[
https://issues.apache.org/jira/browse/LUCENE-6066?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Adrien Grand updated LUCENE-6066:
---------------------------------
Attachment: LUCENE-6066.patch
Hi Mark,
I played with your patch to see if removing the code duplication of
PriorityQueue would hurt the benchmark and everything looks ok:
{code}
TaskQPS baseline StdDev QPS patch StdDev
Pct diff
LowSloppyPhrase 69.77 (4.5%) 69.25 (3.8%)
-0.7% ( -8% - 7%)
PKLookup 259.92 (3.2%) 258.50 (2.0%)
-0.5% ( -5% - 4%)
HighSloppyPhrase 13.96 (5.1%) 13.92 (4.8%)
-0.3% ( -9% - 10%)
OrNotHighLow 1135.87 (6.6%) 1132.89 (5.5%)
-0.3% ( -11% - 12%)
AndHighLow 1075.94 (5.2%) 1073.63 (4.6%)
-0.2% ( -9% - 10%)
LowPhrase 124.58 (2.0%) 124.49 (1.8%)
-0.1% ( -3% - 3%)
MedPhrase 78.58 (1.5%) 78.56 (1.6%)
-0.0% ( -3% - 3%)
Prefix3 77.58 (4.7%) 77.59 (3.6%)
0.0% ( -7% - 8%)
HighPhrase 14.14 (1.4%) 14.16 (1.6%)
0.1% ( -2% - 3%)
AndHighMed 248.72 (3.9%) 249.23 (3.5%)
0.2% ( -6% - 7%)
Fuzzy1 72.16 (5.6%) 72.32 (6.2%)
0.2% ( -10% - 12%)
HighTerm 71.70 (5.3%) 71.91 (5.1%)
0.3% ( -9% - 11%)
OrHighLow 68.70 (5.4%) 68.91 (5.7%)
0.3% ( -10% - 11%)
MedTerm 220.94 (5.8%) 221.62 (5.4%)
0.3% ( -10% - 12%)
Wildcard 20.86 (1.6%) 20.92 (1.4%)
0.3% ( -2% - 3%)
LowSpanNear 16.46 (2.3%) 16.51 (2.3%)
0.3% ( -4% - 5%)
MedSpanNear 18.46 (2.4%) 18.52 (2.1%)
0.3% ( -3% - 4%)
IntNRQ 6.63 (4.0%) 6.65 (4.0%)
0.4% ( -7% - 8%)
OrHighNotHigh 38.52 (1.7%) 38.65 (1.5%)
0.4% ( -2% - 3%)
OrNotHighHigh 79.04 (2.3%) 79.33 (1.8%)
0.4% ( -3% - 4%)
OrHighNotMed 52.77 (1.9%) 52.97 (1.5%)
0.4% ( -2% - 3%)
MedSloppyPhrase 44.24 (2.9%) 44.42 (2.5%)
0.4% ( -4% - 5%)
OrHighMed 47.19 (5.2%) 47.37 (5.4%)
0.4% ( -9% - 11%)
OrHighNotLow 85.13 (2.7%) 85.48 (2.1%)
0.4% ( -4% - 5%)
OrHighHigh 26.42 (5.1%) 26.55 (5.0%)
0.5% ( -9% - 11%)
AndHighHigh 84.14 (3.6%) 84.67 (3.0%)
0.6% ( -5% - 7%)
HighSpanNear 50.80 (1.8%) 51.18 (1.4%)
0.7% ( -2% - 4%)
Fuzzy2 38.02 (8.4%) 38.54 (7.5%)
1.3% ( -13% - 18%)
LowTerm 1395.69 (8.9%) 1420.90 (8.4%)
1.8% ( -14% - 20%)
OrNotHighMed 310.39 (4.4%) 316.65 (3.8%)
2.0% ( -5% - 10%)
Respell 82.66 (4.7%) 84.39 (4.4%)
2.1% ( -6% - 11%)
{code}
I attached the patch that I tested with.
+1 to commit
> Collector that manages diversity in search results
> --------------------------------------------------
>
> Key: LUCENE-6066
> URL: https://issues.apache.org/jira/browse/LUCENE-6066
> Project: Lucene - Core
> Issue Type: Improvement
> Components: core/query/scoring
> Reporter: Mark Harwood
> Priority: Minor
> Fix For: 5.0
>
> Attachments: LUCENE-6066.patch, LUCENE-PQRemoveV8.patch,
> LUCENE-PQRemoveV9.patch
>
>
> This issue provides a new collector for situations where a client doesn't
> want more than N matches for any given key (e.g. no more than 5 products from
> any one retailer in a marketplace). In these circumstances a document that
> was previously thought of as competitive during collection has to be removed
> from the final PQ and replaced with another doc (eg a retailer who already
> has 5 matches in the PQ receives a 6th match which is better than his
> previous ones). This requires a new remove method on the existing
> PriorityQueue class.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
