[
https://issues.apache.org/jira/browse/SOLR-11242?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16151655#comment-16151655
]
David Smiley commented on SOLR-11242:
-------------------------------------
The change to the sources looks good to me.
[~steve_rowe] I suspect you are more familiar with the javacc / build bits.
I'm not sure what to make of the 'build.xml' change in this patch. I applied
this patch and ran "ant javacc" (observing there was such a target) and it made
changes to the CharStream interface (changes we don't want I'm sure; it added
methods), and QueryParserTokenManager.java additional unused import statements
(which I could easily remove). Should I not invoke "ant javacc"... and simply
accept the changes in the patch as good, despite this? Or maybe the patch
needs to do more (though I doubt it).
> QueryParser should avoid creating an LookaheadSuccess(Error) object with
> every instance
> ---------------------------------------------------------------------------------------
>
> Key: SOLR-11242
> URL: https://issues.apache.org/jira/browse/SOLR-11242
> Project: Solr
> Issue Type: Improvement
> Security Level: Public(Default Security Level. Issues are Public)
> Affects Versions: 7.0, 6.6.1, master (8.0)
> Reporter: Michael Braun
> Attachments: SOLR-11242.patch, SOLR-11242.png,
> TestQueryPerfSpeedup.java
>
>
> Sampling of an application which includes a very heavy query load testing
> against an EmbeddedSolrServer reveals a huge amount of the time spent on
> initialization of the QueryParser itself, specifically an internal class
> called LookaheadSuccess which extends from Error. With every instance of
> QueryParser created, it spends time populating the stacktrace for this
> object, which is created as a final instance variable (the variable called
> jj_ls).
> Ideally an exception shouldn't be used as a signaling mechanism, or at the
> very least, not an exception created on every single object instantiation.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
