[
https://issues.apache.org/jira/browse/LUCENE-5628?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13979419#comment-13979419
]
Simon Willnauer commented on LUCENE-5628:
-----------------------------------------
I think it's worth doing the optimization here. A couple of comments
* can we put the exit condition into the while block instead of at the end
with a break I think it can just be while(string.length > 0)
* looking at the impl of State I think we can just use an identity hashset or
maybe even an array since the Ids are within known bounds to check the
pathStates? You could even just us a bitset and mark the state ID as visited?
Hmm now that I wrote it I see your comment :) I will leave it here for
dicsussion.
* Somewhat unrelated but I think the State implementation has a problem since
it doen't override equlas but it should since it has an hashcode impl. I wonder
if we either should remove the hashCode or add equals just for consistency?
* should we rather throw IllegalState than IllegalArgument :D
* just for readability it might be good to s/strings/finiteStrings/ I had a
hard time to see when you do things on the string vs. strings
* is this a leftover ==> // a.getNumberedStates();
> SpecialOperations.getFiniteStrings should not recurse
> -----------------------------------------------------
>
> Key: LUCENE-5628
> URL: https://issues.apache.org/jira/browse/LUCENE-5628
> Project: Lucene - Core
> Issue Type: Bug
> Reporter: Michael McCandless
> Assignee: Michael McCandless
> Fix For: 4.9, 5.0
>
> Attachments: LUCENE-5628.patch
>
>
> Today it consumes one Java stack frame per transition, which when used by
> AnalyzingSuggester is per character in each token. This can lead to stack
> overflows if you have a long suggestion.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
