kbendick commented on PR #5009:
URL: https://github.com/apache/iceberg/pull/5009#issuecomment-1155440960

   This changes some of the existing behavior:
   - `changed-partition-count` does not include "unpartitioned" - so the number 
of changed partitions could potentially go down by 1 in some scenarios (or not 
be present at all).
     -  We can decouple the function that writes the `partition.` stats and the 
count of changed partitions to keep the old behavior.
   - Stops writing partition summaries for unpartitioned tables (the goal of 
the PR). So there won't be a `partitions.` summary in the file for the 
unpartitioned section of the table.
     - Void partition transforms (dropped partitions) in v1 tables will still 
be included in partition summary stats as they transform to `alwaysNull` and 
where the change was made, it's not possible to differentiate between an actual 
partition field with a null value vs the void transform.
     - This does not affect v2 tables which don't use the void transform
   
   The big one is the `changed-partition-count`. We can decouple the logic so 
that `changed-partition-count` is still 1 for the unpartitioned parts of the 
table when they are touched (technically at present, if a V2 table has multiple 
unpartitioned specs, there will only be 1 value in total added to 
`changed-partition-count`, instead of the value being incremented for each 
unpartitioned section of the table).
   
   If anybody has a real-world use case where the original behavior is needed, 
I'm very open to hearing it. My goal is just to remove `partition.` partition 
summary fields caused by the PartitionSpec for `PartitionSpec.unpartitioned()` 
being defined as an empty list of fields.


-- 
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.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to