[SPARK-16005][R] Add randomSplit to SparkR

[dongjoon-hyun]

What changes were proposed in this pull request?

This PR adds randomSplit to SparkR for API parity.

How was this patch tested?

Pass the Jenkins tests (with new testcase.)

[SparkQA]

Test build #60669 has finished for PR 13721 at commit 3fba8b8.

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

[felixcheung]

Could you add #' @note since 2.0.0 - we are trying to add that for 2.0, there should be some example already. I'm going to do a pass on that later - btw, you are welcome to take on that too!

[dongjoon-hyun]

Sure!
And, thank you for review, @felixcheung !

[felixcheung]

looks good!

[dongjoon-hyun]

Thank you, @felixcheung .
For SPARK-14995, I'll do that tonight. It looks good as an exercise for me.
Thank you for let me know that.

[SparkQA]

Test build #60670 has finished for PR 13721 at commit 5eff72b.

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

[SparkQA]

Test build #60671 has finished for PR 13721 at commit 0c0e00b.

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

[SparkQA]

Test build #60673 has finished for PR 13721 at commit 2e1a3cf.

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

[dongjoon-hyun]

Now, it passed all tests and become ready for review again.
Could you review this PR, @shivaram ?

[shivaram]

nit: seed to use for random split will be better here

[dongjoon-hyun]

Yep.

[SparkQA]

Test build #60682 has finished for PR 13721 at commit eed4b1f.

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

[SparkQA]

Test build #60684 has finished for PR 13721 at commit 09a079c.

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

[dongjoon-hyun]

Hi, @shivaram . The followings are updated and become ready for review again.

• The param description is improved.
• The size and ratio of returned list are compared with those of weights.

[dongjoon-hyun]

Hi, @shivaram .
Although it seems to be late for becoming a part of the Spark 2.0.0, could you review again ?

[shivaram]

Could we explain that the dataframes are split by rows ? Its not really clear what randomSplit means without some context otherwise.

[dongjoon-hyun]

Yes. Given 1000 rows, randomSplit(c(2,3,5)) will return three dataframe having approximately 200, 300, 500 rows .

[dongjoon-hyun]

@shivaram . I added the description. Thank you for review!

[shivaram]

Thanks @dongjoon-hyun - LGTM. Will merge once Jenkins passes

[SparkQA]

Test build #60731 has finished for PR 13721 at commit 019cbcf.

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

[dongjoon-hyun]

Thank you for merging, @shivaram .