Github user cloud-fan commented on the issue:
https://github.com/apache/spark/pull/18732
@icexelloss I think as an API, it's a little confusing that `@pandas_udf`
can define both `Series* -> Series` function and `DataFrame -> DataFrame`
function. Besides, to support `StructType` as the return type of `Series* ->
Series` function, I think we have to add an extra flag to `@pandas_udf`. For
the coming `DataFrame -> Scalar` pandas UDAF, we also need extra flags to
represent partial aggregate ability.
From my experience of Java/Scala API design, I think it's a bad idea to
have a method with many parameters as flags. We'd better have more methods. For
this case, `@pandas_udf`, `@pandas_grouped_udf` and `@pandas_udaf` looks better
to me.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
