Details
-
Improvement
-
Status: Open
-
Major
-
Resolution: Unresolved
-
None
-
None
-
None
-
None
Description
The ForkParser is a great option for handling oom/permanent hangs; and from a code/design perspective, IMHO, it is a thing of beauty.
On the user list JimIdle, recently pointed out that the ForkParser can't work with custom parsers that depend on non-serializable components.
It would be great to allow users to specify a TIKA_HOME variable or pass in a directory with the tika-related jars and run the server as a separate process with that dir as the class path. This would also make adding optional jars much easier and could prevent jar hell with the calling application.
Bonus points for enabling the RecursiveParserWrapper to work.