flink-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Robert Metzger <rmetz...@apache.org>
Subject Re: Flink 0.9 built with Scala 2.11
Date Thu, 24 Sep 2015 14:22:24 GMT
Just a note that FLINK-2200
<https://issues.apache.org/jira/browse/FLINK-2200> has been resolved.
All flink artifacts are now two times in the maven snapshot repository: one
without a suffix (scala 2.10) and one with a suffix (_2.11 for scala 2.11).
There is also a nightly _2.11 binary release for Flink:
http://stratosphere-bin.s3.amazonaws.com/flink-0.10-SNAPSHOT-bin-hadoop2_2.11.tgz

There is no hadoop1_2.11 artifact. I'm for now assuming that somebody who's
using the latest scala is not using a legacy Hadoop version ;)


I would appreciate some testing from users who requested this feature to
make sure everything is working as expected for the next major release.
We need to make sure that no 2.10 dependency is coming in anywhere and that
Flink with Scala 2.11 is working properly.


On Tue, Jul 7, 2015 at 1:58 PM, Chiwan Park <chiwanpark@apache.org> wrote:

> I cannot find talking about pure/non-pure java distinction in the
> documentation.
>
> I defined the rules about artifact id to apply modules by only Scala
> version not
> pure/non-pure java. The modules without suffix `_2.11` means that they are
> linked with Scala 2.10 binary.
>
> If I misunderstood your sentence or missed the talking in documentation,
> please
> notify me.
>
> Regards,
> Chiwan Park
>
> > On Jul 6, 2015, at 8:07 PM, Alexander Alexandrov <
> alexander.s.alexandrov@gmail.com> wrote:
> >
> >> Because we are using Scala in our runtime, all modules are Scala
> > dependent module.
> >
> > If all modules will need the suffix after your PR is merged, why would
> you
> > talk about pure/non-pure distinction in the documentation? This adds
> > complexity and may cause confusion which at the moment can be spared. If
> in
> > the future a "pure Java module" (without) is introduced, the
> documentation
> > can be updated accordingly.
> >
> > My two cents regarding the "optional suffix policy" are that thinks
> should
> > be kept simple. If Scala has penetrated 100% percent of the current
> > modules, new modules should be "proactive" and anticipate that this will
> > likely happen to them in the future, even if they start as "pure Java". I
> > prefer a strict policy that makes everything clear for the user (even if
> it
> > is a bit inefficient in terms of packaging and release management) as
> > opposed to a naming convention based on implementation-details.
> >
> > Regards,
> > A.
> >
> >
> > 2015-07-05 13:16 GMT+02:00 Chiwan Park <chiwanpark@apache.org>:
> >
> >> @Stephan: Okay, I’ll find the mentionings in other document. I think
> that
> >> we
> >> can postpone updating downloads page in flink-web until releasing 0.10.
> >>
> >> @Alexandar Thank you for comments. I’ll apply your suggestions.
> >>
> >> In your example, *flink-pure-java* is not pure java module. If there is
> >> any need
> >> of linkage with Scala dependent module in some module, the module is
> also
> >> Scala dependent module. Because we are using Scala in our runtime, all
> >> modules are Scala dependent module.
> >>
> >> So in your example, *flink-some-scala-A*, *flink-some-scala-B*, and
> >> *flink-pure-java* should have a suffix `_2.11` if the user want to run
> in
> >> Flink
> >> with Scala 2.11. (In Scala 2.10, we don’t need it.)
> >>
> >> I agree that it makes too many modules. But it is clear in user
> >> perspective. The
> >> users just decide which Scala version to use their cluster and add a
> >> suffix to
> >> all dependency if the version is 2.11.
> >>
> >> Regards,
> >> Chiwan Park
> >>
> >>> On Jul 3, 2015, at 9:26 PM, Stephan Ewen <sewen@apache.org> wrote:
> >>>
> >>> @Chiwan:
> >>>
> >>> There are a few mentionings of the Scala version in the docs as well.
> For
> >>> example in "docs/index.md" and on the website under "downloads".
> >>>
> >>> We should make sure we explain on these pages that there are downloads
> >> for
> >>> various Scala versions.
> >>>
> >>> Cheers,
> >>> Stephan
> >>>
> >>>
> >>> On Fri, Jul 3, 2015 at 2:01 PM, Alexander Alexandrov <
> >>> alexander.s.alexandrov@gmail.com> wrote:
> >>>
> >>>> Great, I just posted some comments / improvement suggestions.
> >>>>
> >>>> I have to say I'm still not 100% convinced by the strategy not to add
> a
> >>>> suffix to all modules. Here is a small example that illustrates my
> >>>> concerns.
> >>>>
> >>>> Consider the following chained dependency situation. We have pure Java
> >>>> artifact *flink-pure-java* which depends on a Scala artifact
> >>>> *flink-some-scala-A*, which in turn depends on *flink-some-scala-B*.
> >>>>
> >>>> Let's say the user has directly included *flink-pure-java* and
> >>>> *flink-some-scala-B* in the his project and wants to build for Scala
> >> 2.11.
> >>>> We end up with a situation like this
> >>>>
> >>>> - flink-pure-java
> >>>> `- flink-some-scala-A
> >>>>    `- flink-some-scala-B
> >>>> - flink-some-scala-B_2.11
> >>>>
> >>>> We end up having both versions of *flink-some-scala-B* in our project.
> >>>>
> >>>>
> >>>>
> >>>> 2015-07-03 12:24 GMT+02:00 Chiwan Park <chiwanpark@apache.org>:
> >>>>
> >>>>> Hi All,
> >>>>> I created a PR for this issue. [1] Please check and comment about
the
> >> PR.
> >>>>>
> >>>>> Regards,
> >>>>> Chiwan Park
> >>>>>
> >>>>> [1] https://github.com/apache/flink/pull/885
> >>>>>
> >>>>>> On Jul 2, 2015, at 5:59 PM, Chiwan Park <chiwanpark@apache.org>
> >> wrote:
> >>>>>>
> >>>>>> @Alexander I’m happy to hear that you want to help me. If
you help
> me,
> >>>> I
> >>>>> really appreciate. :)
> >>>>>>
> >>>>>> Regards,
> >>>>>> Chiwan Park
> >>>>>>
> >>>>>>
> >>>>>>> On Jul 2, 2015, at 2:57 PM, Alexander Alexandrov <
> >>>>> alexander.s.alexandrov@gmail.com> wrote:
> >>>>>>>
> >>>>>>> @Chiwan: let me know if you need hands-on support. I'll
be more
> then
> >>>>> happy to help (as my downstream project is using Scala 2.11).
> >>>>>>>
> >>>>>>> 2015-07-01 17:43 GMT+02:00 Chiwan Park <chiwanpark@apache.org>:
> >>>>>>> Okay, I will apply this suggestion.
> >>>>>>>
> >>>>>>> Regards,
> >>>>>>> Chiwan Park
> >>>>>>>
> >>>>>>>> On Jul 1, 2015, at 5:41 PM, Ufuk Celebi <uce@apache.org>
wrote:
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> On 01 Jul 2015, at 10:34, Stephan Ewen <sewen@apache.org>
wrote:
> >>>>>>>>
> >>>>>>>>> +1, like that approach
> >>>>>>>>
> >>>>>>>> +1
> >>>>>>>>
> >>>>>>>> I like that this is not breaking for non-Scala users
:-)
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>
> >>
> >>
> >>
> >>
> >>
>
>
>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message