[
https://issues.apache.org/jira/browse/SOLR-7618?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Hoss Man updated SOLR-7618:
---------------------------
Attachment: SOLR-7618.patch
Great point Jacob.
A few updates to your patch...
* fix formatting (we use 2 spaces for indents)
* improved javadocs & linkages
* added equivilence testing between each method version (to future proof
against someone changing one of these methods in the future and making them no
longer behave consistently)
> MultiFunction.anyExists - creating FunctionValues[] objects for every document
> ------------------------------------------------------------------------------
>
> Key: SOLR-7618
> URL: https://issues.apache.org/jira/browse/SOLR-7618
> Project: Solr
> Issue Type: Bug
> Reporter: Jacob Graves
> Priority: Minor
> Labels: easyfix, newbie, patch, performance
> Attachments: SOLR-7618.patch, SOLR-7618.patch
>
> Original Estimate: 2h
> Remaining Estimate: 2h
>
> In the class org.apache.lucene.queries.function.valuesource.MultiFunction
> there is the following method signature (line 52)
> public static boolean allExists(int doc, FunctionValues... values)
> this method is called from the class
> org.apache.lucene.queries.function.valuesource.DualFloatFunction (line 68)
> public boolean exists(int doc) {
> return MultiFunction.allExists(doc, aVals, bVals);
> }
> Because MultiFunction.allExists uses Java varargs syntax ("...") a new
> FunctionValues[] object will be created every time this call takes place.
> The problem is that the call takes place in a document level function, which
> means that it will create new objects in the heap for every document in the
> query results.
> for example if you use the following boost function (where ds and dc1 are
> both TrieDateField)
> bf=min(ms(ds,dc1),604800000)
> You will get extra objects created for each document in the result set, which
> has a big impact on performance and memory usage if you are searching a large
> result set.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
