szehon-ho opened a new pull request, #58439:
URL: https://github.com/apache/spark/pull/58439

   ### What changes were proposed in this pull request?
   
   This follow-up addresses two comments from the post-merge review of #58370:
   
   - Resolve and cache `fullyPushedFilterAttributes()` when 
`runtimeFilterAttrs` is accessed, then reuse the cached result from 
`fullyPushedRuntimeFilterAttrs`. The regression now exercises the primary 
planning entry point ([review 
comment](https://github.com/apache/spark/pull/58370#discussion_r3892275062)).
   - Clarify that V1 filter names use dot-separated paths with each path part 
quoted as needed, for example `parent.\`child.with.dot\`` ([review 
comment](https://github.com/apache/spark/pull/58370#discussion_r3892275058)).
   
   ### Why are the changes needed?
   
   The primary planning path checked whether fully-pushed references were 
top-level but did not resolve them. A missing top-level reference could 
therefore pass validation when no scalar-subquery filter forced 
`fullyPushedRuntimeFilterAttrs`, making validation depend on the query shape.
   
   The V1 Javadoc also described nested paths as entirely unquoted even though 
`FieldReference.toString` quotes individual path parts when needed.
   
   ### Does this PR introduce _any_ user-facing change?
   
   Yes. A `SupportsRuntimeCatalystFiltering` scan that declares a missing 
fully-pushed attribute now fails consistently during planning. Valid scan 
implementations are unchanged. The V1 runtime-filter Javadoc now describes 
nested attribute names accurately.
   
   ### How was this patch tested?
   
   ```
   build/sbt 'sql/testOnly 
org.apache.spark.sql.connector.DataSourceV2CatalystRuntimeFilterSuite -- -z 
"missing fully pushed filter attribute"'
   ```
   
   ### Was this patch authored or co-authored using generative AI tooling?
   
   Generated-by: Codex with GPT-5
   


-- 
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