gsmiller opened a new pull request, #954:
URL: https://github.com/apache/lucene/pull/954
This PR is to migrate the facets module to using the newly-added
`SortedSetDocValues#docValueCount()` for iteration, as described in
LUCENE-10603. It doesn't attempt to move all `SSDV` iteration, just the facets
module.
Benchmark results show a potentially small win, and no regressions, so I
think we should move forward with this.
```
TaskQPS baseline StdDevQPS candidate
StdDev Pct diff p-value
TermDTSort 99.36 (13.9%) 93.27
(12.9%) -6.1% ( -28% - 24%) 0.149
HighTermDayOfYearSort 99.67 (13.9%) 96.80
(12.2%) -2.9% ( -25% - 27%) 0.487
BrowseDayOfYearSSDVFacets 14.07 (18.4%) 13.82
(13.8%) -1.8% ( -28% - 37%) 0.729
HighTermTitleBDVSort 126.78 (21.9%) 124.79
(25.4%) -1.6% ( -40% - 58%) 0.835
IntNRQ 73.42 (4.7%) 72.57
(5.1%) -1.2% ( -10% - 9%) 0.454
OrHighMed 99.39 (3.5%) 98.39
(3.2%) -1.0% ( -7% - 5%) 0.345
OrHighNotMed 1295.86 (3.2%) 1285.26
(5.0%) -0.8% ( -8% - 7%) 0.535
OrHighHigh 46.15 (3.1%) 45.78
(3.0%) -0.8% ( -6% - 5%) 0.400
BrowseMonthSSDVFacets 16.26 (15.1%) 16.13
(9.7%) -0.8% ( -22% - 28%) 0.848
OrHighLow 970.97 (2.8%) 964.35
(1.9%) -0.7% ( -5% - 4%) 0.375
OrNotHighMed 945.22 (3.4%) 939.06
(4.1%) -0.7% ( -7% - 7%) 0.582
MedTerm 2116.21 (5.1%) 2103.04
(4.5%) -0.6% ( -9% - 9%) 0.684
PKLookup 169.76 (3.4%) 168.71
(3.8%) -0.6% ( -7% - 6%) 0.588
AndHighHigh 44.04 (3.0%) 43.78
(5.5%) -0.6% ( -8% - 8%) 0.677
MedIntervalsOrdered 10.35 (5.8%) 10.31
(5.2%) -0.4% ( -10% - 11%) 0.820
OrNotHighHigh 1077.87 (4.1%) 1074.45
(5.0%) -0.3% ( -9% - 9%) 0.827
HighTerm 2923.10 (4.7%) 2914.62
(4.3%) -0.3% ( -8% - 9%) 0.838
LowTerm 1969.85 (4.9%) 1964.63
(5.6%) -0.3% ( -10% - 10%) 0.873
MedSpanNear 59.53 (2.6%) 59.38
(3.1%) -0.2% ( -5% - 5%) 0.784
HighIntervalsOrdered 12.23 (8.2%) 12.20
(7.6%) -0.2% ( -14% - 16%) 0.920
HighSpanNear 5.30 (2.7%) 5.29
(3.2%) -0.1% ( -5% - 5%) 0.902
OrNotHighLow 1213.60 (2.8%) 1212.64
(2.8%) -0.1% ( -5% - 5%) 0.928
LowSloppyPhrase 24.51 (3.3%) 24.49
(3.3%) -0.1% ( -6% - 6%) 0.953
OrHighMedDayTaxoFacets 12.99 (4.9%) 12.98
(6.2%) -0.1% ( -10% - 11%) 0.974
MedTermDayTaxoFacets 23.69 (4.8%) 23.68
(4.1%) -0.1% ( -8% - 9%) 0.971
LowIntervalsOrdered 107.55 (5.3%) 107.51
(3.8%) -0.0% ( -8% - 9%) 0.980
OrHighNotLow 1064.18 (4.6%) 1064.82
(5.6%) 0.1% ( -9% - 10%) 0.970
LowSpanNear 190.49 (3.1%) 190.62
(3.9%) 0.1% ( -6% - 7%) 0.951
AndHighMedDayTaxoFacets 39.56 (2.1%) 39.60
(1.6%) 0.1% ( -3% - 3%) 0.868
MedPhrase 379.28 (2.1%) 379.69
(2.5%) 0.1% ( -4% - 4%) 0.883
HighPhrase 223.12 (2.5%) 223.61
(2.8%) 0.2% ( -4% - 5%) 0.795
HighTermMonthSort 121.98 (16.5%) 122.28
(13.8%) 0.2% ( -25% - 36%) 0.959
LowPhrase 66.70 (2.8%) 66.89
(3.9%) 0.3% ( -6% - 7%) 0.792
Fuzzy1 93.42 (1.8%) 93.69
(1.2%) 0.3% ( -2% - 3%) 0.556
Respell 53.91 (1.8%) 54.07
(1.4%) 0.3% ( -2% - 3%) 0.552
MedSloppyPhrase 16.33 (3.2%) 16.38
(3.3%) 0.3% ( -6% - 7%) 0.763
AndHighMed 90.72 (3.2%) 91.05
(4.0%) 0.4% ( -6% - 7%) 0.753
HighSloppyPhrase 32.08 (4.7%) 32.21
(4.4%) 0.4% ( -8% - 9%) 0.774
OrHighNotHigh 895.72 (4.9%) 899.79
(5.1%) 0.5% ( -9% - 10%) 0.773
AndHighLow 588.99 (2.5%) 591.85
(2.0%) 0.5% ( -3% - 5%) 0.497
Fuzzy2 19.51 (1.8%) 19.61
(1.1%) 0.5% ( -2% - 3%) 0.285
AndHighHighDayTaxoFacets 13.45 (2.7%) 13.53
(2.3%) 0.5% ( -4% - 5%) 0.508
BrowseRandomLabelSSDVFacets 10.00 (5.2%) 10.08
(4.9%) 0.8% ( -8% - 11%) 0.608
BrowseRandomLabelTaxoFacets 17.77 (16.3%) 18.16
(13.7%) 2.2% ( -23% - 38%) 0.638
BrowseMonthTaxoFacets 27.62 (23.3%) 28.42
(18.9%) 2.9% ( -31% - 58%) 0.665
BrowseDateTaxoFacets 20.98 (18.5%) 21.63
(17.7%) 3.1% ( -27% - 48%) 0.590
BrowseDayOfYearTaxoFacets 20.96 (18.7%) 21.63
(18.1%) 3.2% ( -28% - 49%) 0.584
Wildcard 56.56 (8.9%) 58.60
(6.0%) 3.6% ( -10% - 20%) 0.133
Prefix3 521.60 (12.1%) 548.29
(11.9%) 5.1% ( -16% - 33%) 0.178
BrowseDateSSDVFacets 2.63 (11.6%) 2.82
(12.1%) 7.2% ( -14% - 35%) 0.054
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
