[
https://issues.apache.org/jira/browse/UIMA-1771?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12907042#action_12907042
]
Marshall Schor commented on UIMA-1771:
--------------------------------------
This routine has other limitations as well. For instance, a feature structure
can have slots whose value is a reference to other feature structures. These
are not handled by this routine either.
It also replaces certain characters in the document's subject-of-analysis text
fields by blanks (see method replaceInvalidXmlChars). It also presumes that
the CAS's subject of analysis is a text string, which is not always the case.
If the CAS doesn't have a document text, then this method will throw a null
pointer exception.
It also will fail to faithfully write out annotations which are overlapping,
because those cannot be represented as inline XML.
I guess I'm in favor of documenting these issues in the Javadocs.
> CasToInlineXml truncates attributes
> -----------------------------------
>
> Key: UIMA-1771
> URL: https://issues.apache.org/jira/browse/UIMA-1771
> Project: UIMA
> Issue Type: Bug
> Reporter: Jens Grivolla
> Priority: Minor
>
> org.apache.uima.util.CasToInlineXml has a hard coded limit of 64 characters
> for attribute values. This makes inline xml suitable only for debugging, and
> the documentation contains no warning to this effect.
> Given that this component is part of uima-core (and not just example code) I
> would expect such pitfalls to be at least documented. A better solution
> would be to make the truncating optional/configurable, so that inline xml
> could be used as a basis for further processing outside UIMA.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
