sunchao commented on PR #5403:
URL: 
https://github.com/apache/datafusion-comet/pull/5403#issuecomment-5721091005

   Updated in 
[ee57cac60](https://github.com/apache/datafusion-comet/commit/ee57cac609d6e42fbbf136ca43f690e7443a00b5),
 with the main merge in 
[5522ab750](https://github.com/apache/datafusion-comet/commit/5522ab750a33f23e47811f1a9011effc0ac7c734).
   
   @andygrove The `CodegenDispatchFallback` mixin is appropriate here. Both 
extrema serdes now use it for non-default collations, including nested arrays 
and structs. Spark's bound expression preserves the collation IDs through 
closure serialization, and its generated comparator supplies the 
collation-aware ordering. Ordinary binary-string and floating-point inputs 
continue to select the native implementation; `hasConditionalNativeDefault` 
keeps that distinction in the generated docs.
   
   The collation SQL fixtures now require `expect_dispatch(...)`, with 
`expect_native(...)` controls for binary strings and floats. A separate 
regression checks fallback with the dispatcher disabled in both strict modes. 
The new JVM-only regression covers all eight min/max and nested-type 
combinations through serialization and code generation.
   
   @rich7420 The conflicts are resolved against main `3b942e593`, preserving 
the newer benchmark registrations, function imports, and documentation. This 
brings the branch to DataFusion **55.1.0** / Arrow **59.2.0**. The unrelated 
Cargo.toml final-newline change is also removed.
   
   Validation: **64 JVM-only tests passed on Spark 4.1.3**, including the new 
regression; production/test compilation, Spotless, Scalastyle, Rust and 
changed-Markdown formatting, and the PR diff check passed. Spark 4.1 
documentation generation passed; the generated extrema sections correctly 
describe conditional codegen dispatch.
   
   Native execution remains unvalidated on this head: the configured Cargo 
registry lacks DataFusion 55.1.0, so the focused Rust command stopped during 
dependency resolution and ran zero tests. That also blocks rebuilding JNI and 
running the Spark integration fixtures locally. The description now separates 
these current checks from the historical native tests and performance 
measurements.
   


-- 
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