[
https://issues.apache.org/jira/browse/SPARK-8439?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14591678#comment-14591678
]
Eugen Cepoi commented on SPARK-8439:
------------------------------------
"but then what would you do about the inconsistency problem? some methods would
then use one, others use the other. That's a bigger problem"
So basically my opinion is that addInputPaths is more permissive and shouldn't
break any code (except if people where adding paths and expected spark to
override it... ?!). Thus using everywhere addInputPaths would keep consistency
accross the API and preserve compatibility.
> SPARK-7155 - supporting comma separated paths, breaks compatibility
> -------------------------------------------------------------------
>
> Key: SPARK-8439
> URL: https://issues.apache.org/jira/browse/SPARK-8439
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
> Affects Versions: 1.4.0
> Reporter: Eugen Cepoi
>
> The improvement to support comma separated paths, replaced code that was
> adding paths to the config by code that is setting them.
> This breaks user code that added by hand the paths using
> FileInputFormat.addInputPath(...).
> The fix would be to use addInputPaths instead, so people can still add the
> paths using other mechanisms instead of constructing comma separated strings.
> The PR that introduces the regression:
> https://github.com/apache/spark/pull/5708
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
