[SPARK-7499][SPARKR][WIP] Investigate how to specify columns in SparkR without $ or strings.

[sun-rui]

This WIP PR is inspired by https://issues.apache.org/jira/browse/SPARK-11260.

[shivaram]

cc @adrian555

[SparkQA]

Test build #46316 has finished for PR 9835 at commit 8e36661.

• This patch fails R style tests.
• This patch merges cleanly.
• This patch adds no public classes.

[adrian555]

I was going to work on this JIRA but was busy on other deliverable. :) Thanks @sun-rui

[adrian555]

Since 'select' is an external API, should we not put the code for the function after the roxygen codes?

[sun-rui]

this is a WIP, I will update doc and code once all test cases pass

[felixcheung]

I think the goal is to have the same for not only select, but for agg and so on? Is there a way we generalize this? Also we shouldn't export the selectInternal methods

[adrian555]

I'd suggest use 'drop' as the argument name instead since that is more R-like.

[adrian555]

@felixcheung Yes, I agree. We should make this work for all dplyr-like functions.

[felixcheung]

these seem very expensive to have this per each call?

[sun-rui]

yes, will have a cache for the environment per DataFrame.

[sun-rui]

I realize that this takes more effort than I expected as we'd better support specifying columns without $ for all DataFrame methods that take any argument of "Column" type. I will do more investigation.

[SparkQA]

Test build #46600 has finished for PR 9835 at commit 17aa666.

• This patch fails R style tests.
• This patch merges cleanly.
• This patch adds no public classes.

[shivaram]

@sun-rui Could we close this PR for now if its not active ? It'll just clear the PR queue.

[sun-rui]

Close it. Will send new PR later.