Ian et al, I reverted the code changes pending discussion. I left the pom changes; I assert that they are improvement.
On Fri, Jun 27, 2014 at 10:55 AM, Benson Margulies <bimargul...@gmail.com> wrote: > On Fri, Jun 27, 2014 at 10:47 AM, Ian Dickinson <i.j.dickin...@gmail.com> > wrote: >> On Fri, Jun 27, 2014 at 3:31 PM, Benson Margulies <bimargul...@gmail.com> >> wrote: >>> On Fri, Jun 27, 2014 at 10:05 AM, Ian Dickinson <i.j.dickin...@gmail.com> >>> wrote: >>>> Actually, now that I've reviewed it, there's a conflict between this >>>> patch and the current behaviour. At the moment, the way to set the >>>> output directory to something other than target/generated-sources is >>>> to set the output configuration parameter. If this is a directory, >>>> generated files go into that directory. If that option is set in the >>>> defaults section of the schemagen config, it affects all of the >>>> generated files. Tl;dr: there is already a mechanism for changing the >>>> default output directory for all generated files. >>>> >>>> Voting to reject this patch until it's clearer why users need two >>>> mechanisms for achieving the same end. >>> >>> So, I'll remove the redundancy, but have I convinced you about the >>> problem of it not getting compiled? >> Yes, absolutely. I'm not using the plugin day-by-day (I rarely do Java >> dev these days, in fact), so direct experience by someone using the >> tool definitely takes precedence. >> >> I'm actually not sure what the right solution here is. With the >> current mechanism, the output directory can be specified on an >> vocabulary-by-vocabulary basis. Any of those locations might already >> be in the source tree already (e.g. we support the scenario where >> someone generates the .java file directly under src/main/java, so that >> the translated file can be kept under source control). So I'm not sure >> - and you are in a better position to advise on this - that just >> calling addCompileSourceRoot() on each output directory would be >> correct. > > > > >> >> Perhaps the answer is that the location of the output files should >> remain with the current configuration process, but there should be an >> additional means, via the pom.xml, to add one or more directories to >> the sources root. That could have a default value, so *if* the default >> default output is left as target/generated-sources/jena, and *if* the >> user doesn't change anything in the pom, then the output Java files >> will still compile. > > To do exactly what you suggest, all we need to do is revert my change; > the build-helper-maven-plugin is already a way to add a directory to > the source roots via pom config. > > However, I have a more complex idea to offer you: > > Maybe redundancy is not redundancy. > > We tell users: If you want to just produce sources to include in the > build, you use my new top-level property as the single source root, > and we will add it to the build, and all will be well. If you want to > do something more complex, you can use the older scheme, and it's up > to you to arrange for it to compile, perhaps using the > build-helper-maven-plugin. > > To make this really painless, we'd need one more boolean property: > useOutputDirectory. Default is false, preserving old behavior. Set to > true, gets new behavior, respecting new property. > > WDYT? > >> >> With respect to changing the default, if writing to >> target/generated-sources simply doesn't work with Maven 3, then the >> change of default will be less intrusive, because presumably people >> are not doing that at the moment. I still think it would be worth >> raising on users@, to see if anyone would be impacted. >> >> There's enough complexity here that it's probably worth extending the >> current set of integration tests to cover this new scenario. >> > > agreed: > >> Ian
