!({
"average": { "avg": { "field": "score_f64", } }
});
exec_term_with_agg(index, agg_req)
}
fn average_f64_u64(index: &Index) {
let agg_req = json!({
"average_f64": { "avg": { "field": "score_f64" } },
"average": { "avg": { "field": "score" } },
});
exec_term_with_agg(index, agg_req)
}
fn stats_f64(index: &Index) {
let agg_req = json!({
"average_f64": { "stats": { "field": "score_f64", } }
});
exec_term_with_agg(index, agg_req)
}
fn extendedstats_f64(index: &Index) {
let agg_req = json!({
"extendedstats_f64": { "extended_stats": { "field": "score_f64", } }
});
exec_term_with_agg(index, agg_req)
}
fn percentiles_f64(index: &Index) {
let agg_req = json!({
"mypercentiles": {
"percentiles": {
"field": "score_f64",
"percents": [ 95, 99, 99.9 ]
}
}
});
execute_agg(index, agg_req);
}
fn cardinality_agg(index: &Index) {
let agg_req = json!({
"cardinality": {
"cardinality": {
"field": "text_many_terms"
},
}
});
execute_agg(index, agg_req);
}
fn terms_few_with_cardinality_agg(index: &Index) {
let agg_req = json!({
"my_texts": {
"terms": { "field": "text_few_terms" },
"aggs": {
"cardinality": {
"cardinality": {
"field": "text_many_terms"
},
}
}
},
});
execute_agg(index, agg_req);
}
fn terms_few(index: &Index) {
let agg_req = json!({
"my_texts": { "terms": { "field": "text_few_terms" } },
});
execute_agg(index, agg_req);
}
fn terms_many(index: &Index) {
let agg_req = json!({
"my_texts": { "terms": { "field": "text_many_terms" } },
});
execute_agg(index, agg_req);
}
fn terms_many_top_1000(index: &Index) {
let agg_req = json!({
"my_texts": { "terms": { "field": "text_many_terms", "size": 1000 } },
});
execute_agg(index, agg_req);
}
fn terms_many_order_by_term(index: &Index) {
let agg_req = json!({
"my_texts": { "terms": { "field": "text_many_terms", "order": { "_key": "desc" } } },
});
execute_agg(index, agg_req);
}
fn terms_many_with_top_hits(index: &Index) {
let agg_req = json!({
"my_texts": {
"terms": { "field": "text_many_terms" },
"aggs": {
"top_hits": { "top_hits":
{
"sort": [
{ "score": "desc" }
],
"size": 2,
"doc_value_fields": ["score_f64"]
}
}
}
},
});
execute_agg(index, agg_req);
}
fn terms_many_with_avg_sub_agg(index: &Index) {
let agg_req = json!({
"my_texts": {
"terms": { "field": "text_many_terms" },
"aggs": {
"average_f64": { "avg": { "field": "score_f64" } }
}
},
});
execute_agg(index, agg_req);
}
fn terms_many_json_mixed_type_with_avg_sub_agg(index: &Index) {
let agg_req = json!({
"my_texts": {
"terms": { "field": "json.mixed_type" },
"aggs": {
"average_f64": { "avg": { "field": "score_f64" } }
}
},
});
execute_agg(index, agg_req);
}
fn execute_agg(index: &Index, agg_req: serde_json::Value) {
let agg_req: Aggregations = serde_json::from_value(agg_req).unwrap();
let collector = get_collector(agg_req);
let reader = index.reader().unwrap();
let searcher = reader.searcher();
black_box(searcher.search(&AllQuery, &collector).unwrap());
}
fn range_agg(index: &Index) {
let agg_req = json!({
"range_f64": { "range": { "field": "score_f64", "ranges": [
{ "from": 3, "to": 7000 },
{ "from": 7000, "to": 20000 },
{ "from": 20000, "to": 30000 },
{ "from": 30000, "to": 40000 },
{ "from": 40000, "to": 50000 },
{ "from": 50000, "to": 60000 }
] } },
});
execute_agg(index, agg_req);
}
fn range_agg_with_avg_sub_agg(index: &Index) {
let agg_req = json!({
"rangef64": {
"range": {
"field": "score_f64",
"ranges": [
{ "from": 3, "to": 7000 },
{ "from": 7000, "to": 20000 },
{ "from": 20000, "to": 30000 },
{ "from": 30000, "to": 40000 },
{ "from": 40000, "to": 50000 },
{ "from": 50000, "to": 60000 }
]
},
"aggs": {
"average_f64": { "avg": { "field": "score_f64" } }
}
},
});
execute_agg(index, agg_req);
}
fn range_agg_with_term_agg_few(index: &Index) {
let agg_req = json!({
"rangef64": {
"range": {
"field": "score_f64",
"ranges": [
{ "from": 3, "to": 7000 },
{ "from": 7000, "to": 20000 },
{ "from": 20000, "to": 30000 },
{ "from": 30000, "to": 40000 },
{ "from": 40000, "to": 50000 },
{ "from": 50000, "to": 60000 }
]
},
"aggs": {
"my_texts": { "terms": { "field": "text_few_terms" } },
}
},
});
execute_agg(index, agg_req);
}
fn range_agg_with_term_agg_many(index: &Index) {
let agg_req = json!({
"rangef64": {
"range": {
"field": "score_f64",
"ranges": [
{ "from": 3, "to": 7000 },
{ "from": 7000, "to": 20000 },
{ "from": 20000, "to": 30000 },
{ "from": 30000, "to": 40000 },
{ "from": 40000, "to": 50000 },
{ "from": 50000, "to": 60000 }
]
},
"aggs": {
"my_texts": { "terms": { "field": "text_many_terms" } },
}
},
});
execute_agg(index, agg_req);
}
fn histogram(index: &Index) {
let agg_req = json!({
"rangef64": {
"histogram": {
"field": "score_f64",
"interval": 100 // 1000 buckets
},
}
});
execute_agg(index, agg_req);
}
fn histogram_hard_bounds(index: &Index) {
let agg_req = json!({
"rangef64": { "histogram": { "field": "score_f64", "interval": 100, "hard_bounds": { "min": 1000, "max": 300000 } } },
});
execute_agg(index, agg_req);
}
fn histogram_with_avg_sub_agg(index: &Index) {
let agg_req = json!({
"rangef64": {
"histogram": { "field": "score_f64", "interval": 100 },
"aggs": {
"average_f64": { "avg": { "field": "score_f64" } }
}
}
});
execute_agg(index, agg_req);
}
fn avg_and_range_with_avg_sub_agg(index: &Index) {
let agg_req = json!({
"rangef64": {
"range": {
"field": "score_f64",
"ranges": [
{ "from": 3, "to": 7000 },
{ "from": 7000, "to": 20000 },
{ "from": 20000, "to": 60000 }
]
},
"aggs": {
"average_in_range": { "avg": { "field": "score" } }
}
},
"average": { "avg": { "field": "score" } }
});
execute_agg(index, agg_req);
}
#[derive(Clone, Copy, Hash, Default, Debug, PartialEq, Eq, PartialOrd, Ord)]
enum Cardinality {
/// All documents contain exactly one value.
/// `Full` is the default for auto-detecting the Cardinality, since it is the most strict.
#[default]
Full = 0,
/// All documents contain at most one value.
OptionalDense = 1,
/// All documents may contain any number of values.
Multivalued = 2,
/// 1 / 20 documents has a value
OptionalSparse = 3,
}
fn get_collector(agg_req: Aggregations) -> AggregationCollector {
AggregationCollector::from_aggs(agg_req, Default::default())
}
fn get_test_index_bench(cardinality: Cardinality) -> tantivy::Result<Index> {
let mut schema_builder = Schema::builder();
let text_fieldtype = tantivy::schema::TextOptions::default()
.set_indexing_options(
TextFieldIndexing::default().set_index_option(IndexRecordOption::WithFreqs),
)
.set_stored();
let text_field = schema_builder.add_text_field("text", text_fieldtype);
let json_field = schema_builder.add_json_field("json", FAST);
let text_field_many_terms = schema_builder.add_text_field("text_many_terms", STRING | FAST);
let text_field_few_terms = schema_builder.add_text_field("text_few_terms", STRING | FAST);
let score_fieldtype = tantivy::schema::NumericOptions::default().set_fast();
let score_field = schema_builder.add_u64_field("score", score_fieldtype.clone());
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 = ["INFO", "ERROR", "WARN", "DEBUG"];
let lg_norm = rand_distr::LogNormal::new(2.996f64, 0.979f64).unwrap();
let many_terms_data = (0..150_000)
.map(|num| format!("author{num}"))
.collect::<Vec<_>>();
{
let mut rng = StdRng::from_seed([1u8; 32]);
let mut index_writer = index.writer_with_num_threads(1, 200_000_000)?;
// To make the different test cases comparable we just change one doc to force the
// cardinality
if cardinality == Cardinality::OptionalDense {
index_writer.add_document(doc!())?;
}
if cardinality == Cardinality::Multivalued {
index_writer.add_document(doc!(
json_field => json!({"mixed_type": 10.0}),
json_field => json!({"mixed_type": 10.0}),
text_field => "cool",
text_field => "cool",
text_field_many_terms => "cool",
text_field_many_terms => "cool",
text_field_few_terms => "cool",
text_field_few_terms => "cool",
score_field => 1u64,
score_field => 1u64,
score_field_f64 => lg_norm.sample(&mut rng),
score_field_f64 => lg_norm.sample(&mut rng),
score_field_i64 => 1i64,
score_field_i64 => 1i64,
))?;
}
let mut doc_with_value = 1_000_000;
if cardinality == Cardinality::OptionalSparse {
doc_with_value /= 20;
}
let _val_max = 1_000_000.0;
for _ in 0..doc_with_value {
let val: f64 = rng.gen_range(0.0..1_000_000.0);
let json = if rng.gen_bool(0.1) {
// 10% are numeric values
json!({ "mixed_type": val })
} else {
json!({"mixed_type": many_terms_data.choose(&mut rng).unwrap().to_string()})
};
index_writer.add_document(doc!(
text_field => "cool",
json_field => json,
text_field_many_terms => many_terms_data.choose(&mut rng).unwrap().to_string(),
text_field_few_terms => few_terms_data.choose(&mut rng).unwrap().to_string(),
score_field => val as u64,
score_field_f64 => lg_norm.sample(&mut rng),
score_field_i64 => val as i64,
))?;
if cardinality == Cardinality::OptionalSparse {
for _ in 0..20 {
index_writer.add_document(doc!(text_field => "cool"))?;
}
}
}
// writing the segment
index_writer.commit()?;
}
Ok(index)
}

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 +1,98 @@
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::Value = serde_json::from_str(doc_json).unwrap();
tantivy::doc!(json_field=>json_val)
})
}
pub fn hdfs_index_benchmark(c: &mut Criterion) {
@@ -19,7 +103,14 @@ 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 +119,40 @@ 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" };
{
let parse_json = false;
// for parse_json in [false, true] {
let suffix = if parse_json {
format!("{suffix}-with-json-parsing")
} else {
suffix.to_string()
};
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 = "0.8.0"
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"

34
bitpacker/src/bitpacker.rs
View File

@@ -1,4 +1,3 @@
use std::convert::TryInto;
use std::io;
use std::ops::{Range, RangeInclusive};
@@ -66,7 +65,7 @@ impl BitPacker {
#[derive(Clone, Debug, Default, Copy)]
pub struct BitUnpacker {
num_bits: u32,
num_bits: usize,
mask: u64,
}
@@ -84,7 +83,7 @@ impl BitUnpacker {
(1u64 << num_bits) - 1u64
};
BitUnpacker {
num_bits: u32::from(num_bits),
num_bits: usize::from(num_bits),
mask,
}
}
@@ -95,14 +94,14 @@ impl BitUnpacker {
#[inline]
pub fn get(&self, idx: u32, data: &[u8]) -> u64 {
let addr_in_bits = idx * self.num_bits;
let addr = (addr_in_bits >> 3) as usize;
let addr_in_bits = idx as usize * self.num_bits;
let addr = addr_in_bits >> 3;
if addr + 8 > data.len() {
if self.num_bits == 0 {
return 0;
}
let bit_shift = addr_in_bits & 7;
return self.get_slow_path(addr, bit_shift, data);
return self.get_slow_path(addr, bit_shift as u32, data);
}
let bit_shift = addr_in_bits & 7;
let bytes: [u8; 8] = (&data[addr..addr + 8]).try_into().unwrap();
@@ -135,12 +134,13 @@ impl BitUnpacker {
"Bitwidth must be <= 32 to use this method."
);
let end_idx = start_idx + output.len() as u32;
let end_idx: u32 = start_idx + output.len() as u32;
let end_bit_read = end_idx * self.num_bits;
// We use `usize` here to avoid overflow issues.
let end_bit_read = (end_idx as usize) * self.num_bits;
let end_byte_read = (end_bit_read + 7) / 8;
assert!(
end_byte_read as usize <= data.len(),
end_byte_read <= data.len(),
"Requested index is out of bounds."
);
@@ -160,24 +160,24 @@ impl BitUnpacker {
// We want the start of the fast track to start align with bytes.
// A sufficient condition is to start with an idx that is a multiple of 8,
// so highway start is the closest multiple of 8 that is >= start_idx.
let entrance_ramp_len = 8 - (start_idx % 8) % 8;
let entrance_ramp_len: u32 = 8 - (start_idx % 8) % 8;
let highway_start: u32 = start_idx + entrance_ramp_len;
if highway_start + BitPacker1x::BLOCK_LEN as u32 > end_idx {
if highway_start + (BitPacker1x::BLOCK_LEN as u32) > end_idx {
// We don't have enough values to have even a single block of highway.
// Let's just supply the values the simple way.
get_batch_ramp(start_idx, output);
return;
}
let num_blocks: u32 = (end_idx - highway_start) / BitPacker1x::BLOCK_LEN as u32;
let num_blocks: usize = (end_idx - highway_start) as usize / BitPacker1x::BLOCK_LEN;
// Entrance ramp
get_batch_ramp(start_idx, &mut output[..entrance_ramp_len as usize]);
// Highway
let mut offset = (highway_start * self.num_bits) as usize / 8;
let mut offset = (highway_start as usize * self.num_bits) / 8;
let mut output_cursor = (highway_start - start_idx) as usize;
for _ in 0..num_blocks {
offset += BitPacker1x.decompress(
@@ -189,7 +189,7 @@ impl BitUnpacker {
}
// Exit ramp
let highway_end = highway_start + num_blocks * BitPacker1x::BLOCK_LEN as u32;
let highway_end: u32 = highway_start + (num_blocks * BitPacker1x::BLOCK_LEN) as u32;
get_batch_ramp(highway_end, &mut output[output_cursor..]);
}
@@ -367,11 +367,11 @@ 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 {
for (i, output_byte) in output.iter().enumerate() {
let expected = (start_idx + i as u32) & mask;
assert_eq!(output[i], expected);
assert_eq!(*output_byte, expected);
}
}
}

6
bitpacker/src/blocked_bitpacker.rs
View File

@@ -34,7 +34,7 @@ struct BlockedBitpackerEntryMetaData {
impl BlockedBitpackerEntryMetaData {
fn new(offset: u64, num_bits: u8, base_value: u64) -> Self {
let encoded = offset | (num_bits as u64) << (64 - 8);
let encoded = offset | (u64::from(num_bits) << (64 - 8));
Self {
encoded,
base_value,
@@ -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![],
}

4
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;
@@ -35,8 +35,8 @@ const IMPLS: [FilterImplPerInstructionSet; 2] = [
const IMPLS: [FilterImplPerInstructionSet; 1] = [FilterImplPerInstructionSet::Scalar];
impl FilterImplPerInstructionSet {
#[allow(unused_variables)]
#[inline]
#[allow(unused_variables)] // on non-x86_64, code is unused.
fn from(code: u8) -> FilterImplPerInstructionSet {
#[cfg(target_arch = "x86_64")]
if code == FilterImplPerInstructionSet::AVX2 as u8 {

82
cliff.toml
View File

@@ -1,6 +1,10 @@
# configuration file for git-cliff{ pattern = "foo", replace = "bar"}
# see https://github.com/orhun/git-cliff#configuration-file
[remote.github]
owner = "quickwit-oss"
repo = "tantivy"
[changelog]
# changelog header
header = """
@@ -8,15 +12,43 @@ header = """
# template for the changelog body
# https://tera.netlify.app/docs/#introduction
body = """
{% if version %}\
{{ version | trim_start_matches(pat="v") }} ({{ timestamp | date(format="%Y-%m-%d") }})
==================
{% else %}\
## [unreleased]
{% endif %}\
## What's Changed
{%- if version %} in {{ version }}{%- endif -%}
{% for commit in commits %}
- {% if commit.breaking %}[**breaking**] {% endif %}{{ commit.message | split(pat="\n") | first | trim | upper_first }}(@{{ commit.author.name }})\
{% endfor %}
{% if commit.remote.pr_title -%}
{%- set commit_message = commit.remote.pr_title -%}
{%- else -%}
{%- set commit_message = commit.message -%}
{%- endif -%}
- {{ commit_message | split(pat="\n") | first | trim }}\
{% if commit.remote.pr_number %} \
[#{{ commit.remote.pr_number }}]({{ self::remote_url() }}/pull/{{ commit.remote.pr_number }}){% if commit.remote.username %}(@{{ commit.remote.username }}){%- endif -%} \
{%- endif %}
{%- endfor -%}
{% if github.contributors | filter(attribute="is_first_time", value=true) | length != 0 %}
{% raw %}\n{% endraw -%}
## New Contributors
{%- endif %}\
{% for contributor in github.contributors | filter(attribute="is_first_time", value=true) %}
* @{{ contributor.username }} made their first contribution
{%- if contributor.pr_number %} in \
[#{{ contributor.pr_number }}]({{ self::remote_url() }}/pull/{{ contributor.pr_number }}) \
{%- endif %}
{%- endfor -%}
{% if version %}
{% if previous.version %}
**Full Changelog**: {{ self::remote_url() }}/compare/{{ previous.version }}...{{ version }}
{% endif %}
{% else -%}
{% raw %}\n{% endraw %}
{% endif %}
{%- macro remote_url() -%}
https://github.com/{{ remote.github.owner }}/{{ remote.github.repo }}
{%- endmacro -%}
"""
# remove the leading and trailing whitespace from the template
trim = true
@@ -25,52 +57,24 @@ footer = """
"""
postprocessors = [
{ pattern = 'Paul Masurel', replace = "fulmicoton"}, # replace with github user
{ pattern = 'PSeitz', replace = "PSeitz"}, # replace with github user
{ pattern = 'Adam Reichold', replace = "adamreichold"}, # replace with github user
{ pattern = 'trinity-1686a', replace = "trinity-1686a"}, # replace with github user
{ pattern = 'Michael Kleen', replace = "mkleen"}, # replace with github user
{ pattern = 'Adrien Guillo', replace = "guilload"}, # replace with github user
{ pattern = 'François Massot', replace = "fmassot"}, # replace with github user
{ pattern = '', replace = ""}, # replace with github user
]
[git]
# parse the commits based on https://www.conventionalcommits.org
# This is required or commit.message contains the whole commit message and not just the title
conventional_commits = true
conventional_commits = false
# filter out the commits that are not conventional
filter_unconventional = false
filter_unconventional = true
# process each line of a commit as an individual commit
split_commits = false
# regex for preprocessing the commit messages
commit_preprocessors = [
{ pattern = '\((\w+\s)?#([0-9]+)\)', replace = "[#${2}](https://github.com/quickwit-oss/tantivy/issues/${2})"}, # replace issue numbers
{ pattern = '\((\w+\s)?#([0-9]+)\)', replace = ""},
]
#link_parsers = [
#{ pattern = "#(\\d+)", href = "https://github.com/quickwit-oss/tantivy/pulls/$1"},
#]
# regex for parsing and grouping commits
commit_parsers = [
{ message = "^feat", group = "Features"},
{ message = "^fix", group = "Bug Fixes"},
{ message = "^doc", group = "Documentation"},
{ message = "^perf", group = "Performance"},
{ message = "^refactor", group = "Refactor"},
{ message = "^style", group = "Styling"},
{ message = "^test", group = "Testing"},
{ message = "^chore\\(release\\): prepare for", skip = true},
{ message = "(?i)clippy", skip = true},
{ message = "(?i)dependabot", skip = true},
{ message = "(?i)fmt", skip = true},
{ message = "(?i)bump", skip = true},
{ message = "(?i)readme", skip = true},
{ message = "(?i)comment", skip = true},
{ message = "(?i)spelling", skip = true},
{ message = "^chore", group = "Miscellaneous Tasks"},
{ body = ".*security", group = "Security"},
{ message = ".*", group = "Other", default_scope = "other"},
]
# protect breaking changes from being skipped due to matching a skipping commit_parser
protect_breaking_commits = false
# filter out the commits that are not matched by commit parsers

26
columnar/Cargo.toml
View File

@@ -1,28 +1,38 @@
[package]
name = "tantivy-columnar"
version = "0.1.0"
version = "0.5.0"
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 = "0.14.0"
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= "0.5", path = "../stacker", package="tantivy-stacker"}
sstable = { version= "0.5", path = "../sstable", package = "tantivy-sstable" }
common = { version= "0.9", path = "../common", package = "tantivy-common" }
tantivy-bitpacker = { version= "0.8", path = "../bitpacker/" }
serde = "1.0.152"
downcast-rs = "2.0.1"
[dev-dependencies]
proptest = "1"
more-asserts = "0.3.1"
rand = "0.8"
binggan = "0.14.0"
[[bench]]
name = "bench_merge"
harness = false
[[bench]]
name = "bench_access"
harness = false
[features]
unstable = []

2
columnar/README.md
View File

@@ -31,7 +31,7 @@ restriction on 50% of the values (e.g. a 64-bit hash). On the other hand, a lot
# Columnar format
This columnar format may have more than one column (with different types) associated to the same `column_name` (see [Coercion rules](#coercion-rules) above).
The `(column_name, columne_type)` couple however uniquely identifies a column.
The `(column_name, column_type)` couple however uniquely identifies a column.
That couple is serialized as a column `column_key`. The format of that key is:
`[column_name][ZERO_BYTE][column_type_header: u8]`

67
columnar/benches/bench_access.rs Normal file
View File

@@ -0,0 +1,67 @@
use binggan::{black_box, InputGroup};
use common::*;
use tantivy_columnar::Column;
pub mod common;
const NUM_DOCS: u32 = 2_000_000;
pub fn generate_columnar_and_open(card: Card, num_docs: u32) -> Column {
let reader = generate_columnar_with_name(card, num_docs, "price");
reader.read_columns("price").unwrap()[0]
.open_u64_lenient()
.unwrap()
.unwrap()
}
fn main() {
let mut inputs = Vec::new();
let mut add_card = |card1: Card| {
inputs.push((
format!("{card1}"),
generate_columnar_and_open(card1, NUM_DOCS),
));
};
add_card(Card::MultiSparse);
add_card(Card::Multi);
add_card(Card::Sparse);
add_card(Card::Dense);
add_card(Card::Full);
bench_group(InputGroup::new_with_inputs(inputs));
}
fn bench_group(mut runner: InputGroup<Column>) {
runner.register("access_values_for_doc", |column| {
let mut sum = 0;
for i in 0..NUM_DOCS {
for value in column.values_for_doc(i) {
sum += value;
}
}
black_box(sum);
});
runner.register("access_first_vals", |column| {
let mut sum = 0;
const BLOCK_SIZE: usize = 32;
let mut docs = vec![0; BLOCK_SIZE];
let mut buffer = vec![None; BLOCK_SIZE];
for i in (0..NUM_DOCS).step_by(BLOCK_SIZE) {
// fill docs
for idx in 0..BLOCK_SIZE {
docs[idx] = idx as u32 + i;
}
column.first_vals(&docs, &mut buffer);
for val in buffer.iter() {
let Some(val) = val else { continue };
sum += *val;
}
}
black_box(sum);
});
runner.run();
}

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, &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);
}

49
columnar/benches/bench_merge.rs Normal file
View File

@@ -0,0 +1,49 @@
pub mod common;
use binggan::BenchRunner;
use common::{generate_columnar_with_name, Card};
use tantivy_columnar::*;
const NUM_DOCS: u32 = 100_000;
fn main() {
let mut inputs = Vec::new();
let mut add_combo = |card1: Card, card2: Card| {
inputs.push((
format!("merge_{card1}_and_{card2}"),
vec![
generate_columnar_with_name(card1, NUM_DOCS, "price"),
generate_columnar_with_name(card2, NUM_DOCS, "price"),
],
));
};
add_combo(Card::Multi, Card::Multi);
add_combo(Card::MultiSparse, Card::MultiSparse);
add_combo(Card::Dense, Card::Dense);
add_combo(Card::Sparse, Card::Sparse);
add_combo(Card::Sparse, Card::Dense);
add_combo(Card::MultiSparse, Card::Dense);
add_combo(Card::MultiSparse, Card::Sparse);
add_combo(Card::Multi, Card::Dense);
add_combo(Card::Multi, Card::Sparse);
let mut runner: BenchRunner = BenchRunner::new();
let mut group = runner.new_group();
for (input_name, columnar_readers) in inputs.iter() {
group.register_with_input(
input_name,
columnar_readers,
move |columnar_readers: &Vec<ColumnarReader>| {
let mut out = Vec::new();
let columnar_readers = columnar_readers.iter().collect::<Vec<_>>();
let merge_row_order = StackMergeOrder::stack(&columnar_readers[..]);
merge_columnar(&columnar_readers, &[], merge_row_order.into(), &mut out).unwrap();
Some(out.len() as u64)
},
);
}
group.run();
}

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();

59
columnar/benches/common.rs Normal file
View File

@@ -0,0 +1,59 @@
extern crate tantivy_columnar;
use core::fmt;
use std::fmt::{Display, Formatter};
use tantivy_columnar::{ColumnarReader, ColumnarWriter};
pub enum Card {
MultiSparse,
Multi,
Sparse,
Dense,
Full,
}
impl Display for Card {
fn fmt(&self, f: &mut Formatter) -> fmt::Result {
match self {
Card::MultiSparse => write!(f, "multi sparse 1/13"),
Card::Multi => write!(f, "multi 2x"),
Card::Sparse => write!(f, "sparse 1/13"),
Card::Dense => write!(f, "dense 1/12"),
Card::Full => write!(f, "full"),
}
}
}
pub fn generate_columnar_with_name(card: Card, num_docs: u32, column_name: &str) -> ColumnarReader {
let mut columnar_writer = ColumnarWriter::default();
if let Card::MultiSparse = card {
columnar_writer.record_numerical(0, column_name, 10u64);
columnar_writer.record_numerical(0, column_name, 10u64);
}
for i in 0..num_docs {
match card {
Card::MultiSparse | Card::Sparse => {
if i % 13 == 0 {
columnar_writer.record_numerical(i, column_name, i as u64);
}
}
Card::Dense => {
if i % 12 == 0 {
columnar_writer.record_numerical(i, column_name, i as u64);
}
}
Card::Full => {
columnar_writer.record_numerical(i, column_name, i as u64);
}
Card::Multi => {
columnar_writer.record_numerical(i, column_name, i as u64);
columnar_writer.record_numerical(i, column_name, i as u64);
}
}
}
let mut wrt: Vec<u8> = Vec::new();
columnar_writer.serialize(num_docs, &mut wrt).unwrap();
ColumnarReader::open(wrt).unwrap()
}

18
columnar/columnar-cli-inspect/Cargo.toml Normal file
View File

@@ -0,0 +1,18 @@
[package]
name = "tantivy-columnar-inspect"
version = "0.1.0"
edition = "2021"
license = "MIT"
[dependencies]
tantivy = {path="../..", package="tantivy"}
columnar = {path="../", package="tantivy-columnar"}
common = {path="../../common", package="tantivy-common"}
[workspace]
members = []
[profile.release]
debug = true
#debug-assertions = true
#overflow-checks = true

54
columnar/columnar-cli-inspect/src/main.rs Normal file
View File

@@ -0,0 +1,54 @@
use columnar::ColumnarReader;
use common::file_slice::{FileSlice, WrapFile};
use std::io;
use std::path::Path;
use tantivy::directory::footer::Footer;
fn main() -> io::Result<()> {
println!("Opens a columnar file written by tantivy and validates it.");
let path = std::env::args().nth(1).unwrap();
let path = Path::new(&path);
println!("Reading {:?}", path);
let _reader = open_and_validate_columnar(path.to_str().unwrap())?;
Ok(())
}
pub fn validate_columnar_reader(reader: &ColumnarReader) {
let num_rows = reader.num_rows();
println!("num_rows: {}", num_rows);
let columns = reader.list_columns().unwrap();
println!("num columns: {:?}", columns.len());
for (col_name, dynamic_column_handle) in columns {
let col = dynamic_column_handle.open().unwrap();
match col {
columnar::DynamicColumn::Bool(_)
| columnar::DynamicColumn::I64(_)
| columnar::DynamicColumn::U64(_)
| columnar::DynamicColumn::F64(_)
| columnar::DynamicColumn::IpAddr(_)
| columnar::DynamicColumn::DateTime(_)
| columnar::DynamicColumn::Bytes(_) => {}
columnar::DynamicColumn::Str(str_column) => {
let num_vals = str_column.ords().values.num_vals();
let num_terms_dict = str_column.num_terms() as u64;
let max_ord = str_column.ords().values.iter().max().unwrap_or_default();
println!("{col_name:35} num_vals {num_vals:10} \t num_terms_dict {num_terms_dict:8} max_ord: {max_ord:8}",);
for ord in str_column.ords().values.iter() {
assert!(ord < num_terms_dict);
}
}
}
}
}
/// Opens a columnar file that was written by tantivy and validates it.
pub fn open_and_validate_columnar(path: &str) -> io::Result<ColumnarReader> {
let wrap_file = WrapFile::new(std::fs::File::open(path)?)?;
let slice = FileSlice::new(std::sync::Arc::new(wrap_file));
let (_footer, slice) = Footer::extract_footer(slice.clone()).unwrap();
let reader = ColumnarReader::open(slice).unwrap();
validate_columnar_reader(&reader);
Ok(reader)
}

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 = []

BIN
columnar/compat_tests_data/v1.columnar Normal file
View File

Binary file not shown.

BIN
columnar/compat_tests_data/v2.columnar Normal file
View File

Binary file not shown.

6
columnar/src/TODO.md
View File

@@ -10,7 +10,7 @@
# Perf and Size
* remove alloc in `ord_to_term`
+ multivaued range queries restrat frm the beginning all of the time.
+ multivaued range queries restart from the beginning all of the time.
* re-add ZSTD compression for dictionaries
no systematic monotonic mapping
consider removing multilinear
@@ -30,7 +30,7 @@ investigate if should have better errors? io::Error is overused at the moment.
rename rank/select in unit tests
Review the public API via cargo doc
go through TODOs
remove all doc_id occurences -> row_id
remove all doc_id occurrences -> row_id
use the rank & select naming in unit tests branch.
multi-linear -> blockwise
linear codec -> simply a multiplication for the index column
@@ -43,5 +43,5 @@ isolate u128_based and uniform naming
# Other
fix enhance column-cli
# Santa claus
# Santa Claus
autodetect datetime ipaddr, plug customizable tokenizer.

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)> + 'a {
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::<u32>::new());
}
#[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)
}

50
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
@@ -105,11 +131,12 @@ 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))
}
/// Get the docids of values which are in the provided value range.
/// Get the docids of values which are in the provided value and docid range.
#[inline]
pub fn get_docids_for_value_range(
&self,
@@ -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)
}

75
columnar/src/column/serialize.rs
View File

@@ -12,7 +12,7 @@ use crate::column_values::{
CodecType, MonotonicallyMappableToU128, MonotonicallyMappableToU64,
};
use crate::iterable::Iterable;
use crate::StrColumn;
use crate::{StrColumn, Version};
pub fn serialize_column_mappable_to_u128<T: MonotonicallyMappableToU128>(
column_index: SerializableColumnIndex<'_>,
@@ -40,25 +40,9 @@ pub fn serialize_column_mappable_to_u64<T: MonotonicallyMappableToU64>(
Ok(())
}
pub fn open_column_u64<T: MonotonicallyMappableToU64>(bytes: OwnedBytes) -> io::Result<Column<T>> {
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 = load_u64_based_column_values(column_values_data)?;
Ok(Column {
index: column_index,
values: column_values,
})
}
pub fn open_column_u128<T: MonotonicallyMappableToU128>(
pub fn open_column_u64<T: MonotonicallyMappableToU64>(
bytes: OwnedBytes,
format_version: Version,
) -> io::Result<Column<T>> {
let (body, column_index_num_bytes_payload) = bytes.rsplit(4);
let column_index_num_bytes = u32::from_le_bytes(
@@ -68,7 +52,27 @@ pub fn open_column_u128<T: MonotonicallyMappableToU128>(
.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_index = crate::column_index::open_column_index(column_index_data, format_version)?;
let column_values = load_u64_based_column_values(column_values_data)?;
Ok(Column {
index: column_index,
values: column_values,
})
}
pub fn open_column_u128<T: MonotonicallyMappableToU128>(
bytes: OwnedBytes,
format_version: Version,
) -> io::Result<Column<T>> {
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, format_version)?;
let column_values = crate::column_values::open_u128_mapped(column_values_data)?;
Ok(Column {
index: column_index,
@@ -76,19 +80,42 @@ pub fn open_column_u128<T: MonotonicallyMappableToU128>(
})
}
pub fn open_column_bytes(data: OwnedBytes) -> io::Result<BytesColumn> {
/// Open the column as u64.
///
/// See [`open_u128_as_compact_u64`] for more details.
pub fn open_column_u128_as_compact_u64(
bytes: OwnedBytes,
format_version: Version,
) -> 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, format_version)?;
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, format_version: Version) -> 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());
let (dictionary_bytes, column_bytes) = body.split(dictionary_len as usize);
let dictionary = Arc::new(Dictionary::from_bytes(dictionary_bytes)?);
let term_ord_column = crate::column::open_column_u64::<u64>(column_bytes)?;
let term_ord_column = crate::column::open_column_u64::<u64>(column_bytes, format_version)?;
Ok(BytesColumn {
dictionary,
term_ord_column,
})
}
pub fn open_column_str(data: OwnedBytes) -> io::Result<StrColumn> {
let bytes_column = open_column_bytes(data)?;
pub fn open_column_str(data: OwnedBytes, format_version: Version) -> io::Result<StrColumn> {
let bytes_column = open_column_bytes(data, format_version)?;
Ok(StrColumn::wrap(bytes_column))
}

29
columnar/src/column_index/merge/mod.rs
View File

@@ -95,8 +95,12 @@ pub fn merge_column_index<'a>(
#[cfg(test)]
mod tests {
use common::OwnedBytes;
use crate::column_index::merge::detect_cardinality;
use crate::column_index::multivalued_index::MultiValueIndex;
use crate::column_index::multivalued_index::{
open_multivalued_index, serialize_multivalued_index, MultiValueIndex,
};
use crate::column_index::{merge_column_index, OptionalIndex, SerializableColumnIndex};
use crate::{
Cardinality, ColumnIndex, MergeRowOrder, RowAddr, RowId, ShuffleMergeOrder, StackMergeOrder,
@@ -168,9 +172,14 @@ 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 start_indexes: Vec<RowId> = start_index_iterable.boxed_iter().collect();
let SerializableColumnIndex::Multivalued(start_index_iterable) = merged_column_index else {
panic!("Expected a multivalued index")
};
let mut output = Vec::new();
serialize_multivalued_index(&start_index_iterable, &mut output).unwrap();
let multivalue =
open_multivalued_index(OwnedBytes::new(output), crate::Version::V2).unwrap();
let start_indexes: Vec<RowId> = multivalue.get_start_index_column().iter().collect();
assert_eq!(&start_indexes, &[0, 3, 5]);
}
@@ -199,10 +208,16 @@ 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 start_indexes: Vec<RowId> = start_index_iterable.boxed_iter().collect();
let SerializableColumnIndex::Multivalued(start_index_iterable) = merged_column_index else {
panic!("Expected a multivalued index")
};
let mut output = Vec::new();
serialize_multivalued_index(&start_index_iterable, &mut output).unwrap();
let multivalue =
open_multivalued_index(OwnedBytes::new(output), crate::Version::V2).unwrap();
let start_indexes: Vec<RowId> = multivalue.get_start_index_column().iter().collect();
assert_eq!(&start_indexes, &[0, 3, 5, 6]);
}
}

54
columnar/src/column_index/merge/shuffled.rs
View File

@@ -1,6 +1,8 @@
use std::iter;
use crate::column_index::{SerializableColumnIndex, Set};
use crate::column_index::{
SerializableColumnIndex, SerializableMultivalueIndex, SerializableOptionalIndex, Set,
};
use crate::iterable::Iterable;
use crate::{Cardinality, ColumnIndex, RowId, ShuffleMergeOrder};
@@ -14,15 +16,24 @@ pub fn merge_column_index_shuffled<'a>(
Cardinality::Optional => {
let non_null_row_ids =
merge_column_index_shuffled_optional(column_indexes, shuffle_merge_order);
SerializableColumnIndex::Optional {
SerializableColumnIndex::Optional(SerializableOptionalIndex {
non_null_row_ids,
num_rows: shuffle_merge_order.num_rows(),
}
})
}
Cardinality::Multivalued => {
let multivalue_start_index =
merge_column_index_shuffled_multivalued(column_indexes, shuffle_merge_order);
SerializableColumnIndex::Multivalued(multivalue_start_index)
let non_null_row_ids =
merge_column_index_shuffled_optional(column_indexes, shuffle_merge_order);
SerializableColumnIndex::Multivalued(SerializableMultivalueIndex {
doc_ids_with_values: SerializableOptionalIndex {
non_null_row_ids,
num_rows: shuffle_merge_order.num_rows(),
},
start_offsets: merge_column_index_shuffled_multivalued(
column_indexes,
shuffle_merge_order,
),
})
}
}
}
@@ -47,7 +58,7 @@ struct ShuffledIndex<'a> {
merge_order: &'a ShuffleMergeOrder,
}
impl<'a> Iterable<u32> for ShuffledIndex<'a> {
impl Iterable<u32> for ShuffledIndex<'_> {
fn boxed_iter(&self) -> Box<dyn Iterator<Item = u32> + '_> {
Box::new(
self.merge_order
@@ -102,14 +113,21 @@ fn iter_num_values<'a>(
/// Transforms an iterator containing the number of vals per row (with `num_rows` elements)
/// into a `start_offset` iterator starting at 0 and (with `num_rows + 1` element)
///
/// This will filter values with 0 values as these are covered by the optional index in the
/// multivalue index.
fn integrate_num_vals(num_vals: impl Iterator<Item = u32>) -> impl Iterator<Item = RowId> {
iter::once(0u32).chain(num_vals.scan(0, |state, num_vals| {
*state += num_vals;
Some(*state)
}))
iter::once(0u32).chain(
num_vals
.filter(|num_vals| *num_vals != 0)
.scan(0, |state, num_vals| {
*state += num_vals;
Some(*state)
}),
)
}
impl<'a> Iterable<u32> for ShuffledMultivaluedIndex<'a> {
impl Iterable<u32> for ShuffledMultivaluedIndex<'_> {
fn boxed_iter(&self) -> Box<dyn Iterator<Item = u32> + '_> {
let num_vals_per_row = iter_num_values(self.column_indexes, self.merge_order);
Box::new(integrate_num_vals(num_vals_per_row))
@@ -134,13 +152,13 @@ mod tests {
#[test]
fn test_integrate_num_vals_several() {
assert!(integrate_num_vals([3, 0, 10, 20].into_iter()).eq([0, 3, 3, 13, 33].into_iter()));
assert!(integrate_num_vals([3, 0, 10, 20].into_iter()).eq([0, 3, 13, 33].into_iter()));
}
#[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 +175,13 @@ mod tests {
Cardinality::Optional,
&shuffle_merge_order,
);
let SerializableColumnIndex::Optional { non_null_row_ids, num_rows } = serializable_index else { panic!() };
let SerializableColumnIndex::Optional(SerializableOptionalIndex {
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]);

235
columnar/src/column_index/merge/stacked.rs
View File

@@ -1,6 +1,8 @@
use std::iter;
use std::ops::Range;
use crate::column_index::{SerializableColumnIndex, Set};
use crate::column_index::multivalued_index::{MultiValueIndex, SerializableMultivalueIndex};
use crate::column_index::serialize::SerializableOptionalIndex;
use crate::column_index::SerializableColumnIndex;
use crate::iterable::Iterable;
use crate::{Cardinality, ColumnIndex, RowId, StackMergeOrder};
@@ -15,23 +17,149 @@ pub fn merge_column_index_stacked<'a>(
) -> SerializableColumnIndex<'a> {
match cardinality_after_merge {
Cardinality::Full => SerializableColumnIndex::Full,
Cardinality::Optional => SerializableColumnIndex::Optional {
Cardinality::Optional => SerializableColumnIndex::Optional(SerializableOptionalIndex {
non_null_row_ids: Box::new(StackedOptionalIndex {
columns,
stack_merge_order,
}),
num_rows: stack_merge_order.num_rows(),
},
}),
Cardinality::Multivalued => {
let stacked_multivalued_index = StackedMultivaluedIndex {
columns,
stack_merge_order,
};
SerializableColumnIndex::Multivalued(Box::new(stacked_multivalued_index))
let serializable_multivalue_index =
make_serializable_multivalued_index(columns, stack_merge_order);
SerializableColumnIndex::Multivalued(serializable_multivalue_index)
}
}
}
struct StackedDocIdsWithValues<'a> {
column_indexes: &'a [ColumnIndex],
stack_merge_order: &'a StackMergeOrder,
}
impl Iterable<u32> for StackedDocIdsWithValues<'_> {
fn boxed_iter(&self) -> Box<dyn Iterator<Item = u32> + '_> {
Box::new((0..self.column_indexes.len()).flat_map(|i| {
let column_index = &self.column_indexes[i];
let doc_range = self.stack_merge_order.columnar_range(i);
get_doc_ids_with_values(column_index, doc_range)
}))
}
}
fn get_doc_ids_with_values<'a>(
column_index: &'a ColumnIndex,
doc_range: Range<u32>,
) -> Box<dyn Iterator<Item = u32> + 'a> {
match column_index {
ColumnIndex::Empty { .. } => Box::new(0..0),
ColumnIndex::Full => Box::new(doc_range),
ColumnIndex::Optional(optional_index) => Box::new(
optional_index
.iter_docs()
.map(move |row| row + doc_range.start),
),
ColumnIndex::Multivalued(multivalued_index) => match multivalued_index {
MultiValueIndex::MultiValueIndexV1(multivalued_index) => {
Box::new((0..multivalued_index.num_docs()).filter_map(move |docid| {
let range = multivalued_index.range(docid);
if range.is_empty() {
None
} else {
Some(docid + doc_range.start)
}
}))
}
MultiValueIndex::MultiValueIndexV2(multivalued_index) => Box::new(
multivalued_index
.optional_index
.iter_docs()
.map(move |row| row + doc_range.start),
),
},
}
}
fn stack_doc_ids_with_values<'a>(
column_indexes: &'a [ColumnIndex],
stack_merge_order: &'a StackMergeOrder,
) -> SerializableOptionalIndex<'a> {
let num_rows = stack_merge_order.num_rows();
SerializableOptionalIndex {
non_null_row_ids: Box::new(StackedDocIdsWithValues {
column_indexes,
stack_merge_order,
}),
num_rows,
}
}
struct StackedStartOffsets<'a> {
column_indexes: &'a [ColumnIndex],
stack_merge_order: &'a StackMergeOrder,
}
fn get_num_values_iterator<'a>(
column_index: &'a ColumnIndex,
num_docs: u32,
) -> Box<dyn Iterator<Item = u32> + 'a> {
match column_index {
ColumnIndex::Empty { .. } => Box::new(std::iter::empty()),
ColumnIndex::Full => Box::new(std::iter::repeat(1u32).take(num_docs as usize)),
ColumnIndex::Optional(optional_index) => {
Box::new(std::iter::repeat(1u32).take(optional_index.num_non_nulls() as usize))
}
ColumnIndex::Multivalued(multivalued_index) => Box::new(
multivalued_index
.get_start_index_column()
.iter()
.scan(0u32, |previous_start_offset, current_start_offset| {
let num_vals = current_start_offset - *previous_start_offset;
*previous_start_offset = current_start_offset;
Some(num_vals)
})
.skip(1),
),
}
}
impl Iterable<u32> for StackedStartOffsets<'_> {
fn boxed_iter(&self) -> Box<dyn Iterator<Item = u32> + '_> {
let num_values_it = (0..self.column_indexes.len()).flat_map(|columnar_id| {
let num_docs = self.stack_merge_order.columnar_range(columnar_id).len() as u32;
let column_index = &self.column_indexes[columnar_id];
get_num_values_iterator(column_index, num_docs)
});
Box::new(std::iter::once(0u32).chain(num_values_it.into_iter().scan(
0u32,
|cumulated, el| {
*cumulated += el;
Some(*cumulated)
},
)))
}
}
fn stack_start_offsets<'a>(
column_indexes: &'a [ColumnIndex],
stack_merge_order: &'a StackMergeOrder,
) -> Box<dyn Iterable<u32> + 'a> {
Box::new(StackedStartOffsets {
column_indexes,
stack_merge_order,
})
}
fn make_serializable_multivalued_index<'a>(
columns: &'a [ColumnIndex],
stack_merge_order: &'a StackMergeOrder,
) -> SerializableMultivalueIndex<'a> {
SerializableMultivalueIndex {
doc_ids_with_values: stack_doc_ids_with_values(columns, stack_merge_order),
start_offsets: stack_start_offsets(columns, stack_merge_order),
}
}
struct StackedOptionalIndex<'a> {
columns: &'a [ColumnIndex],
stack_merge_order: &'a StackMergeOrder,
@@ -49,7 +177,7 @@ impl<'a> Iterable<RowId> for StackedOptionalIndex<'a> {
ColumnIndex::Full => Box::new(columnar_row_range),
ColumnIndex::Optional(optional_index) => Box::new(
optional_index
.iter_rows()
.iter_docs()
.map(move |row_id: RowId| columnar_row_range.start + row_id),
),
ColumnIndex::Multivalued(_) => {
@@ -62,90 +190,3 @@ impl<'a> Iterable<RowId> for StackedOptionalIndex<'a> {
)
}
}
#[derive(Clone, Copy)]
struct StackedMultivaluedIndex<'a> {
columns: &'a [ColumnIndex],
stack_merge_order: &'a StackMergeOrder,
}
fn convert_column_opt_to_multivalued_index<'a>(
column_index_opt: &'a ColumnIndex,
num_rows: RowId,
) -> Box<dyn Iterator<Item = RowId> + 'a> {
match column_index_opt {
ColumnIndex::Empty { .. } => Box::new(iter::repeat(0u32).take(num_rows as usize + 1)),
ColumnIndex::Full => Box::new(0..num_rows + 1),
ColumnIndex::Optional(optional_index) => {
Box::new(
(0..num_rows)
// TODO optimize
.map(|row_id| optional_index.rank(row_id))
.chain(std::iter::once(optional_index.num_non_nulls())),
)
}
ColumnIndex::Multivalued(multivalued_index) => multivalued_index.start_index_column.iter(),
}
}
impl<'a> Iterable<RowId> for StackedMultivaluedIndex<'a> {
fn boxed_iter(&self) -> Box<dyn Iterator<Item = RowId> + '_> {
let multivalued_indexes =
self.columns
.iter()
.enumerate()
.map(|(columnar_id, column_opt)| {
let num_rows =
self.stack_merge_order.columnar_range(columnar_id).len() as RowId;
convert_column_opt_to_multivalued_index(column_opt, num_rows)
});
stack_multivalued_indexes(multivalued_indexes)
}
}
// Refactor me
fn stack_multivalued_indexes<'a>(
mut multivalued_indexes: impl Iterator<Item = Box<dyn Iterator<Item = RowId> + 'a>> + 'a,
) -> Box<dyn Iterator<Item = RowId> + 'a> {
let mut offset = 0;
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() {
last_row_id = offset + row_id;
return Some(last_row_id);
}
offset = last_row_id;
loop {
current_it = multivalued_indexes.next();
if current_it.as_mut()?.next().is_some() {
break;
}
}
}))
}
#[cfg(test)]
mod tests {
use crate::RowId;
fn it<'a>(row_ids: &'a [RowId]) -> Box<dyn Iterator<Item = RowId> + 'a> {
Box::new(row_ids.iter().copied())
}
#[test]
fn test_stack() {
let columns = [
it(&[0u32, 0u32]),
it(&[0u32, 1u32, 1u32, 4u32]),
it(&[0u32, 3u32, 5u32]),
it(&[0u32, 4u32]),
]
.into_iter();
let start_offsets: Vec<RowId> = super::stack_multivalued_indexes(columns).collect();
assert_eq!(start_offsets, &[0, 0, 1, 1, 4, 7, 9, 13]);
}
}

70
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;
@@ -6,8 +11,11 @@ mod serialize;
use std::ops::Range;
pub use merge::merge_column_index;
pub(crate) use multivalued_index::SerializableMultivalueIndex;
pub use optional_index::{OptionalIndex, Set};
pub use serialize::{open_column_index, serialize_column_index, SerializableColumnIndex};
pub use serialize::{
open_column_index, serialize_column_index, SerializableColumnIndex, SerializableOptionalIndex,
};
use crate::column_index::multivalued_index::MultiValueIndex;
use crate::{Cardinality, DocId, RowId};
@@ -20,7 +28,7 @@ pub enum ColumnIndex {
Full,
Optional(OptionalIndex),
/// In addition, at index num_rows, an extra value is added
/// containing the overal number of values.
/// containing the overall number of values.
Multivalued(MultiValueIndex),
}
@@ -37,10 +45,10 @@ 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 {
@@ -117,24 +125,50 @@ 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);
ColumnIndex::Multivalued(multivalued_index) => match multivalued_index {
MultiValueIndex::MultiValueIndexV1(index) => {
let row_start = index.start_index_column.get_val(doc_id_range.start);
let row_end = index.start_index_column.get_val(doc_id_range.end);
row_start..row_end
}
MultiValueIndex::MultiValueIndexV2(index) => {
// In this case we will use the optional_index select the next values
// that are valid. There are different cases to consider:
// Not exists below means does not exist in the optional
// index, because it has no values.
// * doc_id_range may cover a range of docids which are non existent
// => rank
// will give us the next document outside the range with a value. They both
// get the same rank and therefore return a zero range
//
// * doc_id_range.start and doc_id_range.end may not exist, but docids in
// between may have values
// => rank will give us the next document outside the range with a value.
//
// * doc_id_range.start may be not existent but doc_id_range.end may exist
// * doc_id_range.start may exist but doc_id_range.end may not exist
// * doc_id_range.start and doc_id_range.end may exist
// => rank on doc_id_range.end will give use the next value, which matches
// how the `start_index_column` works, so we get the value start of the next
// docid which we use to create the exclusive range.
//
let rank_start = index.optional_index.rank(doc_id_range.start);
let row_start = index.start_index_column.get_val(rank_start);
let rank_end = index.optional_index.rank(doc_id_range.end);
let row_end = index.start_index_column.get_val(rank_end);
let row_start = multivalued_index.start_index_column.get_val(start_docid);
let row_end = multivalued_index.start_index_column.get_val(end_docid);
row_start..row_end
}
row_start..row_end
}
},
}
}

318
columnar/src/column_index/multivalued_index.rs
View File

@@ -3,64 +3,98 @@ use std::io::Write;
use std::ops::Range;
use std::sync::Arc;
use common::OwnedBytes;
use common::{CountingWriter, OwnedBytes};
use super::optional_index::{open_optional_index, serialize_optional_index};
use super::{OptionalIndex, SerializableOptionalIndex, Set};
use crate::column_values::{
load_u64_based_column_values, serialize_u64_based_column_values, CodecType, ColumnValues,
};
use crate::iterable::Iterable;
use crate::{DocId, RowId};
use crate::{DocId, RowId, Version};
pub struct SerializableMultivalueIndex<'a> {
pub doc_ids_with_values: SerializableOptionalIndex<'a>,
pub start_offsets: Box<dyn Iterable<u32> + 'a>,
}
pub fn serialize_multivalued_index(
multivalued_index: &dyn Iterable<RowId>,
multivalued_index: &SerializableMultivalueIndex,
output: &mut impl Write,
) -> io::Result<()> {
let SerializableMultivalueIndex {
doc_ids_with_values,
start_offsets,
} = multivalued_index;
let mut count_writer = CountingWriter::wrap(output);
let SerializableOptionalIndex {
non_null_row_ids,
num_rows,
} = doc_ids_with_values;
serialize_optional_index(&**non_null_row_ids, *num_rows, &mut count_writer)?;
let optional_len = count_writer.written_bytes() as u32;
let output = count_writer.finish();
serialize_u64_based_column_values(
multivalued_index,
&**start_offsets,
&[CodecType::Bitpacked, CodecType::Linear],
output,
)?;
output.write_all(&optional_len.to_le_bytes())?;
Ok(())
}
pub fn open_multivalued_index(bytes: OwnedBytes) -> io::Result<MultiValueIndex> {
let start_index_column: Arc<dyn ColumnValues<RowId>> = load_u64_based_column_values(bytes)?;
Ok(MultiValueIndex { start_index_column })
pub fn open_multivalued_index(
bytes: OwnedBytes,
format_version: Version,
) -> io::Result<MultiValueIndex> {
match format_version {
Version::V1 => {
let start_index_column: Arc<dyn ColumnValues<RowId>> =
load_u64_based_column_values(bytes)?;
Ok(MultiValueIndex::MultiValueIndexV1(MultiValueIndexV1 {
start_index_column,
}))
}
Version::V2 => {
let (body_bytes, optional_index_len) = bytes.rsplit(4);
let optional_index_len =
u32::from_le_bytes(optional_index_len.as_slice().try_into().unwrap());
let (optional_index_bytes, start_index_bytes) =
body_bytes.split(optional_index_len as usize);
let optional_index = open_optional_index(optional_index_bytes)?;
let start_index_column: Arc<dyn ColumnValues<RowId>> =
load_u64_based_column_values(start_index_bytes)?;
Ok(MultiValueIndex::MultiValueIndexV2(MultiValueIndexV2 {
optional_index,
start_index_column,
}))
}
}
}
#[derive(Clone)]
/// Index to resolve value range for given doc_id.
/// Starts at 0.
pub struct MultiValueIndex {
pub enum MultiValueIndex {
MultiValueIndexV1(MultiValueIndexV1),
MultiValueIndexV2(MultiValueIndexV2),
}
#[derive(Clone)]
/// Index to resolve value range for given doc_id.
/// Starts at 0.
pub struct MultiValueIndexV1 {
pub start_index_column: Arc<dyn crate::ColumnValues<RowId>>,
}
impl std::fmt::Debug for MultiValueIndex {
fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
f.debug_struct("MultiValuedIndex")
.field("num_rows", &self.start_index_column.num_vals())
.finish_non_exhaustive()
}
}
impl From<Arc<dyn ColumnValues<RowId>>> for MultiValueIndex {
fn from(start_index_column: Arc<dyn ColumnValues<RowId>>) -> Self {
MultiValueIndex { start_index_column }
}
}
impl MultiValueIndex {
pub fn for_test(start_offsets: &[RowId]) -> MultiValueIndex {
let mut buffer = Vec::new();
serialize_multivalued_index(&start_offsets, &mut buffer).unwrap();
let bytes = OwnedBytes::new(buffer);
open_multivalued_index(bytes).unwrap()
}
impl MultiValueIndexV1 {
/// Returns `[start, end)`, such that the values associated with
/// the given document are `start..end`.
#[inline]
pub(crate) fn range(&self, doc_id: DocId) -> Range<RowId> {
if doc_id >= self.num_docs() {
return 0..0;
}
let start = self.start_index_column.get_val(doc_id);
let end = self.start_index_column.get_val(doc_id + 1);
start..end
@@ -83,7 +117,6 @@ impl MultiValueIndex {
///
/// TODO: Instead of a linear scan we can employ a exponential search into binary search to
/// match a docid to its value position.
#[allow(clippy::bool_to_int_with_if)]
pub(crate) fn select_batch_in_place(&self, docid_start: DocId, ranks: &mut Vec<u32>) {
if ranks.is_empty() {
return;
@@ -111,11 +144,170 @@ impl MultiValueIndex {
}
}
#[derive(Clone)]
/// Index to resolve value range for given doc_id.
/// Starts at 0.
pub struct MultiValueIndexV2 {
pub optional_index: OptionalIndex,
pub start_index_column: Arc<dyn crate::ColumnValues<RowId>>,
}
impl std::fmt::Debug for MultiValueIndex {
fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
let index = match self {
MultiValueIndex::MultiValueIndexV1(idx) => &idx.start_index_column,
MultiValueIndex::MultiValueIndexV2(idx) => &idx.start_index_column,
};
f.debug_struct("MultiValuedIndex")
.field("num_rows", &index.num_vals())
.finish_non_exhaustive()
}
}
impl MultiValueIndex {
pub fn for_test(start_offsets: &[RowId]) -> MultiValueIndex {
assert!(!start_offsets.is_empty());
assert_eq!(start_offsets[0], 0);
let mut doc_with_values = Vec::new();
let mut compact_start_offsets: Vec<u32> = vec![0];
for doc in 0..start_offsets.len() - 1 {
if start_offsets[doc] < start_offsets[doc + 1] {
doc_with_values.push(doc as RowId);
compact_start_offsets.push(start_offsets[doc + 1]);
}
}
let serializable_multivalued_index = SerializableMultivalueIndex {
doc_ids_with_values: SerializableOptionalIndex {
non_null_row_ids: Box::new(&doc_with_values[..]),
num_rows: start_offsets.len() as u32 - 1,
},
start_offsets: Box::new(&compact_start_offsets[..]),
};
let mut buffer = Vec::new();
serialize_multivalued_index(&serializable_multivalued_index, &mut buffer).unwrap();
let bytes = OwnedBytes::new(buffer);
open_multivalued_index(bytes, Version::V2).unwrap()
}
pub fn get_start_index_column(&self) -> &Arc<dyn crate::ColumnValues<RowId>> {
match self {
MultiValueIndex::MultiValueIndexV1(idx) => &idx.start_index_column,
MultiValueIndex::MultiValueIndexV2(idx) => &idx.start_index_column,
}
}
/// Returns `[start, end)` values range, such that the values associated with
/// the given document are `start..end`.
#[inline]
pub(crate) fn range(&self, doc_id: DocId) -> Range<RowId> {
match self {
MultiValueIndex::MultiValueIndexV1(idx) => idx.range(doc_id),
MultiValueIndex::MultiValueIndexV2(idx) => idx.range(doc_id),
}
}
/// Returns the number of documents in the index.
#[inline]
pub fn num_docs(&self) -> u32 {
match self {
MultiValueIndex::MultiValueIndexV1(idx) => idx.start_index_column.num_vals() - 1,
MultiValueIndex::MultiValueIndexV2(idx) => idx.optional_index.num_docs(),
}
}
/// Converts a list of ranks (row ids of values) in a 1:n index to the corresponding list of
/// docids. Positions are converted inplace to docids.
///
/// Since there is no index for value pos -> docid, but docid -> value pos range, we scan the
/// index.
///
/// Correctness: positions needs to be sorted. idx_reader needs to contain monotonically
/// increasing positions.
///
/// TODO: Instead of a linear scan we can employ a exponential search into binary search to
/// match a docid to its value position.
pub(crate) fn select_batch_in_place(&self, docid_start: DocId, ranks: &mut Vec<u32>) {
match self {
MultiValueIndex::MultiValueIndexV1(idx) => {
idx.select_batch_in_place(docid_start, ranks)
}
MultiValueIndex::MultiValueIndexV2(idx) => {
idx.select_batch_in_place(docid_start, ranks)
}
}
}
}
impl MultiValueIndexV2 {
/// Returns `[start, end)`, such that the values associated with
/// the given document are `start..end`.
#[inline]
pub(crate) fn range(&self, doc_id: DocId) -> Range<RowId> {
let Some(rank) = self.optional_index.rank_if_exists(doc_id) else {
return 0..0;
};
let start = self.start_index_column.get_val(rank);
let end = self.start_index_column.get_val(rank + 1);
start..end
}
/// Returns the number of documents in the index.
#[inline]
pub fn num_docs(&self) -> u32 {
self.optional_index.num_docs()
}
/// Converts a list of ranks (row ids of values) in a 1:n index to the corresponding list of
/// docids. Positions are converted inplace to docids.
///
/// Since there is no index for value pos -> docid, but docid -> value pos range, we scan the
/// index.
///
/// Correctness: positions needs to be sorted. idx_reader needs to contain monotonically
/// increasing positions.
///
/// TODO: Instead of a linear scan we can employ a exponential search into binary search to
/// match a docid to its value position.
pub(crate) fn select_batch_in_place(&self, docid_start: DocId, ranks: &mut Vec<u32>) {
if ranks.is_empty() {
return;
}
let mut cur_pos_in_idx = self.optional_index.rank(docid_start);
let mut last_doc = None;
assert!(cur_pos_in_idx <= ranks[0]);
let mut write_doc_pos = 0;
for i in 0..ranks.len() {
let pos = ranks[i];
loop {
let end = self.start_index_column.get_val(cur_pos_in_idx + 1);
if end > pos {
ranks[write_doc_pos] = cur_pos_in_idx;
write_doc_pos += if last_doc == Some(cur_pos_in_idx) {
0
} else {
1
};
last_doc = Some(cur_pos_in_idx);
break;
}
cur_pos_in_idx += 1;
}
}
ranks.truncate(write_doc_pos);
for rank in ranks.iter_mut() {
*rank = self.optional_index.select(*rank);
}
}
}
#[cfg(test)]
mod tests {
use std::ops::Range;
use super::MultiValueIndex;
use crate::{ColumnarReader, DynamicColumn};
fn index_to_pos_helper(
index: &MultiValueIndex,
@@ -134,6 +326,7 @@ mod tests {
let positions = &[10u32, 11, 15, 20, 21, 22];
assert_eq!(index_to_pos_helper(&index, 0..5, positions), vec![1, 3, 4]);
assert_eq!(index_to_pos_helper(&index, 1..5, positions), vec![1, 3, 4]);
assert_eq!(index_to_pos_helper(&index, 0..5, &[9]), vec![0]);
assert_eq!(index_to_pos_helper(&index, 1..5, &[10]), vec![1]);
assert_eq!(index_to_pos_helper(&index, 1..5, &[11]), vec![1]);
@@ -141,4 +334,67 @@ mod tests {
assert_eq!(index_to_pos_helper(&index, 2..5, &[12, 14]), vec![2]);
assert_eq!(index_to_pos_helper(&index, 2..5, &[12, 14, 15]), vec![2, 3]);
}
#[test]
fn test_range_to_rowids() {
use crate::ColumnarWriter;
let mut columnar_writer = ColumnarWriter::default();
// This column gets coerced to u64
columnar_writer.record_numerical(1, "full", u64::MAX);
columnar_writer.record_numerical(1, "full", u64::MAX);
columnar_writer.record_numerical(5, "full", u64::MAX);
columnar_writer.record_numerical(5, "full", u64::MAX);
let mut wrt: Vec<u8> = Vec::new();
columnar_writer.serialize(7, &mut wrt).unwrap();
let reader = ColumnarReader::open(wrt).unwrap();
// Open the column as u64
let column = reader.read_columns("full").unwrap()[0]
.open()
.unwrap()
.coerce_numerical(crate::NumericalType::U64)
.unwrap();
let DynamicColumn::U64(column) = column else {
panic!();
};
let row_id_range = column.index.docid_range_to_rowids(1..2);
assert_eq!(row_id_range, 0..2);
let row_id_range = column.index.docid_range_to_rowids(0..2);
assert_eq!(row_id_range, 0..2);
let row_id_range = column.index.docid_range_to_rowids(0..4);
assert_eq!(row_id_range, 0..2);
let row_id_range = column.index.docid_range_to_rowids(3..4);
assert_eq!(row_id_range, 2..2);
let row_id_range = column.index.docid_range_to_rowids(1..6);
assert_eq!(row_id_range, 0..4);
let row_id_range = column.index.docid_range_to_rowids(3..6);
assert_eq!(row_id_range, 2..4);
let row_id_range = column.index.docid_range_to_rowids(0..6);
assert_eq!(row_id_range, 0..4);
let row_id_range = column.index.docid_range_to_rowids(0..6);
assert_eq!(row_id_range, 0..4);
let check = |range, expected| {
let full_range = 0..=u64::MAX;
let mut docids = Vec::new();
column.get_docids_for_value_range(full_range, range, &mut docids);
assert_eq!(docids, expected);
};
// check(0..1, vec![]);
// check(0..2, vec![1]);
check(1..2, vec![1]);
}
}

62
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,
@@ -82,17 +80,23 @@ impl BlockVariant {
/// index is the block index. For each block `byte_start` and `offset` is computed.
#[derive(Clone)]
pub struct OptionalIndex {
num_rows: RowId,
num_non_null_rows: RowId,
num_docs: RowId,
num_non_null_docs: RowId,
block_data: OwnedBytes,
block_metas: Arc<[BlockMeta]>,
}
impl Iterable<u32> for &OptionalIndex {
fn boxed_iter(&self) -> Box<dyn Iterator<Item = u32> + '_> {
Box::new(self.iter_docs())
}
}
impl std::fmt::Debug for OptionalIndex {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
f.debug_struct("OptionalIndex")
.field("num_rows", &self.num_rows)
.field("num_non_null_rows", &self.num_non_null_rows)
.field("num_docs", &self.num_docs)
.field("num_non_null_docs", &self.num_non_null_docs)
.finish_non_exhaustive()
}
}
@@ -109,8 +113,8 @@ 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,
}
}
@@ -119,7 +123,7 @@ enum BlockSelectCursor<'a> {
Sparse(<SparseBlock<'a> as Set<u16>>::SelectCursor<'a>),
}
impl<'a> BlockSelectCursor<'a> {
impl BlockSelectCursor<'_> {
fn select(&mut self, rank: u16) -> u16 {
match self {
BlockSelectCursor::Dense(dense_select_cursor) => dense_select_cursor.select(rank),
@@ -137,7 +141,7 @@ pub struct OptionalIndexSelectCursor<'a> {
num_null_rows_before_block: RowId,
}
impl<'a> OptionalIndexSelectCursor<'a> {
impl OptionalIndexSelectCursor<'_> {
fn search_and_load_block(&mut self, rank: RowId) {
if rank < self.current_block_end_rank {
// we are already in the right block
@@ -161,7 +165,7 @@ impl<'a> OptionalIndexSelectCursor<'a> {
}
}
impl<'a> SelectCursor<RowId> for OptionalIndexSelectCursor<'a> {
impl SelectCursor<RowId> for OptionalIndexSelectCursor<'_> {
fn select(&mut self, rank: RowId) -> RowId {
self.search_and_load_block(rank);
let index_in_block = (rank - self.num_null_rows_before_block) as u16;
@@ -170,7 +174,9 @@ impl<'a> SelectCursor<RowId> for OptionalIndexSelectCursor<'a> {
}
impl Set<RowId> for OptionalIndex {
type SelectCursor<'b> = OptionalIndexSelectCursor<'b> where Self: 'b;
type SelectCursor<'b>
= OptionalIndexSelectCursor<'b>
where Self: 'b;
// Check if value at position is not null.
#[inline]
fn contains(&self, row_id: RowId) -> bool {
@@ -185,14 +191,20 @@ 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,
} = row_addr_from_row_id(doc_id);
let block_meta = self.block_metas[block_id as usize];
let block = self.block(block_meta);
let block_offset_row_id = match block {
Block::Dense(dense_block) => dense_block.rank(in_block_row_id),
Block::Sparse(sparse_block) => sparse_block.rank(in_block_row_id),
@@ -200,13 +212,15 @@ 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),
@@ -257,17 +271,17 @@ impl OptionalIndex {
}
pub fn num_docs(&self) -> RowId {
self.num_rows
self.num_docs
}
pub fn num_non_nulls(&self) -> RowId {
self.num_non_null_rows
self.num_non_null_docs
}
pub fn iter_rows(&self) -> impl Iterator<Item = RowId> + '_ {
pub fn iter_docs(&self) -> impl Iterator<Item = RowId> + '_ {
// TODO optimize
let mut select_batch = self.select_cursor();
(0..self.num_non_null_rows).map(move |rank| select_batch.select(rank))
(0..self.num_non_null_docs).map(move |rank| select_batch.select(rank))
}
pub fn select_batch(&self, ranks: &mut [RowId]) {
let mut select_cursor = self.select_cursor();
@@ -491,7 +505,7 @@ 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.div_ceil(ELEMENTS_PER_BLOCK) as usize,
BlockMeta {
non_null_rows_before_block,
start_byte_offset,
@@ -505,15 +519,15 @@ pub fn open_optional_index(bytes: OwnedBytes) -> io::Result<OptionalIndex> {
let (mut bytes, num_non_empty_blocks_bytes) = bytes.rsplit(2);
let num_non_empty_block_bytes =
u16::from_le_bytes(num_non_empty_blocks_bytes.as_slice().try_into().unwrap());
let num_rows = VInt::deserialize_u64(&mut bytes)? as u32;
let num_docs = VInt::deserialize_u64(&mut bytes)? as u32;
let block_metas_num_bytes =
num_non_empty_block_bytes as usize * SERIALIZED_BLOCK_META_NUM_BYTES;
let (block_data, block_metas) = bytes.rsplit(block_metas_num_bytes);
let (block_metas, num_non_null_rows) =
deserialize_optional_index_block_metadatas(block_metas.as_slice(), num_rows);
let (block_metas, num_non_null_docs) =
deserialize_optional_index_block_metadatas(block_metas.as_slice(), num_docs);
let optional_index = OptionalIndex {
num_rows,
num_non_null_rows,
num_docs,
num_non_null_docs,
block_data,
block_metas: block_metas.into(),
};

8
columnar/src/column_index/optional_index/set.rs
View File

@@ -28,10 +28,11 @@ pub trait Set<T> {
/// Returns true if the elements is contained in the Set
fn contains(&self, el: T) -> bool;
/// Returns the number of rows in the set that are < `el`
/// Returns the element's rank (its position in the set).
/// If the set does not contain the element, it will return the next existing elements rank.
fn rank(&self, el: T) -> T;
/// If the set contains `el` returns the element rank.
/// If the set contains `el`, returns the element's rank (its position in the set).
/// If the set does not contain the element, it returns `None`.
fn rank_if_exists(&self, el: T) -> Option<T>;
@@ -39,7 +40,8 @@ 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.

10
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;
@@ -24,7 +23,6 @@ fn set_bit_at(input: &mut u64, n: u16) {
///
/// When translating a dense index to the original index, we can use the offset to find the correct
/// block. Direct computation is not possible, but we can employ a linear or binary search.
const ELEMENTS_PER_MINI_BLOCK: u16 = 64;
const MINI_BLOCK_BITVEC_NUM_BYTES: usize = 8;
const MINI_BLOCK_OFFSET_NUM_BYTES: usize = 2;
@@ -110,7 +108,7 @@ pub struct DenseBlockSelectCursor<'a> {
dense_block: DenseBlock<'a>,
}
impl<'a> SelectCursor<u16> for DenseBlockSelectCursor<'a> {
impl SelectCursor<u16> for DenseBlockSelectCursor<'_> {
#[inline]
fn select(&mut self, rank: u16) -> u16 {
self.block_id = self
@@ -124,7 +122,9 @@ impl<'a> SelectCursor<u16> for DenseBlockSelectCursor<'a> {
}
impl<'a> Set<u16> for DenseBlock<'a> {
type SelectCursor<'b> = DenseBlockSelectCursor<'a> where Self: 'b;
type SelectCursor<'b>
= DenseBlockSelectCursor<'a>
where Self: 'b;
#[inline(always)]
fn contains(&self, el: u16) -> bool {
@@ -174,7 +174,7 @@ impl<'a> Set<u16> for DenseBlock<'a> {
}
}
impl<'a> DenseBlock<'a> {
impl DenseBlock<'_> {
#[inline]
fn mini_block(&self, mini_block_id: u16) -> DenseMiniBlock {
let data_start_pos = mini_block_id as usize * MINI_BLOCK_NUM_BYTES;

10
columnar/src/column_index/optional_index/set_block/sparse.rs
View File

@@ -31,8 +31,10 @@ impl<'a> SelectCursor<u16> for SparseBlock<'a> {
}
}
impl<'a> Set<u16> for SparseBlock<'a> {
type SelectCursor<'b> = Self where Self: 'b;
impl Set<u16> for SparseBlock<'_> {
type SelectCursor<'b>
= Self
where Self: 'b;
#[inline(always)]
fn contains(&self, el: u16) -> bool {
@@ -67,7 +69,7 @@ fn get_u16(data: &[u8], byte_position: usize) -> u16 {
u16::from_le_bytes(bytes)
}
impl<'a> SparseBlock<'a> {
impl SparseBlock<'_> {
#[inline(always)]
fn value_at_idx(&self, data: &[u8], idx: u16) -> u16 {
let start_offset: usize = idx as usize * 2;
@@ -80,7 +82,7 @@ impl<'a> SparseBlock<'a> {
}
#[inline]
#[allow(clippy::comparison_chain)]
#[expect(clippy::comparison_chain)]
// Looks for the element in the block. Returns the positions if found.
fn binary_search(&self, target: u16) -> Result<u16, u16> {
let data = &self.0;

42
columnar/src/column_index/optional_index/set_block/tests.rs
View File

@@ -22,8 +22,8 @@ fn test_set_helper<C: SetCodec<Item = u16>>(vals: &[u16]) -> usize {
vals.iter().cloned().take_while(|v| *v < val).count() as u16
);
}
for rank in 0..vals.len() {
assert_eq!(tested_set.select(rank as u16), vals[rank]);
for (rank, val) in vals.iter().enumerate() {
assert_eq!(tested_set.select(rank as u16), *val);
}
buffer.len()
}
@@ -107,3 +107,41 @@ fn test_simple_translate_codec_idx_to_original_idx_dense() {
assert_eq!(i, select_cursor.select(i));
}
}
#[test]
fn test_simple_translate_idx_to_value_idx_dense() {
let mut buffer = Vec::new();
DenseBlockCodec::serialize([1, 10].iter().copied(), &mut buffer).unwrap();
let tested_set = DenseBlockCodec::open(buffer.as_slice());
assert!(tested_set.contains(1));
assert!(!tested_set.contains(2));
assert_eq!(tested_set.rank(0), 0);
assert_eq!(tested_set.rank(1), 0);
for rank in 2..10 {
// ranks that don't exist select the next highest one
assert_eq!(tested_set.rank_if_exists(rank), None);
assert_eq!(tested_set.rank(rank), 1);
}
assert_eq!(tested_set.rank(10), 1);
}
#[test]
fn test_simple_translate_idx_to_value_idx_sparse() {
let mut buffer = Vec::new();
SparseBlockCodec::serialize([1, 10].iter().copied(), &mut buffer).unwrap();
let tested_set = SparseBlockCodec::open(buffer.as_slice());
assert!(tested_set.contains(1));
assert!(!tested_set.contains(2));
assert_eq!(tested_set.rank(0), 0);
assert_eq!(tested_set.select(tested_set.rank(0)), 1);
assert_eq!(tested_set.rank(1), 0);
assert_eq!(tested_set.select(tested_set.rank(1)), 1);
for rank in 2..10 {
// ranks that don't exist select the next highest one
assert_eq!(tested_set.rank_if_exists(rank), None);
assert_eq!(tested_set.rank(rank), 1);
assert_eq!(tested_set.select(tested_set.rank(rank)), 10);
}
assert_eq!(tested_set.rank(10), 1);
assert_eq!(tested_set.select(tested_set.rank(10)), 10);
}

47
columnar/src/column_index/optional_index/tests.rs
View File

@@ -1,8 +1,29 @@
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, &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 +56,7 @@ 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();
@@ -89,8 +110,8 @@ fn test_null_index(data: &[bool]) {
.map(|(pos, _val)| pos as u32)
.collect();
let mut select_iter = null_index.select_cursor();
for i in 0..orig_idx_with_value.len() {
assert_eq!(select_iter.select(i as u32), orig_idx_with_value[i]);
for (i, expected) in orig_idx_with_value.iter().enumerate() {
assert_eq!(select_iter.select(i as u32), *expected);
}
let step_size = (orig_idx_with_value.len() / 100).max(1);
@@ -143,7 +164,7 @@ fn test_optional_index_large() {
fn test_optional_index_iter_aux(row_ids: &[RowId], num_rows: RowId) {
let optional_index = OptionalIndex::for_test(num_rows, row_ids);
assert_eq!(optional_index.num_docs(), num_rows);
assert!(optional_index.iter_rows().eq(row_ids.iter().copied()));
assert!(optional_index.iter_docs().eq(row_ids.iter().copied()));
}
#[test]
@@ -171,7 +192,7 @@ 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 +206,8 @@ 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 +236,12 @@ 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 +263,7 @@ 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)

47
columnar/src/column_index/serialize.rs
View File

@@ -3,33 +3,45 @@ use std::io::Write;
use common::{CountingWriter, OwnedBytes};
use super::multivalued_index::SerializableMultivalueIndex;
use super::OptionalIndex;
use crate::column_index::multivalued_index::serialize_multivalued_index;
use crate::column_index::optional_index::serialize_optional_index;
use crate::column_index::ColumnIndex;
use crate::iterable::Iterable;
use crate::{Cardinality, RowId};
use crate::{Cardinality, RowId, Version};
pub struct SerializableOptionalIndex<'a> {
pub non_null_row_ids: Box<dyn Iterable<RowId> + 'a>,
pub num_rows: RowId,
}
impl<'a> From<&'a OptionalIndex> for SerializableOptionalIndex<'a> {
fn from(optional_index: &'a OptionalIndex) -> Self {
SerializableOptionalIndex {
non_null_row_ids: Box::new(optional_index),
num_rows: optional_index.num_docs(),
}
}
}
pub enum SerializableColumnIndex<'a> {
Full,
Optional {
non_null_row_ids: Box<dyn Iterable<RowId> + 'a>,
num_rows: RowId,
},
// TODO remove the Arc<dyn> apart from serialization this is not
// dynamic at all.
Multivalued(Box<dyn Iterable<RowId> + 'a>),
Optional(SerializableOptionalIndex<'a>),
Multivalued(SerializableMultivalueIndex<'a>),
}
impl<'a> SerializableColumnIndex<'a> {
impl SerializableColumnIndex<'_> {
pub fn get_cardinality(&self) -> Cardinality {
match self {
SerializableColumnIndex::Full => Cardinality::Full,
SerializableColumnIndex::Optional { .. } => Cardinality::Optional,
SerializableColumnIndex::Optional(_) => Cardinality::Optional,
SerializableColumnIndex::Multivalued(_) => Cardinality::Multivalued,
}
}
}
/// Serialize a column index.
pub fn serialize_column_index(
column_index: SerializableColumnIndex,
output: &mut impl Write,
@@ -39,19 +51,23 @@ pub fn serialize_column_index(
output.write_all(&[cardinality])?;
match column_index {
SerializableColumnIndex::Full => {}
SerializableColumnIndex::Optional {
SerializableColumnIndex::Optional(SerializableOptionalIndex {
non_null_row_ids,
num_rows,
} => serialize_optional_index(non_null_row_ids.as_ref(), num_rows, &mut output)?,
}) => serialize_optional_index(non_null_row_ids.as_ref(), num_rows, &mut output)?,
SerializableColumnIndex::Multivalued(multivalued_index) => {
serialize_multivalued_index(&*multivalued_index, &mut output)?
serialize_multivalued_index(&multivalued_index, &mut output)?
}
}
let column_index_num_bytes = output.written_bytes() as u32;
Ok(column_index_num_bytes)
}
pub fn open_column_index(mut bytes: OwnedBytes) -> io::Result<ColumnIndex> {
/// Open a serialized column index.
pub fn open_column_index(
mut bytes: OwnedBytes,
format_version: Version,
) -> io::Result<ColumnIndex> {
if bytes.is_empty() {
return Err(io::Error::new(
io::ErrorKind::UnexpectedEof,
@@ -68,7 +84,8 @@ pub fn open_column_index(mut bytes: OwnedBytes) -> io::Result<ColumnIndex> {
Ok(ColumnIndex::Optional(optional_index))
}
Cardinality::Multivalued => {
let multivalue_index = super::multivalued_index::open_multivalued_index(bytes)?;
let multivalue_index =
super::multivalued_index::open_multivalued_index(bytes, format_version)?;
Ok(ColumnIndex::Multivalued(multivalue_index))
}
}

6
columnar/src/column_values/bench.rs
View File

@@ -34,6 +34,7 @@ fn compute_stats(vals: impl Iterator<Item = u64>) -> ColumnStats {
fn value_iter() -> impl Iterator<Item = u64> {
0..20_000
}
fn get_reader_for_bench<Codec: ColumnCodec>(data: &[u64]) -> Codec::ColumnValues {
let mut bytes = Vec::new();
let stats = compute_stats(data.iter().cloned());
@@ -41,10 +42,13 @@ fn get_reader_for_bench<Codec: ColumnCodec>(data: &[u64]) -> Codec::ColumnValues
for val in data {
codec_serializer.collect(*val);
}
codec_serializer.serialize(&stats, Box::new(data.iter().copied()).as_mut(), &mut bytes);
codec_serializer
.serialize(&stats, Box::new(data.iter().copied()).as_mut(), &mut bytes)
.unwrap();
Codec::load(OwnedBytes::new(bytes)).unwrap()
}
fn bench_get<Codec: ColumnCodec>(b: &mut Bencher, data: &[u64]) {
let col = get_reader_for_bench::<Codec>(data);
b.iter(|| {

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<T: Copy + PartialOrd + Debug + 'static> Iterable<T> for MergedColumnValues<'_, 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))
}
}

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

@@ -184,7 +184,7 @@ impl CompactSpaceBuilder {
let mut covered_space = Vec::with_capacity(self.blanks.len());
// begining of the blanks
// beginning of the blanks
if let Some(first_blank_start) = self.blanks.first().map(RangeInclusive::start) {
if *first_blank_start != 0 {
covered_space.push(0..=first_blank_start - 1);

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)
}

23
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::{
@@ -108,8 +110,25 @@ 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 {
pub(crate) mod tests {
use super::*;
use crate::column_values::u64_based::{
serialize_and_load_u64_based_column_values, serialize_u64_based_column_values,

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;

9
columnar/src/column_values/u64_based/blockwise_linear.rs
View File

@@ -39,7 +39,7 @@ impl BinarySerializable for Block {
}
fn compute_num_blocks(num_vals: u32) -> u32 {
(num_vals + BLOCK_SIZE - 1) / BLOCK_SIZE
num_vals.div_ceil(BLOCK_SIZE)
}
pub struct BlockwiseLinearEstimator {
@@ -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());

9
columnar/src/column_values/u64_based/line.rs
View File

@@ -122,12 +122,11 @@ impl Line {
line
}
/// Returns a line that attemps to approximate a function
/// Returns a line that attempts to approximate a function
/// f: i in 0..[ys.num_vals()) -> ys[i].
///
/// - The approximation is always lower than the actual value.
/// Or more rigorously, formally `f(i).wrapping_sub(ys[i])` is small
/// for any i in [0..ys.len()).
/// - The approximation is always lower than the actual value. Or more rigorously, formally
/// `f(i).wrapping_sub(ys[i])` is small for any i in [0..ys.len()).
/// - It computes without panicking for any value of it.
///
/// This function is only invariable by translation if all of the
@@ -184,7 +183,7 @@ 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
}
}

25
columnar/src/columnar/format_version.rs
View File

@@ -1,3 +1,6 @@
use core::fmt;
use std::fmt::{Display, Formatter};
use crate::InvalidData;
pub const VERSION_FOOTER_NUM_BYTES: usize = MAGIC_BYTES.len() + std::mem::size_of::<u32>();
@@ -8,7 +11,7 @@ const MAGIC_BYTES: [u8; 4] = [2, 113, 119, 66];
pub fn footer() -> [u8; VERSION_FOOTER_NUM_BYTES] {
let mut footer_bytes = [0u8; VERSION_FOOTER_NUM_BYTES];
footer_bytes[0..4].copy_from_slice(&Version::V1.to_bytes());
footer_bytes[0..4].copy_from_slice(&CURRENT_VERSION.to_bytes());
footer_bytes[4..8].copy_from_slice(&MAGIC_BYTES[..]);
footer_bytes
}
@@ -20,10 +23,22 @@ pub fn parse_footer(footer_bytes: [u8; VERSION_FOOTER_NUM_BYTES]) -> Result<Vers
Version::try_from_bytes(footer_bytes[0..4].try_into().unwrap())
}
pub const CURRENT_VERSION: Version = Version::V2;
#[derive(Debug, Copy, Clone, Eq, PartialEq)]
#[repr(u32)]
pub enum Version {
V1 = 1u32,
V2 = 2u32,
}
impl Display for Version {
fn fmt(&self, f: &mut Formatter) -> fmt::Result {
match self {
Version::V1 => write!(f, "v1"),
Version::V2 => write!(f, "v2"),
}
}
}
impl Version {
@@ -35,6 +50,7 @@ impl Version {
let code = u32::from_le_bytes(bytes);
match code {
1u32 => Ok(Version::V1),
2u32 => Ok(Version::V2),
_ => Err(InvalidData),
}
}
@@ -47,9 +63,9 @@ mod tests {
use super::*;
#[test]
fn test_footer_dserialization() {
fn test_footer_deserialization() {
let parsed_version: Version = parse_footer(footer()).unwrap();
assert_eq!(Version::V1, parsed_version);
assert_eq!(Version::V2, parsed_version);
}
#[test]
@@ -63,11 +79,10 @@ mod tests {
for &i in &version_to_tests {
let version_res = Version::try_from_bytes(i.to_le_bytes());
if let Ok(version) = version_res {
assert_eq!(version, Version::V1);
assert_eq!(version.to_bytes(), i.to_le_bytes());
valid_versions.insert(i);
}
}
assert_eq!(valid_versions.len(), 1);
assert_eq!(valid_versions.len(), 2);
}
}

18
columnar/src/columnar/merge/merge_dict_column.rs
View File

@@ -3,7 +3,7 @@ use std::io::{self, Write};
use common::{BitSet, CountingWriter, ReadOnlyBitSet};
use sstable::{SSTable, Streamer, TermOrdinal, VoidSSTable};
use super::term_merger::TermMerger;
use super::term_merger::{TermMerger, TermsWithSegmentOrd};
use crate::column::serialize_column_mappable_to_u64;
use crate::column_index::SerializableColumnIndex;
use crate::iterable::Iterable;
@@ -39,7 +39,7 @@ struct RemappedTermOrdinalsValues<'a> {
merge_row_order: &'a MergeRowOrder,
}
impl<'a> Iterable for RemappedTermOrdinalsValues<'a> {
impl Iterable for RemappedTermOrdinalsValues<'_> {
fn boxed_iter(&self) -> Box<dyn Iterator<Item = u64> + '_> {
match self.merge_row_order {
MergeRowOrder::Stack(_) => self.boxed_iter_stacked(),
@@ -50,7 +50,7 @@ impl<'a> Iterable for RemappedTermOrdinalsValues<'a> {
}
}
impl<'a> RemappedTermOrdinalsValues<'a> {
impl RemappedTermOrdinalsValues<'_> {
fn boxed_iter_stacked(&self) -> Box<dyn Iterator<Item = u64> + '_> {
let iter = self
.bytes_columns
@@ -126,14 +126,17 @@ fn serialize_merged_dict(
let mut term_ord_mapping = TermOrdinalMapping::default();
let mut field_term_streams = Vec::new();
for column_opt in bytes_columns.iter() {
for (segment_ord, column_opt) in bytes_columns.iter().enumerate() {
if let Some(column) = column_opt {
term_ord_mapping.add_segment(column.dictionary.num_terms());
let terms: Streamer<VoidSSTable> = column.dictionary.stream()?;
field_term_streams.push(terms);
field_term_streams.push(TermsWithSegmentOrd { terms, segment_ord });
} else {
term_ord_mapping.add_segment(0);
field_term_streams.push(Streamer::empty());
field_term_streams.push(TermsWithSegmentOrd {
terms: Streamer::empty(),
segment_ord,
});
}
}
@@ -191,6 +194,7 @@ fn serialize_merged_dict(
#[derive(Default, Debug)]
struct TermOrdinalMapping {
/// Contains the new term ordinals for each segment.
per_segment_new_term_ordinals: Vec<Vec<TermOrdinal>>,
}
@@ -205,6 +209,6 @@ impl TermOrdinalMapping {
}
fn get_segment(&self, segment_ord: u32) -> &[TermOrdinal] {
&(self.per_segment_new_term_ordinals[segment_ord as usize])[..]
&self.per_segment_new_term_ordinals[segment_ord as usize]
}
}

6
columnar/src/columnar/merge/merge_mapping.rs
View File

@@ -26,7 +26,7 @@ impl StackMergeOrder {
let mut cumulated_row_ids: Vec<RowId> = Vec::with_capacity(columnars.len());
let mut cumulated_row_id = 0;
for columnar in columnars {
cumulated_row_id += columnar.num_rows();
cumulated_row_id += columnar.num_docs();
cumulated_row_ids.push(cumulated_row_id);
}
StackMergeOrder { cumulated_row_ids }
@@ -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),

244
columnar/src/columnar/merge/mod.rs
View File

@@ -2,12 +2,11 @@ mod merge_dict_column;
mod merge_mapping;
mod term_merger;
use std::collections::{BTreeMap, HashMap, HashSet};
use std::collections::{BTreeMap, HashSet};
use std::io;
use std::net::Ipv6Addr;
use std::sync::Arc;
use itertools::Itertools;
pub use merge_mapping::{MergeRowOrder, ShuffleMergeOrder, StackMergeOrder};
use super::writer::ColumnarSerializer;
@@ -18,24 +17,27 @@ 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.
/// After merge, all columns belonging to the same category are coerced to
/// the same column type.
///
/// In practise, today, only Numerical colummns are coerced into one type today.
/// In practise, today, only Numerical columns 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 {
@@ -61,11 +63,10 @@ impl From<ColumnType> for ColumnTypeCategory {
/// `require_columns` makes it possible to ensure that some columns will be present in the
/// resulting columnar. When a required column is a numerical column type, one of two things can
/// happen:
/// - If the required column type is compatible with all of the input columnar, the resulsting
/// merged
/// columnar will simply coerce the input column and use the required column type.
/// - If the required column type is incompatible with one of the input columnar, the merged
/// will fail with an InvalidData error.
/// - If the required column type is compatible with all of the input columnar, the resulting merged
/// columnar will simply coerce the input column and use the required column type.
/// - If the required column type is incompatible with one of the input columnar, the merged will
/// fail with an InvalidData error.
///
/// `merge_row_order` makes it possible to remove or reorder row in the resulting
/// `Columnar` table.
@@ -79,24 +80,38 @@ pub fn merge_columnar(
output: &mut impl io::Write,
) -> io::Result<()> {
let mut serializer = ColumnarSerializer::new(output);
let num_rows_per_columnar = columnar_readers
let num_docs_per_columnar = columnar_readers
.iter()
.map(|reader| reader.num_rows())
.map(|reader| reader.num_docs())
.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 {
let columns_to_merge = group_columns_for_merge(columnar_readers, required_columns)?;
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_after_merge = grouped_columns.column_type_after_merge();
let mut columns = grouped_columns.columns;
// Make sure the number of columns is the same as the number of columnar readers.
// Or num_docs_per_columnar would be incorrect.
assert_eq!(columns.len(), columnar_readers.len());
coerce_columns(column_type_after_merge, &mut columns)?;
let mut column_serializer =
serializer.start_serialize_column(column_name.as_bytes(), column_type);
serializer.start_serialize_column(column_name.as_bytes(), column_type_after_merge);
merge_column(
column_type,
&num_rows_per_columnar,
column_type_after_merge,
&num_docs_per_columnar,
columns,
&merge_row_order,
&mut column_serializer,
)?;
column_serializer.finalize()?;
}
serializer.finalize(merge_row_order.num_rows())?;
Ok(())
}
@@ -115,7 +130,7 @@ fn dynamic_column_to_u64_monotonic(dynamic_column: DynamicColumn) -> Option<Colu
fn merge_column(
column_type: ColumnType,
num_docs_per_column: &[u32],
columns: Vec<Option<DynamicColumn>>,
columns_to_merge: Vec<Option<DynamicColumn>>,
merge_row_order: &MergeRowOrder,
wrt: &mut impl io::Write,
) -> io::Result<()> {
@@ -125,10 +140,10 @@ fn merge_column(
| ColumnType::F64
| ColumnType::DateTime
| ColumnType::Bool => {
let mut column_indexes: Vec<ColumnIndex> = Vec::with_capacity(columns.len());
let mut column_indexes: Vec<ColumnIndex> = Vec::with_capacity(columns_to_merge.len());
let mut column_values: Vec<Option<Arc<dyn ColumnValues>>> =
Vec::with_capacity(columns.len());
for (i, dynamic_column_opt) in columns.into_iter().enumerate() {
Vec::with_capacity(columns_to_merge.len());
for (i, dynamic_column_opt) in columns_to_merge.into_iter().enumerate() {
if let Some(Column { index: idx, values }) =
dynamic_column_opt.and_then(dynamic_column_to_u64_monotonic)
{
@@ -151,10 +166,10 @@ fn merge_column(
serialize_column_mappable_to_u64(merged_column_index, &merge_column_values, wrt)?;
}
ColumnType::IpAddr => {
let mut column_indexes: Vec<ColumnIndex> = Vec::with_capacity(columns.len());
let mut column_indexes: Vec<ColumnIndex> = Vec::with_capacity(columns_to_merge.len());
let mut column_values: Vec<Option<Arc<dyn ColumnValues<Ipv6Addr>>>> =
Vec::with_capacity(columns.len());
for (i, dynamic_column_opt) in columns.into_iter().enumerate() {
Vec::with_capacity(columns_to_merge.len());
for (i, dynamic_column_opt) in columns_to_merge.into_iter().enumerate() {
if let Some(DynamicColumn::IpAddr(Column { index: idx, values })) =
dynamic_column_opt
{
@@ -179,9 +194,10 @@ fn merge_column(
serialize_column_mappable_to_u128(merged_column_index, &merge_column_values, wrt)?;
}
ColumnType::Bytes | ColumnType::Str => {
let mut column_indexes: Vec<ColumnIndex> = Vec::with_capacity(columns.len());
let mut bytes_columns: Vec<Option<BytesColumn>> = Vec::with_capacity(columns.len());
for (i, dynamic_column_opt) in columns.into_iter().enumerate() {
let mut column_indexes: Vec<ColumnIndex> = Vec::with_capacity(columns_to_merge.len());
let mut bytes_columns: Vec<Option<BytesColumn>> =
Vec::with_capacity(columns_to_merge.len());
for (i, dynamic_column_opt) in columns_to_merge.into_iter().enumerate() {
match dynamic_column_opt {
Some(DynamicColumn::Str(str_column)) => {
column_indexes.push(str_column.term_ord_column.index.clone());
@@ -210,20 +226,80 @@ 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 {
/// 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.
///
/// This method does not check if the column types can actually be coerced to
/// this type.
fn column_type_after_merge(&self) -> ColumnType {
if let Some(required_type) = self.required_column_type {
return required_type;
}
let column_type: HashSet<ColumnType> = self
.columns
.iter()
.flatten()
.map(|column| column.column_type())
.collect();
if column_type.len() == 1 {
return column_type.into_iter().next().unwrap();
}
// At the moment, only the numerical column type category has more than one possible
// column type.
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],
column_category,
}
}
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: DynamicColumn) {
fn set_column(&mut self, columnar_id: usize, column: DynamicColumnHandle) {
self.columns[columnar_id] = Some(column);
}
@@ -245,29 +321,6 @@ impl GroupedColumns {
self.required_column_type = Some(required_type);
Ok(())
}
/// Returns the column type after merge.
///
/// This method does not check if the column types can actually be coerced to
/// this type.
fn column_type_after_merge(&self) -> ColumnType {
if let Some(required_type) = self.required_column_type {
return required_type;
}
let column_type: HashSet<ColumnType> = self
.columns
.iter()
.flatten()
.map(|column| column.column_type())
.collect();
if column_type.len() == 1 {
return column_type.into_iter().next().unwrap();
}
// At the moment, only the numerical categorical column type has more than one possible
// column type.
assert_eq!(self.column_category, ColumnTypeCategory::Numerical);
merged_numerical_columns_type(self.columns.iter().flatten()).into()
}
}
/// Returns the type of the merged numerical column.
@@ -293,7 +346,7 @@ 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.
@@ -305,13 +358,13 @@ 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,
ColumnIndex::Full => alive_bitset.len() == 0,
ColumnIndex::Optional(optional_index) => {
for doc in optional_index.iter_rows() {
for doc in optional_index.iter_docs() {
if alive_bitset.contains(doc) {
return false;
}
@@ -319,20 +372,8 @@ fn is_empty_after_merge(
true
}
ColumnIndex::Multivalued(multivalued_index) => {
for (doc_id, (start_index, end_index)) in multivalued_index
.start_index_column
.iter()
.tuple_windows()
.enumerate()
{
let doc_id = doc_id as u32;
if start_index == end_index {
// There are no values in this document
continue;
}
// The document contains values and is present in the alive bitset.
// The column is therefore not empty.
if alive_bitset.contains(doc_id) {
for alive_docid in alive_bitset.iter() {
if !multivalued_index.range(alive_docid).is_empty() {
return false;
}
}
@@ -348,56 +389,33 @@ 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)],
) -> 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(

53
columnar/src/columnar/merge/term_merger.rs
View File

@@ -5,28 +5,29 @@ use sstable::TermOrdinal;
use crate::Streamer;
pub struct HeapItem<'a> {
pub streamer: Streamer<'a>,
/// The terms of a column with the ordinal of the segment.
pub struct TermsWithSegmentOrd<'a> {
pub terms: Streamer<'a>,
pub segment_ord: usize,
}
impl<'a> PartialEq for HeapItem<'a> {
impl PartialEq for TermsWithSegmentOrd<'_> {
fn eq(&self, other: &Self) -> bool {
self.segment_ord == other.segment_ord
}
}
impl<'a> Eq for HeapItem<'a> {}
impl Eq for TermsWithSegmentOrd<'_> {}
impl<'a> PartialOrd for HeapItem<'a> {
fn partial_cmp(&self, other: &HeapItem<'a>) -> Option<Ordering> {
impl<'a> PartialOrd for TermsWithSegmentOrd<'a> {
fn partial_cmp(&self, other: &TermsWithSegmentOrd<'a>) -> Option<Ordering> {
Some(self.cmp(other))
}
}
impl<'a> Ord for HeapItem<'a> {
fn cmp(&self, other: &HeapItem<'a>) -> Ordering {
(&other.streamer.key(), &other.segment_ord).cmp(&(&self.streamer.key(), &self.segment_ord))
impl<'a> Ord for TermsWithSegmentOrd<'a> {
fn cmp(&self, other: &TermsWithSegmentOrd<'a>) -> Ordering {
(&other.terms.key(), &other.segment_ord).cmp(&(&self.terms.key(), &self.segment_ord))
}
}
@@ -35,42 +36,34 @@ impl<'a> Ord for HeapItem<'a> {
///
/// The item yield is actually a pair with
/// - the term
/// - a slice with the ordinal of the segments containing
/// the terms.
/// - a slice with the ordinal of the segments containing the terms.
pub struct TermMerger<'a> {
heap: BinaryHeap<HeapItem<'a>>,
current_streamers: Vec<HeapItem<'a>>,
heap: BinaryHeap<TermsWithSegmentOrd<'a>>,
term_streams_with_segment: Vec<TermsWithSegmentOrd<'a>>,
}
impl<'a> TermMerger<'a> {
/// Stream of merged term dictionary
pub fn new(streams: Vec<Streamer<'a>>) -> TermMerger<'a> {
pub fn new(term_streams_with_segment: Vec<TermsWithSegmentOrd<'a>>) -> TermMerger<'a> {
TermMerger {
heap: BinaryHeap::new(),
current_streamers: streams
.into_iter()
.enumerate()
.map(|(ord, streamer)| HeapItem {
streamer,
segment_ord: ord,
})
.collect(),
term_streams_with_segment,
}
}
pub(crate) fn matching_segments<'b: 'a>(
&'b self,
) -> impl 'b + Iterator<Item = (usize, TermOrdinal)> {
self.current_streamers
self.term_streams_with_segment
.iter()
.map(|heap_item| (heap_item.segment_ord, heap_item.streamer.term_ord()))
.map(|heap_item| (heap_item.segment_ord, heap_item.terms.term_ord()))
}
fn advance_segments(&mut self) {
let streamers = &mut self.current_streamers;
let streamers = &mut self.term_streams_with_segment;
let heap = &mut self.heap;
for mut heap_item in streamers.drain(..) {
if heap_item.streamer.advance() {
if heap_item.terms.advance() {
heap.push(heap_item);
}
}
@@ -82,13 +75,13 @@ impl<'a> TermMerger<'a> {
pub fn advance(&mut self) -> bool {
self.advance_segments();
if let Some(head) = self.heap.pop() {
self.current_streamers.push(head);
self.term_streams_with_segment.push(head);
while let Some(next_streamer) = self.heap.peek() {
if self.current_streamers[0].streamer.key() != next_streamer.streamer.key() {
if self.term_streams_with_segment[0].terms.key() != next_streamer.terms.key() {
break;
}
let next_heap_it = self.heap.pop().unwrap(); // safe : we peeked beforehand
self.current_streamers.push(next_heap_it);
self.term_streams_with_segment.push(next_heap_it);
}
true
} else {
@@ -102,6 +95,6 @@ impl<'a> TermMerger<'a> {
/// if and only if advance() has been called before
/// and "true" was returned.
pub fn key(&self) -> &[u8] {
self.current_streamers[0].streamer.key()
self.term_streams_with_segment[0].terms.key()
}
}

292
columnar/src/columnar/merge/tests.rs
View File

@@ -1,7 +1,10 @@
use itertools::Itertools;
use proptest::collection::vec;
use proptest::prelude::*;
use super::*;
use crate::{Cardinality, ColumnarWriter, HasAssociatedColumnType, RowId};
use crate::columnar::{merge_columnar, ColumnarReader, MergeRowOrder, StackMergeOrder};
use crate::{Cardinality, ColumnarWriter, DynamicColumn, HasAssociatedColumnType, RowId};
fn make_columnar<T: Into<NumericalValue> + HasAssociatedColumnType + Copy>(
column_name: &str,
@@ -14,7 +17,7 @@ fn make_columnar<T: Into<NumericalValue> + HasAssociatedColumnType + Copy>(
}
let mut buffer: Vec<u8> = Vec::new();
dataframe_writer
.serialize(vals.len() as RowId, None, &mut buffer)
.serialize(vals.len() as RowId, &mut buffer)
.unwrap();
ColumnarReader::open(buffer).unwrap()
}
@@ -26,23 +29,10 @@ fn test_column_coercion_to_u64() {
// u64 type
let columnar2 = make_columnar("numbers", &[u64::MAX]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(columnars, &[], &merge_order).unwrap();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(columnars, &[]).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]
@@ -50,41 +40,34 @@ fn test_column_coercion_to_i64() {
let columnar1 = make_columnar("numbers", &[-1i64]);
let columnar2 = make_columnar("numbers", &[2u64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(columnars, &[], &merge_order).unwrap();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(columnars, &[]).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() {
let columnar1 = make_columnar("numbers", &[1i64]);
let columnar2 = make_columnar("numbers", &[2u64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(
&[&columnar1, &columnar2],
&[("numbers".to_string(), ColumnType::U64)],
&merge_order,
)
.unwrap();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(columnars, &[("numbers".to_string(), ColumnType::U64)]).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]
@@ -92,18 +75,14 @@ fn test_group_columns_required_column_with_no_existing_columns() {
let columnar1 = make_columnar("numbers", &[2u64]);
let columnar2 = make_columnar("numbers", &[2u64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(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)])
.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());
@@ -114,16 +93,10 @@ fn test_group_columns_required_column_is_above_all_columns_have_the_same_type_ru
let columnar1 = make_columnar("numbers", &[2i64]);
let columnar2 = make_columnar("numbers", &[2i64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(
columnars,
&[("numbers".to_string(), ColumnType::U64)],
&merge_order,
)
.unwrap();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(columnars, &[("numbers".to_string(), ColumnType::U64)]).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]
@@ -131,22 +104,23 @@ fn test_missing_column() {
let columnar1 = make_columnar("numbers", &[-1i64]);
let columnar2 = make_columnar("numbers2", &[2u64]);
let columnars = &[&columnar1, &columnar2];
let merge_order = StackMergeOrder::stack(columnars).into();
let column_map: BTreeMap<(String, ColumnType), Vec<Option<DynamicColumn>>> =
group_columns_for_merge(columnars, &[], &merge_order).unwrap();
let column_map: BTreeMap<(String, ColumnTypeCategory), GroupedColumnsHandle> =
group_columns_for_merge(columnars, &[]).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());
}
@@ -169,9 +143,7 @@ fn make_numerical_columnar_multiple_columns(
.max()
.unwrap_or(0u32);
let mut buffer: Vec<u8> = Vec::new();
dataframe_writer
.serialize(num_rows, None, &mut buffer)
.unwrap();
dataframe_writer.serialize(num_rows, &mut buffer).unwrap();
ColumnarReader::open(buffer).unwrap()
}
@@ -194,9 +166,7 @@ fn make_byte_columnar_multiple_columns(
}
}
let mut buffer: Vec<u8> = Vec::new();
dataframe_writer
.serialize(num_rows, None, &mut buffer)
.unwrap();
dataframe_writer.serialize(num_rows, &mut buffer).unwrap();
ColumnarReader::open(buffer).unwrap()
}
@@ -215,9 +185,7 @@ fn make_text_columnar_multiple_columns(columns: &[(&str, &[&[&str]])]) -> Column
.max()
.unwrap_or(0u32);
let mut buffer: Vec<u8> = Vec::new();
dataframe_writer
.serialize(num_rows, None, &mut buffer)
.unwrap();
dataframe_writer.serialize(num_rows, &mut buffer).unwrap();
ColumnarReader::open(buffer).unwrap()
}
@@ -240,11 +208,13 @@ fn test_merge_columnar_numbers() {
)
.unwrap();
let columnar_reader = ColumnarReader::open(buffer).unwrap();
assert_eq!(columnar_reader.num_rows(), 3);
assert_eq!(columnar_reader.num_docs(), 3);
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);
@@ -266,11 +236,13 @@ fn test_merge_columnar_texts() {
)
.unwrap();
let columnar_reader = ColumnarReader::open(buffer).unwrap();
assert_eq!(columnar_reader.num_rows(), 3);
assert_eq!(columnar_reader.num_docs(), 3);
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| {
@@ -313,11 +285,13 @@ fn test_merge_columnar_byte() {
)
.unwrap();
let columnar_reader = ColumnarReader::open(buffer).unwrap();
assert_eq!(columnar_reader.num_rows(), 4);
assert_eq!(columnar_reader.num_docs(), 4);
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();
@@ -367,11 +341,13 @@ fn test_merge_columnar_byte_with_missing() {
)
.unwrap();
let columnar_reader = ColumnarReader::open(buffer).unwrap();
assert_eq!(columnar_reader.num_rows(), 3 + 2 + 3);
assert_eq!(columnar_reader.num_docs(), 3 + 2 + 3);
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();
@@ -417,23 +393,27 @@ fn test_merge_columnar_different_types() {
)
.unwrap();
let columnar_reader = ColumnarReader::open(buffer).unwrap();
assert_eq!(columnar_reader.num_rows(), 4);
assert_eq!(columnar_reader.num_docs(), 4);
assert_eq!(columnar_reader.num_columns(), 2);
let cols = columnar_reader.read_columns("mixed").unwrap();
// 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![]);
assert_eq!(vals.values_for_doc(2).collect_vec(), vec![]);
assert_eq!(vals.values_for_doc(0).collect_vec(), Vec::<i64>::new());
assert_eq!(vals.values_for_doc(1).collect_vec(), Vec::<i64>::new());
assert_eq!(vals.values_for_doc(2).collect_vec(), Vec::<i64>::new());
assert_eq!(vals.values_for_doc(3).collect_vec(), vec![1]);
assert_eq!(vals.values_for_doc(4).collect_vec(), vec![]);
assert_eq!(vals.values_for_doc(4).collect_vec(), Vec::<i64>::new());
// 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();
@@ -478,7 +458,7 @@ fn test_merge_columnar_different_empty_cardinality() {
)
.unwrap();
let columnar_reader = ColumnarReader::open(buffer).unwrap();
assert_eq!(columnar_reader.num_rows(), 2);
assert_eq!(columnar_reader.num_docs(), 2);
assert_eq!(columnar_reader.num_columns(), 2);
let cols = columnar_reader.read_columns("mixed").unwrap();
@@ -490,3 +470,119 @@ fn test_merge_columnar_different_empty_cardinality() {
let dynamic_column = cols[1].open().unwrap();
assert_eq!(dynamic_column.get_cardinality(), Cardinality::Optional);
}
#[derive(Debug, Clone)]
struct ColumnSpec {
column_name: String,
/// (row_id, term)
terms: Vec<(RowId, Vec<u8>)>,
}
#[derive(Clone, Debug)]
struct ColumnarSpec {
columns: Vec<ColumnSpec>,
}
/// Generate a random (row_id, term) pair:
/// - row_id in [0..10]
/// - term is either from POSSIBLE_TERMS or random bytes
fn rowid_and_term_strategy() -> impl Strategy<Value = (RowId, Vec<u8>)> {
const POSSIBLE_TERMS: &[&[u8]] = &[b"a", b"b", b"allo"];
let term_strat = prop_oneof![
// pick from the fixed list
(0..POSSIBLE_TERMS.len()).prop_map(|i| POSSIBLE_TERMS[i].to_vec()),
// or random bytes (length 0..10)
prop::collection::vec(any::<u8>(), 0..10),
];
(0u32..11, term_strat)
}
/// Generate one ColumnSpec, with a random name and a random list of (row_id, term).
/// We sort it by row_id so that data is in ascending order.
fn column_spec_strategy() -> impl Strategy<Value = ColumnSpec> {
let column_name = prop_oneof![
Just("col".to_string()),
Just("col2".to_string()),
"col.*".prop_map(|s| s),
];
// We'll produce 0..8 (rowid,term) entries for this column
let data_strat = vec(rowid_and_term_strategy(), 0..8).prop_map(|mut pairs| {
// Sort by row_id
pairs.sort_by_key(|(row_id, _)| *row_id);
pairs
});
(column_name, data_strat).prop_map(|(name, data)| ColumnSpec {
column_name: name,
terms: data,
})
}
/// Strategy to generate an ColumnarSpec
fn columnar_strategy() -> impl Strategy<Value = ColumnarSpec> {
vec(column_spec_strategy(), 0..3).prop_map(|columns| ColumnarSpec { columns })
}
/// Strategy to generate multiple ColumnarSpecs, each of which we will treat
/// as one "columnar" to be merged together.
fn columnars_strategy() -> impl Strategy<Value = Vec<ColumnarSpec>> {
vec(columnar_strategy(), 1..4)
}
/// Build a `ColumnarReader` from a `ColumnarSpec`
fn build_columnar(spec: &ColumnarSpec) -> ColumnarReader {
let mut writer = ColumnarWriter::default();
let mut max_row_id = 0;
for col in &spec.columns {
for &(row_id, ref term) in &col.terms {
writer.record_bytes(row_id, &col.column_name, term);
max_row_id = max_row_id.max(row_id);
}
}
let mut buffer = Vec::new();
writer.serialize(max_row_id + 1, &mut buffer).unwrap();
ColumnarReader::open(buffer).unwrap()
}
proptest! {
// We just test that the merge_columnar function doesn't crash.
#![proptest_config(ProptestConfig::with_cases(256))]
#[test]
fn test_merge_columnar_bytes_no_crash(columnars in columnars_strategy(), second_merge_columnars in columnars_strategy()) {
let columnars: Vec<ColumnarReader> = columnars.iter()
.map(build_columnar)
.collect();
let mut out = Vec::new();
let columnar_refs: Vec<&ColumnarReader> = columnars.iter().collect();
let stack_merge_order = StackMergeOrder::stack(&columnar_refs);
merge_columnar(
&columnar_refs,
&[],
MergeRowOrder::Stack(stack_merge_order),
&mut out,
).unwrap();
let merged_reader = ColumnarReader::open(out).unwrap();
// Merge the second set of columnars with the result of the first merge
let mut columnars: Vec<ColumnarReader> = second_merge_columnars.iter()
.map(build_columnar)
.collect();
columnars.push(merged_reader);
let mut out = Vec::new();
let columnar_refs: Vec<&ColumnarReader> = columnars.iter().collect();
let stack_merge_order = StackMergeOrder::stack(&columnar_refs);
merge_columnar(
&columnar_refs,
&[],
MergeRowOrder::Stack(stack_merge_order),
&mut out,
).unwrap();
}
}

1
columnar/src/columnar/mod.rs
View File

@@ -5,6 +5,7 @@ mod reader;
mod writer;
pub use column_type::{ColumnType, HasAssociatedColumnType};
pub use format_version::{Version, CURRENT_VERSION};
#[cfg(test)]
pub(crate) use merge::ColumnTypeCategory;
pub use merge::{merge_columnar, MergeRowOrder, ShuffleMergeOrder, StackMergeOrder};

208
columnar/src/columnar/reader/mod.rs
View File

@@ -1,12 +1,13 @@
use std::{fmt, io, mem};
use common::file_slice::FileSlice;
use common::json_path_writer::JSON_PATH_SEGMENT_SEP;
use common::BinarySerializable;
use sstable::{Dictionary, RangeSSTable};
use crate::columnar::{format_version, ColumnType};
use crate::dynamic_column::DynamicColumnHandle;
use crate::RowId;
use crate::{RowId, Version};
fn io_invalid_data(msg: String) -> io::Error {
io::Error::new(io::ErrorKind::InvalidData, msg)
@@ -18,12 +19,13 @@ fn io_invalid_data(msg: String) -> io::Error {
pub struct ColumnarReader {
column_dictionary: Dictionary<RangeSSTable>,
column_data: FileSlice,
num_rows: RowId,
num_docs: RowId,
format_version: Version,
}
impl fmt::Debug for ColumnarReader {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
let num_rows = self.num_rows();
let num_rows = self.num_docs();
let columns = self.list_columns().unwrap();
let num_cols = columns.len();
let mut debug_struct = f.debug_struct("Columnar");
@@ -53,6 +55,7 @@ impl fmt::Debug for ColumnarReader {
fn read_all_columns_in_stream(
mut stream: sstable::Streamer<'_, RangeSSTable>,
column_data: &FileSlice,
format_version: Version,
) -> io::Result<Vec<DynamicColumnHandle>> {
let mut results = Vec::new();
while stream.advance() {
@@ -67,12 +70,26 @@ fn read_all_columns_in_stream(
let dynamic_column_handle = DynamicColumnHandle {
file_slice,
column_type,
format_version,
};
results.push(dynamic_column_handle);
}
Ok(results)
}
fn column_dictionary_prefix_for_column_name(column_name: &str) -> String {
// Each column is a associated to a given `column_key`,
// that starts by `column_name\0column_header`.
//
// Listing the columns associated to the given column name is therefore equivalent to
// listing `column_key` with the prefix `column_name\0`.
format!("{}{}", column_name, '\0')
}
fn column_dictionary_prefix_for_subpath(root_path: &str) -> String {
format!("{}{}", root_path, JSON_PATH_SEGMENT_SEP as char)
}
impl ColumnarReader {
/// Opens a new Columnar file.
pub fn open<F>(file_slice: F) -> io::Result<ColumnarReader>
@@ -88,75 +105,70 @@ impl ColumnarReader {
let num_rows = u32::deserialize(&mut &footer_bytes[8..12])?;
let version_footer_bytes: [u8; format_version::VERSION_FOOTER_NUM_BYTES] =
footer_bytes[12..].try_into().unwrap();
let _version = format_version::parse_footer(version_footer_bytes)?;
let format_version = format_version::parse_footer(version_footer_bytes)?;
let (column_data, sstable) =
file_slice_without_sstable_len.split_from_end(sstable_len as usize);
let column_dictionary = Dictionary::open(sstable)?;
Ok(ColumnarReader {
column_dictionary,
column_data,
num_rows,
num_docs: num_rows,
format_version,
})
}
pub fn num_rows(&self) -> RowId {
self.num_rows
pub fn num_docs(&self) -> RowId {
self.num_docs
}
// 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,
format_version: self.format_version,
};
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)
}
fn stream_for_column_range(&self, column_name: &str) -> sstable::StreamerBuilder<RangeSSTable> {
// Each column is a associated to a given `column_key`,
// that starts by `column_name\0column_header`.
//
// Listing the columns associated to the given column name is therefore equivalent to
// listing `column_key` with the prefix `column_name\0`.
//
// This is in turn equivalent to searching for the range
// `[column_name,\0`..column_name\1)`.
// TODO can we get some more generic `prefix(..)` logic in the dictionary.
let mut start_key = column_name.to_string();
start_key.push('\0');
let mut end_key = column_name.to_string();
end_key.push(1u8 as char);
self.column_dictionary
.range()
.ge(start_key.as_bytes())
.lt(end_key.as_bytes())
Ok(self.iter_columns()?.collect())
}
pub async fn read_columns_async(
&self,
column_name: &str,
) -> io::Result<Vec<DynamicColumnHandle>> {
let prefix = column_dictionary_prefix_for_column_name(column_name);
let stream = self
.stream_for_column_range(column_name)
.column_dictionary
.prefix_range(prefix)
.into_stream_async()
.await?;
read_all_columns_in_stream(stream, &self.column_data)
read_all_columns_in_stream(stream, &self.column_data, self.format_version)
}
/// Get all columns for the given column name.
@@ -164,8 +176,36 @@ impl ColumnarReader {
/// There can be more than one column associated to a given column name, provided they have
/// different types.
pub fn read_columns(&self, column_name: &str) -> io::Result<Vec<DynamicColumnHandle>> {
let stream = self.stream_for_column_range(column_name).into_stream()?;
read_all_columns_in_stream(stream, &self.column_data)
let prefix = column_dictionary_prefix_for_column_name(column_name);
let stream = self.column_dictionary.prefix_range(prefix).into_stream()?;
read_all_columns_in_stream(stream, &self.column_data, self.format_version)
}
pub async fn read_subpath_columns_async(
&self,
root_path: &str,
) -> io::Result<Vec<DynamicColumnHandle>> {
let prefix = column_dictionary_prefix_for_subpath(root_path);
let stream = self
.column_dictionary
.prefix_range(prefix)
.into_stream_async()
.await?;
read_all_columns_in_stream(stream, &self.column_data, self.format_version)
}
/// Get all inner columns for a given JSON prefix, i.e columns for which the name starts
/// with the prefix then contain the [`JSON_PATH_SEGMENT_SEP`].
///
/// There can be more than one column associated to each path within the JSON structure,
/// provided they have different types.
pub fn read_subpath_columns(&self, root_path: &str) -> io::Result<Vec<DynamicColumnHandle>> {
let prefix = column_dictionary_prefix_for_subpath(root_path);
let stream = self
.column_dictionary
.prefix_range(prefix.as_bytes())
.into_stream()?;
read_all_columns_in_stream(stream, &self.column_data, self.format_version)
}
/// Return the number of columns in the columnar.
@@ -176,6 +216,8 @@ impl ColumnarReader {
#[cfg(test)]
mod tests {
use common::json_path_writer::JSON_PATH_SEGMENT_SEP;
use crate::{ColumnType, ColumnarReader, ColumnarWriter};
#[test]
@@ -184,7 +226,7 @@ mod tests {
columnar_writer.record_column_type("col1", ColumnType::Str, false);
columnar_writer.record_column_type("col2", ColumnType::U64, false);
let mut buffer = Vec::new();
columnar_writer.serialize(1, None, &mut buffer).unwrap();
columnar_writer.serialize(1, &mut buffer).unwrap();
let columnar = ColumnarReader::open(buffer).unwrap();
let columns = columnar.list_columns().unwrap();
assert_eq!(columns.len(), 2);
@@ -200,7 +242,7 @@ mod tests {
columnar_writer.record_column_type("count", ColumnType::U64, false);
columnar_writer.record_numerical(1, "count", 1u64);
let mut buffer = Vec::new();
columnar_writer.serialize(2, None, &mut buffer).unwrap();
columnar_writer.serialize(2, &mut buffer).unwrap();
let columnar = ColumnarReader::open(buffer).unwrap();
let columns = columnar.list_columns().unwrap();
assert_eq!(columns.len(), 1);
@@ -208,6 +250,64 @@ mod tests {
assert_eq!(columns[0].1.column_type(), ColumnType::U64);
}
#[test]
fn test_read_columns() {
let mut columnar_writer = ColumnarWriter::default();
columnar_writer.record_column_type("col", ColumnType::U64, false);
columnar_writer.record_numerical(1, "col", 1u64);
let mut buffer = Vec::new();
columnar_writer.serialize(2, &mut buffer).unwrap();
let columnar = ColumnarReader::open(buffer).unwrap();
{
let columns = columnar.read_columns("col").unwrap();
assert_eq!(columns.len(), 1);
assert_eq!(columns[0].column_type(), ColumnType::U64);
}
{
let columns = columnar.read_columns("other").unwrap();
assert_eq!(columns.len(), 0);
}
}
#[test]
fn test_read_subpath_columns() {
let mut columnar_writer = ColumnarWriter::default();
columnar_writer.record_str(
0,
&format!("col1{}subcol1", JSON_PATH_SEGMENT_SEP as char),
"hello",
);
columnar_writer.record_numerical(
0,
&format!("col1{}subcol2", JSON_PATH_SEGMENT_SEP as char),
1i64,
);
columnar_writer.record_str(1, "col1", "hello");
columnar_writer.record_str(0, "col2", "hello");
let mut buffer = Vec::new();
columnar_writer.serialize(2, &mut buffer).unwrap();
let columnar = ColumnarReader::open(buffer).unwrap();
{
let columns = columnar.read_subpath_columns("col1").unwrap();
assert_eq!(columns.len(), 2);
assert_eq!(columns[0].column_type(), ColumnType::Str);
assert_eq!(columns[1].column_type(), ColumnType::I64);
}
{
let columns = columnar.read_subpath_columns("col1.subcol1").unwrap();
assert_eq!(columns.len(), 0);
}
{
let columns = columnar.read_subpath_columns("col2").unwrap();
assert_eq!(columns.len(), 0);
}
{
let columns = columnar.read_subpath_columns("other").unwrap();
assert_eq!(columns.len(), 0);
}
}
#[test]
#[should_panic(expected = "Input type forbidden")]
fn test_list_columns_strict_typing_panics_on_wrong_types() {

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

@@ -87,7 +87,7 @@ impl<V: SymbolValue> ColumnOperation<V> {
minibuf
}
/// Deserialize a colummn operation.
/// Deserialize a column operation.
/// Returns None if the buffer is empty.
///
/// Panics if the payload is invalid:
@@ -122,7 +122,6 @@ impl<T> From<T> for ColumnOperation<T> {
// In order to limit memory usage, and in order
// to benefit from the stacker, we do this by serialization our data
// as "Symbols".
#[allow(clippy::from_over_into)]
pub(super) trait SymbolValue: Clone + Copy {
// Serializes the symbol into the given buffer.
// Returns the number of bytes written into the buffer.

32
columnar/src/columnar/writer/column_writers.rs
View File

@@ -41,31 +41,10 @@ impl ColumnWriter {
pub(super) fn operation_iterator<'a, V: SymbolValue>(
&self,
arena: &MemoryArena,
old_to_new_ids_opt: Option<&[RowId]>,
buffer: &'a mut Vec<u8>,
) -> impl Iterator<Item = ColumnOperation<V>> + 'a {
buffer.clear();
self.values.read_to_end(arena, buffer);
if let Some(old_to_new_ids) = old_to_new_ids_opt {
// TODO avoid the extra deserialization / serialization.
let mut sorted_ops: Vec<(RowId, ColumnOperation<V>)> = Vec::new();
let mut new_doc = 0u32;
let mut cursor = &buffer[..];
for op in std::iter::from_fn(|| ColumnOperation::<V>::deserialize(&mut cursor)) {
if let ColumnOperation::NewDoc(doc) = &op {
new_doc = old_to_new_ids[*doc as usize];
sorted_ops.push((new_doc, ColumnOperation::NewDoc(new_doc)));
} else {
sorted_ops.push((new_doc, op));
}
}
// stable sort is crucial here.
sorted_ops.sort_by_key(|(new_doc_id, _)| *new_doc_id);
buffer.clear();
for (_, op) in sorted_ops {
buffer.extend_from_slice(op.serialize().as_ref());
}
}
let mut cursor: &[u8] = &buffer[..];
std::iter::from_fn(move || ColumnOperation::deserialize(&mut cursor))
}
@@ -231,11 +210,9 @@ impl NumericalColumnWriter {
pub(super) fn operation_iterator<'a>(
self,
arena: &MemoryArena,
old_to_new_ids: Option<&[RowId]>,
buffer: &'a mut Vec<u8>,
) -> impl Iterator<Item = ColumnOperation<NumericalValue>> + 'a {
self.column_writer
.operation_iterator(arena, old_to_new_ids, buffer)
self.column_writer.operation_iterator(arena, buffer)
}
}
@@ -269,18 +246,17 @@ 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);
}
pub(super) fn operation_iterator<'a>(
&self,
arena: &MemoryArena,
old_to_new_ids: Option<&[RowId]>,
byte_buffer: &'a mut Vec<u8>,
) -> impl Iterator<Item = ColumnOperation<UnorderedId>> + 'a {
self.column_writer
.operation_iterator(arena, old_to_new_ids, byte_buffer)
self.column_writer.operation_iterator(arena, byte_buffer)
}
}

278
columnar/src/columnar/writer/mod.rs
View File

@@ -8,14 +8,13 @@ use std::net::Ipv6Addr;
use column_operation::ColumnOperation;
pub(crate) use column_writers::CompatibleNumericalTypes;
use common::json_path_writer::JSON_END_OF_PATH;
use common::CountingWriter;
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_index::{SerializableColumnIndex, SerializableOptionalIndex};
use crate::column_values::{MonotonicallyMappableToU128, MonotonicallyMappableToU64};
use crate::columnar::column_type::ColumnType;
use crate::columnar::writer::column_writers::{
ColumnWriter, NumericalColumnWriter, StrOrBytesColumnWriter,
@@ -45,7 +44,7 @@ struct SpareBuffers {
/// columnar_writer.record_str(1u32 /* doc id */, "product_name", "Apple");
/// columnar_writer.record_numerical(0u32 /* doc id */, "price", 10.5f64); //< uh oh we ended up mixing integer and floats.
/// let mut wrt: Vec<u8> = Vec::new();
/// columnar_writer.serialize(2u32, None, &mut wrt).unwrap();
/// columnar_writer.serialize(2u32, &mut wrt).unwrap();
/// ```
#[derive(Default)]
pub struct ColumnarWriter {
@@ -61,25 +60,8 @@ pub struct ColumnarWriter {
buffers: SpareBuffers,
}
#[inline]
fn mutate_or_create_column<V, TMutator>(
arena_hash_map: &mut ArenaHashMap,
column_name: &str,
updater: TMutator,
) where
V: Copy + 'static,
TMutator: FnMut(Option<V>) -> V,
{
assert!(
!column_name.as_bytes().contains(&0u8),
"key may not contain the 0 byte"
);
arena_hash_map.mutate_or_create(column_name.as_bytes(), updater);
}
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()
@@ -87,57 +69,11 @@ impl ColumnarWriter {
+ self.str_field_hash_map.mem_usage()
+ self.ip_addr_field_hash_map.mem_usage()
+ self.datetime_field_hash_map.mem_usage()
}
/// Returns the list of doc ids from 0..num_docs sorted by the `sort_field`
/// column.
///
/// If the column is multivalued, use the first value for scoring.
/// If no value is associated to a specific row, the document is assigned
/// the lowest possible score.
///
/// 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 mut symbols_buffer = Vec::new();
let mut values = Vec::new();
let mut start_doc_check_fill = 0;
let mut current_doc_opt: Option<RowId> = None;
// Assumption: NewDoc will never call the same doc twice and is strictly increasing between
// calls
for op in numerical_col_writer.operation_iterator(&self.arena, None, &mut symbols_buffer) {
match op {
ColumnOperation::NewDoc(doc) => {
current_doc_opt = Some(doc);
}
ColumnOperation::Value(numerical_value) => {
if let Some(current_doc) = current_doc_opt {
// Fill up with 0.0 since last doc
values.extend((start_doc_check_fill..current_doc).map(|doc| (0.0, doc)));
start_doc_check_fill = current_doc + 1;
// handle multi values
current_doc_opt = None;
let score: f32 = f64::coerce(numerical_value) as f32;
values.push((score, current_doc));
}
}
}
}
for doc in values.len() as u32..num_docs {
values.push((0.0f32, doc));
}
values.sort_by(|(left_score, _), (right_score, _)| {
if reversed {
right_score.total_cmp(left_score)
} else {
left_score.total_cmp(right_score)
}
});
values.into_iter().map(|(_score, doc)| doc).collect()
+ self
.dictionaries
.iter()
.map(|dict| dict.mem_usage())
.sum::<usize>()
}
/// Records a column type. This is useful to bypass the coercion process,
@@ -167,9 +103,8 @@ impl ColumnarWriter {
},
&mut self.dictionaries,
);
mutate_or_create_column(
hash_map,
column_name,
hash_map.mutate_or_create(
column_name.as_bytes(),
|column_opt: Option<StrOrBytesColumnWriter>| {
let mut column_writer = if let Some(column_writer) = column_opt {
column_writer
@@ -184,24 +119,21 @@ impl ColumnarWriter {
);
}
ColumnType::Bool => {
mutate_or_create_column(
&mut self.bool_field_hash_map,
column_name,
self.bool_field_hash_map.mutate_or_create(
column_name.as_bytes(),
|column_opt: Option<ColumnWriter>| column_opt.unwrap_or_default(),
);
}
ColumnType::DateTime => {
mutate_or_create_column(
&mut self.datetime_field_hash_map,
column_name,
self.datetime_field_hash_map.mutate_or_create(
column_name.as_bytes(),
|column_opt: Option<ColumnWriter>| column_opt.unwrap_or_default(),
);
}
ColumnType::I64 | ColumnType::F64 | ColumnType::U64 => {
let numerical_type = column_type.numerical_type().unwrap();
mutate_or_create_column(
&mut self.numerical_field_hash_map,
column_name,
self.numerical_field_hash_map.mutate_or_create(
column_name.as_bytes(),
|column_opt: Option<NumericalColumnWriter>| {
let mut column: NumericalColumnWriter = column_opt.unwrap_or_default();
column.force_numerical_type(numerical_type);
@@ -209,9 +141,8 @@ impl ColumnarWriter {
},
);
}
ColumnType::IpAddr => mutate_or_create_column(
&mut self.ip_addr_field_hash_map,
column_name,
ColumnType::IpAddr => self.ip_addr_field_hash_map.mutate_or_create(
column_name.as_bytes(),
|column_opt: Option<ColumnWriter>| column_opt.unwrap_or_default(),
),
}
@@ -224,9 +155,8 @@ impl ColumnarWriter {
numerical_value: T,
) {
let (hash_map, arena) = (&mut self.numerical_field_hash_map, &mut self.arena);
mutate_or_create_column(
hash_map,
column_name,
hash_map.mutate_or_create(
column_name.as_bytes(),
|column_opt: Option<NumericalColumnWriter>| {
let mut column: NumericalColumnWriter = column_opt.unwrap_or_default();
column.record_numerical_value(doc, numerical_value.into(), arena);
@@ -236,10 +166,6 @@ impl ColumnarWriter {
}
pub fn record_ip_addr(&mut self, doc: RowId, column_name: &str, ip_addr: Ipv6Addr) {
assert!(
!column_name.as_bytes().contains(&0u8),
"key may not contain the 0 byte"
);
let (hash_map, arena) = (&mut self.ip_addr_field_hash_map, &mut self.arena);
hash_map.mutate_or_create(
column_name.as_bytes(),
@@ -253,24 +179,30 @@ impl ColumnarWriter {
pub fn record_bool(&mut self, doc: RowId, column_name: &str, val: bool) {
let (hash_map, arena) = (&mut self.bool_field_hash_map, &mut self.arena);
mutate_or_create_column(hash_map, column_name, |column_opt: Option<ColumnWriter>| {
let mut column: ColumnWriter = column_opt.unwrap_or_default();
column.record(doc, val, arena);
column
});
hash_map.mutate_or_create(
column_name.as_bytes(),
|column_opt: Option<ColumnWriter>| {
let mut column: ColumnWriter = column_opt.unwrap_or_default();
column.record(doc, val, arena);
column
},
);
}
pub fn record_datetime(&mut self, doc: RowId, column_name: &str, datetime: common::DateTime) {
let (hash_map, arena) = (&mut self.datetime_field_hash_map, &mut self.arena);
mutate_or_create_column(hash_map, column_name, |column_opt: Option<ColumnWriter>| {
let mut column: ColumnWriter = column_opt.unwrap_or_default();
column.record(
doc,
NumericalValue::I64(datetime.into_timestamp_nanos()),
arena,
);
column
});
hash_map.mutate_or_create(
column_name.as_bytes(),
|column_opt: Option<ColumnWriter>| {
let mut column: ColumnWriter = column_opt.unwrap_or_default();
column.record(
doc,
NumericalValue::I64(datetime.into_timestamp_nanos()),
arena,
);
column
},
);
}
pub fn record_str(&mut self, doc: RowId, column_name: &str, value: &str) {
@@ -295,10 +227,6 @@ impl ColumnarWriter {
}
pub fn record_bytes(&mut self, doc: RowId, column_name: &str, value: &[u8]) {
assert!(
!column_name.as_bytes().contains(&0u8),
"key may not contain the 0 byte"
);
let (hash_map, arena, dictionaries) = (
&mut self.bytes_field_hash_map,
&mut self.arena,
@@ -318,17 +246,13 @@ impl ColumnarWriter {
},
);
}
pub fn serialize(
&mut self,
num_docs: RowId,
old_to_new_row_ids: Option<&[RowId]>,
wrt: &mut dyn io::Write,
) -> io::Result<()> {
pub fn serialize(&mut self, num_docs: RowId, wrt: &mut dyn io::Write) -> io::Result<()> {
let mut serializer = ColumnarSerializer::new(wrt);
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();
@@ -338,33 +262,38 @@ impl ColumnarWriter {
columns.extend(
self.bytes_field_hash_map
.iter()
.map(|(term, addr, _)| (term, ColumnType::Bytes, addr)),
.map(|(column_name, addr)| (column_name, 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));
let (arena, buffers, dictionaries) = (&self.arena, &mut self.buffers, &self.dictionaries);
let mut symbol_byte_buffer: Vec<u8> = Vec::new();
for (column_name, column_type, addr) in columns {
if column_name.contains(&JSON_END_OF_PATH) {
// Tantivy uses b'0' as a separator for nested fields in JSON.
// Column names with a b'0' are not simply ignored by the columnar (and the inverted
// index).
continue;
}
match column_type {
ColumnType::Bool => {
let column_writer: ColumnWriter = self.bool_field_hash_map.read(addr);
@@ -374,11 +303,7 @@ impl ColumnarWriter {
serialize_bool_column(
cardinality,
num_docs,
column_writer.operation_iterator(
arena,
old_to_new_row_ids,
&mut symbol_byte_buffer,
),
column_writer.operation_iterator(arena, &mut symbol_byte_buffer),
buffers,
&mut column_serializer,
)?;
@@ -392,11 +317,7 @@ impl ColumnarWriter {
serialize_ip_addr_column(
cardinality,
num_docs,
column_writer.operation_iterator(
arena,
old_to_new_row_ids,
&mut symbol_byte_buffer,
),
column_writer.operation_iterator(arena, &mut symbol_byte_buffer),
buffers,
&mut column_serializer,
)?;
@@ -421,12 +342,10 @@ impl ColumnarWriter {
num_docs,
str_or_bytes_column_writer.sort_values_within_row,
dictionary_builder,
str_or_bytes_column_writer.operation_iterator(
arena,
old_to_new_row_ids,
&mut symbol_byte_buffer,
),
str_or_bytes_column_writer
.operation_iterator(arena, &mut symbol_byte_buffer),
buffers,
&self.arena,
&mut column_serializer,
)?;
column_serializer.finalize()?;
@@ -442,11 +361,7 @@ impl ColumnarWriter {
cardinality,
num_docs,
numerical_type,
numerical_column_writer.operation_iterator(
arena,
old_to_new_row_ids,
&mut symbol_byte_buffer,
),
numerical_column_writer.operation_iterator(arena, &mut symbol_byte_buffer),
buffers,
&mut column_serializer,
)?;
@@ -461,11 +376,7 @@ impl ColumnarWriter {
cardinality,
num_docs,
NumericalType::I64,
column_writer.operation_iterator(
arena,
old_to_new_row_ids,
&mut symbol_byte_buffer,
),
column_writer.operation_iterator(arena, &mut symbol_byte_buffer),
buffers,
&mut column_serializer,
)?;
@@ -480,6 +391,7 @@ impl ColumnarWriter {
// Serialize [Dictionary, Column, dictionary num bytes U32::LE]
// Column: [Column Index, Column Values, column index num bytes U32::LE]
#[expect(clippy::too_many_arguments)]
fn serialize_bytes_or_str_column(
cardinality: Cardinality,
num_docs: RowId,
@@ -487,6 +399,7 @@ 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 +408,8 @@ 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 +545,7 @@ 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 {
@@ -650,16 +561,16 @@ where
let optional_index_builder = value_index_builders.borrow_optional_index_builder();
consume_operation_iterator(op_iterator, optional_index_builder, values);
let optional_index = optional_index_builder.finish(num_rows);
SerializableColumnIndex::Optional {
SerializableColumnIndex::Optional(SerializableOptionalIndex {
num_rows,
non_null_row_ids: Box::new(optional_index),
}
})
}
Cardinality::Multivalued => {
let multivalued_index_builder = value_index_builders.borrow_multivalued_index_builder();
consume_operation_iterator(op_iterator, multivalued_index_builder, values);
let multivalued_index = multivalued_index_builder.finish(num_rows);
SerializableColumnIndex::Multivalued(Box::new(multivalued_index))
let serializable_multivalued_index = multivalued_index_builder.finish(num_rows);
SerializableColumnIndex::Multivalued(serializable_multivalued_index)
}
};
crate::column::serialize_column_mappable_to_u128(
@@ -670,15 +581,6 @@ where
Ok(())
}
fn sort_values_within_row_in_place(multivalued_index: &[RowId], values: &mut [u64]) {
let mut start_index: usize = 0;
for end_index in multivalued_index.iter().copied() {
let end_index = end_index as usize;
values[start_index..end_index].sort_unstable();
start_index = end_index;
}
}
fn send_to_serialize_column_mappable_to_u64(
op_iterator: impl Iterator<Item = ColumnOperation<u64>>,
cardinality: Cardinality,
@@ -687,10 +589,7 @@ 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 => {
@@ -705,19 +604,22 @@ where
let optional_index_builder = value_index_builders.borrow_optional_index_builder();
consume_operation_iterator(op_iterator, optional_index_builder, values);
let optional_index = optional_index_builder.finish(num_rows);
SerializableColumnIndex::Optional {
SerializableColumnIndex::Optional(SerializableOptionalIndex {
non_null_row_ids: Box::new(optional_index),
num_rows,
}
})
}
Cardinality::Multivalued => {
let multivalued_index_builder = value_index_builders.borrow_multivalued_index_builder();
consume_operation_iterator(op_iterator, multivalued_index_builder, values);
let multivalued_index = multivalued_index_builder.finish(num_rows);
let serializable_multivalued_index = multivalued_index_builder.finish(num_rows);
if sort_values_within_row {
sort_values_within_row_in_place(multivalued_index, values);
sort_values_within_row_in_place(
serializable_multivalued_index.start_offsets.boxed_iter(),
values,
);
}
SerializableColumnIndex::Multivalued(Box::new(multivalued_index))
SerializableColumnIndex::Multivalued(serializable_multivalued_index)
}
};
crate::column::serialize_column_mappable_to_u64(
@@ -728,6 +630,18 @@ where
Ok(())
}
fn sort_values_within_row_in_place(
multivalued_index: impl Iterator<Item = RowId>,
values: &mut [u64],
) {
let mut start_index: usize = 0;
for end_index in multivalued_index {
let end_index = end_index as usize;
values[start_index..end_index].sort_unstable();
start_index = end_index;
}
}
fn coerce_numerical_symbol<T>(
operation_iterator: impl Iterator<Item = ColumnOperation<NumericalValue>>,
) -> impl Iterator<Item = ColumnOperation<u64>>
@@ -775,7 +689,7 @@ mod tests {
assert_eq!(column_writer.get_cardinality(3), Cardinality::Full);
let mut buffer = Vec::new();
let symbols: Vec<ColumnOperation<NumericalValue>> = column_writer
.operation_iterator(&arena, None, &mut buffer)
.operation_iterator(&arena, &mut buffer)
.collect();
assert_eq!(symbols.len(), 6);
assert!(matches!(symbols[0], ColumnOperation::NewDoc(0u32)));
@@ -804,7 +718,7 @@ mod tests {
assert_eq!(column_writer.get_cardinality(3), Cardinality::Optional);
let mut buffer = Vec::new();
let symbols: Vec<ColumnOperation<NumericalValue>> = column_writer
.operation_iterator(&arena, None, &mut buffer)
.operation_iterator(&arena, &mut buffer)
.collect();
assert_eq!(symbols.len(), 4);
assert!(matches!(symbols[0], ColumnOperation::NewDoc(1u32)));
@@ -827,7 +741,7 @@ mod tests {
assert_eq!(column_writer.get_cardinality(2), Cardinality::Optional);
let mut buffer = Vec::new();
let symbols: Vec<ColumnOperation<NumericalValue>> = column_writer
.operation_iterator(&arena, None, &mut buffer)
.operation_iterator(&arena, &mut buffer)
.collect();
assert_eq!(symbols.len(), 2);
assert!(matches!(symbols[0], ColumnOperation::NewDoc(0u32)));
@@ -846,7 +760,7 @@ mod tests {
assert_eq!(column_writer.get_cardinality(1), Cardinality::Multivalued);
let mut buffer = Vec::new();
let symbols: Vec<ColumnOperation<NumericalValue>> = column_writer
.operation_iterator(&arena, None, &mut buffer)
.operation_iterator(&arena, &mut buffer)
.collect();
assert_eq!(symbols.len(), 3);
assert!(matches!(symbols[0], ColumnOperation::NewDoc(0u32)));

23
columnar/src/columnar/writer/serializer.rs
View File

@@ -1,6 +1,7 @@
use std::io;
use std::io::Write;
use common::json_path_writer::JSON_END_OF_PATH;
use common::{BinarySerializable, CountingWriter};
use sstable::value::RangeValueWriter;
use sstable::RangeSSTable;
@@ -19,7 +20,7 @@ pub struct ColumnarSerializer<W: io::Write> {
fn prepare_key(key: &[u8], column_type: ColumnType, buffer: &mut Vec<u8>) {
buffer.clear();
buffer.extend_from_slice(key);
buffer.push(0u8);
buffer.push(JSON_END_OF_PATH);
buffer.push(column_type.to_code());
}
@@ -66,7 +67,7 @@ pub struct ColumnSerializer<'a, W: io::Write> {
start_offset: u64,
}
impl<'a, W: io::Write> ColumnSerializer<'a, W> {
impl<W: io::Write> ColumnSerializer<'_, W> {
pub fn finalize(self) -> io::Result<()> {
let end_offset: u64 = self.columnar_serializer.wrt.written_bytes();
let byte_range = self.start_offset..end_offset;
@@ -79,7 +80,7 @@ impl<'a, W: io::Write> ColumnSerializer<'a, W> {
}
}
impl<'a, W: io::Write> io::Write for ColumnSerializer<'a, W> {
impl<W: io::Write> io::Write for ColumnSerializer<'_, W> {
fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
self.columnar_serializer.wrt.write(buf)
}
@@ -92,19 +93,3 @@ impl<'a, W: io::Write> io::Write for ColumnSerializer<'a, W> {
self.columnar_serializer.wrt.write_all(buf)
}
}
#[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], 0u8);
assert_eq!(buffer[11], ColumnType::Str.to_code());
}
}

81
columnar/src/columnar/writer/value_index.rs
View File

@@ -1,3 +1,4 @@
use crate::column_index::{SerializableMultivalueIndex, SerializableOptionalIndex};
use crate::iterable::Iterable;
use crate::RowId;
@@ -59,31 +60,47 @@ impl IndexBuilder for OptionalIndexBuilder {
#[derive(Default)]
pub struct MultivaluedIndexBuilder {
start_offsets: Vec<RowId>,
doc_with_values: Vec<RowId>,
start_offsets: Vec<u32>,
total_num_vals_seen: u32,
current_row: RowId,
current_row_has_value: bool,
}
impl MultivaluedIndexBuilder {
pub fn finish(&mut self, num_docs: RowId) -> &[u32] {
self.start_offsets
.resize(num_docs as usize + 1, self.total_num_vals_seen);
&self.start_offsets[..]
pub fn finish(&mut self, num_docs: RowId) -> SerializableMultivalueIndex<'_> {
self.start_offsets.push(self.total_num_vals_seen);
let non_null_row_ids: Box<dyn Iterable<RowId>> = Box::new(&self.doc_with_values[..]);
SerializableMultivalueIndex {
doc_ids_with_values: SerializableOptionalIndex {
non_null_row_ids,
num_rows: num_docs,
},
start_offsets: Box::new(&self.start_offsets[..]),
}
}
fn reset(&mut self) {
self.doc_with_values.clear();
self.start_offsets.clear();
self.start_offsets.push(0u32);
self.total_num_vals_seen = 0;
self.current_row = 0;
self.current_row_has_value = false;
}
}
impl IndexBuilder for MultivaluedIndexBuilder {
fn record_row(&mut self, row_id: RowId) {
self.start_offsets
.resize(row_id as usize + 1, self.total_num_vals_seen);
self.current_row = row_id;
self.current_row_has_value = false;
}
fn record_value(&mut self) {
if !self.current_row_has_value {
self.current_row_has_value = true;
self.doc_with_values.push(self.current_row);
self.start_offsets.push(self.total_num_vals_seen);
}
self.total_num_vals_seen += 1;
}
}
@@ -141,6 +158,32 @@ mod tests {
);
}
#[test]
fn test_multivalued_value_index_builder_simple() {
let mut multivalued_value_index_builder = MultivaluedIndexBuilder::default();
{
multivalued_value_index_builder.record_row(0u32);
multivalued_value_index_builder.record_value();
multivalued_value_index_builder.record_value();
let serialized_multivalue_index = multivalued_value_index_builder.finish(1u32);
let start_offsets: Vec<u32> = serialized_multivalue_index
.start_offsets
.boxed_iter()
.collect();
assert_eq!(&start_offsets, &[0, 2]);
}
multivalued_value_index_builder.reset();
multivalued_value_index_builder.record_row(0u32);
multivalued_value_index_builder.record_value();
multivalued_value_index_builder.record_value();
let serialized_multivalue_index = multivalued_value_index_builder.finish(1u32);
let start_offsets: Vec<u32> = serialized_multivalue_index
.start_offsets
.boxed_iter()
.collect();
assert_eq!(&start_offsets, &[0, 2]);
}
#[test]
fn test_multivalued_value_index_builder() {
let mut multivalued_value_index_builder = MultivaluedIndexBuilder::default();
@@ -149,17 +192,15 @@ mod tests {
multivalued_value_index_builder.record_value();
multivalued_value_index_builder.record_row(2u32);
multivalued_value_index_builder.record_value();
assert_eq!(
multivalued_value_index_builder.finish(4u32).to_vec(),
vec![0, 0, 2, 3, 3]
);
multivalued_value_index_builder.reset();
multivalued_value_index_builder.record_row(2u32);
multivalued_value_index_builder.record_value();
multivalued_value_index_builder.record_value();
assert_eq!(
multivalued_value_index_builder.finish(4u32).to_vec(),
vec![0, 0, 0, 2, 2]
);
let SerializableMultivalueIndex {
doc_ids_with_values,
start_offsets,
} = multivalued_value_index_builder.finish(4u32);
assert_eq!(doc_ids_with_values.num_rows, 4u32);
let doc_ids_with_values: Vec<u32> =
doc_ids_with_values.non_null_row_ids.boxed_iter().collect();
assert_eq!(&doc_ids_with_values, &[1u32, 2u32]);
let start_offsets: Vec<u32> = start_offsets.boxed_iter().collect();
assert_eq!(&start_offsets[..], &[0, 2, 3]);
}
}

183
columnar/src/compat_tests.rs Normal file
View File

@@ -0,0 +1,183 @@
use std::path::PathBuf;
use itertools::Itertools;
use crate::{
merge_columnar, Cardinality, Column, ColumnarReader, DynamicColumn, StackMergeOrder,
CURRENT_VERSION,
};
const NUM_DOCS: u32 = u16::MAX as u32;
fn generate_columnar(num_docs: u32, value_offset: u64) -> Vec<u8> {
use crate::ColumnarWriter;
let mut columnar_writer = ColumnarWriter::default();
for i in 0..num_docs {
if i % 100 == 0 {
columnar_writer.record_numerical(i, "sparse", value_offset + i as u64);
}
if i % 5 == 0 {
columnar_writer.record_numerical(i, "dense", value_offset + i as u64);
}
columnar_writer.record_numerical(i, "full", value_offset + i as u64);
columnar_writer.record_numerical(i, "multi", value_offset + i as u64);
columnar_writer.record_numerical(i, "multi", value_offset + i as u64);
}
let mut wrt: Vec<u8> = Vec::new();
columnar_writer.serialize(num_docs, &mut wrt).unwrap();
wrt
}
#[test]
/// Writes a columnar for the CURRENT_VERSION to disk.
fn create_format() {
let version = CURRENT_VERSION.to_string();
let file_path = path_for_version(&version);
if PathBuf::from(file_path.clone()).exists() {
return;
}
let columnar = generate_columnar(NUM_DOCS, 0);
std::fs::write(file_path, columnar).unwrap();
}
fn path_for_version(version: &str) -> String {
format!("./compat_tests_data/{}.columnar", version)
}
#[test]
fn test_format_v1() {
let path = path_for_version("v1");
test_format(&path);
}
#[test]
fn test_format_v2() {
let path = path_for_version("v2");
test_format(&path);
}
fn test_format(path: &str) {
let file_content = std::fs::read(path).unwrap();
let reader = ColumnarReader::open(file_content).unwrap();
check_columns(&reader);
// Test merge
let reader2 = ColumnarReader::open(generate_columnar(NUM_DOCS, NUM_DOCS as u64)).unwrap();
let columnar_readers = vec![&reader, &reader2];
let merge_row_order = StackMergeOrder::stack(&columnar_readers[..]);
let mut out = Vec::new();
merge_columnar(&columnar_readers, &[], merge_row_order.into(), &mut out).unwrap();
let reader = ColumnarReader::open(out).unwrap();
check_columns(&reader);
}
fn check_columns(reader: &ColumnarReader) {
let column = open_column(reader, "full");
check_column(&column, |doc_id| vec![(doc_id, doc_id as u64).into()]);
assert_eq!(column.get_cardinality(), Cardinality::Full);
let column = open_column(reader, "multi");
check_column(&column, |doc_id| {
vec![
(doc_id * 2, doc_id as u64).into(),
(doc_id * 2 + 1, doc_id as u64).into(),
]
});
assert_eq!(column.get_cardinality(), Cardinality::Multivalued);
let column = open_column(reader, "sparse");
check_column(&column, |doc_id| {
if doc_id % 100 == 0 {
vec![(doc_id / 100, doc_id as u64).into()]
} else {
vec![]
}
});
assert_eq!(column.get_cardinality(), Cardinality::Optional);
let column = open_column(reader, "dense");
check_column(&column, |doc_id| {
if doc_id % 5 == 0 {
vec![(doc_id / 5, doc_id as u64).into()]
} else {
vec![]
}
});
assert_eq!(column.get_cardinality(), Cardinality::Optional);
}
struct RowIdAndValue {
row_id: u32,
value: u64,
}
impl From<(u32, u64)> for RowIdAndValue {
fn from((row_id, value): (u32, u64)) -> Self {
Self { row_id, value }
}
}
fn check_column<F: Fn(u32) -> Vec<RowIdAndValue>>(column: &Column<u64>, expected: F) {
let num_docs = column.num_docs();
let test_doc = |doc: u32| {
if expected(doc).is_empty() {
assert_eq!(column.first(doc), None);
} else {
assert_eq!(column.first(doc), Some(expected(doc)[0].value));
}
let values = column.values_for_doc(doc).collect_vec();
assert_eq!(values, expected(doc).iter().map(|x| x.value).collect_vec());
let mut row_ids = Vec::new();
column.row_ids_for_docs(&[doc], &mut vec![], &mut row_ids);
assert_eq!(
row_ids,
expected(doc).iter().map(|x| x.row_id).collect_vec()
);
let values = column.values_for_doc(doc).collect_vec();
assert_eq!(values, expected(doc).iter().map(|x| x.value).collect_vec());
// Docid rowid conversion
let mut row_ids = Vec::new();
let safe_next_doc = |doc: u32| (doc + 1).min(num_docs - 1);
column
.index
.docids_to_rowids(&[doc, safe_next_doc(doc)], &mut vec![], &mut row_ids);
let expected_rowids = expected(doc)
.iter()
.map(|x| x.row_id)
.chain(expected(safe_next_doc(doc)).iter().map(|x| x.row_id))
.collect_vec();
assert_eq!(row_ids, expected_rowids);
let rowid_range = column
.index
.docid_range_to_rowids(doc..safe_next_doc(doc) + 1);
if expected_rowids.is_empty() {
assert!(rowid_range.is_empty());
} else {
assert_eq!(
rowid_range,
expected_rowids[0]..expected_rowids.last().unwrap() + 1
);
}
};
test_doc(0);
test_doc(num_docs - 1);
test_doc(num_docs - 2);
test_doc(65000);
}
fn open_column(reader: &ColumnarReader, name: &str) -> Column<u64> {
let column = reader.read_columns(name).unwrap()[0]
.open()
.unwrap()
.coerce_numerical(crate::NumericalType::U64)
.unwrap();
let DynamicColumn::U64(column) = column else {
panic!();
};
column
}

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));

65
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, Version};
#[derive(Clone)]
pub enum DynamicColumn {
@@ -228,10 +228,11 @@ 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,
pub(crate) format_version: Version,
}
impl DynamicColumnHandle {
@@ -247,7 +248,12 @@ 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.
@@ -255,13 +261,24 @@ impl DynamicColumnHandle {
let column_bytes = self.file_slice.read_bytes()?;
match self.column_type {
ColumnType::Str | ColumnType::Bytes => {
let column: BytesColumn = crate::column::open_column_bytes(column_bytes)?;
let column: BytesColumn =
crate::column::open_column_bytes(column_bytes, self.format_version)?;
Ok(Some(column.term_ord_column))
}
ColumnType::Bool => Ok(None),
ColumnType::IpAddr => Ok(None),
ColumnType::I64 | ColumnType::U64 | ColumnType::F64 | ColumnType::DateTime => {
let column = crate::column::open_column_u64::<u64>(column_bytes)?;
ColumnType::IpAddr => {
let column = crate::column::open_column_u128_as_compact_u64(
column_bytes,
self.format_version,
)?;
Ok(Some(column))
}
ColumnType::Bool
| ColumnType::I64
| ColumnType::U64
| ColumnType::F64
| ColumnType::DateTime => {
let column =
crate::column::open_column_u64::<u64>(column_bytes, self.format_version)?;
Ok(Some(column))
}
}
@@ -269,15 +286,31 @@ impl DynamicColumnHandle {
fn open_internal(&self, column_bytes: OwnedBytes) -> io::Result<DynamicColumn> {
let dynamic_column: DynamicColumn = match self.column_type {
ColumnType::Bytes => crate::column::open_column_bytes(column_bytes)?.into(),
ColumnType::Str => crate::column::open_column_str(column_bytes)?.into(),
ColumnType::I64 => crate::column::open_column_u64::<i64>(column_bytes)?.into(),
ColumnType::U64 => crate::column::open_column_u64::<u64>(column_bytes)?.into(),
ColumnType::F64 => crate::column::open_column_u64::<f64>(column_bytes)?.into(),
ColumnType::Bool => crate::column::open_column_u64::<bool>(column_bytes)?.into(),
ColumnType::IpAddr => crate::column::open_column_u128::<Ipv6Addr>(column_bytes)?.into(),
ColumnType::Bytes => {
crate::column::open_column_bytes(column_bytes, self.format_version)?.into()
}
ColumnType::Str => {
crate::column::open_column_str(column_bytes, self.format_version)?.into()
}
ColumnType::I64 => {
crate::column::open_column_u64::<i64>(column_bytes, self.format_version)?.into()
}
ColumnType::U64 => {
crate::column::open_column_u64::<u64>(column_bytes, self.format_version)?.into()
}
ColumnType::F64 => {
crate::column::open_column_u64::<f64>(column_bytes, self.format_version)?.into()
}
ColumnType::Bool => {
crate::column::open_column_u64::<bool>(column_bytes, self.format_version)?.into()
}
ColumnType::IpAddr => {
crate::column::open_column_u128::<Ipv6Addr>(column_bytes, self.format_version)?
.into()
}
ColumnType::DateTime => {
crate::column::open_column_u64::<DateTime>(column_bytes)?.into()
crate::column::open_column_u64::<DateTime>(column_bytes, self.format_version)?
.into()
}
};
Ok(dynamic_column)

11
columnar/src/iterable.rs
View File

@@ -1,10 +1,13 @@
use std::ops::Range;
use std::sync::Arc;
use crate::{ColumnValues, RowId};
pub trait Iterable<T = u64> {
fn boxed_iter(&self) -> Box<dyn Iterator<Item = T> + '_>;
}
impl<'a, T: Copy> Iterable<T> for &'a [T] {
impl<T: Copy> Iterable<T> for &[T] {
fn boxed_iter(&self) -> Box<dyn Iterator<Item = T> + '_> {
Box::new(self.iter().copied())
}
@@ -17,3 +20,9 @@ where Range<T>: Iterator<Item = T>
Box::new(self.clone())
}
}
impl Iterable for Arc<dyn crate::ColumnValues<RowId>> {
fn boxed_iter(&self) -> Box<dyn Iterator<Item = u64> + '_> {
Box::new(self.iter().map(|row_id| row_id as u64))
}
}

29
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;
@@ -29,7 +48,7 @@ pub use column_values::{
};
pub use columnar::{
merge_columnar, ColumnType, ColumnarReader, ColumnarWriter, HasAssociatedColumnType,
MergeRowOrder, ShuffleMergeOrder, StackMergeOrder,
MergeRowOrder, ShuffleMergeOrder, StackMergeOrder, Version, CURRENT_VERSION,
};
use sstable::VoidSSTable;
pub use value::{NumericalType, NumericalValue};
@@ -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
}
@@ -109,3 +131,6 @@ impl Cardinality {
#[cfg(test)]
mod tests;
#[cfg(test)]
mod compat_tests;

278
columnar/src/tests.rs
View File

@@ -21,12 +21,12 @@ fn test_dataframe_writer_str() {
dataframe_writer.record_str(1u32, "my_string", "hello");
dataframe_writer.record_str(3u32, "my_string", "helloeee");
let mut buffer: Vec<u8> = Vec::new();
dataframe_writer.serialize(5, None, &mut buffer).unwrap();
dataframe_writer.serialize(5, &mut buffer).unwrap();
let columnar = ColumnarReader::open(buffer).unwrap();
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]
@@ -35,12 +35,12 @@ fn test_dataframe_writer_bytes() {
dataframe_writer.record_bytes(1u32, "my_string", b"hello");
dataframe_writer.record_bytes(3u32, "my_string", b"helloeee");
let mut buffer: Vec<u8> = Vec::new();
dataframe_writer.serialize(5, None, &mut buffer).unwrap();
dataframe_writer.serialize(5, &mut buffer).unwrap();
let columnar = ColumnarReader::open(buffer).unwrap();
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]
@@ -49,7 +49,7 @@ fn test_dataframe_writer_bool() {
dataframe_writer.record_bool(1u32, "bool.value", false);
dataframe_writer.record_bool(3u32, "bool.value", true);
let mut buffer: Vec<u8> = Vec::new();
dataframe_writer.serialize(5, None, &mut buffer).unwrap();
dataframe_writer.serialize(5, &mut buffer).unwrap();
let columnar = ColumnarReader::open(buffer).unwrap();
assert_eq!(columnar.num_columns(), 1);
let cols: Vec<DynamicColumnHandle> = columnar.read_columns("bool.value").unwrap();
@@ -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,]);
}
@@ -72,14 +74,16 @@ fn test_dataframe_writer_u64_multivalued() {
dataframe_writer.record_numerical(6u32, "divisor", 2u64);
dataframe_writer.record_numerical(6u32, "divisor", 3u64);
let mut buffer: Vec<u8> = Vec::new();
dataframe_writer.serialize(7, None, &mut buffer).unwrap();
dataframe_writer.serialize(7, &mut buffer).unwrap();
let columnar = ColumnarReader::open(buffer).unwrap();
assert_eq!(columnar.num_columns(), 1);
let cols: Vec<DynamicColumnHandle> = columnar.read_columns("divisor").unwrap();
assert_eq!(cols.len(), 1);
assert_eq!(cols[0].num_bytes(), 29);
assert_eq!(cols[0].num_bytes(), 50);
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
@@ -93,7 +97,7 @@ fn test_dataframe_writer_ip_addr() {
dataframe_writer.record_ip_addr(1, "ip_addr", Ipv6Addr::from_u128(1001));
dataframe_writer.record_ip_addr(3, "ip_addr", Ipv6Addr::from_u128(1050));
let mut buffer: Vec<u8> = Vec::new();
dataframe_writer.serialize(5, None, &mut buffer).unwrap();
dataframe_writer.serialize(5, &mut buffer).unwrap();
let columnar = ColumnarReader::open(buffer).unwrap();
assert_eq!(columnar.num_columns(), 1);
let cols: Vec<DynamicColumnHandle> = columnar.read_columns("ip_addr").unwrap();
@@ -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,
@@ -122,7 +128,7 @@ fn test_dataframe_writer_numerical() {
dataframe_writer.record_numerical(2u32, "srical.value", NumericalValue::U64(13u64));
dataframe_writer.record_numerical(4u32, "srical.value", NumericalValue::U64(15u64));
let mut buffer: Vec<u8> = Vec::new();
dataframe_writer.serialize(6, None, &mut buffer).unwrap();
dataframe_writer.serialize(6, &mut buffer).unwrap();
let columnar = ColumnarReader::open(buffer).unwrap();
assert_eq!(columnar.num_columns(), 1);
let cols: Vec<DynamicColumnHandle> = columnar.read_columns("srical.value").unwrap();
@@ -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));
@@ -145,46 +153,6 @@ fn test_dataframe_writer_numerical() {
assert_eq!(column_i64.first(6), None); //< we can change the spec for that one.
}
#[test]
fn test_dataframe_sort_by_full() {
let mut dataframe_writer = ColumnarWriter::default();
dataframe_writer.record_numerical(0u32, "value", NumericalValue::U64(1));
dataframe_writer.record_numerical(1u32, "value", NumericalValue::U64(2));
let data = dataframe_writer.sort_order("value", 2, false);
assert_eq!(data, vec![0, 1]);
}
#[test]
fn test_dataframe_sort_by_opt() {
let mut dataframe_writer = ColumnarWriter::default();
dataframe_writer.record_numerical(1u32, "value", NumericalValue::U64(3));
dataframe_writer.record_numerical(3u32, "value", NumericalValue::U64(2));
let data = dataframe_writer.sort_order("value", 5, false);
// 0, 2, 4 is 0.0
assert_eq!(data, vec![0, 2, 4, 3, 1]);
let data = dataframe_writer.sort_order("value", 5, true);
assert_eq!(
data,
vec![4, 2, 0, 3, 1].into_iter().rev().collect::<Vec<_>>()
);
}
#[test]
fn test_dataframe_sort_by_multi() {
let mut dataframe_writer = ColumnarWriter::default();
// valid for sort
dataframe_writer.record_numerical(1u32, "value", NumericalValue::U64(2));
// those are ignored for sort
dataframe_writer.record_numerical(1u32, "value", NumericalValue::U64(4));
dataframe_writer.record_numerical(1u32, "value", NumericalValue::U64(4));
// valid for sort
dataframe_writer.record_numerical(3u32, "value", NumericalValue::U64(3));
// ignored, would change sort order
dataframe_writer.record_numerical(3u32, "value", NumericalValue::U64(1));
let data = dataframe_writer.sort_order("value", 4, false);
assert_eq!(data, vec![0, 2, 1, 3]);
}
#[test]
fn test_dictionary_encoded_str() {
let mut buffer = Vec::new();
@@ -193,12 +161,14 @@ fn test_dictionary_encoded_str() {
columnar_writer.record_str(3, "my.column", "c");
columnar_writer.record_str(3, "my.column2", "different_column!");
columnar_writer.record_str(4, "my.column", "b");
columnar_writer.serialize(5, None, &mut buffer).unwrap();
columnar_writer.serialize(5, &mut buffer).unwrap();
let columnar_reader = ColumnarReader::open(buffer).unwrap();
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);
@@ -225,12 +195,14 @@ fn test_dictionary_encoded_bytes() {
columnar_writer.record_bytes(3, "my.column", b"c");
columnar_writer.record_bytes(3, "my.column2", b"different_column!");
columnar_writer.record_bytes(4, "my.column", b"b");
columnar_writer.serialize(5, None, &mut buffer).unwrap();
columnar_writer.serialize(5, &mut buffer).unwrap();
let columnar_reader = ColumnarReader::open(buffer).unwrap();
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 +290,9 @@ 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,
@@ -331,8 +303,8 @@ fn column_value_strategy() -> impl Strategy<Value = ColumnValue> {
0,
ip_addr_byte
))),
1 => any::<bool>().prop_map(|b| ColumnValue::Bool(b)),
1 => (0_679_723_993i64..1_679_723_995i64)
1 => any::<bool>().prop_map(ColumnValue::Bool),
1 => (679_723_993i64..1_679_723_995i64)
.prop_map(|val| { ColumnValue::DateTime(DateTime::from_timestamp_secs(val)) })
]
}
@@ -357,26 +329,12 @@ fn columnar_docs_strategy() -> impl Strategy<Value = Vec<Vec<(&'static str, Colu
.prop_flat_map(|num_docs| proptest::collection::vec(doc_strategy(), num_docs))
}
fn columnar_docs_and_mapping_strategy(
) -> impl Strategy<Value = (Vec<Vec<(&'static str, ColumnValue)>>, Vec<RowId>)> {
columnar_docs_strategy().prop_flat_map(|docs| {
permutation_strategy(docs.len()).prop_map(move |permutation| (docs.clone(), permutation))
})
}
fn permutation_strategy(n: usize) -> impl Strategy<Value = Vec<RowId>> {
Just((0u32..n as RowId).collect()).prop_shuffle()
}
fn permutation_and_subset_strategy(n: usize) -> impl Strategy<Value = Vec<usize>> {
let vals: Vec<usize> = (0..n).collect();
subsequence(vals, 0..=n).prop_shuffle()
}
fn build_columnar_with_mapping(
docs: &[Vec<(&'static str, ColumnValue)>],
old_to_new_row_ids_opt: Option<&[RowId]>,
) -> ColumnarReader {
fn build_columnar_with_mapping(docs: &[Vec<(&'static str, ColumnValue)>]) -> ColumnarReader {
let num_docs = docs.len() as u32;
let mut buffer = Vec::new();
let mut columnar_writer = ColumnarWriter::default();
@@ -404,15 +362,13 @@ 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
columnar_writer.serialize(num_docs, &mut buffer).unwrap();
ColumnarReader::open(buffer).unwrap()
}
fn build_columnar(docs: &[Vec<(&'static str, ColumnValue)>]) -> ColumnarReader {
build_columnar_with_mapping(docs, None)
build_columnar_with_mapping(docs)
}
fn assert_columnar_eq_strict(left: &ColumnarReader, right: &ColumnarReader) {
@@ -424,7 +380,7 @@ fn assert_columnar_eq(
right: &ColumnarReader,
lenient_on_numerical_value: bool,
) {
assert_eq!(left.num_rows(), right.num_rows());
assert_eq!(left.num_docs(), right.num_docs());
let left_columns = left.list_columns().unwrap();
let right_columns = right.list_columns().unwrap();
assert_eq!(left_columns.len(), right_columns.len());
@@ -436,6 +392,7 @@ fn assert_columnar_eq(
}
}
#[track_caller]
fn assert_column_eq<T: Copy + PartialOrd + Debug + Send + Sync + 'static>(
left: &Column<T>,
right: &Column<T>,
@@ -533,28 +490,36 @@ 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);
}
}
@@ -623,7 +588,7 @@ proptest! {
#[test]
fn test_single_columnar_builder_proptest(docs in columnar_docs_strategy()) {
let columnar = build_columnar(&docs[..]);
assert_eq!(columnar.num_rows() as usize, docs.len());
assert_eq!(columnar.num_docs() as usize, docs.len());
let mut expected_columns: HashMap<(&str, ColumnTypeCategory), HashMap<u32, Vec<&ColumnValue>> > = Default::default();
for (doc_id, doc_vals) in docs.iter().enumerate() {
for (col_name, col_val) in doc_vals {
@@ -663,54 +628,6 @@ proptest! {
}
}
// Same as `test_single_columnar_builder_proptest` but with a shuffling mapping.
proptest! {
#![proptest_config(ProptestConfig::with_cases(500))]
#[test]
fn test_single_columnar_builder_with_shuffle_proptest((docs, mapping) in columnar_docs_and_mapping_strategy()) {
let columnar = build_columnar_with_mapping(&docs[..], Some(&mapping));
assert_eq!(columnar.num_rows() as usize, docs.len());
let mut expected_columns: HashMap<(&str, ColumnTypeCategory), HashMap<u32, Vec<&ColumnValue>> > = Default::default();
for (doc_id, doc_vals) in docs.iter().enumerate() {
for (col_name, col_val) in doc_vals {
expected_columns
.entry((col_name, col_val.column_type_category()))
.or_default()
.entry(mapping[doc_id])
.or_default()
.push(col_val);
}
}
let column_list = columnar.list_columns().unwrap();
assert_eq!(expected_columns.len(), column_list.len());
for (column_name, column) in column_list {
let dynamic_column = column.open().unwrap();
let col_category: ColumnTypeCategory = dynamic_column.column_type().into();
let expected_col_values: &HashMap<u32, Vec<&ColumnValue>> = expected_columns.get(&(column_name.as_str(), col_category)).unwrap();
for _doc_id in 0..columnar.num_rows() {
match &dynamic_column {
DynamicColumn::Bool(col) =>
assert_column_values(col, expected_col_values),
DynamicColumn::I64(col) =>
assert_column_values(col, expected_col_values),
DynamicColumn::U64(col) =>
assert_column_values(col, expected_col_values),
DynamicColumn::F64(col) =>
assert_column_values(col, expected_col_values),
DynamicColumn::IpAddr(col) =>
assert_column_values(col, expected_col_values),
DynamicColumn::DateTime(col) =>
assert_column_values(col, expected_col_values),
DynamicColumn::Bytes(col) =>
assert_bytes_column_values(col, expected_col_values, false),
DynamicColumn::Str(col) =>
assert_bytes_column_values(col, expected_col_values, true),
}
}
}
}
}
// This tests create 2 or 3 random small columnar and attempts to merge them.
// It compares the resulting merged dataframe with what would have been obtained by building the
// dataframe from the concatenated rows to begin with.
@@ -726,7 +643,7 @@ 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 +669,7 @@ 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 +706,7 @@ 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);
}
@@ -798,6 +715,7 @@ fn test_columnar_merging_number_columns() {
// TODO test required_columns
// TODO document edge case: required_columns incompatible with values.
#[allow(clippy::type_complexity)]
fn columnar_docs_and_remap(
) -> impl Strategy<Value = (Vec<Vec<Vec<(&'static str, ColumnValue)>>>, Vec<RowAddr>)> {
proptest::collection::vec(columnar_docs_strategy(), 2..=3).prop_flat_map(
@@ -824,24 +742,68 @@ fn columnar_docs_and_remap(
proptest! {
#![proptest_config(ProptestConfig::with_cases(1000))]
#[test]
fn test_columnar_merge_and_remap_proptest((columnar_docs, shuffle_merge_order) in columnar_docs_and_remap()) {
let shuffled_rows: Vec<Vec<(&'static str, ColumnValue)>> = shuffle_merge_order.iter()
.map(|row_addr| columnar_docs[row_addr.segment_ord as usize][row_addr.row_id as usize].clone())
.collect();
let expected_merged_columnar = build_columnar(&shuffled_rows[..]);
let columnar_readers: Vec<ColumnarReader> = columnar_docs.iter()
.map(|docs| build_columnar(&docs[..]))
.collect::<Vec<_>>();
let columnar_readers_arr: Vec<&ColumnarReader> = columnar_readers.iter().collect();
let mut output: Vec<u8> = Vec::new();
let segment_num_rows: Vec<RowId> = columnar_docs.iter().map(|docs| docs.len() as RowId).collect();
let shuffle_merge_order = ShuffleMergeOrder::for_test(&segment_num_rows, shuffle_merge_order);
crate::merge_columnar(&columnar_readers_arr[..], &[], shuffle_merge_order.into(), &mut output).unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
assert_columnar_eq(&merged_columnar, &expected_merged_columnar, true);
fn test_columnar_merge_and_remap_proptest((columnar_docs, shuffle_merge_order) in
columnar_docs_and_remap()) {
test_columnar_merge_and_remap(columnar_docs, shuffle_merge_order);
}
}
fn test_columnar_merge_and_remap(
columnar_docs: Vec<Vec<Vec<(&'static str, ColumnValue)>>>,
shuffle_merge_order: Vec<RowAddr>,
) {
let shuffled_rows: Vec<Vec<(&'static str, ColumnValue)>> = shuffle_merge_order
.iter()
.map(|row_addr| {
columnar_docs[row_addr.segment_ord as usize][row_addr.row_id as usize].clone()
})
.collect();
let expected_merged_columnar = build_columnar(&shuffled_rows[..]);
let columnar_readers: Vec<ColumnarReader> = columnar_docs
.iter()
.map(|docs| build_columnar(&docs[..]))
.collect::<Vec<_>>();
let columnar_readers_ref: Vec<&ColumnarReader> = columnar_readers.iter().collect();
let mut output: Vec<u8> = Vec::new();
let segment_num_rows: Vec<RowId> = columnar_docs
.iter()
.map(|docs| docs.len() as RowId)
.collect();
let shuffle_merge_order = ShuffleMergeOrder::for_test(&segment_num_rows, shuffle_merge_order);
crate::merge_columnar(
&columnar_readers_ref[..],
&[],
shuffle_merge_order.into(),
&mut output,
)
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
assert_columnar_eq(&merged_columnar, &expected_merged_columnar, true);
}
#[test]
fn test_columnar_merge_and_remap_bug_1() {
let columnar_docs = vec![vec![
vec![
("c1", ColumnValue::Numerical(NumericalValue::U64(0))),
("c1", ColumnValue::Numerical(NumericalValue::U64(0))),
],
vec![],
]];
let shuffle_merge_order: Vec<RowAddr> = vec![
RowAddr {
segment_ord: 0,
row_id: 1,
},
RowAddr {
segment_ord: 0,
row_id: 0,
},
];
test_columnar_merge_and_remap(columnar_docs, shuffle_merge_order);
}
#[test]
fn test_columnar_merge_empty() {
let columnar_reader_1 = build_columnar(&[]);
@@ -858,7 +820,7 @@ fn test_columnar_merge_empty() {
)
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
assert_eq!(merged_columnar.num_rows(), 0);
assert_eq!(merged_columnar.num_docs(), 0);
assert_eq!(merged_columnar.num_columns(), 0);
}
@@ -884,7 +846,7 @@ fn test_columnar_merge_single_str_column() {
)
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
assert_eq!(merged_columnar.num_rows(), 1);
assert_eq!(merged_columnar.num_docs(), 1);
assert_eq!(merged_columnar.num_columns(), 1);
}
@@ -916,7 +878,7 @@ fn test_delete_decrease_cardinality() {
)
.unwrap();
let merged_columnar = ColumnarReader::open(output).unwrap();
assert_eq!(merged_columnar.num_rows(), 1);
assert_eq!(merged_columnar.num_docs(), 1);
assert_eq!(merged_columnar.num_columns(), 1);
let cols = merged_columnar.read_columns("c").unwrap();
assert_eq!(cols.len(), 1);

25
columnar/src/value.rs
View File

@@ -17,6 +17,31 @@ impl NumericalValue {
NumericalValue::F64(_) => NumericalType::F64,
}
}
/// Tries to normalize the numerical value in the following priorities:
/// i64, i64, f64
pub fn normalize(self) -> Self {
match self {
NumericalValue::U64(val) => {
if val <= i64::MAX as u64 {
NumericalValue::I64(val as i64)
} else {
NumericalValue::F64(val as f64)
}
}
NumericalValue::I64(val) => NumericalValue::I64(val),
NumericalValue::F64(val) => {
let fract = val.fract();
if fract == 0.0 && val >= i64::MIN as f64 && val <= i64::MAX as f64 {
NumericalValue::I64(val as i64)
} else if fract == 0.0 && val >= u64::MIN as f64 && val <= u64::MAX as f64 {
NumericalValue::U64(val as u64)
} else {
NumericalValue::F64(val)
}
}
}
}
}
impl From<u64> for NumericalValue {

7
common/Cargo.toml
View File

@@ -1,6 +1,6 @@
[package]
name = "tantivy-common"
version = "0.5.0"
version = "0.9.0"
authors = ["Paul Masurel <paul@quickwit.io>", "Pascal Seitz <pascal@quickwit.io>"]
license = "MIT"
edition = "2021"
@@ -9,16 +9,17 @@ documentation = "https://docs.rs/tantivy_common/"
homepage = "https://github.com/quickwit-oss/tantivy"
repository = "https://github.com/quickwit-oss/tantivy"
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
[dependencies]
byteorder = "1.4.3"
ownedbytes = { version= "0.5", path="../ownedbytes" }
ownedbytes = { version= "0.9", path="../ownedbytes" }
async-trait = "0.1"
time = { version = "0.3.10", features = ["serde-well-known"] }
serde = { version = "1.0.136", features = ["derive"] }
[dev-dependencies]
binggan = "0.14.0"
proptest = "1.0.0"
rand = "0.8.4"

95
common/benches/bench.rs
View File

@@ -1,39 +1,64 @@
#![feature(test)]
use binggan::{black_box, BenchRunner};
use rand::seq::IteratorRandom;
use rand::thread_rng;
use tantivy_common::{serialize_vint_u32, BitSet, TinySet};
extern crate test;
fn bench_vint() {
let mut runner = BenchRunner::new();
#[cfg(test)]
mod tests {
use rand::seq::IteratorRandom;
use rand::thread_rng;
use tantivy_common::serialize_vint_u32;
use test::Bencher;
let vals: Vec<u32> = (0..20_000).collect();
runner.bench_function("bench_vint", move |_| {
let mut out = 0u64;
for val in vals.iter().cloned() {
let mut buf = [0u8; 8];
serialize_vint_u32(val, &mut buf);
out += u64::from(buf[0]);
}
black_box(out);
});
#[bench]
fn bench_vint(b: &mut Bencher) {
let vals: Vec<u32> = (0..20_000).collect();
b.iter(|| {
let mut out = 0u64;
for val in vals.iter().cloned() {
let mut buf = [0u8; 8];
serialize_vint_u32(val, &mut buf);
out += u64::from(buf[0]);
}
out
});
}
#[bench]
fn bench_vint_rand(b: &mut Bencher) {
let vals: Vec<u32> = (0..20_000).choose_multiple(&mut thread_rng(), 100_000);
b.iter(|| {
let mut out = 0u64;
for val in vals.iter().cloned() {
let mut buf = [0u8; 8];
serialize_vint_u32(val, &mut buf);
out += u64::from(buf[0]);
}
out
});
}
let vals: Vec<u32> = (0..20_000).choose_multiple(&mut thread_rng(), 100_000);
runner.bench_function("bench_vint_rand", move |_| {
let mut out = 0u64;
for val in vals.iter().cloned() {
let mut buf = [0u8; 8];
serialize_vint_u32(val, &mut buf);
out += u64::from(buf[0]);
}
black_box(out);
});
}
fn bench_bitset() {
let mut runner = BenchRunner::new();
runner.bench_function("bench_tinyset_pop", move |_| {
let mut tinyset = TinySet::singleton(black_box(31u32));
tinyset.pop_lowest();
tinyset.pop_lowest();
tinyset.pop_lowest();
tinyset.pop_lowest();
tinyset.pop_lowest();
tinyset.pop_lowest();
black_box(tinyset);
});
let tiny_set = TinySet::empty().insert(10u32).insert(14u32).insert(21u32);
runner.bench_function("bench_tinyset_sum", move |_| {
assert_eq!(black_box(tiny_set).into_iter().sum::<u32>(), 45u32);
});
let v = [10u32, 14u32, 21u32];
runner.bench_function("bench_tinyarr_sum", move |_| {
black_box(v.iter().cloned().sum::<u32>());
});
runner.bench_function("bench_bitset_initialize", move |_| {
black_box(BitSet::with_max_value(1_000_000));
});
}
fn main() {
bench_vint();
bench_bitset();
}

43
common/src/bitset.rs
View File

@@ -1,6 +1,5 @@
use std::convert::TryInto;
use std::io::Write;
use std::{fmt, io, u64};
use std::{fmt, io};
use ownedbytes::OwnedBytes;
@@ -697,43 +696,3 @@ mod tests {
}
}
}
#[cfg(all(test, feature = "unstable"))]
mod bench {
use test;
use super::{BitSet, TinySet};
#[bench]
fn bench_tinyset_pop(b: &mut test::Bencher) {
b.iter(|| {
let mut tinyset = TinySet::singleton(test::black_box(31u32));
tinyset.pop_lowest();
tinyset.pop_lowest();
tinyset.pop_lowest();
tinyset.pop_lowest();
tinyset.pop_lowest();
tinyset.pop_lowest();
});
}
#[bench]
fn bench_tinyset_sum(b: &mut test::Bencher) {
let tiny_set = TinySet::empty().insert(10u32).insert(14u32).insert(21u32);
b.iter(|| {
assert_eq!(test::black_box(tiny_set).into_iter().sum::<u32>(), 45u32);
});
}
#[bench]
fn bench_tinyarr_sum(b: &mut test::Bencher) {
let v = [10u32, 14u32, 21u32];
b.iter(|| test::black_box(v).iter().cloned().sum::<u32>());
}
#[bench]
fn bench_bitset_initialize(b: &mut test::Bencher) {
b.iter(|| BitSet::with_max_value(1_000_000));
}
}

130
common/src/bounds.rs Normal file
View File

@@ -0,0 +1,130 @@
use std::io;
use std::ops::Bound;
#[derive(Clone, Debug)]
pub struct BoundsRange<T> {
pub lower_bound: Bound<T>,
pub upper_bound: Bound<T>,
}
impl<T> BoundsRange<T> {
pub fn new(lower_bound: Bound<T>, upper_bound: Bound<T>) -> Self {
BoundsRange {
lower_bound,
upper_bound,
}
}
pub fn is_unbounded(&self) -> bool {
matches!(self.lower_bound, Bound::Unbounded) && matches!(self.upper_bound, Bound::Unbounded)
}
pub fn map_bound<TTo>(&self, transform: impl Fn(&T) -> TTo) -> BoundsRange<TTo> {
BoundsRange {
lower_bound: map_bound(&self.lower_bound, &transform),
upper_bound: map_bound(&self.upper_bound, &transform),
}
}
pub fn map_bound_res<TTo, Err>(
&self,
transform: impl Fn(&T) -> Result<TTo, Err>,
) -> Result<BoundsRange<TTo>, Err> {
Ok(BoundsRange {
lower_bound: map_bound_res(&self.lower_bound, &transform)?,
upper_bound: map_bound_res(&self.upper_bound, &transform)?,
})
}
pub fn transform_inner<TTo>(
&self,
transform_lower: impl Fn(&T) -> TransformBound<TTo>,
transform_upper: impl Fn(&T) -> TransformBound<TTo>,
) -> BoundsRange<TTo> {
BoundsRange {
lower_bound: transform_bound_inner(&self.lower_bound, &transform_lower),
upper_bound: transform_bound_inner(&self.upper_bound, &transform_upper),
}
}
/// Returns the first set inner value
pub fn get_inner(&self) -> Option<&T> {
inner_bound(&self.lower_bound).or(inner_bound(&self.upper_bound))
}
}
pub enum TransformBound<T> {
/// Overwrite the bounds
NewBound(Bound<T>),
/// Use Existing bounds with new value
Existing(T),
}
/// Takes a bound and transforms the inner value into a new bound via a closure.
/// The bound variant may change by the value returned value from the closure.
pub fn transform_bound_inner_res<TFrom, TTo>(
bound: &Bound<TFrom>,
transform: impl Fn(&TFrom) -> io::Result<TransformBound<TTo>>,
) -> io::Result<Bound<TTo>> {
use self::Bound::*;
Ok(match bound {
Excluded(ref from_val) => match transform(from_val)? {
TransformBound::NewBound(new_val) => new_val,
TransformBound::Existing(new_val) => Excluded(new_val),
},
Included(ref from_val) => match transform(from_val)? {
TransformBound::NewBound(new_val) => new_val,
TransformBound::Existing(new_val) => Included(new_val),
},
Unbounded => Unbounded,
})
}
/// Takes a bound and transforms the inner value into a new bound via a closure.
/// The bound variant may change by the value returned value from the closure.
pub fn transform_bound_inner<TFrom, TTo>(
bound: &Bound<TFrom>,
transform: impl Fn(&TFrom) -> TransformBound<TTo>,
) -> Bound<TTo> {
use self::Bound::*;
match bound {
Excluded(ref from_val) => match transform(from_val) {
TransformBound::NewBound(new_val) => new_val,
TransformBound::Existing(new_val) => Excluded(new_val),
},
Included(ref from_val) => match transform(from_val) {
TransformBound::NewBound(new_val) => new_val,
TransformBound::Existing(new_val) => Included(new_val),
},
Unbounded => Unbounded,
}
}
/// Returns the inner value of a `Bound`
pub fn inner_bound<T>(val: &Bound<T>) -> Option<&T> {
match val {
Bound::Included(term) | Bound::Excluded(term) => Some(term),
Bound::Unbounded => None,
}
}
pub fn map_bound<TFrom, TTo>(
bound: &Bound<TFrom>,
transform: impl Fn(&TFrom) -> TTo,
) -> Bound<TTo> {
use self::Bound::*;
match bound {
Excluded(ref from_val) => Bound::Excluded(transform(from_val)),
Included(ref from_val) => Bound::Included(transform(from_val)),
Unbounded => Unbounded,
}
}
pub fn map_bound_res<TFrom, TTo, Err>(
bound: &Bound<TFrom>,
transform: impl Fn(&TFrom) -> Result<TTo, Err>,
) -> Result<Bound<TTo>, Err> {
use self::Bound::*;
Ok(match bound {
Excluded(ref from_val) => Excluded(transform(from_val)?),
Included(ref from_val) => Included(transform(from_val)?),
Unbounded => Unbounded,
})
}

22
common/src/datetime.rs
View File

@@ -1,11 +1,12 @@
#![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(
@@ -24,9 +25,6 @@ pub enum DateTimePrecision {
Nanoseconds,
}
#[deprecated(since = "0.20.0", note = "Use `DateTimePrecision` instead")]
pub type DatePrecision = DateTimePrecision;
/// A date/time value with nanoseconds precision.
///
/// This timestamp does not carry any explicit time zone information.
@@ -37,7 +35,7 @@ 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 +162,15 @@ 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))
}
}

88
common/src/file_slice.rs
View File

@@ -1,4 +1,6 @@
use std::fs::File;
use std::ops::{Deref, Range, RangeBounds};
use std::path::Path;
use std::sync::Arc;
use std::{fmt, io};
@@ -32,6 +34,62 @@ 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 +125,30 @@ 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`.
@@ -96,6 +178,12 @@ fn combine_ranges<R: RangeBounds<usize>>(orig_range: Range<usize>, rel_range: R)
}
impl FileSlice {
/// Creates a FileSlice from a path.
pub fn open(path: &Path) -> io::Result<FileSlice> {
let wrap_file = WrapFile::new(File::open(path)?)?;
Ok(FileSlice::new(Arc::new(wrap_file)))
}
/// Wraps a FileHandle.
pub fn new(file_handle: Arc<dyn FileHandle>) -> Self {
let num_bytes = file_handle.len();

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 {

144
common/src/json_path_writer.rs Normal file
View File

@@ -0,0 +1,144 @@
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]) };
/// Separates the json path and the value in
/// a JSON term binary representation.
pub const JSON_END_OF_PATH: u8 = 0u8;
pub const JSON_END_OF_PATH_STR: &str =
unsafe { std::str::from_utf8_unchecked(&[JSON_END_OF_PATH]) };
/// 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 with_expand_dots(expand_dots: bool) -> Self {
JsonPathWriter {
path: String::new(),
indices: Vec::new(),
expand_dots,
}
}
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.indices.len() > 1 {
self.path.push(JSON_PATH_SEGMENT_SEP as char);
}
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())
};
}
}
/// Set the end of JSON path marker.
#[inline]
pub fn set_end(&mut self) {
self.path.push_str(JSON_END_OF_PATH_STR);
}
/// 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.set_expand_dots(false);
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");
}
#[test]
fn test_json_path_expand_dots_enabled_pop_segment() {
let mut json_writer = JsonPathWriter::with_expand_dots(true);
json_writer.push("hello");
assert_eq!(json_writer.as_str(), "hello");
json_writer.push("color.hue");
assert_eq!(json_writer.as_str(), "hello\x01color\x01hue");
json_writer.pop();
assert_eq!(json_writer.as_str(), "hello");
}
}

16
common/src/lib.rs
View File

@@ -5,19 +5,20 @@ use std::ops::Deref;
pub use byteorder::LittleEndian as Endianness;
mod bitset;
pub mod bounds;
mod byte_count;
mod datetime;
pub mod file_slice;
mod group_by;
pub mod json_path_writer;
mod serialize;
mod vint;
mod writer;
pub use bitset::*;
pub use byte_count::ByteCount;
#[allow(deprecated)]
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 +117,7 @@ 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;
@@ -128,11 +130,11 @@ pub fn replace_in_place(needle: u8, replacement: u8, bytes: &mut [u8]) {
}
#[cfg(test)]
pub mod test {
pub(crate) mod test {
use proptest::prelude::*;
use super::{f64_to_u64, i64_to_u64, u64_to_f64, u64_to_i64, BinarySerializable, FixedSize};
use super::{f64_to_u64, i64_to_u64, u64_to_f64, u64_to_i64};
fn test_i64_converter_helper(val: i64) {
assert_eq!(u64_to_i64(i64_to_u64(val)), val);
@@ -142,12 +144,6 @@ pub mod test {
assert_eq!(u64_to_f64(f64_to_u64(val)), val);
}
pub fn fixed_size_test<O: BinarySerializable + FixedSize + Default>() {
let mut buffer = Vec::new();
O::default().serialize(&mut buffer).unwrap();
assert_eq!(buffer.len(), O::SIZE_IN_BYTES);
}
proptest! {
#[test]
fn test_f64_converter_monotonicity_proptest((left, right) in (proptest::num::f64::NORMAL, proptest::num::f64::NORMAL)) {

49
common/src/serialize.rs
View File

@@ -1,3 +1,4 @@
use std::borrow::Cow;
use std::io::{Read, Write};
use std::{fmt, io};
@@ -73,14 +74,14 @@ impl FixedSize for () {
impl<T: BinarySerializable> BinarySerializable for Vec<T> {
fn serialize<W: Write + ?Sized>(&self, writer: &mut W) -> io::Result<()> {
VInt(self.len() as u64).serialize(writer)?;
BinarySerializable::serialize(&VInt(self.len() as u64), writer)?;
for it in self {
it.serialize(writer)?;
}
Ok(())
}
fn deserialize<R: Read>(reader: &mut R) -> io::Result<Vec<T>> {
let num_items = VInt::deserialize(reader)?.val();
let num_items = <VInt as BinarySerializable>::deserialize(reader)?.val();
let mut items: Vec<T> = Vec::with_capacity(num_items as usize);
for _ in 0..num_items {
let item = T::deserialize(reader)?;
@@ -235,12 +236,12 @@ impl FixedSize for bool {
impl BinarySerializable for String {
fn serialize<W: Write + ?Sized>(&self, writer: &mut W) -> io::Result<()> {
let data: &[u8] = self.as_bytes();
VInt(data.len() as u64).serialize(writer)?;
BinarySerializable::serialize(&VInt(data.len() as u64), writer)?;
writer.write_all(data)
}
fn deserialize<R: Read>(reader: &mut R) -> io::Result<String> {
let string_length = VInt::deserialize(reader)?.val() as usize;
let string_length = <VInt as BinarySerializable>::deserialize(reader)?.val() as usize;
let mut result = String::with_capacity(string_length);
reader
.take(string_length as u64)
@@ -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();
BinarySerializable::serialize(&VInt(data.len() as u64), writer)?;
writer.write_all(data)
}
fn deserialize<R: Read>(reader: &mut R) -> io::Result<Cow<'a, str>> {
let string_length = <VInt as BinarySerializable>::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<()> {
BinarySerializable::serialize(&VInt(self.len() as u64), writer)?;
for it in self.iter() {
BinarySerializable::serialize(it, writer)?;
}
Ok(())
}
fn deserialize<R: Read>(reader: &mut R) -> io::Result<Cow<'a, [u8]>> {
let num_items = <VInt as BinarySerializable>::deserialize(reader)?.val();
let mut items: Vec<u8> = Vec::with_capacity(num_items as usize);
for _ in 0..num_items {
let item = <u8 as BinarySerializable>::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();

2
common/src/vint.rs
View File

@@ -151,7 +151,7 @@ pub fn read_u32_vint_no_advance(data: &[u8]) -> (u32, usize) {
(result, vlen)
}
/// Write a `u32` as a vint payload.
pub fn write_u32_vint<W: io::Write>(val: u32, writer: &mut W) -> io::Result<()> {
pub fn write_u32_vint<W: io::Write + ?Sized>(val: u32, writer: &mut W) -> io::Result<()> {
let mut buf = [0u8; 8];
let data = serialize_vint_u32(val, &mut buf);
writer.write_all(data)

2
common/src/writer.rs
View File

@@ -87,7 +87,7 @@ impl<W: TerminatingWrite> TerminatingWrite for BufWriter<W> {
}
}
impl<'a> TerminatingWrite for &'a mut Vec<u8> {
impl TerminatingWrite for &mut Vec<u8> {
fn terminate_ref(&mut self, _a: AntiCallToken) -> io::Result<()> {
self.flush()
}

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

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 30 KiB

6
doc/src/avant-propos.md
View File

@@ -2,7 +2,7 @@
> Tantivy is a **search** engine **library** for Rust.
If you are familiar with Lucene, it's an excellent approximation to consider tantivy as Lucene for rust. tantivy is heavily inspired by Lucene's design and
If you are familiar with Lucene, it's an excellent approximation to consider tantivy as Lucene for Rust. Tantivy is heavily inspired by Lucene's design and
they both have the same scope and targeted use cases.
If you are not familiar with Lucene, let's break down our little tagline.
@@ -17,7 +17,7 @@ relevancy, collapsing, highlighting, spatial search.
experience. But keep in mind this is just a toolbox.
Which bring us to the second keyword...
- **Library** means that you will have to write code. tantivy is not an *all-in-one* server solution like elastic search for instance.
- **Library** means that you will have to write code. Tantivy is not an *all-in-one* server solution like Elasticsearch for instance.
Sometimes a functionality will not be available in tantivy because it is too
specific to your use case. By design, tantivy should make it possible to extend
@@ -31,4 +31,4 @@ relevancy, collapsing, highlighting, spatial search.
index from a different format.
Tantivy exposes a lot of low level API to do all of these things.

2
doc/src/basis.md
View File

@@ -11,7 +11,7 @@ directory shipped with tantivy is the `MmapDirectory`.
While this design has some downsides, this greatly simplifies the source code of
tantivy. Caching is also entirely delegated to the OS.
`tantivy` works entirely (or almost) by directly reading the datastructures as they are laid on disk. As a result, the act of opening an indexing does not involve loading different datastructures from the disk into random access memory : starting a process, opening an index, and performing your first query can typically be done in a matter of milliseconds.
Tantivy works entirely (or almost) by directly reading the datastructures as they are laid on disk. As a result, the act of opening an indexing does not involve loading different datastructures from the disk into random access memory : starting a process, opening an index, and performing your first query can typically be done in a matter of milliseconds.
This is an interesting property for a command line search engine, or for some multi-tenant log search engine : spawning a new process for each new query can be a perfectly sensible solution in some use case.

11
doc/src/index_sorting.md
View File

@@ -7,6 +7,11 @@
- [Other](#other)
- [Usage](#usage)
# Index Sorting has been removed!
More infos here:
https://github.com/quickwit-oss/tantivy/issues/2352
# Index Sorting
Tantivy allows you to sort the index according to a property.
@@ -26,13 +31,13 @@ Compression ratio is mainly affected on the fast field of the sorted property, e
When data is presorted by a field and search queries request sorting by the same field, we can leverage the natural order of the documents.
E.g. if the data is sorted by timestamp and want the top n newest docs containing a term, we can simply leveraging the order of the docids.
Note: Tantivy 0.16 does not do this optimization yet.
Note: tantivy 0.16 does not do this optimization yet.
### Pruning
Let's say we want all documents and want to apply the filter `>= 2010-08-11`. When the data is sorted, we could make a lookup in the fast field to find the docid range and use this as the filter.
Note: Tantivy 0.16 does not do this optimization yet.
Note: tantivy 0.16 does not do this optimization yet.
### Other?
@@ -40,7 +45,7 @@ In principle there are many algorithms possible that exploit the monotonically i
## Usage
The index sorting can be configured setting [`sort_by_field`](https://github.com/quickwit-oss/tantivy/blob/000d76b11a139a84b16b9b95060a1c93e8b9851c/src/core/index_meta.rs#L238) on `IndexSettings` and passing it to a `IndexBuilder`. As of Tantivy 0.16 only fast fields are allowed to be used.
The index sorting can be configured setting [`sort_by_field`](https://github.com/quickwit-oss/tantivy/blob/000d76b11a139a84b16b9b95060a1c93e8b9851c/src/core/index_meta.rs#L238) on `IndexSettings` and passing it to a `IndexBuilder`. As of tantivy 0.16 only fast fields are allowed to be used.
```rust
let settings = IndexSettings {

2
doc/src/json.md
View File

@@ -39,7 +39,7 @@ Its representation is done by separating segments by a unicode char `\x01`, and
- `value`: The value representation is just the regular Value representation.
This representation is designed to align the natural sort of Terms with the lexicographical sort
of their binary representation (Tantivy's dictionary (whether fst or sstable) is sorted and does prefix encoding).
of their binary representation (tantivy's dictionary (whether fst or sstable) is sorted and does prefix encoding).
In the example above, the terms will be sorted as

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(())
}

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