[SPARK-17177][SQL] Make grouping columns accessible from RelationalGroupedDataset

[NarineK]

What changes were proposed in this pull request?

Currently, once we create RelationalGroupedDataset, we cannot access the grouping columns from its instance.
Analog to Dataset we can have a public method which returns the list of grouping columns.
https://github.com/apache/spark/blob/master/sql/core/src/main/scala/org/apache/spark/sql/Dataset.scala#L457

This can be useful for instance in SparkR when we want to have certain logic associated with the grouping columns, accessible from RelationalGroupedDataset.

---

Similar to Dataset.columns I've added RelationalGroupedDataset.columns method which makes grouping column names accessible.

How was this patch tested?

Unit tests in DataFrameAggregateSuite

[SparkQA]

Test build #64169 has finished for PR 14742 at commit 02d6bab.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[NarineK]

cc: @shivaram, @liancheng

[NarineK]

@liancheng, @rxin, Do you think adding columns to RelationalGroupedDataset is reasonable or should we find a workaround on R side ?

[gatorsmile]

This might not be a column (a named expression). Thus, this might not be useful to end users.

[gatorsmile]

Could we please close this PR now?

[SparkQA]

Test build #78035 has started for PR 14742 at commit 02d6bab.

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/SparkPullRequestBuilder/78035/
Test FAILed.

[NarineK]

Hi @gatorsmile, #14431 depends on this. Is there a way I can access the grouping columns from RelationalGroupedDataset ?

[gatorsmile]

Let us close this one. We can discuss how to resolve the issue in #14431

[NarineK]

yes, we can close this, but it would be great if you could help us to find a way to access the grouping columns from SparkR in #14431