Project to abstract the use of tika-server in a seperate process/JVM
This project runs a background process/JVM for tika-server to isolate its execution. Files/InputStreams can be passed to this class for parsing, and they are sent to the tika-server via JAXRS. The return values are populated in to the Metadata object that is passed to the parse() method, just like a normal call to Tika.
A copy of a tika-server jar will be contained within this jar and is copied to a temporary folder for background execution.
If a call to the tika-server takes longer than 10 seconds (by default) the tika-server is restarted.
A tika-server is already running in the background and there is no JVM start cost. Communication with tika-server is encapsulated within this class. Speed should be close to pure-Java combined with networking overhead for JAXRS.
A crash of the tika-server should not take out the controlling JVM.
-
To change Tika version you just need to change it once in the pom. However, if you are reusing this as a library and this jar's manifest is changes then you need to ensure that the version in the class matches the pom.
-
Also - this library blocks until the tika-server is started (or it has waited 120s). During this time unwanted stack traces are printed to the log at WARN level but we set JAXRS to ERROR (see log4j.properties)