[SPARK-8371][SQL] improve unit test for MaxOf and MinOf and fix bugs

[cloud-fan]

a follow up of #6813

[cloud-fan]

This test will fail at checkEvaluationWithGeneratedProjection for hash code not equal. I'm not sure why we want to compare the hash code, and I looked into the hashCode implementation of GenericRow and Row, they are different. I'm trying to figure it out.

[cloud-fan]

cc @rxin @davies

[cloud-fan]

The origin code is not right as super here refers to BinaryArithmetic, not Expression that we wanted.

[lianhuiwang]

yes,i think you are right. because their symbol is empty.

[SparkQA]

Test build #34923 has finished for PR 6825 at commit cd51a7d.

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

[SparkQA]

Test build #34933 has finished for PR 6825 at commit 03021bc.

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

[lianhuiwang]

i think at first it needs to check whether eval1 or eval2 isNull. and then calculate primitive.

[cloud-fan]

the code result is inserted after null check.

[cloud-fan]

retest it please.

[davies]

@cloud-fan There are some changes related to Max/Min in #6726 , is it useful for you? It will be good if you could merge them here.

[SparkQA]

Test build #34987 has finished for PR 6825 at commit f153380.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
  • case class IsNull(child: Expression) extends UnaryExpression with Predicate
  • case class IsNotNull(child: Expression) extends UnaryExpression with Predicate

[davies]

All expressions only work with InternalRow, which only contain internal types (int, long, UTF8String), no Date or Timestamp. so we don't need to test Date/Timestamp, or use DateUtils.fromDate() to do the conversion before calling expression on them.

[cloud-fan]

This test will fail at checkEvaluationWithGeneratedProjection for hash code not equal. I'm not sure why we want to compare the hash code, and I looked into the hashCode implementation of GenericRow and Row, they are different. Does anyone know why we want to compare the hash code there?

[SparkQA]

Test build #35044 has finished for PR 6825 at commit 97066a6.

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

[cloud-fan]

@davies , we need the conversion here, as we use String instead of UTF8String at many tests...

[davies]

Sound reasonable, it's anonying to have so many UTF8String.fromString in test cases.

[SparkQA]

Test build #35045 has finished for PR 6825 at commit 8c3fbde.

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

[davies]

@cloud-fan Now #6876 is merged, could you update this PR?

[SparkQA]

Test build #35636 has finished for PR 6825 at commit 43170cc.

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

[davies]

Good catch!

[davies]

LGTM, the failed PySpark test is not related, merging this into master, thanks!

[SparkQA]

Test build #955 has finished for PR 6825 at commit 43170cc.

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