[SPARK-14639][PYTHON][R] Add bround function in Python/R.

[dongjoon-hyun]

What changes were proposed in this pull request?

This issue aims to expose Scala bround function in Python/R API.
bround function is implemented in SPARK-14614 by extending current round function.
We used the following semantics from Hive.

public static double bround(double input, int scale) {
    if (Double.isNaN(input) || Double.isInfinite(input)) {
      return input;
    }
    return BigDecimal.valueOf(input).setScale(scale, RoundingMode.HALF_EVEN).doubleValue();
}

After this PR, pyspark and sparkR also support bround function.

PySpark

>>> from pyspark.sql.functions import bround
>>> sqlContext.createDataFrame([(2.5,)], ['a']).select(bround('a', 0).alias('r')).collect()
[Row(r=2.0)]

SparkR

> df = createDataFrame(sqlContext, data.frame(x = c(2.5, 3.5)))
> head(collect(select(df, bround(df$x, 0))))
  bround(x, 0)
1            2
2            4

How was this patch tested?

Pass the Jenkins tests (including new testcases).

[SparkQA]

Test build #56283 has finished for PR 12509 at commit e63e1fa.

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

[felixcheung]

is this changing the existing behavior?

[dongjoon-hyun]

Thank you, @felixcheung !
I just tried to match the example between round and bround.
round does not change anything.

[SparkQA]

Test build #56286 has finished for PR 12509 at commit d5c0b3d.

• This patch fails SparkR unit tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #56291 has finished for PR 12509 at commit fef1bb2.

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

[dongjoon-hyun]

Hi, @davies .
Could you review this PR? After your comments, I studied and finally learned how to expose function in Python and R.

And, if you don't mind, may I include some code of minor PR #12329 here too? That PR is so minor and needed to be merged other PR like this. If then, I will just close that PR.

[felixcheung]

should scale be default to 0 to be consistent with Python?
to do that, do

           signature(x = "Column"),
           function(x, scale = 0),

[dongjoon-hyun]

Oh, thank you for sharing know-how. I didn't know that how to do!
I will fix right now.

[shivaram]

Could you give a more detailed example of where it gets rounded down and where it gets rounded up ?

[dongjoon-hyun]

Sure. Thank you for review. I'll add more example.

[SparkQA]

Test build #56298 has finished for PR 12509 at commit 58a0558.

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

[davies]

LGTM

[dongjoon-hyun]

Thank you, @davies !
Oh, may I add the minor PR to here? Last time, I remember that you prefer that way.

[dongjoon-hyun]

I mean #12329 . I want to close that one in order to reduce committers' review & merge cost.

[SparkQA]

Test build #56319 has finished for PR 12509 at commit 077633f.

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

[davies]

Merging this into master, thanks!