Hi Ismaël,
Thanks for the update, No problem at all, please take your time and let me
know if my assistance is needed, The virus has affected everyone's
timetables. I hope you are safe.
Best Regards,
Pulasthi
On Fri, Apr 3, 2020 at 12:14 PM Ismaël Mejía wrote:
> Hello Pulasthi,
>
> Please
Hello Pulasthi,
Please excuse me for my delay, I have probably 1/3 of my common
available time since the coronavirus lockdown so I have not advanced
as expected. I hope to catch up rapidly and ping you. Our expected
target of merging it before the 2.21.0 release seems to be hard to get
at this
Hi Ismaël
Did you get some free time to perform a code review on the pull request
Best Regards
Pulasthi
On Tue, Mar 10, 2020 at 3:30 PM Luke Cwik wrote:
> I have to disagree. Allowing for runners within the Apache Beam repo and
> SDKs that reach into the implementation details of each other
I have to disagree. Allowing for runners within the Apache Beam repo and
SDKs that reach into the implementation details of each other are
usability, feature development, maintenance and complexity problems.
The usability issue comes from our public core facing APIs exposing methods
that runners
I do support all the efforts to get Dataflow, Flink, and Spark to 3 (Fn
API). But I disagree with it as a requirement; the whole point of
ptransforms with URNs is that if the runner can figure out how to execute
it according to semantics, then it is fine. A runner meets (1) and (2) but
can only
I would like to move away from having runners access APIs that are related
to pipeline construction and other internal SDK APIs and I would like for
SDKs to not inspect internal runner APIs. This would enable the community
to improve each independently without needing to fix the world all the time
There are a lot of different meanings to "portable runner". Here are some:
(1) A runner that accepts a pipeline proto and either runs it or says it
cannot run it
(2) A runner that accepts jobs via the job management APIs
(3) A runner that executes UDFs via the Fn API
(4) A runner that can execute
+1
On Tue, Mar 10, 2020 at 12:59 AM Alex Van Boxel wrote:
> One last thing, for any runner after this one... wouldn't it be a good
> acceptance criteria to only accept portable implementations anymore?
>
> _/
> _/ Alex Van Boxel
>
>
> On Mon, Mar 9, 2020 at 10:42 PM Ismaël Mejía wrote:
>
>>
One last thing, for any runner after this one... wouldn't it be a good
acceptance criteria to only accept portable implementations anymore?
_/
_/ Alex Van Boxel
On Mon, Mar 9, 2020 at 10:42 PM Ismaël Mejía wrote:
> Good points Kenn. I think we mostly agree on what has been discussed in
>
Good points Kenn. I think we mostly agree on what has been discussed in this
thread the pros/cons of having runners on our repository, but this is probably
not the best moment in time to change any policy in that aspect.
So if nobody objects I think we can proceed. I am OOO this week so with less
I haven't heard anyone suggest that we need a vote. I haven't heard anyone
object to this being merged to master. Some time ago, we mostly decided to
favor master instead of branches, because it is so much smoother for
contributors and users.
So I am poking this thread one last time and otherwise
I understand that the discussion is on a more broad level than the Twister2
runner. From my experience developing the runner the main advantage of
being inside the beam project was the easy access to the wide range of
tests and other core/utility code as Kyle pointed out. Unmerging runners
that
I think we will get to a point where it makes sense for runners to
live in their own repositories, with their own release cadence, but
we're not at that point yet. One prerequisite is a stable API--we're
closing in on that with the portability protos, but many (java)
runners actually share the
I agree with both of you, mostly :-)
The monorepo approach doesn't work/scale well for shipped libraries (name a
Google library that silently just works and never causes any dependency
problems) and the pain we feel has been constant and increasing, but I
don't think we are at the breaking point.
> Should runners, current and future, be in the same repository as Beam
> core?
In the distant past, runners lived in their own repositories, and then were
donated to Beam. But Beam's current uber-repo setup allows a lot of
convenience. For example, a ton of code (including core functionality and
Generic question without commenting on Twister2 specifically:
Should runners, current and future, be in the same repository as Beam
core? Can or should they be completely separate products with their
own release cycles?
Generally, loose coupling leads to more maintainable, reliable
projects.
Hi
I believe the pull request is pretty complete now with the help of Ismaël.
Kenn, would you be able to take a look at it and suggest any changes if
needed?. The build checks and validations tests are passing at the moment.
I will start working on the documentation that you mentioned in an
Hi All,
I have created the initial pull request [1] to contribute the Twister2 Beam
runner to the Apache Beam codebase. More information on Twister2 can be
found here[2] and the Twister2 codebase is available here[3]. At the moment
only batch mode is supported in the runner, but we are planning
18 matches
Mail list logo