Skip to content

feat: soft delete optimize #12339

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 6 commits into from
Jun 9, 2023
Merged

feat: soft delete optimize #12339

merged 6 commits into from
Jun 9, 2023

Conversation

@fudongyingluck
Copy link
Contributor

@fudongyingluck fudongyingluck commented May 30, 2023 •
edited
Loading

as es issuse metioned when soft delete enable the numDeletesToMerge function is very time consume part. As the following picture show, there is actually calculate duplicate in there.
image

This change want to reuse the numDeletesToMerge result to reduce the time used
@fudongyingluck
Copy link
Contributor Author

fudongyingluck commented Jun 5, 2023 •
edited
Loading

This is the esrally result. The command is likeesrally race --track=http_logs --target-hosts=*:9201 --pipeline=benchmark-only --offline --user-tag=softdelete:baseline --challenge=update

| Metric | Task | Baseline | Contender | Diff | Unit | Diff % | |--------------------------------------------------------------:|-------:|----------------:|----------------:|------------:|-------:|---------:| | Cumulative indexing time of primary shards | | 515.49 | 504.15 | -11.3398 | min | -2.20% | | Min cumulative indexing time across primary shard | | 0 | 0 | 0 | min | 0.00% | | Median cumulative indexing time across primary shard | | 17.7529 | 17.9699 | 0.2169 | min | +1.22% | | Max cumulative indexing time across primary shard | | 404.723 | 393.369 | -11.3536 | min | -2.81% | | Cumulative indexing throttle time of primary shards | | 0 | 0 | 0 | min | 0.00% | | Min cumulative indexing throttle time across primary shard | | 0 | 0 | 0 | min | 0.00% | | Median cumulative indexing throttle time across primary shard | | 0 | 0 | 0 | min | 0.00% | | Max cumulative indexing throttle time across primary shard | | 0 | 0 | 0 | min | 0.00% | | Cumulative merge time of primary shards | | 133.81 | 127.489 | -6.32017 | min | -4.72% | | Cumulative merge count of primary shards | | 173 | 172 | -1 | | -0.58% | | Min cumulative merge time across primary shard | | 0 | 0 | 0 | min | 0.00% | | Median cumulative merge time across primary shard | | 2.61536 | 2.96084 | 0.34548 | min | +13.21% | | Max cumulative merge time across primary shard | | 118.648 | 110.923 | -7.7245 | min | -6.51% | | Cumulative merge throttle time of primary shards | | 57.0305 | 55.1042 | -1.92633 | min | -3.38% | | Min cumulative merge throttle time across primary shard | | 0 | 0 | 0 | min | 0.00% | | Median cumulative merge throttle time across primary shard | | 0.215533 | 0.307242 | 0.09171 | min | +42.55% | | Max cumulative merge throttle time across primary shard | | 55.2842 | 53.1749 | -2.10932 | min | -3.82% | | Cumulative refresh time of primary shards | | 21.5803 | 20.5713 | -1.009 | min | -4.68% | | Cumulative refresh count of primary shards | | 668 | 674 | 6 | | +0.90% | | Min cumulative refresh time across primary shard | | 0 | 0 | 0 | min | 0.00% | | Median cumulative refresh time across primary shard | | 0.542333 | 0.508642 | -0.03369 | min | -6.21% | | Max cumulative refresh time across primary shard | | 18.1363 | 17.4352 | -0.70113 | min | -3.87% | | Cumulative flush time of primary shards | | 9.37332 | 10.4646 | 1.09132 | min | +11.64% | | Cumulative flush count of primary shards | | 63 | 64 | 1 | | +1.59% | | Min cumulative flush time across primary shard | | 0.00296667 | 0.0001 | -0.00287 | min | -96.63% | | Median cumulative flush time across primary shard | | 0.0971583 | 0.0769667 | -0.02019 | min | -20.78% | | Max cumulative flush time across primary shard | | 8.6855 | 9.83638 | 1.15088 | min | +13.25% | | Total Young Gen GC time | | 1070.97 | 1065.08 | -5.889 | s | -0.55% | | Total Young Gen GC count | | 8254 | 8187 | -67 | | -0.81% | | Total Old Gen GC time | | 0.586 | 0 | -0.586 | s | -100.00% | | Total Old Gen GC count | | 3 | 0 | -3 | | -100.00% | | Store size | | 17.0535 | 16.9082 | -0.14531 | GB | -0.85% | | Translog size | | 4.09782e-07 | 4.09782e-07 | 0 | GB | 0.00% | | Heap used for segments | | 0 | 0 | 0 | MB | 0.00% | | Heap used for doc values | | 0 | 0 | 0 | MB | 0.00% | | Heap used for terms | | 0 | 0 | 0 | MB | 0.00% | | Heap used for norms | | 0 | 0 | 0 | MB | 0.00% | | Heap used for points | | 0 | 0 | 0 | MB | 0.00% | | Heap used for stored fields | | 0 | 0 | 0 | MB | 0.00% | | Segment count | | 158 | 163 | 5 | | +3.16% | | Total Ingest Pipeline count | | 0 | 0 | 0 | | 0.00% | | Total Ingest Pipeline time | | 0 | 0 | 0 | ms | 0.00% | | Total Ingest Pipeline failed | | 0 | 0 | 0 | | 0.00% | | Min Throughput | update | 23056.7 | 23029.1 | -27.5735 | docs/s | -0.12% | | Mean Throughput | update | 29585.3 | 29794 | 208.699 | docs/s | +0.71% | | Median Throughput | update | 28990.2 | 29011.7 | 21.4849 | docs/s | +0.07% | | Max Throughput | update | 36131.5 | 36197.3 | 65.8749 | docs/s | +0.18% | | 50th percentile latency | update | 1421.89 | 1437.74 | 15.8507 | ms | +1.11% | | 90th percentile latency | update | 2410.13 | 2420.23 | 10.1008 | ms | +0.42% | | 99th percentile latency | update | 7076.3 | 7045.81 | -30.4936 | ms | -0.43% | | 99.9th percentile latency | update | 11033.5 | 10406.9 | -626.525 | ms | -5.68% | | 99.99th percentile latency | update | 14342.9 | 13304.1 | -1038.85 | ms | -7.24% | | 100th percentile latency | update | 21652.9 | 21399.9 | -253 | ms | -1.17% | | 50th percentile service time | update | 1421.89 | 1437.74 | 15.8507 | ms | +1.11% | | 90th percentile service time | update | 2410.13 | 2420.23 | 10.1008 | ms | +0.42% | | 99th percentile service time | update | 7076.3 | 7045.81 | -30.4936 | ms | -0.43% | | 99.9th percentile service time | update | 11033.5 | 10406.9 | -626.525 | ms | -5.68% | | 99.99th percentile service time | update | 14342.9 | 13304.1 | -1038.85 | ms | -7.24% | | 100th percentile service time | update | 21652.9 | 21399.9 | -253 | ms | -1.17% | | error rate | update | 0 | 0 | 0 | % | 0.00% |```
@fudongyingluck
Copy link
Contributor Author

fudongyingluck commented Jun 6, 2023 •
edited
Loading

lucene benchmark result, python3.10 src/python/localrun.py -source wikimediumall

 BrowseDateSSDVFacets 1.54 (11.4%) 1.46 (16.1%) -5.2% ( -29% - 25%) 0.242 OrHighMedDayTaxoFacets 5.38 (5.6%) 5.24 (5.0%) -2.6% ( -12% - 8%) 0.127 PKLookup 279.48 (3.0%) 273.06 (3.1%) -2.3% ( -8% - 3%) 0.018 MedTermDayTaxoFacets 35.78 (2.2%) 35.10 (1.8%) -1.9% ( -5% - 2%) 0.002 BrowseDateTaxoFacets 7.23 (22.3%) 7.10 (23.8%) -1.8% ( -39% - 56%) 0.802 HighIntervalsOrdered 10.59 (8.9%) 10.42 (8.6%) -1.6% ( -17% - 17%) 0.568 BrowseDayOfYearTaxoFacets 7.30 (21.8%) 7.19 (23.9%) -1.6% ( -38% - 56%) 0.829 LowIntervalsOrdered 4.55 (7.1%) 4.48 (7.1%) -1.5% ( -14% - 13%) 0.495 MedIntervalsOrdered 6.90 (8.1%) 6.81 (7.3%) -1.4% ( -15% - 15%) 0.565 Fuzzy2 118.84 (2.2%) 117.28 (2.5%) -1.3% ( -5% - 3%) 0.078 Respell 82.74 (3.1%) 81.79 (4.0%) -1.2% ( -7% - 6%) 0.308 HighTermMonthSort 3093.29 (5.8%) 3057.85 (6.7%) -1.1% ( -12% - 12%) 0.562 BrowseRandomLabelTaxoFacets 6.40 (38.8%) 6.33 (40.9%) -1.1% ( -58% - 128%) 0.930 HighTerm 791.45 (5.1%) 783.46 (4.7%) -1.0% ( -10% - 9%) 0.517 HighPhrase 30.44 (2.3%) 30.16 (2.2%) -0.9% ( -5% - 3%) 0.190 Fuzzy1 108.68 (2.7%) 107.67 (3.6%) -0.9% ( -7% - 5%) 0.359 OrHighNotMed 320.94 (6.6%) 318.02 (5.3%) -0.9% ( -11% - 11%) 0.629 OrNotHighHigh 468.36 (5.3%) 464.33 (4.2%) -0.9% ( -9% - 9%) 0.568 LowSloppyPhrase 34.97 (4.1%) 34.69 (4.2%) -0.8% ( -8% - 7%) 0.534 MedPhrase 242.27 (2.5%) 240.32 (1.9%) -0.8% ( -5% - 3%) 0.248 AndHighMed 77.34 (6.0%) 76.76 (5.7%) -0.8% ( -11% - 11%) 0.686 OrHighNotLow 744.00 (6.5%) 738.66 (5.8%) -0.7% ( -12% - 12%) 0.711 AndHighLow 586.58 (3.5%) 582.51 (4.2%) -0.7% ( -8% - 7%) 0.573 HighSloppyPhrase 3.91 (4.5%) 3.89 (3.9%) -0.6% ( -8% - 8%) 0.670 MedSpanNear 37.46 (2.1%) 37.26 (2.5%) -0.6% ( -5% - 4%) 0.441 LowPhrase 153.02 (2.2%) 152.17 (2.1%) -0.6% ( -4% - 3%) 0.417 OrNotHighLow 1030.00 (3.2%) 1025.40 (3.5%) -0.4% ( -6% - 6%) 0.675 Wildcard 35.75 (3.2%) 35.59 (4.5%) -0.4% ( -7% - 7%) 0.723 MedTerm 761.12 (5.8%) 757.86 (6.0%) -0.4% ( -11% - 12%) 0.819 AndHighHigh 22.42 (6.5%) 22.33 (5.7%) -0.4% ( -11% - 12%) 0.830 LowTerm 689.41 (3.9%) 686.65 (4.6%) -0.4% ( -8% - 8%) 0.768 HighSpanNear 2.47 (4.2%) 2.46 (5.0%) -0.4% ( -9% - 9%) 0.789 AndHighHighDayTaxoFacets 7.97 (1.6%) 7.94 (1.9%) -0.4% ( -3% - 3%) 0.522 OrHighNotHigh 352.84 (6.6%) 351.68 (4.9%) -0.3% ( -11% - 11%) 0.859 AndHighMedDayTaxoFacets 48.80 (1.6%) 48.65 (2.3%) -0.3% ( -4% - 3%) 0.611 MedSloppyPhrase 24.12 (2.4%) 24.04 (2.5%) -0.3% ( -5% - 4%) 0.684 OrHighMed 37.82 (6.3%) 37.72 (5.5%) -0.3% ( -11% - 12%) 0.891 HighTermTitleBDVSort 7.13 (8.7%) 7.11 (8.1%) -0.2% ( -15% - 18%) 0.927 LowSpanNear 26.13 (3.7%) 26.08 (3.3%) -0.2% ( -6% - 7%) 0.866 Prefix3 408.84 (1.3%) 408.62 (2.1%) -0.1% ( -3% - 3%) 0.923 OrNotHighMed 469.82 (4.2%) 470.09 (3.6%) 0.1% ( -7% - 8%) 0.963 HighTermTitleSort 105.40 (2.8%) 105.54 (4.6%) 0.1% ( -7% - 7%) 0.914 OrHighHigh 13.48 (5.0%) 13.51 (4.7%) 0.2% ( -9% - 10%) 0.905 TermDTSort 241.07 (3.6%) 242.46 (4.9%) 0.6% ( -7% - 9%) 0.671 OrHighLow 235.33 (5.1%) 237.04 (5.2%) 0.7% ( -9% - 11%) 0.655 BrowseRandomLabelSSDVFacets 4.96 (3.8%) 5.00 (11.6%) 0.9% ( -13% - 16%) 0.746 HighTermDayOfYearSort 290.03 (3.5%) 292.75 (3.7%) 0.9% ( -6% - 8%) 0.408 IntNRQ 52.81 (18.2%) 54.52 (15.7%) 3.2% ( -25% - 45%) 0.546 BrowseDayOfYearSSDVFacets 6.11 (4.2%) 6.32 (10.6%) 3.4% ( -10% - 19%) 0.186 BrowseMonthTaxoFacets 9.69 (33.2%) 10.17 (34.1%) 5.0% ( -46% - 108%) 0.641 BrowseMonthSSDVFacets 6.35 (2.9%) 6.68 (10.3%) 5.2% ( -7% - 18%) 0.030

and the part of cpu profile result

PERCENT CPU SAMPLES STACK 4.71% 49024 org.apache.lucene.codecs.lucene90.Lucene90DocValuesProducer$VaryingBPVReader#getLongValue() 3.88% 40306 org.apache.lucene.facet.sortedset.SortedSetDocValuesFacetCounts#countOneSegmentNHLD() 3.50% 36379 java.nio.Buffer#scope() 3.35% 34860 org.apache.lucene.queries.intervals.OrderedIntervalsSource$OrderedIntervalIterator#nextInterval() 3.18% 33041 org.apache.lucene.facet.taxonomy.FastTaxonomyFacetCounts#countAll() CPU merged search profile for baseline: PERCENT CPU SAMPLES STACK 6.19% 63449 org.apache.lucene.codecs.lucene90.Lucene90DocValuesProducer$VaryingBPVReader#getLongValue() 3.63% 37149 org.apache.lucene.facet.taxonomy.FastTaxonomyFacetCounts#countAll() 3.58% 36660 org.apache.lucene.facet.sortedset.SortedSetDocValuesFacetCounts#countOneSegmentNHLD() 3.46% 35483 org.apache.lucene.queries.intervals.OrderedIntervalsSource$OrderedIntervalIterator#nextInterval() 3.19% 32707 org.apache.lucene.util.packed.DirectReader$DirectPackedReader20#get()```
@luyuncheng luyuncheng mentioned this pull request Jun 6, 2023
Closed
jpountz
jpountz reviewed Jun 8, 2023
View reviewed changes
Copy link
Contributor

@jpountz jpountz left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No computing the number of deletes twice makes sense to me. What I'm not super happy about is that it's a bit trappy for merge policies, they need to be very careful to call the right methods to not compute it twice. E.g. I believe that LogMergePolicy needs a similar fix to the one that you made to TieredMergePolicy.

As a potential alternative, I wonder if IndexWriter could use a wrapper around the MergeContext which would memoize the number of deletes of every SegmentCommitInfo in a hash map when calling the merge policy. This way, if you happen to call numDeletesToMerge twice on the same SegmentCommitInfo, the second one would be served from the cache?
@fudongyingluck
Copy link
Contributor Author

fudongyingluck commented Jun 9, 2023 •
edited
Loading

Thanks @jpountz for your time. I really think this is a good idea, much better than I do. I wonder if the newest commit implement your idea. I'll commit some test cases and post benchmark result later.
@fudongyingluck fudongyingluck reopened this Jun 9, 2023
@fudongyingluck
Copy link
Contributor Author

fudongyingluck commented Jun 9, 2023 •
edited
Loading

lucene benchmark result, python3.10 src/python/localrun.py -source wikimediumall

 TaskQPS baseline StdDevQPS my_modified_version StdDev Pct diff p-value BrowseDayOfYearSSDVFacets 6.29 (7.5%) 6.16 (9.4%) -2.1% ( -17% - 15%) 0.428 HighTermTitleBDVSort 10.68 (8.0%) 10.51 (4.3%) -1.6% ( -12% - 11%) 0.442 IntNRQ 47.94 (1.8%) 47.23 (8.1%) -1.5% ( -11% - 8%) 0.422 OrHighMedDayTaxoFacets 6.09 (6.9%) 6.02 (5.2%) -1.1% ( -12% - 11%) 0.563 BrowseMonthSSDVFacets 6.56 (8.0%) 6.50 (8.3%) -0.9% ( -15% - 16%) 0.719 MedIntervalsOrdered 27.91 (4.9%) 27.67 (5.1%) -0.9% ( -10% - 9%) 0.579 HighSloppyPhrase 3.61 (4.3%) 3.58 (4.4%) -0.7% ( -9% - 8%) 0.596 BrowseDayOfYearTaxoFacets 7.06 (6.1%) 7.04 (5.2%) -0.3% ( -10% - 11%) 0.850 OrHighLow 351.52 (2.9%) 350.34 (3.4%) -0.3% ( -6% - 6%) 0.737 OrHighHigh 20.17 (3.4%) 20.11 (3.7%) -0.3% ( -7% - 7%) 0.776 MedSloppyPhrase 4.99 (4.0%) 4.97 (4.7%) -0.3% ( -8% - 8%) 0.822 BrowseRandomLabelTaxoFacets 6.20 (7.1%) 6.19 (5.3%) -0.2% ( -11% - 13%) 0.914 OrHighMed 106.18 (3.5%) 105.99 (3.4%) -0.2% ( -6% - 6%) 0.866 BrowseDateTaxoFacets 7.00 (5.9%) 6.99 (4.8%) -0.1% ( -10% - 11%) 0.947 Fuzzy1 88.03 (3.3%) 87.96 (1.9%) -0.1% ( -5% - 5%) 0.925 HighIntervalsOrdered 12.93 (4.5%) 12.92 (4.4%) -0.0% ( -8% - 9%) 0.979 Prefix3 224.61 (2.8%) 224.69 (2.3%) 0.0% ( -4% - 5%) 0.966 LowSloppyPhrase 22.07 (4.2%) 22.10 (4.1%) 0.1% ( -7% - 8%) 0.928 OrNotHighMed 403.85 (2.4%) 404.49 (2.9%) 0.2% ( -4% - 5%) 0.851 OrHighNotLow 468.49 (5.1%) 469.62 (4.9%) 0.2% ( -9% - 10%) 0.879 HighTermMonthSort 3512.17 (6.9%) 3523.19 (7.5%) 0.3% ( -13% - 15%) 0.890 OrHighNotMed 532.57 (4.3%) 534.39 (3.6%) 0.3% ( -7% - 8%) 0.786 MedTerm 1019.27 (4.5%) 1022.80 (4.3%) 0.3% ( -8% - 9%) 0.805 AndHighHighDayTaxoFacets 7.35 (2.8%) 7.38 (1.7%) 0.4% ( -4% - 5%) 0.633 AndHighHigh 32.60 (3.8%) 32.72 (4.2%) 0.4% ( -7% - 8%) 0.776 AndHighLow 662.12 (3.5%) 664.62 (3.9%) 0.4% ( -6% - 7%) 0.745 Fuzzy2 91.31 (4.0%) 91.66 (2.3%) 0.4% ( -5% - 6%) 0.709 OrNotHighHigh 675.72 (3.3%) 679.20 (3.6%) 0.5% ( -6% - 7%) 0.636 AndHighMed 96.86 (5.9%) 97.41 (6.3%) 0.6% ( -11% - 13%) 0.771 PKLookup 281.91 (3.6%) 283.54 (2.6%) 0.6% ( -5% - 7%) 0.566 Wildcard 183.21 (4.9%) 184.35 (2.8%) 0.6% ( -6% - 8%) 0.619 OrHighNotHigh 500.76 (3.4%) 504.24 (3.4%) 0.7% ( -5% - 7%) 0.513 LowPhrase 183.76 (2.7%) 185.24 (2.5%) 0.8% ( -4% - 6%) 0.326 LowTerm 732.82 (3.0%) 738.99 (3.0%) 0.8% ( -4% - 6%) 0.368 MedTermDayTaxoFacets 38.20 (2.9%) 38.53 (1.9%) 0.9% ( -3% - 5%) 0.273 MedPhrase 85.79 (2.5%) 86.54 (2.3%) 0.9% ( -3% - 5%) 0.250 HighTerm 678.62 (4.8%) 684.64 (4.4%) 0.9% ( -7% - 10%) 0.544 AndHighMedDayTaxoFacets 34.42 (2.5%) 34.73 (1.5%) 0.9% ( -2% - 4%) 0.164 LowIntervalsOrdered 16.93 (3.6%) 17.09 (3.2%) 1.0% ( -5% - 7%) 0.373 MedSpanNear 25.65 (3.4%) 25.89 (4.3%) 1.0% ( -6% - 9%) 0.440 HighSpanNear 9.16 (3.9%) 9.25 (4.7%) 1.0% ( -7% - 9%) 0.473 HighPhrase 136.81 (2.8%) 138.28 (2.8%) 1.1% ( -4% - 6%) 0.231 Respell 67.25 (4.6%) 68.00 (3.4%) 1.1% ( -6% - 9%) 0.377 BrowseRandomLabelSSDVFacets 5.26 (7.4%) 5.32 (7.2%) 1.1% ( -12% - 16%) 0.627 LowSpanNear 7.90 (3.7%) 7.99 (3.9%) 1.1% ( -6% - 9%) 0.347 HighTermDayOfYearSort 400.43 (2.6%) 405.41 (2.7%) 1.2% ( -3% - 6%) 0.137 OrNotHighLow 818.63 (3.1%) 828.86 (3.0%) 1.2% ( -4% - 7%) 0.199 HighTermTitleSort 62.96 (2.5%) 63.77 (3.1%) 1.3% ( -4% - 7%) 0.149 BrowseMonthTaxoFacets 10.34 (33.0%) 10.47 (33.7%) 1.3% ( -49% - 101%) 0.902 TermDTSort 239.19 (4.0%) 242.85 (6.9%) 1.5% ( -8% - 12%) 0.390 BrowseDateSSDVFacets 1.49 (12.4%) 1.54 (11.4%) 3.4% ( -18% - 31%) 0.362```
jpountz
jpountz approved these changes Jun 9, 2023
View reviewed changes
Copy link
Contributor

@jpountz jpountz left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks great. Can you add a CHANGES entry under 9.7?
lucene/core/src/java/org/apache/lucene/index/CachingMergeContext.java Outdated
* a wrapper of IndexWriter MergeContext. Try to cache the {@link
* #numDeletesToMerge(SegmentCommitInfo)} result in merge phase, to avoid duplicate calculation
*/
public class CachingMergeContext implements MergePolicy.MergeContext {
Copy link
Contributor

@jpountz jpountz Jun 9, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you make it pkg-private instead of public?
Copy link
Contributor Author

@fudongyingluck fudongyingluck Jun 9, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, I've done this ~
@jpountz
Copy link
Contributor

jpountz commented Jun 9, 2023

I'll note that there is still room for improvement, as this change doesn't cache the number of soft deletes across calls to findMerges. But the fix is so simple and contained, this looks to me like a good case of progress over perfection.
@jpountz jpountz added this to the 9.7.0 milestone Jun 9, 2023
@jpountz jpountz merged commit 2934899 into apache:main Jun 9, 2023
jpountz pushed a commit that referenced this pull request Jun 9, 2023
@fudongyingluck @jpountz
feat: soft delete optimize (#12339)
1107aa2
hiteshk25 pushed a commit to cowpaths/lucene that referenced this pull request Jul 18, 2023
@patsonluk @romseygeek
@javanna @uschindler @original-brownbear @mkhludnev @MarcusSorealheis @benwtrent @kaivalnp @tang-hi @zhaih @gsmiller @JerryChin @AndreyBozhko @ChrisHegarty @rmuir @dantuzi @alessandrobenedetti @gashutos @JarvisCraft @tflobbe @iverase @fudongyingluck @cfournie @gf2121 @jpountz @eliaporciani @donnerpeter
Merge branch_9x into fs/branch_9x (up to `ccf4b198ec328095d45d2746189…
a7f4238 
…dc8ca633e8bcf`) (#20) * Add next minor version 9.7.0 * Fix SynonymQuery equals implementation (apache#12260) The term member of TermAndBoost used to be a Term instance and became a BytesRef with apache#11941, which means its equals impl won't take the field name into account. The SynonymQuery equals impl needs to be updated accordingly to take the field into account as well, otherwise synonym queries with same term and boost across different fields are equal which is a bug. * Fix MMapDirectory documentation for Java 20 (apache#12265) * Don't generate stacktrace in CollectionTerminatedException (apache#12270) CollectionTerminatedException is always caught and never exposed to users so there's no point in filling in a stack-trace for it. * add missing changelog entry for apache#12260 * Add missing author to changelog entry for apache#12220 * Make query timeout members final in ExitableDirectoryReader (apache#12274) There's a couple of places in the Exitable wrapper classes where queryTimeout is set within the constructor and never modified. This commit makes such members final. * Update javadocs for QueryTimeout (apache#12272) QueryTimeout was introduced together with ExitableDirectoryReader but is now also optionally set to the IndexSearcher to wrap the bulk scorer with a TimeLimitingBulkScorer. Its javadocs needs updating. * Make TimeExceededException members final (apache#12271) TimeExceededException has three members that are set within its constructor and never modified. They can be made final. * DOAP changes for release 9.6.0 * Add back-compat indices for 9.6.0 * `ToParentBlockJoinQuery` Explain Support Score Mode (apache#12245) (apache#12283) * `ToParentBlockJoinQuery` Explain Support Score Mode --------- Co-authored-by: Marcus <marcuseagan@gmail.com> * Simplify SliceExecutor and QueueSizeBasedExecutor (apache#12285) The only behaviour that QueueSizeBasedExecutor overrides from SliceExecutor is when to execute on the caller thread. There is no need to override the whole invokeAll method for that. Instead, this commit introduces a shouldExecuteOnCallerThread method that can be overridden. * [Backport] GITHUB-11838 Add api to allow concurrent query rewrite (apache#12197) * GITHUB-11838 Change API to allow concurrent query rewrite (apache#11840) Replace Query#rewrite(IndexReader) with Query#rewrite(IndexSearcher) Co-authored-by: Patrick Zhai <zhaih@users.noreply.github.com> Co-authored-by: Adrien Grand <jpountz@gmail.com> Backport of apache#11840 Changes from original: - Query keeps `rewrite(IndexReader)`, but it is now deprecated - VirtualMethod is used to correct delegate to the overridden methods - The changes to `RewriteMethod` type classes are reverted, this increased the backwards compatibility impact. ------------------------------ ### Description Issue: apache#11838 #### Updated Proposal * Change signature of rewrite to `rewrite(IndexSearcher)` * How did I migrate the usage: * Use Intellij to do preliminary refactoring for me * For test usage, use searcher whenever is available, otherwise create one using `newSearcher(reader)` * For very few non-test classes which doesn't have IndexSearcher available but called rewrite, create a searcher using `new IndexSearcher(reader)`, tried my best to avoid creating it recurrently (Especially in `FieldQuery`) * For queries who have implemented the rewrite and uses some part of reader's functionality, use shortcut method when possible, otherwise pull out the reader from indexSearcher. * Backport: Concurrent rewrite for KnnVectorQuery (apache#12160) (apache#12288) * Concurrent rewrite for KnnVectorQuery (apache#12160) - Reduce overhead of non-concurrent search by preserving original execution - Improve readability by factoring into separate functions --------- Co-authored-by: Kaival Parikh <kaivalp2000@gmail.com> * adjusting for backport --------- Co-authored-by: Kaival Parikh <46070017+kaivalnp@users.noreply.github.com> Co-authored-by: Kaival Parikh <kaivalp2000@gmail.com> * toposort use iterator to avoid stackoverflow (apache#12286) Co-authored-by: tangdonghai <tangdonghai@meituan.com> # Conflicts: #	lucene/CHANGES.txt * Fix test to compile with Java 11 after backport of apache#12286 * Update Javadoc for topoSortStates method after apache#12286 (apache#12292) * Optimize HNSW diversity calculation (apache#12235) * Minor cleanup and improvements to DaciukMihovAutomatonBuilder (apache#12305) * GITHUB-12291: Skip blank lines from stopwords list. (apache#12299) * Wrap Query rewrite backwards layer with AccessController (apache#12308) * Make sure APIJAR reproduces with different timezone (unfortunately java encodes the date using local timezone) (apache#12315) * Add multi-thread searchability to OnHeapHnswGraph (apache#12257) * Fix backport error * [MINOR] Update javadoc in Query class (apache#12233) - add a few missing full stops - update wording in the description of Query#equals method * [Backport] Integrate the Incubating Panama Vector API apache#12311 (apache#12327) Leverage accelerated vector hardware instructions in Vector Search. Lucene already has a mechanism that enables the use of non-final JDK APIs, currently used for the Previewing Pamana Foreign API. This change expands this mechanism to include the Incubating Pamana Vector API. When the jdk.incubator.vector module is present at run time the Panamaized version of the low-level primitives used by Vector Search is enabled. If not present, the default scalar version of these low-level primitives is used (as it was previously). Currently, we're only targeting support for JDK 20. A subsequent PR should evaluate JDK 21. --------- Co-authored-by: Uwe Schindler <uschindler@apache.org> Co-authored-by: Robert Muir <rmuir@apache.org> * Parallelize knn query rewrite across slices rather than segments (apache#12325) The concurrent query rewrite for knn vectory query introduced with apache#12160 requests one thread per segment to the executor. To align this with the IndexSearcher parallel behaviour, we should rather parallelize across slices. Also, we can reuse the same slice executor instance that the index searcher already holds, in that way we are using a QueueSizeBasedExecutor when a thread pool executor is provided. * Optimize ConjunctionDISI.createConjunction (apache#12328) This method is showing up as a little hot when profiling some queries. Almost all the time spent in this method is just burnt on ceremony around stream indirections that don't inline. Moving this to iterators, simplifying the check for same doc id and also saving one iteration (for the min cost) makes this method far cheaper and easier to read. * Update changes to be correct with ARM (it is called NEON there) * GH#12321: Marked DaciukMihovAutomatonBuilder as deprecated (apache#12332) Preparing to reduce visibility of this class in a future release * add BitSet.clear() (apache#12268) # Conflicts: #	lucene/CHANGES.txt * Clenaup and update changes and synchronize with 9.x * Update TestVectorUtilProviders.java (apache#12338) * Don't generate stacktrace for TimeExceededException (apache#12335) The exception is package private and never rethrown, we can avoid generating a stacktrace for it. * Introduced the Word2VecSynonymFilter (apache#12169) Co-authored-by: Alessandro Benedetti <a.benedetti@sease.io> * Word2VecSynonymFilter constructor null check (apache#12169) * Use thread-safe search version of HnswGraphSearcher (apache#12246) Addressing comment received in the PR apache#12246 * Word2VecSynonymProvider to use standard Integer max value for hnsw searches (apache#12235) We observed this change was not ported previously from main in an old cherry-pick * Fix searchafter high latency when after value is out of range for segment (apache#12334) * Make memory fence in `ByteBufferGuard` explicit (apache#12290) * Add "direct to binary" option for DaciukMihovAutomatonBuilder and use it in TermInSetQuery#visit (apache#12320) * Add updateDocuments API which accept a query (reopen) (apache#12346) * GITHUB#11350: Handle backward compatibility when merging segments with different FieldInfo This commits restores Lucene 9's ability to handle indices created with Lucene 8 where there are discrepancies in FieldInfos, such as different IndexOptions * [Tessellator] Improve the checks that validate the diagonal between two polygon nodes (apache#12353) # Conflicts: #	lucene/CHANGES.txt * feat: soft delete optimize (apache#12339) * Better paging when random reads go backwards (apache#12357) When reading data from outside the buffer, BufferedIndexInput always resets its buffer to start at the new read position. If we are reading backwards (for example, using an OffHeapFSTStore for a terms dictionary) then this can have the effect of re-reading the same data over and over again. This commit changes BufferedIndexInput to use paging when reading backwards, so that if we ask for a byte that is before the current buffer, we read a block of data of bufferSize that ends at the previous buffer start. Fixes apache#12356 * Work around SecurityManager issues during initialization of vector api (JDK-8309727) (apache#12362) * Restrict GraphTokenStreamFiniteStrings#articulationPointsRecurse recursion depth (apache#12249) * Implement MMapDirectory with Java 21 Project Panama Preview API (apache#12294) Backport incl JDK21 apijar file with java.util.Objects regenerated * remove relic in apijar folder caused by vector additions * Speed up IndexedDISI Sparse #AdvanceExactWithinBlock for tiny step advance (apache#12324) * Add checks in KNNVectorField / KNNVectorQuery to only allow non-null, non-empty and finite vectors (apache#12281) --------- Co-authored-by: Uwe Schindler <uschindler@apache.org> * Implement VectorUtilProvider with Java 21 Project Panama Vector API (apache#12363) (apache#12365) This commit enables the Panama Vector API for Java 21. The version of VectorUtilPanamaProvider for Java 21 is identical to that of Java 20. As such, there is no specific 21 version - the Java 20 version will be loaded from the MRJAR. * Add CHANGES.txt for apache#12334 Honor after value for skipping documents even if queue is not full for PagingFieldCollector (apache#12368) Signed-off-by: gashutos <gashutos@amazon.com> * Move TermAndBoost back to its original location. (apache#12366) PR apache#12169 accidentally moved the `TermAndBoost` class to a different location, which would break custom sub-classes of `QueryBuilder`. This commit moves it back to its original location. * GITHUB-12252: Add function queries for computing similarity scores between knn vectors (apache#12253) Co-authored-by: Alessandro Benedetti <a.benedetti@sease.io> * hunspell (minor): reduce allocations when processing compound rules (apache#12316) (cherry picked from commit a454388) * hunspell (minor): reduce allocations when reading the dictionary's morphological data (apache#12323) there can be many entries with morph data, so we'd better avoid compiling and matching regexes and even stream allocation (cherry picked from commit 4bf1b94) * TestHunspell: reduce the flakiness probability (apache#12351) * TestHunspell: reduce the flakiness probability We need to check how the timeout interacts with custom exception-throwing checkCanceled. The default timeout seems not enough for some CI agents, so let's increase it. Co-authored-by: Dawid Weiss <dawid.weiss@gmail.com> (cherry picked from commit 5b63a18) * This allows VectorUtilProvider tests to be executed although hardware may not fully support vectorization or if C2 is not enabled (apache#12376) --------- Signed-off-by: gashutos <gashutos@amazon.com> Co-authored-by: Alan Woodward <romseygeek@apache.org> Co-authored-by: Luca Cavanna <javanna@apache.org> Co-authored-by: Uwe Schindler <uschindler@apache.org> Co-authored-by: Armin Braun <me@obrown.io> Co-authored-by: Mikhail Khludnev <mkhludnev@users.noreply.github.com> Co-authored-by: Marcus <marcuseagan@gmail.com> Co-authored-by: Benjamin Trent <ben.w.trent@gmail.com> Co-authored-by: Kaival Parikh <46070017+kaivalnp@users.noreply.github.com> Co-authored-by: Kaival Parikh <kaivalp2000@gmail.com> Co-authored-by: tang donghai <tangdhcs@gmail.com> Co-authored-by: Patrick Zhai <zhaih@users.noreply.github.com> Co-authored-by: Greg Miller <gsmiller@gmail.com> Co-authored-by: Jerry Chin <metrxqin@gmail.com> Co-authored-by: Patrick Zhai <zhai7631@gmail.com> Co-authored-by: Andrey Bozhko <andybozhko@gmail.com> Co-authored-by: Chris Hegarty <62058229+ChrisHegarty@users.noreply.github.com> Co-authored-by: Robert Muir <rmuir@apache.org> Co-authored-by: Jonathan Ellis <jbellis@datastax.com> Co-authored-by: Daniele Antuzi <daniele.antuzi@gmail.com> Co-authored-by: Alessandro Benedetti <a.benedetti@sease.io> Co-authored-by: Chaitanya Gohel <104654647+gashutos@users.noreply.github.com> Co-authored-by: Petr Portnov | PROgrm_JARvis <pportnov@ozon.ru> Co-authored-by: Tomas Eduardo Fernandez Lobbe <tflobbe@apache.org> Co-authored-by: Ignacio Vera <ivera@apache.org> Co-authored-by: fudongying <30896830+fudongyingluck@users.noreply.github.com> Co-authored-by: Chris Fournier <chris.fournier@shopify.com> Co-authored-by: gf2121 <52390227+gf2121@users.noreply.github.com> Co-authored-by: Adrien Grand <jpountz@gmail.com> Co-authored-by: Elia Porciani <e.porciani@sease.io> Co-authored-by: Peter Gromov <peter@jetbrains.com>
@dnhatn dnhatn mentioned this pull request May 1, 2024
Closed
@cgejian cgejian mentioned this pull request Feb 18, 2025
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

2 participants

@fudongyingluck @jpountz