Hi Aaron -
When running Pears contained in an aggregate, the Java classpath used to
load the JCas (and other) classes is switched to be what the Pear specifies.
The classpath used by the PEAR is built using the jars in the "lib"
directory, plus any additional paths that are specified in the PEAR
installation descriptor's "set_env_variable" action, done by specifying
the "CLASSPATH" environmental variable. (see
http://incubator.apache.org/uima/downloads/releaseDocs/2.2.2-incubating/docs/html/references/references.html#ugr.ref.pear.installation_descriptor
and scroll down to the section 6.1.4.5 )
Can you verify that the JCas cover classes are either in a Jar file that
is in the Pear's lib directory, or that you have added a environment
variable setting action for the Pear specifying the CLASSPATH
environmental variable and having it point to where the JCas classes are
in the Pear structure?
Thanks. -Marshall
Aaron Kaplan wrote:
I have two annotators that share a type system. Each is packaged in a
separate pear with its own copy of the type system descriptor and the
associated JCas classes. When I build an aggregate annotator using
the two pear descriptors and run it in the CVD, I get a run-time
exception in the process() method of the second annotator saying that
it can't find one of the cover classes. I have verified that the
.class file for this class is present in the bin/ subdirectory of the
annotator's pear installation directory, so it seems like the
classpath must not be getting set properly.
The following works fine: load the first annotator's pear descriptor
in the CVD, run the annotator on an input text file, save the result
in an XMI CAS file; load the second annotator's pear descriptor in the
CVD, load the XMI CAS file, run the annotator on that. It's when I
combine the two annotators into an aggregate and try to run it in one
go that I have a problem.
Does it sound like I'm doing something wrong? How can I debug this?
Thanks
-Aaron
