vustef commented on issue #10381:
URL: https://github.com/apache/arrow-rs/issues/10381#issuecomment-5043827101

   Thanks all, the plan sounds good to me as well. `RowNumberReader` will error 
out in case missing ordinals, which will include mixed row-groups, so erroring 
out in `OrdinalAssigner` is not needed.
   
   The only caveat, and I forgot how things actually work, but if it works by 
reusing deserialized data, then this might be a problem:
   1. Use a reader without row number. It's a file without rowgroup ordinal, on 
all row groups. OrdinalAssigner is skipped.
   2. Now reuse the same metadata, but with row number virtual column. This 
fails, because there are no row group ordinals. It would've failed if this was 
the first read that deserialized metadata.
   
   So perhaps this constraint should be lifted:
   ```
   Ordinals are assigned only when needed for row numbering (or already set in 
metadata)
   ```
   It's enough that it doesn't error out, but we could keep ordinal assigning 
the best effort regardless of whether row numbering is used?


-- 
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]

Reply via email to