[ https://issues.apache.org/jira/browse/LUCENE-1903?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12752813#action_12752813 ]
Uwe Schindler commented on LUCENE-1903: --------------------------------------- Chris: Could you test this patch, if it works as exspected also for you. Maybe I found a fix only valid for your testcase but not other cases. In my opinion, the code works now identical to the 2.4.1 one (without the output buffer). Unigrams are simply detected by shingleBufferPosition==0. The position increments are also tested by your code and the implementation also looks right. In principle, there was only missing the increment of the shingleBufferPosition if no unigrams are provided. Mark: If you want to build RC3 soon, just assign yourself and commit this fix. I will go to bed now. I will commit this tomorrow, if you hadn't. A CHANGES.txt entry is not needed in my opinion, as this is not a new feature or a bug from 2.4.1. > Incorrect ShingleFilter behavior when outputUnigrams == false > ------------------------------------------------------------- > > Key: LUCENE-1903 > URL: https://issues.apache.org/jira/browse/LUCENE-1903 > Project: Lucene - Java > Issue Type: Bug > Components: contrib/analyzers > Affects Versions: 2.9 > Reporter: Chris Harris > Fix For: 2.9 > > Attachments: LUCENE-1903.patch, LUCENE-1903_testcases.patch, > LUCENE-1903_testcases_lucene2_4_1_version.patch, > TEST-org.apache.lucene.analysis.shingle.ShingleFilterTest.xml > > > ShingleFilter isn't working as expected when outputUnigrams == false. In > particular, it is outputting unigrams at least some of the time when > outputUnigrams==false. > I'll attach a patch to ShingleFilterTest.java that adds some test cases that > demonstrate the problem. > I haven't checked this, but I hypothesize that the behavior for > outputUnigrams == false got changed when the class was upgraded to the new > TokenStream API? -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online. --------------------------------------------------------------------- To unsubscribe, e-mail: java-dev-unsubscr...@lucene.apache.org For additional commands, e-mail: java-dev-h...@lucene.apache.org