[
https://issues.apache.org/jira/browse/LUCENE-6388?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14392222#comment-14392222
]
Robert Muir commented on LUCENE-6388:
-------------------------------------
{noformat}
Report after iter 10:
Chart saved to out.png... (wd: /home/rmuir/workspace/util/src/python)
Task QPS trunk StdDev QPS patch StdDev
Pct diff
MedSpanNear 75.69 (2.0%) 80.58 (3.9%)
6.5% ( 0% - 12%)
LowSpanNear 233.30 (3.8%) 259.44 (6.5%)
11.2% ( 0% - 22%)
HighSpanNear 9.43 (3.6%) 10.76 (7.5%)
14.0% ( 2% - 25%)
{noformat}
> Optimize SpanNearQuery
> ----------------------
>
> Key: LUCENE-6388
> URL: https://issues.apache.org/jira/browse/LUCENE-6388
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Robert Muir
> Attachments: LUCENE-6388.patch
>
>
> After the big spans overhaul in LUCENE-6308, we can speed up SpanNearQuery a
> little more:
> * SpanNearQuery defaults to collectPayloads=true, but this requires a slower
> implementation, for an uncommon case. Use the faster no-payloads impl if the
> field doesn't actually have any payloads.
> * Use a simple array of Spans rather than List in NearSpans classes. This is
> iterated over often in the logic.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
