Eric Riebling wrote: > Marshall Schor wrote: > > - contents edited - > >> The UCR repository code might do something with the environmental >> variable information from the PEAR prior to doing any install or running >> of the PEAR - Eric - can you say if it does? > > > The repository doesn't do anything with the PEAR environment variables > beyond this: > It does perform edits to setenv.txt to localize paths for setting the > CLASSPATH prior to calling UIMAFramework.produceAnalysisEngine(). > (or produceCasConsumer(), produceCollectionReader() etc. as appropriate) > > Using System.getProperty(name) instead of System.getenv(name): great idea! > > -er
I don't konw how your PEAR installation in your repository works, but you don't have to edit setenv.txt and update any classpath settings. There is a easier way. There is a PEAR installer API that can install and verify/run PEAR packages. Please refer for details and for an example to: http://incubator.apache.org/uima/downloads/releaseDocs/2.2.2-incubating/docs/html/references/references.html#ugr.ref.pear.installing_pear_using_API Additionally during the PEAR installation, independent of how (API, UI) you installed the PEAR, a PEAR package descriptor is generated automatically. This generated descriptor can be used similar to an AE descriptor. When using this descriptor, you don't have to take care about classpath, datapath and "environemnt" variables (System.getProperty(name)). The UIMA framework will take care of that. For details about the PEAR package descriptor please refer to: http://incubator.apache.org/uima/downloads/releaseDocs/2.2.2-incubating/docs/html/references/references.html#ugr.ref.pear.specifier Hope that helps. -- Michael
