[
https://issues.apache.org/jira/browse/BEAM-463?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15892939#comment-15892939
]
Kenneth Knowles commented on BEAM-463:
--------------------------------------
What is motivating this? BoundedHeapCoder (and a few similar classes) all have
a serialized Java comparator inside them.
So if we did this, then we could get into a situation that is a new variation
on a well-defined coder that has some SDK-defined stuff inside it. The other
common case is like a KvCoder that has a custom coder for the key or value.
Only in this case it isn't a component coder but one of the parameters in the
coder's spec.
[~robertwb] and [~tgroh]] this touches on things we've talked about, regarding
where SDK-specific code should go
> BoundedHeapCoder should be a StandardCoder and not a CustomCoder
> ----------------------------------------------------------------
>
> Key: BEAM-463
> URL: https://issues.apache.org/jira/browse/BEAM-463
> Project: Beam
> Issue Type: Bug
> Components: sdk-java-core
> Reporter: Luke Cwik
> Assignee: Aviem Zur
> Priority: Minor
> Labels: backward-incompatible
>
> The issue is that BoundedHeapCoder does not report component encodings which
> prevents effective runner inspection of the components.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
