+1 from separating the models/resources from source. I think separating them out as maven artifacts hosted outside ASF seems like a decent idea, but had a pragmatic question though: Recall that some items like Lucene requires directories to exist as File and Directory instead of inside an archive (jar/zip). Wonder if there's an easy way around this (other package types?)
--Pei > -----Original Message----- > From: Coarr, Matt [mailto:[email protected]] > Sent: Monday, November 05, 2012 10:34 AM > To: [email protected] > Subject: Re: [DISCUSS] What should we do with cTAKES resources? > > +1 for Steve's suggestion! > > I really like Steve's suggestion of a separate project each large resource > (and > having that project produce a jar file that is loaded into the maven central > repo). > > I'm taking this to mean that we would do this for both ASL-compatible > resources and non-compatible resources. I think this is the way to go. > > As an added benefit, I think this would allow someone to checkout a single > module (say ctakes-assertion or ctakes-relation-extractor) and work on only > that module, without having to check out all of ctakes. > > A quick suggestion: > > If the models or data to build these models are large, can we keep these in a > separate subversion root from the main source code modules? This would > make our svn repo for our main source code much smaller. (I've done a > couple fresh checkouts lately, and I think they took around 90 minutes to > checkout. My current working copy is 2.3GB after running "mvn clean".) > > I'm really just looking for a way to checkout all of the main ctakes sources > from svn without grabbing the models (they'll come from maven in Steve's > scheme). If there are other ways of accomplishing that source-only checkout, > I welcome those ideas. > > Matt
