asolimando commented on issue #24461:
URL: https://github.com/apache/datafusion/issues/24461#issuecomment-6669385079

   > > 26. [#27108](https://github.com/apache/datafusion/issues/27108) is a 
notable regression with potential to be a release blocker, 
[@devanbenz](https://github.com/devanbenz) pinned the regression to 
[#25790](https://github.com/apache/datafusion/pull/25790), given the timeline I 
suggest to revert it, and file an issue to look into it after cutting 56.
   > 
   > yes let's revert it. Thank you [@asolimando](https://github.com/asolimando)
   > 
   > Cardinality estimation is known to be super tricky and this is why I want 
to try and get a API in place so people can add fancy formuals outside of the 
core
   
   Agreed, that's the whole idea behind the 
[`StatisticsRegistry`](https://github.com/apache/datafusion/issues/21443) and 
[`ExpressionAnalyzer/ExpressionSynopsis`](https://github.com/apache/datafusion/issues/21120),
 so we can keep the core simple, and provide more advanced formulas as opt-in 
providers, and people can override/add more downstream (as I agree with you 
that's hard/impossible to get it right for everyone with a single 
implementation).
   
   Re. #27108's revert, I have opened 
https://github.com/apache/datafusion/pull/27145 as a revert PR and filed 
https://github.com/apache/datafusion/issues/27146 to track a proper fix before 
re-integrating that change, happy to back-port to `branch-56` once the revert 
PR lands on `main`.


-- 
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