[SPARK-9549][SQL] fix bugs in expressions

[yjshen]

JIRA: https://issues.apache.org/jira/browse/SPARK-9549

This PR fix the following bugs:

1. UnaryMinus's codegen version would fail to compile when the input is Long.MinValue
2. BinaryComparison would fail to compile in codegen mode when comparing Boolean types.
3. AddMonth would fail if passed a huge negative month, which would lead accessing negative index of monthDays array.
4. Nanvl with different type operands.

[rxin]

add one for int/short/byte type too

[rxin]

could this throw a bad exception if the input is some other type that is not Double/Float?

[yjshen]

I have this in Nanvl:

override def inputTypes: Seq[AbstractDataType] =
    Seq(TypeCollection(DoubleType, FloatType), TypeCollection(DoubleType, FloatType))

[rxin]

can you double check that FunctionArgumentConversion is not called before the implicit cast rule?

[SparkQA]

Test build #39500 has finished for PR 7882 at commit 4fa5de0.

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

[SparkQA]

Test build #39505 has finished for PR 7882 at commit 3dee204.

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

[rxin]

LGTM

[SparkQA]

Test build #39514 has finished for PR 7882 at commit 41bbd2c.

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

[rxin]

Thanks - I've merged this.

[tedyu]

The above two statements can be replaced with:
val (currentYear, currentMonthInYear) = nonNegativeMonth /% 12

[rxin]

How is that possible?

[tedyu]

/% is not defined for Int.

I read the notion in a Scala book which I have returned. I will read more once I have that book back.