[SPARK-17108][SQL]: Fix BIGINT and INT comparison failure in spark sql

Author: weiqingy

sql/catalyst/src/main/scala/org/apache/spark/sql/types/DataType.scala

@@ -78,6 +78,17 @@ abstract class DataType extends AbstractDataType {
   private[spark] def sameType(other: DataType): Boolean =
     DataType.equalsIgnoreNullability(this, other)
 
+  /**
+   * Check if two integers are compatible. Returns true if `right` can be converted to `left`.
+   */
+  private[spark] def isCompatibleIntegralType(left: DataType, right: DataType): Boolean = {
+    (left, right) match {
+      case (l, r) => l.isInstanceOf[IntegralType] && r.isInstanceOf[IntegralType] &&
+        l.defaultSize >= r.defaultSize
+      case _ => false
+    }
+  }
+
   /**
    * Returns the same data type but set all nullability fields are true
    * (`StructField.nullable`, `ArrayType.containsNull`, and `MapType.valueContainsNull`).
@@ -91,7 +102,8 @@ abstract class DataType extends AbstractDataType {
 
   override private[sql] def defaultConcreteType: DataType = this
 
-  override private[sql] def acceptsType(other: DataType): Boolean = sameType(other)
+  override private[sql] def acceptsType(other: DataType): Boolean = sameType(other) ||
+    isCompatibleIntegralType(this, other)
 }