[SPARK-29535][SQL] ADD some aggregate functions for Column in RelationalGroupedDataset.scala

[TomokoKomiyama]

What changes were proposed in this pull request?

Add five aggregation functions with Column type parameters.

mean(Column, Column*)
max(Column, Column*)
avg(Column, Column*)
min(Column, Column*)
sum(Column, Column*)

Why are the changes needed?

If we want pass Column type parameters to some aggregation functions with agg(), but it's redundant.

df.groupBy("_c0").agg(max($"_c1"))

Other aggregation functions such as pivot() can use Column arguments, but these aggregation functions can't use it.
df.groupBy("_c0").max($"_c1")

Does this PR introduce any user-facing change?

Yes.

We will be able to pass Column type parameters to aggregation functions(mean, max, avg, min, sum) without agg().
df.groupBy("_c0").max($"_c1")

How was this patch tested?

Manually tested.

[AmplabJenkins]

Can one of the admins verify this patch?

[HyukjinKwon]

I won't add APIs just for consistency. You can use agg, right?

[TomokoKomiyama]

@HyukjinKwon
We can use agg, but I think it would be easier for users to use these aggregation functions with Colmun type in the similar way with String one.
df.groupBy("_c0").max("_c1")
df.groupBy("_c0").max($"_c1")
Other functions can pass String type parameters because of the legacy reason and these will be removed someday, right?
I think it's better to pass Column type parameters to these aggregation functions(min, max...) so that the user is not confused at the time when legacy functions removed.

[dongjoon-hyun]

Sorry, @TomokoKomiyama . I also agree with @HyukjinKwon .

cc @sarutak

[sarutak]

@dongjoon-hyun She doesn't mind closing this PR.

[maropu]

+1, too. I'll close this. Thanks.

[dongjoon-hyun]

Thank you all! Especially, @TomokoKomiyama and @sarutak .
IIRC, there was a discussion before similarly and we made the same decision (to minimize the API.)