[SPARK-6452] [SQL] Checks for missing attributes and unresolved operator for all types of operator

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

@@ -63,7 +63,7 @@ class CheckAnalysis {
               s"filter expression '${f.condition.prettyString}' " +
                 s"of type ${f.condition.dataType.simpleString} is not a boolean.")
 
-          case aggregatePlan@Aggregate(groupingExprs, aggregateExprs, child) =>
+          case Aggregate(groupingExprs, aggregateExprs, child) =>
             def checkValidAggregateExpression(expr: Expression): Unit = expr match {
               case _: AggregateExpression => // OK
               case e: Attribute if !groupingExprs.contains(e) =>
@@ -85,13 +85,18 @@ class CheckAnalysis {
 
             cleaned.foreach(checkValidAggregateExpression)
 
+          case _ => // Fallbacks to the following checks
+        }
+
+        operator match {
           case o if o.children.nonEmpty && o.missingInput.nonEmpty =>
-            val missingAttributes = o.missingInput.map(_.prettyString).mkString(",")
-            val input = o.inputSet.map(_.prettyString).mkString(",")
+            val missingAttributes = o.missingInput.mkString(",")
+            val input = o.inputSet.mkString(",")
 
-            failAnalysis(s"resolved attributes $missingAttributes missing from $input")
+            failAnalysis(
+              s"resolved attribute(s) $missingAttributes missing from $input " +
+                s"in operator ${operator.simpleString}")
 
-          // Catch all
           case o if !o.resolved =>
             failAnalysis(
               s"unresolved operator ${operator.simpleString}")

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/QueryPlan.scala

@@ -47,9 +47,12 @@ abstract class QueryPlan[PlanType <: TreeNode[PlanType]] extends TreeNode[PlanTy
    * Attributes that are referenced by expressions but not provided by this nodes children.
    * Subclasses should override this method if they produce attributes internally as it is used by
    * assertions designed to prevent the construction of invalid plans.
+   *
+   * Note that virtual columns should be excluded. Currently, we only support the grouping ID
+   * virtual column.
    */
-  def missingInput: AttributeSet = (references -- inputSet)
-    .filter(_.name != VirtualColumn.groupingIdName)
+  def missingInput: AttributeSet =
+    (references -- inputSet).filter(_.name != VirtualColumn.groupingIdName)
 
   /**
    * Runs [[transform]] with `rule` on all expressions present in this query operator.

sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/analysis/AnalysisSuite.scala

@@ -199,4 +199,22 @@ class AnalysisSuite extends FunSuite with BeforeAndAfter {
     assert(pl(3).dataType == DecimalType.Unlimited)
     assert(pl(4).dataType == DoubleType)
   }
+
+  test("SPARK-6452 regression test") {
+    // CheckAnalysis should throw AnalysisException when Aggregate contains missing attribute(s)
+    val plan =
+      Aggregate(
+        Nil,
+        Alias(Sum(AttributeReference("a", StringType)(exprId = ExprId(1))), "b")() :: Nil,
+        LocalRelation(
+          AttributeReference("a", StringType)(exprId = ExprId(2))))
+
+    assert(plan.resolved)
+
+    val message = intercept[AnalysisException] {
+      caseSensitiveAnalyze(plan)
+    }.getMessage
+
+    assert(message.contains("resolved attribute(s) a#1 missing from a#2"))
+  }
 }