sunchao commented on issue #5419:
URL: 
https://github.com/apache/datafusion-comet/issues/5419#issuecomment-5464375735

   Thanks for the additional reproducer. This is the same shuffle-fallback path 
that #5421 addresses. Its check runs after conversion and restores incompatible 
native partials when the final aggregate must run in Spark, including with the 
ordinary `CometShuffleManager`.
   
   The current #5421 also handles the three test-expectation questions. It 
separates compatibility by direction: native `COUNT` partials can still feed 
Spark finals, while decimal `SUM` and `AVG` fall back across that boundary, 
with the affected expectations updated. So those failures do not all call for 
changing the expected native aggregate count to zero.
   
   The existing regressions cover `collect_list`/`collect_set` with shuffle 
enabled but native hash partitioning ineligible, under AQE on and off. Your 
exact array-key percentile query would provide useful additional coverage.


-- 
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