pvary commented on pull request #1095:
URL: https://github.com/apache/hive/pull/1095#issuecomment-642930948
Main comment:
- If I understand correctly this gigapatch has 2 mostly distinct part:
- Run DDL queries in transactions, and generate writeIds for them
- Optimize writeId handling, so locally committed changes do not cause us
to refresh data from HMS
I would vote for separating these changes out to 2 standalone patches, so it
is easier to understand/review them.
Smaller comments:
- Commit message should contain the HIVE jira number. Check other pull
requests for help :)
- I think if we decide that we open/commit transaction for DDLs, we should
do this on the same place where we do it for DMLs, if possible - I hate that we
do this in arbitrary methods inside the Hive class
- I guess that the test changes highlight some basic behavioral change -
like somehow we start from writeId 2 after the patch. Is this a planned /
required change?
- Please remove formatting only changes from this jira
Thanks,
Peter
----------------------------------------------------------------
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
