[
https://issues.apache.org/jira/browse/LUCENE-5279?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Michael McCandless updated LUCENE-5279:
---------------------------------------
Attachment: LUCENE-5279.patch
Patch.
However, it seems to be slower, testing on full Wikpedia en:
{noformat}
Report after iter 10:
Task QPS base StdDev QPS comp StdDev
Pct diff
OrHighLow 14.44 (7.7%) 12.48 (4.7%)
-13.6% ( -24% - -1%)
OrHighHigh 5.56 (6.2%) 4.86 (4.4%)
-12.6% ( -21% - -2%)
OrHighMed 18.62 (6.7%) 16.29 (4.4%)
-12.5% ( -22% - -1%)
AndHighLow 398.09 (1.6%) 390.34 (2.3%)
-1.9% ( -5% - 1%)
OrNotHighLow 374.60 (1.7%) 369.61 (1.7%)
-1.3% ( -4% - 2%)
Fuzzy1 67.10 (2.1%) 66.41 (2.2%)
-1.0% ( -5% - 3%)
OrNotHighMed 51.68 (1.7%) 51.37 (1.5%)
-0.6% ( -3% - 2%)
Fuzzy2 46.73 (2.8%) 46.45 (2.6%)
-0.6% ( -5% - 4%)
OrHighNotLow 20.05 (3.5%) 19.96 (5.0%)
-0.5% ( -8% - 8%)
OrHighNotMed 27.15 (3.2%) 27.05 (4.8%)
-0.3% ( -8% - 7%)
OrNotHighHigh 7.72 (3.2%) 7.70 (4.7%)
-0.3% ( -7% - 7%)
OrHighNotHigh 9.81 (3.0%) 9.79 (4.5%)
-0.1% ( -7% - 7%)
LowSloppyPhrase 43.83 (1.9%) 43.89 (2.1%)
0.2% ( -3% - 4%)
IntNRQ 3.49 (4.5%) 3.50 (4.1%)
0.2% ( -8% - 9%)
Prefix3 70.74 (2.7%) 71.01 (2.4%)
0.4% ( -4% - 5%)
HighTerm 65.33 (3.0%) 65.62 (13.5%)
0.4% ( -15% - 17%)
MedSloppyPhrase 3.47 (3.5%) 3.49 (4.7%)
0.6% ( -7% - 9%)
LowPhrase 13.06 (1.5%) 13.14 (2.0%)
0.6% ( -2% - 4%)
Wildcard 16.71 (2.9%) 16.82 (2.2%)
0.7% ( -4% - 5%)
MedTerm 100.90 (2.5%) 101.71 (10.4%)
0.8% ( -11% - 14%)
LowTerm 311.85 (1.4%) 314.53 (6.4%)
0.9% ( -6% - 8%)
HighSpanNear 8.06 (5.1%) 8.13 (5.9%)
0.9% ( -9% - 12%)
Respell 48.00 (2.3%) 48.45 (2.8%)
0.9% ( -4% - 6%)
HighSloppyPhrase 3.40 (4.1%) 3.43 (6.6%)
1.0% ( -9% - 12%)
AndHighMed 34.14 (1.6%) 34.52 (1.7%)
1.1% ( -2% - 4%)
AndHighHigh 28.15 (1.7%) 28.48 (1.7%)
1.2% ( -2% - 4%)
MedSpanNear 30.62 (2.8%) 31.07 (3.2%)
1.5% ( -4% - 7%)
LowSpanNear 10.30 (2.6%) 10.48 (2.9%)
1.7% ( -3% - 7%)
MedPhrase 195.60 (5.1%) 201.44 (6.6%)
3.0% ( -8% - 15%)
HighPhrase 4.17 (5.6%) 4.34 (6.9%)
4.0% ( -8% - 17%)
{noformat}
So ... I don't plan on pursuing it any further, but wanted to open the issue in
case anybody wants to try ...
> Don't use recursion in DisjunctionSumScorer.countMatches
> --------------------------------------------------------
>
> Key: LUCENE-5279
> URL: https://issues.apache.org/jira/browse/LUCENE-5279
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Michael McCandless
> Attachments: LUCENE-5279.patch
>
>
> I noticed the TODO in there, to not use recursion, so I fixed it to just use
> a private queue ...
--
This message was sent by Atlassian JIRA
(v6.1#6144)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
