Yeah Drew has a good point on JCL -- it uses this unorthodox but clever system wherein you compile against generic SLF4J APIs (that's the -api.jar) and then add an implementation .jar of your choice which actually binds to a particular logging system (that's the other one). The latter is not needed at compile time but needed at runtime. If we can express that in Maven, good. If it's necessary to call it a compile-time dependency for it to be included, OK.
I don't know about the others. In theory if they are needed indirectly (by one of our direct dependencies) then that dependency will have it included. Could we recklessly try taking it out and see what happens, unless someone's sure it's not going to fly? On Mon, Feb 8, 2010 at 2:33 PM, Robin Anil <robin.a...@gmail.com> wrote: > But the whole thing is compiling without those jars. Are you sure. Let me > dig deep into the code. Instantiating class purely by reflection without > even a single import will cause errors then > > Robin > On Mon, Feb 8, 2010 at 7:24 PM, Drew Farris <drew.far...@gmail.com> wrote: > >> The dependency:analyze reports can't be trusted 100% because some of >> the jars are used at runtime. >> >> For example, I recently discovered that >> org.slf4j:slf4j-jcl:jar:1.5.8:compile is necessary for proper logging >> when running in mvn exec:java -- I suspect that commons-logging is >> needed as well because the slf4jcl (jcl = jakarta commons logging) is >> a wrapper for commons logging. >> >> The alternative would be to use another wrapper entirely, but I >> believe some of our deps use commons-logging so that could likely lead >> to a mess. >> >> Also, IIRC, The jets3t stuff was added back into examples recently to >> support MAHOUT-249 >> >> The google collections stuff was recently added too, I can't remember >> what for however. >> >> On Mon, Feb 8, 2010 at 7:41 AM, Robin Anil <robin.a...@gmail.com> wrote: >> > Kicked direct dependencies. Current Job jar size is 12MB as compared to >> 14MB >> > >> > On Mon, Feb 8, 2010 at 5:35 PM, Sean <sro...@gmail.com> wrote: >> > >> >> Seems OK to me. I mean, if it doesn't work, we'll know immediately, so >> >> little harm. >> >> >> >> On Mon, Feb 8, 2010 at 11:58 AM, Robin Anil <robin.a...@gmail.com> >> wrote: >> >> > Take a look at the maven dependency:analyze output >> >> > >> >> >> > >> >
