zacharymorn commented on code in PR #1018:
URL: https://github.com/apache/lucene/pull/1018#discussion_r922641944
##########
lucene/core/src/java/org/apache/lucene/search/BooleanWeight.java:
##########
@@ -191,6 +191,69 @@ public long cost() {
// or null if it is not applicable
// pkg-private for forcing use of BooleanScorer in tests
BulkScorer optionalBulkScorer(LeafReaderContext context) throws IOException {
+ if (scoreMode == ScoreMode.TOP_SCORES) {
+ if (query.getMinimumNumberShouldMatch() > 1 || weightedClauses.size() >
2) {
+ return null;
+ }
+
+ List<ScorerSupplier> optional = new ArrayList<>();
+ for (WeightedBooleanClause wc : weightedClauses) {
+ Weight w = wc.weight;
+ BooleanClause c = wc.clause;
+ if (c.getOccur() != Occur.SHOULD) {
+ continue;
+ }
+ ScorerSupplier scorer = w.scorerSupplier(context);
+ if (scorer != null) {
+ optional.add(scorer);
+ }
+ }
+
+ if (optional.size() <= 1) {
+ return null;
+ }
+
+ List<Scorer> optionalScorers = new ArrayList<>();
+ for (ScorerSupplier ss : optional) {
+ optionalScorers.add(ss.get(Long.MAX_VALUE));
+ }
+
+ return new BulkScorer() {
Review Comment:
Thanks for the suggestion! I gave that a try and it did work, but it would
reduce the performance boost for OrHighMed from around 110+% to 70+%, most
likely due to the extra logic inside `DefaultBulkScorer`. I guess my preference
would be to use the anonymous bulk scorer to maintain the performance
advantage, but I'm also good with using `DefaultBulkScorer` if reducing
potentially duplicated code and keeping things consistent are preferred?
```
TaskQPS baseline StdDevQPS
my_modified_version StdDev Pct diff p-value
TermBGroup1M1P 55.89 (7.1%) 54.01
(6.1%) -3.4% ( -15% - 10%) 0.108
TermDateFacets 34.46 (5.8%) 33.58
(5.0%) -2.5% ( -12% - 8%) 0.138
AndHighOrMedMed 90.90 (5.6%) 88.59
(4.6%) -2.5% ( -12% - 8%) 0.115
BrowseDayOfYearSSDVFacets 28.63 (12.5%) 28.01
(14.5%) -2.2% ( -25% - 28%) 0.612
MedTermDayTaxoFacets 79.74 (5.1%) 78.04
(4.3%) -2.1% ( -10% - 7%) 0.150
TermGroup100 36.28 (3.5%) 35.54
(3.1%) -2.1% ( -8% - 4%) 0.050
TermBGroup1M 30.37 (3.7%) 29.87
(3.8%) -1.6% ( -8% - 6%) 0.165
TermGroup10K 41.33 (3.2%) 40.66
(3.3%) -1.6% ( -7% - 5%) 0.117
PKLookup 330.84 (5.1%) 326.20
(4.3%) -1.4% ( -10% - 8%) 0.349
SloppyPhrase 13.56 (2.8%) 13.39
(2.5%) -1.2% ( -6% - 4%) 0.139
TermGroup1M 39.76 (3.2%) 39.32
(3.2%) -1.1% ( -7% - 5%) 0.272
AndMedOrHighHigh 88.13 (5.5%) 87.22
(4.4%) -1.0% ( -10% - 9%) 0.511
BrowseDateSSDVFacets 4.17 (29.0%) 4.13
(29.0%) -0.8% ( -45% - 80%) 0.933
SpanNear 169.70 (2.6%) 168.59
(2.0%) -0.7% ( -5% - 4%) 0.369
Fuzzy2 83.59 (2.4%) 83.12
(2.2%) -0.6% ( -5% - 4%) 0.442
Respell 96.22 (3.1%) 95.85
(2.6%) -0.4% ( -5% - 5%) 0.672
IntervalsOrdered 23.02 (4.3%) 22.94
(4.2%) -0.3% ( -8% - 8%) 0.799
Wildcard 231.51 (4.5%) 230.74
(5.0%) -0.3% ( -9% - 9%) 0.827
AndHighMed 143.73 (5.8%) 143.48
(4.0%) -0.2% ( -9% - 10%) 0.914
AndHighHighDayTaxoFacets 54.80 (1.3%) 54.71
(1.5%) -0.2% ( -2% - 2%) 0.717
Fuzzy1 154.41 (2.8%) 154.44
(1.9%) 0.0% ( -4% - 4%) 0.981
BrowseMonthSSDVFacets 28.06 (10.4%) 28.06
(13.3%) 0.0% ( -21% - 26%) 0.995
OrHighMedDayTaxoFacets 7.38 (5.7%) 7.39
(5.4%) 0.0% ( -10% - 11%) 0.981
AndHighMedDayTaxoFacets 134.39 (2.0%) 134.59
(1.9%) 0.2% ( -3% - 4%) 0.809
Phrase 38.80 (1.8%) 38.86
(2.5%) 0.2% ( -4% - 4%) 0.823
TermMonthSort 357.58 (5.2%) 359.31
(6.7%) 0.5% ( -10% - 13%) 0.801
TermTitleSort 274.32 (5.1%) 275.77
(6.8%) 0.5% ( -10% - 13%) 0.781
TermDayOfYearSort 259.44 (2.5%) 261.32
(5.9%) 0.7% ( -7% - 9%) 0.615
TermDTSort 208.81 (2.6%) 210.65
(5.8%) 0.9% ( -7% - 9%) 0.532
Term 2535.91 (3.4%) 2562.20
(3.1%) 1.0% ( -5% - 7%) 0.317
AndHighHigh 32.15 (3.3%) 32.55
(3.1%) 1.2% ( -5% - 7%) 0.231
BrowseMonthTaxoFacets 32.02 (34.2%) 32.57
(35.8%) 1.7% ( -50% - 109%) 0.876
Prefix3 404.00 (9.7%) 411.80
(8.5%) 1.9% ( -14% - 22%) 0.505
IntNRQ 86.95 (9.5%) 88.74
(8.7%) 2.1% ( -14% - 22%) 0.475
BrowseRandomLabelSSDVFacets 19.82 (7.6%) 20.34
(9.2%) 2.7% ( -13% - 21%) 0.318
BrowseDayOfYearTaxoFacets 28.80 (33.0%) 31.25
(33.8%) 8.5% ( -43% - 112%) 0.420
BrowseDateTaxoFacets 28.68 (33.1%) 31.15
(33.9%) 8.6% ( -43% - 113%) 0.415
BrowseRandomLabelTaxoFacets 28.81 (45.1%) 31.60
(47.4%) 9.7% ( -57% - 186%) 0.508
OrHighHigh 23.12 (3.5%) 29.67
(6.8%) 28.3% ( 17% - 39%) 0.000
OrHighMed 103.75 (3.7%) 180.03
(11.5%) 73.5% ( 56% - 92%) 0.000
```
```
TaskQPS baseline StdDevQPS
my_modified_version StdDev Pct diff p-value
BrowseDateSSDVFacets 4.27 (27.4%) 3.92
(26.3%) -8.2% ( -48% - 62%) 0.333
BrowseRandomLabelTaxoFacets 32.05 (49.5%) 30.51
(47.1%) -4.8% ( -67% - 181%) 0.753
BrowseDateTaxoFacets 30.22 (32.6%) 29.15
(32.4%) -3.6% ( -51% - 91%) 0.730
BrowseDayOfYearTaxoFacets 30.27 (32.4%) 29.24
(32.4%) -3.4% ( -51% - 90%) 0.738
TermDateFacets 34.74 (4.2%) 33.81
(5.9%) -2.7% ( -12% - 7%) 0.101
Prefix3 475.23 (5.7%) 462.80
(5.7%) -2.6% ( -13% - 9%) 0.146
TermBGroup1M1P 55.99 (7.5%) 54.91
(8.1%) -1.9% ( -16% - 14%) 0.434
MedTermDayTaxoFacets 77.78 (3.6%) 76.33
(4.7%) -1.9% ( -9% - 6%) 0.160
BrowseMonthTaxoFacets 30.33 (37.8%) 29.78
(35.6%) -1.8% ( -54% - 115%) 0.877
PKLookup 332.94 (5.3%) 327.77
(5.3%) -1.6% ( -11% - 9%) 0.351
BrowseRandomLabelSSDVFacets 21.30 (9.6%) 21.01
(8.8%) -1.3% ( -18% - 18%) 0.648
AndHighMedDayTaxoFacets 140.47 (1.7%) 138.62
(4.1%) -1.3% ( -6% - 4%) 0.186
IntervalsOrdered 130.34 (5.8%) 128.98
(6.8%) -1.0% ( -12% - 12%) 0.602
SloppyPhrase 97.23 (7.7%) 96.30
(6.1%) -1.0% ( -13% - 13%) 0.667
AndMedOrHighHigh 35.81 (4.0%) 35.48
(3.3%) -0.9% ( -7% - 6%) 0.431
TermGroup100 36.34 (4.1%) 36.02
(3.9%) -0.9% ( -8% - 7%) 0.481
AndHighOrMedMed 92.43 (6.3%) 91.64
(5.2%) -0.9% ( -11% - 11%) 0.635
AndHighMed 151.18 (5.3%) 150.06
(5.0%) -0.7% ( -10% - 10%) 0.651
AndHighHighDayTaxoFacets 13.90 (2.1%) 13.82
(3.0%) -0.6% ( -5% - 4%) 0.461
Phrase 53.96 (4.9%) 53.66
(4.6%) -0.5% ( -9% - 9%) 0.716
Wildcard 382.71 (4.6%) 380.67
(4.5%) -0.5% ( -9% - 8%) 0.710
TermDTSort 142.16 (3.3%) 141.51
(3.2%) -0.5% ( -6% - 6%) 0.659
Term 3700.77 (3.7%) 3686.00
(3.5%) -0.4% ( -7% - 7%) 0.726
TermBGroup1M 36.19 (4.7%) 36.07
(3.5%) -0.3% ( -8% - 8%) 0.808
TermTitleSort 191.76 (6.3%) 191.17
(7.4%) -0.3% ( -13% - 14%) 0.888
TermDayOfYearSort 211.60 (3.5%) 210.97
(4.6%) -0.3% ( -8% - 8%) 0.819
TermMonthSort 224.12 (6.1%) 223.53
(7.6%) -0.3% ( -13% - 14%) 0.902
TermGroup1M 23.75 (3.9%) 23.72
(3.0%) -0.1% ( -6% - 7%) 0.924
OrHighMedDayTaxoFacets 15.65 (5.1%) 15.64
(5.3%) -0.0% ( -9% - 10%) 0.985
IntNRQ 110.02 (0.4%) 109.99
(0.4%) -0.0% ( 0% - 0%) 0.823
Fuzzy1 164.30 (1.3%) 164.42
(2.0%) 0.1% ( -3% - 3%) 0.889
Fuzzy2 83.64 (1.4%) 83.85
(2.0%) 0.2% ( -3% - 3%) 0.649
AndHighHigh 44.91 (3.5%) 45.09
(3.7%) 0.4% ( -6% - 7%) 0.724
Respell 117.19 (2.2%) 117.69
(2.7%) 0.4% ( -4% - 5%) 0.585
SpanNear 28.79 (2.9%) 29.02
(3.4%) 0.8% ( -5% - 7%) 0.426
TermGroup10K 43.33 (3.8%) 43.73
(2.4%) 0.9% ( -5% - 7%) 0.357
BrowseDayOfYearSSDVFacets 26.67 (12.7%) 26.94
(8.8%) 1.0% ( -18% - 25%) 0.767
BrowseMonthSSDVFacets 27.30 (8.6%) 28.47
(10.1%) 4.3% ( -13% - 25%) 0.149
OrHighHigh 25.20 (4.7%) 39.24
(4.8%) 55.7% ( 44% - 68%) 0.000
OrHighMed 103.99 (5.1%) 184.11
(5.5%) 77.0% ( 63% - 92%) 0.000
```
```
TaskQPS baseline StdDevQPS
my_modified_version StdDev Pct diff p-value
BrowseMonthSSDVFacets 29.26 (10.2%) 28.24
(5.9%) -3.5% ( -17% - 13%) 0.185
BrowseRandomLabelSSDVFacets 20.62 (6.1%) 20.36
(7.4%) -1.2% ( -13% - 13%) 0.565
Prefix3 475.09 (6.1%) 470.79
(6.7%) -0.9% ( -12% - 12%) 0.655
IntervalsOrdered 13.21 (3.1%) 13.12
(2.9%) -0.7% ( -6% - 5%) 0.461
TermDateFacets 48.80 (6.2%) 48.56
(6.2%) -0.5% ( -12% - 12%) 0.800
PKLookup 329.55 (4.9%) 328.09
(3.9%) -0.4% ( -8% - 8%) 0.750
AndHighOrMedMed 117.25 (3.9%) 116.74
(4.9%) -0.4% ( -8% - 8%) 0.759
MedTermDayTaxoFacets 78.41 (5.5%) 78.16
(5.5%) -0.3% ( -10% - 11%) 0.855
Phrase 232.22 (2.8%) 231.49
(2.3%) -0.3% ( -5% - 4%) 0.694
TermTitleSort 224.25 (6.4%) 223.78
(6.3%) -0.2% ( -12% - 13%) 0.916
TermBGroup1M 39.69 (3.2%) 39.60
(3.9%) -0.2% ( -6% - 7%) 0.852
TermMonthSort 277.13 (6.3%) 276.61
(6.4%) -0.2% ( -12% - 13%) 0.925
TermDayOfYearSort 259.60 (3.2%) 259.15
(3.6%) -0.2% ( -6% - 6%) 0.872
AndHighMed 185.96 (3.9%) 185.74
(5.3%) -0.1% ( -8% - 9%) 0.936
TermGroup100 51.51 (3.3%) 51.52
(3.8%) 0.0% ( -6% - 7%) 0.976
TermGroup1M 39.83 (2.7%) 39.85
(3.4%) 0.1% ( -5% - 6%) 0.957
TermGroup10K 41.38 (2.9%) 41.40
(3.1%) 0.1% ( -5% - 6%) 0.950
Wildcard 238.58 (4.1%) 238.92
(3.4%) 0.1% ( -7% - 7%) 0.904
SloppyPhrase 4.75 (4.1%) 4.76
(3.0%) 0.2% ( -6% - 7%) 0.834
IntNRQ 265.85 (1.0%) 266.60
(0.7%) 0.3% ( -1% - 1%) 0.298
AndHighHighDayTaxoFacets 54.33 (1.6%) 54.49
(1.5%) 0.3% ( -2% - 3%) 0.554
TermBGroup1M1P 62.07 (4.4%) 62.41
(5.6%) 0.5% ( -9% - 11%) 0.735
AndHighMedDayTaxoFacets 202.86 (2.6%) 204.19
(2.6%) 0.7% ( -4% - 6%) 0.434
Respell 111.78 (3.4%) 112.52
(3.1%) 0.7% ( -5% - 7%) 0.519
Fuzzy1 155.04 (3.6%) 156.09
(3.4%) 0.7% ( -6% - 7%) 0.541
SpanNear 13.68 (1.7%) 13.77
(1.3%) 0.7% ( -2% - 3%) 0.142
Fuzzy2 144.74 (2.9%) 145.81
(2.8%) 0.7% ( -4% - 6%) 0.419
TermDTSort 214.37 (7.3%) 216.04
(8.8%) 0.8% ( -14% - 18%) 0.761
AndMedOrHighHigh 35.72 (3.2%) 36.03
(3.8%) 0.9% ( -5% - 8%) 0.441
AndHighHigh 44.40 (2.8%) 44.83
(4.7%) 1.0% ( -6% - 8%) 0.429
OrHighMedDayTaxoFacets 4.34 (5.5%) 4.38
(7.7%) 1.1% ( -11% - 15%) 0.621
BrowseRandomLabelTaxoFacets 29.93 (46.8%) 30.38
(47.5%) 1.5% ( -63% - 179%) 0.921
BrowseDateSSDVFacets 3.98 (33.0%) 4.04
(32.0%) 1.6% ( -47% - 99%) 0.879
Term 2674.87 (3.8%) 2721.55
(4.1%) 1.7% ( -5% - 10%) 0.164
BrowseDayOfYearTaxoFacets 29.20 (31.2%) 29.75
(33.2%) 1.9% ( -47% - 96%) 0.853
BrowseDateTaxoFacets 29.07 (31.3%) 29.63
(33.1%) 1.9% ( -47% - 96%) 0.851
BrowseDayOfYearSSDVFacets 26.74 (9.6%) 27.76
(9.4%) 3.8% ( -13% - 25%) 0.205
BrowseMonthTaxoFacets 30.73 (33.3%) 32.84
(35.6%) 6.9% ( -46% - 113%) 0.528
OrHighHigh 117.70 (4.1%) 161.58
(10.4%) 37.3% ( 21% - 54%) 0.000
OrHighMed 105.09 (4.3%) 186.95
(10.2%) 77.9% ( 60% - 96%) 0.000
```
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]