FelixYBW commented on issue #13014:
URL: https://github.com/apache/gluten/issues/13014#issuecomment-5671695167

   The initial design is to make Gluten exactly compatible with Vanilla Spark's 
SQL or pyspark code. Gluten's first design principle is that user needn't 
change any line of Spark SQL or pyspark code.
   While your new proposal is to create a new use case for native Velox 
offloading only.
   
   I think it's OK that we have both explicit declaration and inferring the 
Java signature from the Velox C++ UDF. We just need to make sure
   
   1. Legacy Spark code `CREATE TEMPORARY FUNCTION ` doesn't break. Internally, 
we may reuse Spark's registration or overwrite it as a nop then inferring. But 
functionally it should behave exactly like Spark does.
   2. Document that if customers use only native UDF functions, they won't have 
the fallback path. After all, if they didn't create their Java implementation, 
Gluten definitely won't fall back. We can raise exceptions directly.


-- 
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