BryanCutler commented on pull request #7702:
URL: https://github.com/apache/arrow/pull/7702#issuecomment-657288029
>The only place in Arrow/Java where the metadata version gets checked is in
MessageSerializer.deserializeMessageBatch, which only gets used in a unit test
and isn't used in Spark, so I suppose we got lucky ("lucky") here.
Yeah, you're right. Looks like the `MetadataVersion` does not get checked
when reading file or stream formats. That should be fixed right? The correct
behavior is to not allow reading metadata from a version other than the
current, except in this case where we allow v4 if it doesn't have unions. Is
that what is done in C++?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
