[SPARK-25816][SQL] Fix attribute resolution in nested extractors

## What changes were proposed in this pull request?

Extractors are made of 2 expressions, one of them defines the the value to be extract from (called `child`) and the other defines the way of extraction (called `extraction`). In this term extractors have 2 children so they shouldn't be `UnaryExpression`s.

`ResolveReferences` was changed in this commit: apache@36b826f which resulted a regression with nested extractors. An extractor need to define its children as the set of both `child` and `extraction`; and should try to resolve both in `ResolveReferences`.

This PR changes `UnresolvedExtractValue` to a `BinaryExpression`.

## How was this patch tested?

added UT

Closes apache#22817 from peter-toth/SPARK-25816.

Authored-by: Peter Toth <[email protected]>
Signed-off-by: gatorsmile <[email protected]>
(cherry picked from commit ca2fca1)
Signed-off-by: gatorsmile <[email protected]>

Author: peter-toth

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/unresolved.scala

@@ -407,7 +407,10 @@ case class ResolvedStar(expressions: Seq[NamedExpression]) extends Star with Une
  *                   can be key of Map, index of Array, field name of Struct.
  */
 case class UnresolvedExtractValue(child: Expression, extraction: Expression)
-  extends UnaryExpression with Unevaluable {
+  extends BinaryExpression with Unevaluable {
+
+  override def left: Expression = child
+  override def right: Expression = extraction
 
   override def dataType: DataType = throw new UnresolvedException(this, "dataType")
   override def foldable: Boolean = throw new UnresolvedException(this, "foldable")

sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala

@@ -2590,4 +2590,11 @@ class DataFrameSuite extends QueryTest with SharedSQLContext {
         Row ("abc", 1))
     }
   }
+
+  test("SPARK-25816 ResolveReferences works with nested extractors") {
+    val df = Seq((1, Map(1 -> "a")), (2, Map(2 -> "b"))).toDF("key", "map")
+    val swappedDf = df.select($"key".as("map"), $"map".as("key"))
+
+    checkAnswer(swappedDf.filter($"key"($"map") > "a"), Row(2, Map(2 -> "b")))
+  }
 }