[jira] [Closed] (UIMA-4535) Flatten indexes need to be invalidated when switching to JCas

["Marshall Schor (JIRA)" <de...@uima.apache.org>]

     [ https://issues.apache.org/jira/browse/UIMA-4535?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Marshall Schor closed UIMA-4535.
--------------------------------
    Resolution: Fixed

> Flatten indexes need to be invalidated when switching to JCas
> -------------------------------------------------------------
>
>                 Key: UIMA-4535
>                 URL: https://issues.apache.org/jira/browse/UIMA-4535
>             Project: UIMA
>          Issue Type: Bug
>          Components: Core Java Framework
>    Affects Versions: 2.8.0SDK
>            Reporter: Marshall Schor
>            Assignee: Marshall Schor
>             Fix For: 2.8.1SDK
>
>
> uima 2.8.0 introduced automatically-generated "flattened-indexes", created after measurements of use indicate it would be beneficial to replace iteration over a type with (lots of) subtypes, with a flattened version of that index.  These flattened versions hold the Java cover class instantiation of the instances.  
> If these are created before JCas is used, then the cover classes are the non-JCas ones, e.g. FeatureStructureImpl or AnnotationImpl.  After the JCas is used, the generators are switched to use the JCas ones.
> A problem occurs in this case: JCas is not used, a flattened index is created, JCas is activated, and then an iteration is done over the flattened index.  This iteration will return the non-JCas versions of the cover classes, potentially causing class cast exceptions.
> The fix, I think, is to invalidate any flattened indexes that might exist in all views, the first time JCas use is started. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)