[
https://issues.apache.org/jira/browse/LUCENE-5959?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14137449#comment-14137449
]
Markus Heiden edited comment on LUCENE-5959 at 9/17/14 4:19 PM:
----------------------------------------------------------------
A better approach would be to create the Automaton with the correct size in
Builder.finish() instead of creating it in the constructor with a default size.
But I can't do that, because the states have already been added to the
automaton at that point.
I may directly allocate
a.transitions = new int[numTransitions * 3];
in Builder.finish(), but this looks a bit dirty to me. The overhead of the
approach in the patch should be negligible.
was (Author: markus_heiden):
A better approach would be to create the Automaton with the correct size in
Builder.finish() instead of creating it in the constructor with a default size.
But I can't do that, because the states have already been added to the
automaton at that point.
I may directly allocate
a.transitions = new int[numTransitions * 4];
in Builder.finish(), but this looks a bit dirty to me. The overhead of the
approach in the patch should be negligible.
> Optimized memory management in Automaton.Builder.finish()
> ---------------------------------------------------------
>
> Key: LUCENE-5959
> URL: https://issues.apache.org/jira/browse/LUCENE-5959
> Project: Lucene - Core
> Issue Type: Improvement
> Components: core/other
> Affects Versions: 4.10
> Reporter: Markus Heiden
> Priority: Minor
> Labels: patch
> Attachments: finish.patch
>
>
> Reworked Automaton.Builder.finish() to not allocate memory stepwise. Added
> growTransitions(int numTransitions) to be able to resize the transistions
> array just once.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
