[SPARK-28243][PYSPARK][ML][FOLLOW-UP] Move Python DecisionTreeParams to regression.py

[huaxingao]

What changes were proposed in this pull request?

Leave shared.py untouched. Move Python DecisionTreeParams to regression.py

How was this patch tested?

Use existing tests

[huaxingao]

I move DecisionTreeParams to regression.py for now. May need to move again after we re-org the python code.

[SparkQA]

Test build #108927 has finished for PR 25406 at commit 45b9575.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
• class DecisionTreeParams(Params):

[srowen]

Maybe @mengxr or @brkyvz have a thought on this, having originally set up this mechanism. Is it reasonable to move to just defining these in the normal class hierarchy?

[huaxingao]

@zhengruifeng

[zhengruifeng]

@huaxingao Do you update the shared.py by run the cmd?

[huaxingao]

Yes, shared.py was generated by running the cmd. @zhengruifeng

[zhengruifeng]

I am Ok with this modification, however the DecisionTreeParams in the py side is still not sync with the scala side. We can leave this confilict alone, and make them in sync in the future.

[srowen]

Eh, yeah this is my fault for not realizing this is how the parameters are defined in some cases. I imagine there are other sync problems. It'd be great to re-sync them, though I do wonder whether this mechanism is worth it. That can happen later if you're willing. I think this is OK.

[srowen]

Merged to master

[huaxingao]

Thanks @srowen @zhengruifeng
I will be more careful next time.