[
https://issues.apache.org/jira/browse/BEAM-1820?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Eugene Kirpichov closed BEAM-1820.
----------------------------------
Resolution: Won't Fix
Fix Version/s: Not applicable
The consensus on dev@ is that instead, PTransform's should always provide a
Coder for their output; and the natural generalization to Source is that
sources also always should provide a Coder for their output; letting the user
configure the Source with an explicit Coder if necessary.
> Source.getDefaultOutputCoder() should be @Nullable
> --------------------------------------------------
>
> Key: BEAM-1820
> URL: https://issues.apache.org/jira/browse/BEAM-1820
> Project: Beam
> Issue Type: Improvement
> Components: sdk-java-core
> Reporter: Eugene Kirpichov
> Assignee: Łukasz Gajowy
> Labels: easyfix, starter
> Fix For: Not applicable
>
>
> Source.getDefaultOutputCoder() returns a coder for elements produced by the
> source.
> However, the Source objects are nearly always hidden from the user and
> instead encapsulated in a transform. Often, an enclosing transform has a
> better idea of what coder should be used to encode these elements (e.g. a
> user supplied a Coder to that transform's configuration). In that case, it'd
> be good if Source.getDefaultOutputCoder() could just return null, and coder
> would have to be handled by the enclosing transform or perhaps specified on
> the output of that transform explicitly.
> Right now there's a bunch of code in the SDK and runners that assumes
> Source.getDefaultOutputCoder() returns non-null. That code would need to be
> fixed to instead use the coder set on the collection produced by
> Read.from(source).
> It all appears pretty easy to fix, so this is a good starter item.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
