[SPARK-7952][SQL] use internal Decimal instead of java.math.BigDecimal

[cloud-fan]

This PR fixes a bug introduced in #6505.
Decimal literal's value is not java.math.BigDecimal, but Spark SQL internal type: Decimal.

[cloud-fan]

cc @marmbrus @rxin

[rxin]

LGTM

[marmbrus]

LGTM

[rxin]

btw we should rename this BooleanEquality instead of "Equalization".

[cloud-fan]

Just realized decimal literal's type is not always DecimalType.Unlimited. User can do something like Literal.create(Decimal(1), DecimalType(8, 0)). Should we handle this case too? If so, I think we still need to compare the wrapped value for literals.

[rxin]

We should handle those as well ...

[SparkQA]

Test build #33960 has finished for PR 6574 at commit 4d35e58.

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

[SparkQA]

Test build #33963 has finished for PR 6574 at commit 90eeb25.

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

[cloud-fan]

We only care about boolean type and literal type here, not left and right, so I use bool instead.

[SparkQA]

Test build #34392 has finished for PR 6574 at commit b0e3549.

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

[rxin]

Thanks. Merging this in.

[rxin]

btw I see the following compilation warning for master branch just now. Can you take a look?

[warn] /scratch/rxin/spark/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/HiveTypeCoercion.scala:448: fruitless type test: a value of type org.apache.spark.sql.catalyst.expressions.Expression cannot also be a org.apache.spark.sql.types.DecimalType
[warn]         case b @ BinaryComparison(e1 @ DecimalType.Fixed(_, _), e2)
[warn]                                                         ^
[warn] /scratch/rxin/spark/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/HiveTypeCoercion.scala:451: fruitless type test: a value of type org.apache.spark.sql.catalyst.expressions.Expression cannot also be a org.apache.spark.sql.types.DecimalType
[warn]         case b @ BinaryComparison(e1, e2 @ DecimalType.Fixed(_, _))

[cloud-fan]

It's a mistake made in #6405, we should use DecimalType.Expression instead of DecimalType.Fixed to match expression. Sorry about it...

[SparkQA]

Test build #34406 has finished for PR 6574 at commit d7b60c8.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
  • case class HasCachedBlocks(executorId: String) extends ToBlockManagerMaster
  • class KernelDensity(object):

[rxin]

Can you submit a separate pull request? This patch has actually been merged already, but ASF's github sync is behind so it is not showing up here.

[rxin]

Please close this pull request also (since it's already been merged).

You can see the up-to-date merges here: https://git-wip-us.apache.org/repos/asf?p=spark.git