I’m planning on porting the existing Java coder to Python. Any objections
to that?
-chad
On Sun, Sep 29, 2019 at 1:02 PM Robert Burke wrote:
> +1
>
> I'm happy to whip together the Go SDK version once the encoding has been
> concretely decided.
>
> On Fri, Sep 27, 2019, 6:07 PM Chad Dombrova
+1
I'm happy to whip together the Go SDK version once the encoding has been
concretely decided.
On Fri, Sep 27, 2019, 6:07 PM Chad Dombrova wrote:
>
> It would still be a standard coder - the distinction I'm proposing is that
>> there are certain coders that _must_ be implemented by a new
Jan,
The fact that the annotation on the ParDo "changes" the GroupByKey
implementation is very specific to the Spark runner implementation. You can
imagine another runner that simply writes out files in HDFS to implement a
GroupByKey - this GroupByKey implementation is agnostic whether the result
Hi,
Thanx for the reply.
So – re-iteration on grouped elements is a runner-dependent. Flink & DataFlow
allows it, while Spark isn’t.
Since we investigating here the runners also, Does anyone have a list which
runner allow\not-allow re-iteration?
Noam
From: [apache.org] Kenneth Knowles
