Just an added note on the topic of potential package name changes. In the past, we've had to move some AST transformations from groovy.lang to groovy.transforms, and we managed to make that transition without too much harm. Similarly to what we did on that occasion, we could offer a compatibility module that would just delegate classes from old packages to classes of the new packages. And once you've had the time to make the migration, you'd just remove the compatibility module. We could also have more complex solutions involving bytecode rewriting.
But at some point, it'll really look ridiculous to still have org.codehaus here and there, although codehaus' already been long gone already. So package name changes are not all black'n white. There are possible shades of gray :-) Guillaume On Mon, Mar 27, 2017 at 9:03 PM, Winnebeck, Jason < jason.winneb...@windstream.com> wrote: > I don’t know if it was totally clear from my previous mails, but I agree > on not changing the package names, unless breaking changes are required and > the package names need to change to preserve the ability of Groovy 1.x/2.x > code to co-exist, avoiding the “Python 3 Effect”. If the only breaking > change would be the package change, then there’s no sense to change the > name just to change the name. > > > > I think it is OK to change the Maven coordinates in any case. While it is > used sometimes as a starting point to look at a class and try to figure out > what library it comes from based on matching the package name to the group > ID, that’s not a common operation and modern IDEs (and search.maven.org) > can easily answer the question if needed. The only drawback to changing > Maven coordinates is it might make it harder for people to know there is a > newer package, that is, to search for upgrades we check for more recent > versions of current dependencies. However, with a project as big as Groovy > I think it will be clear when Groovy 3 comes out that users will know. > > > > Jason > > > > *From:* Keith Suderman [mailto:suder...@anc.org] > *Sent:* Monday, March 27, 2017 2:49 PM > *To:* users@groovy.apache.org > *Subject:* Re: Maven coordinates going forward > > > > +1 for changing Maven coordinates. > > > > -1 for changing package names. Sure, new code can use the new package > names, but changing existing packages is just breaking changes for the sake > of breaking things with no real benefit. I hope the Groovy team tries to > break as little as possible to avoid the "Python3 Effect", whether real or > imagined. > > > > Having said that, how much public facing code is in org.codehaus.groovy > packages? I don't think I've typed "import org.codehaus.groovy..." in my > life, but IntelliJ may have inserted a few for me. > > > > Keith > > > > > > On Mar 27, 2017, at 2:20 PM, Jochen Theodorou <blackd...@gmx.org> wrote: > > > > On 27.03.2017 20:08, Winnebeck, Jason wrote: > > The key thing in my mind is that you can't make a change that breaks > 100% of libraries at one time without fracturing the community or at > least introducing a major hindrance to upgrade that will mean > maintaining 2.x series for a very long time. Even if the upgrade > process is as easy as a recompile, there are a lot of published > libraries no longer maintained. Even for the ones that are > maintained, there are people who might not want to be forced to > upgrade every library. I'm not a Grails user, but my impression is > that the framework relies on a lot of plugins and is one of the (if > not the most) active Groovy communities and I have a hard time > envisioning how that upgrade path will take. You'd have to maintain > Groovy 2 and Groovy 3 versions of each plugin, and to upgrade you'd > have to upgrade everything at one time to the (most likely) latest > version. > > What is the possibility that the package names are changed, the > parser, metaprogramming model, etc. that all break in Groovy 3 > change, but yet still have a compatibility JAR implementing the > minimal Groovy 2.x classes in a way that allows interoperability with > Groovy 3 code? Theoretically at a worst case, Groovy 3 should be able > to view Groovy 2 classes the same way as any other Java class. I > think many concerns would be lifted if Groovy 2 and 3 could co-exist > at the same time, allowing you to upgrade incrementally. > > > If you see the new metaprogramming model as chance, then it would make > sense to implement that in the new packages instead of transferring old and > to be deprecated classes. The goal would the to be able to run old and new > system at the same time, where the Groovy 1.x/2.x classes would use Groovy > 3.x/4.x classes as implementation. > > The problem with this approach is simply manpower and of course some > conceptual problems still to be solved. > > bye Jochen > > > > ---------------------- > > Keith Suderman > > Research Associate > > Department of Computer Science > > Vassar College, Poughkeepsie NY > > suder...@cs.vassar.edu > > > > > > > > > This email message and any attachments are for the sole use of the > intended recipient(s). Any unauthorized review, use, disclosure or > distribution is prohibited. If you are not the intended recipient, please > contact the sender by reply email and destroy all copies of the original > message and any attachments. > -- Guillaume Laforge Apache Groovy committer & PMC Vice-President Developer Advocate @ Google Cloud Platform Blog: http://glaforge.appspot.com/ Social: @glaforge <http://twitter.com/glaforge> / Google+ <https://plus.google.com/u/0/114130972232398734985/posts>
