[
https://issues.apache.org/jira/browse/S2GRAPH-11?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15101133#comment-15101133
]
ASF GitHub Bot commented on S2GRAPH-11:
---------------------------------------
Github user hsleep commented on a diff in the pull request:
https://github.com/apache/incubator-s2graph/pull/5#discussion_r49817258
--- Diff: CHANGES ---
@@ -5,6 +5,8 @@ Release 0.12.1 - unreleased
NEW FEATURES
IMPROVEMENT
+
+ S2GRAPH-11: Performance tuning on sort by scoreSum on PostProcess
--- End diff --
I think that CHANGES file should not be changed in this PR.
> Performance tuning on sort by scoreSum on PostProcess
> -----------------------------------------------------
>
> Key: S2GRAPH-11
> URL: https://issues.apache.org/jira/browse/S2GRAPH-11
> Project: S2Graph
> Issue Type: Improvement
> Reporter: DOYUNG YOON
> Assignee: DOYUNG YOON
> Labels: performance
> Original Estimate: 72h
> Remaining Estimate: 72h
>
> when query has group by, current implementation first build array of json
> object with scoreSum. after that sort this array by looking up json path with
> “scoreSum” key which is unnecessary. worse part is after look up, JsValue
> need to be casted as double which is very slow. profiling with visualvm shows
> that this JsValue.as[Double] use 7% of cpu cycle.
> we already know what is scoreSum before sorting and costly operation like
> JsValue.as[Double] should be avoided for performance tuning.
> this affect query performance a lot since JsValue.as[Double] should be run on
> every result edges.
> for example, 10k query per second with average result size 100, then 10k x
> 100 call per second on as[Double] will be called.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
