[SPARK-49086][CONNECT] Move ML function registration to SparkSessionExtensions
What changes were proposed in this pull request?
This PR moves ML function registration from the SparkConnectPlanner to the internal function registry. This registration is done using the SparkSessionExtensions mechanism.
Why are the changes needed?
Unification of Connect and Classic Column API. This PR decouples the ML functions from Catalyst expressions.
Does this PR introduce any user-facing change?
No.
How was this patch tested?
Existing tests.
Was this patch authored or co-authored using generative AI tooling?
No.
@zhengruifeng this is failing org.apache.spark.ml.FunctionsLoadingSuite. Can I remove this? I can't make it error out anymore after this.
@hvanhovell I think we can remove test org.apache.spark.ml.FunctionsLoadingSuite introduced in https://github.com/apache/spark/pull/43739, because we don't need objects vectorToArrayUdf and vectorToArrayFloatUdf any more
also cc @WeichenXu123 and @zsxwing who are the reviewers of https://github.com/apache/spark/pull/43739
@hvanhovell wanted to merge this but there's a conflict ..
Merging to master!