[SPARK-7562][SPARK-6444][SQL] Improve error reporting for expression data type mismatch

It seems hard to find a common pattern of checking types in `Expression`. Sometimes we know what input types we need(like `And`, we know we need two booleans), sometimes we just have some rules(like `Add`, we need 2 numeric types which are equal). So I defined a general interface `checkInputDataTypes` in `Expression` which returns a `TypeCheckResult`. `TypeCheckResult` can tell whether this expression passes the type checking or what the type mismatch is.

This PR mainly works on apply input types checking for arithmetic and predicate expressions.

TODO: apply type checking interface to more expressions.

Author: Wenchen Fan <cloud0fan@outlook.com>

Closes #6405 from cloud-fan/6444 and squashes the following commits:

b5ff31b [Wenchen Fan] address comments
b917275 [Wenchen Fan] rebase
39929d9 [Wenchen Fan] add todo
0808fd2 [Wenchen Fan] make constrcutor of TypeCheckResult private
3bee157 [Wenchen Fan] and decimal type coercion rule for binary comparison
8883025 [Wenchen Fan] apply type check interface to CaseWhen
cffb67c [Wenchen Fan] to have resolved call the data type check function
6eaadff [Wenchen Fan] add equal type constraint to EqualTo
3affbd8 [Wenchen Fan] more fixes
654d46a [Wenchen Fan] improve tests
e0a3628 [Wenchen Fan] improve error message
1524ff6 [Wenchen Fan] fix style
69ca3fe [Wenchen Fan] add error message and tests
c71d02c [Wenchen Fan] fix hive tests
6491721 [Wenchen Fan] use value class TypeCheckResult
7ae76b9 [Wenchen Fan] address comments
cb77e4f [Wenchen Fan] Improve error reporting for expression data type mismatch

2 files changed, 2 insertions, 2 deletions

diff --git a/sql/core/src/main/scala/org/apache/spark/sql/json/InferSchema.scala b/sql/core/src/main/scala/org/apache/spark/sql/json/InferSchema.scala
index 06aa19ef09..565d10247f 100644
--- a/sql/core/src/main/scala/org/apache/spark/sql/json/InferSchema.scala
+++ b/sql/core/src/main/scala/org/apache/spark/sql/json/InferSchema.scala
@@ -147,7 +147,7 @@ private[sql] object InferSchema {
    * Returns the most general data type for two given data types.
    */
   private[json] def compatibleType(t1: DataType, t2: DataType): DataType = {
-    HiveTypeCoercion.findTightestCommonType(t1, t2).getOrElse {
+    HiveTypeCoercion.findTightestCommonTypeOfTwo(t1, t2).getOrElse {
       // t1 or t2 is a StructType, ArrayType, or an unexpected type.
       (t1, t2) match {
         case (other: DataType, NullType) => other
diff --git a/sql/core/src/main/scala/org/apache/spark/sql/json/JsonRDD.scala b/sql/core/src/main/scala/org/apache/spark/sql/json/JsonRDD.scala
index 95eb1174b1..7e1e21f5fb 100644
--- a/sql/core/src/main/scala/org/apache/spark/sql/json/JsonRDD.scala
+++ b/sql/core/src/main/scala/org/apache/spark/sql/json/JsonRDD.scala
@@ -155,7 +155,7 @@ private[sql] object JsonRDD extends Logging {
    * Returns the most general data type for two given data types.
    */
   private[json] def compatibleType(t1: DataType, t2: DataType): DataType = {
-    HiveTypeCoercion.findTightestCommonType(t1, t2) match {
+    HiveTypeCoercion.findTightestCommonTypeOfTwo(t1, t2) match {
       case Some(commonType) => commonType
       case None =>
         // t1 or t2 is a StructType, ArrayType, or an unexpected type.