Hi fellow UIMA users--
I'm hoping we can share some thoughts on how to structure a UIMA application installation. In other words, what are folks doing for directory structures, classpaths, datapaths, etc? The UIMA docs don't appear to offer any policy suggestions, just mechanisms. For example, there is the ResourceManager where one can set the data path, set up classloaders, and resolve paths. Another mechanism is that the description files of sub-ordinate analysis engines are searched for relative to the location of the parent description file. And then there are PEAR files and the PEAR installer, of course. And, if I remember correctly, one can use some variables in paths in PEAR files. I haven't quite waded through all of the documentation yet, but I'm wondering how people are using these mechanisms to structure the installation of an application. For example, do you just have a known directory in your application, and each PEAR file is in a sub-directory of that? Then all import paths start with "../"? Is it possible to have a path-resolving strategy that works both in Eclipse and stand-alone, which may have different structures? I assume that each analysis engine is a separate Eclipse project in order to produce its PEAR file. I'm using maven, which wants a different project directory structure than the PEAR structure. Has anyone tried to address this? Has anyone managed to configure maven to use the PEAR structure? By the way, does anyone else find it curious that none of the annotators in the UIMA distribution use PEAR files? Perhaps it's because the UIMA source code uses maven, making it difficult to produce the PEAR structure. Does anyone run in Tomcat? If so, what's your experience combining the Tomcat classloading system with the UIMA classloading system? So many questions... Can anyone share their experiences? Thanks, Greg Holmberg
