blinding-pixels commented on issue #24120:
URL: https://github.com/apache/datafusion/issues/24120#issuecomment-5389202638
So,I ran the example from this issue on the latest `main`, and the reported
behavior no longer reproduces. The narrow-schema filter and physical-schema
control both scanned 152 bytes, while reading the complete nested value scanned
319 bytes. This suggests #24125 already addressed the specific example,
although this does not prove that every possible nested shape is covered. I
haven't checked for other shapes yet.
I also tested the runtime failure path by making
`RowFilterGenerator::build()` return no row filter. Planning still removed the
outer filter, and all three rows were returned instead of the single matching
row. This confirms there is currently no general correctness fallback if
runtime row-filter construction fails.
Before starting work, could you clarify whether this issue still has
specific nested cases that need optimization, whether the planning/runtime
correctness fallback should be implemented first or included here, or whether
the issue is now fully addressed?
--
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]
