[
https://issues.apache.org/jira/browse/BEAM-8616?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16973438#comment-16973438
]
Ismaël Mejía commented on BEAM-8616:
------------------------------------
It is, we have a similar change already on Kafka, for details: BEAM-3620
However we need to make this explicit in the release notes I think we have a
label for this. Impact for end users is the need to explicitly add the
hadoop-client dependency in their projects if they already don't have it in
their clusters.
> ParquetIO should have Hadoop dependencies as provided
> -----------------------------------------------------
>
> Key: BEAM-8616
> URL: https://issues.apache.org/jira/browse/BEAM-8616
> Project: Beam
> Issue Type: Sub-task
> Components: io-java-parquet
> Affects Versions: 2.16.0
> Reporter: Ismaël Mejía
> Assignee: Ismaël Mejía
> Priority: Major
> Time Spent: 50m
> Remaining Estimate: 0h
>
> ParquetIO has the hadoop-client dependency as a compile-time dependency
> however this dependency should be provided by the user as defined in
> parquet-hadoop. By pinning a hadoop version we are limiting users from
> providing different Hadoop jars (as they can with native Parquet), it also
> limits us from providing different hadoop versions to test that the Parquet
> module is compatible with Hadoop 3 (when it will).
> Note this is a 'backwards incompatible' change in the sense that users might
> need to explicitly provide the dependency from now on.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
