errors in text filters can cause indexing to fail without warning the client

• Key: JCR-299
• Type: Bug
• Status: Closed
• Resolution: Fixed
• Priority: Minor
• Assignee: Jukka Zitting
• Reporter: Brian Moseley
• Component/s: indexing
• Affects Version/s: 0.9, 1.0, 1.0.1
• Fix Version/s: 1.1

Description

i've opened this issue to track the discussion at <http://thread.gmane.org/gmane.comp.apache.jackrabbit.devel/5086>. briefly, exceptions thrown by text filters are logged and swallowed by jackrabbit; there's no way for a text filter to signal to the jcr client that indexing failed.

some solutions have been proposed, including throwing an unchecked exception, which doesn't allow jackrabbit to maintain transactional integrity, and giving filters veto power over the observed repository operation. depending on the difficulty of the solution that is eventually determined to be correct, it may be sufficient for 1.0 to document the issue and perhaps improve the warning/error logging.

Comments

[Jukka Zitting]

Fixed in revision 423547 by catching and logging all index filter exceptions (even unchecked ones). After considering this issue I don't think it makes sense to give the index filters veto powers over saving content. They are not validators, so if a binary property cannot be parsed for text content because of a file format error or some similar problem, the indexer should return an empty reader instead of trying to revert the save or somehow signal the client application.

The requested functionality would be better implemented by extending the value constraint support in Jackrabbit. Please file a separate feature request if you still need this functionality.