I don't know of a Maven plugin for this, but it sounds like a good idea. There is a Maven plugin for doing pear packaging in the sandbox, which might serve as a guide if you were considering doing a plugin.
One thing to note: there are 2 versions of JCasGen - the normal one requires Eclipse to be present because it runs as an Eclipse application, in order to pick up the EMF "code merging" capability. There is a way to run without this - in which case no merging is done. See http://incubator.apache.org/uima/downloads/releaseDocs/2.2.2-incubating/docs/html/tools/tools.html#ugr.tools.jcasgen.running_without_eclipse -Marshall Александър Л. Димитров wrote: > Hello, > > I'm currently working on a distributed project involving UIMA. At this stage, > sometimes modifications to the UIMA descriptor files are absolutely necessary > and, unfortunately, this confuses our VCS. Whenever JCasGen is run from within > Eclipse or similar, it updates *all* generated class files, which leads the > VCS > to treating them as new commits. The problem here is, when editing the > descriptors on different branches, while the merging of the descriptor files > goes well most of the time, the merging of the class files fails most of the > time. This leaves the merger with several dozen conflicts. > > The current policy is to then just delete those files and run JCasGen on the > merged branch. Of course, the actual problem here is, that one shouldn't > really > track those files in the first place. > > But the matter is slightly more delicate. After checking out the project from > version control, the project should be readily compilable, without further > setup. That's why we're using Maven to fetch all dependencies and automate the > build process. Without having these files in source control, they have to be > generated first. > > So the real question is: is there any way to run a JCasGen from within the > Maven > build? Probably a plugin? Of course, one could just spawn a process to call > org.apache.uima.tools.jcasgen.Jg, but I wanted to ask if such a thing already > existed. > > Thanks, > Aleks >
