[SPARK-18814][SQL] CheckAnalysis rejects TPCDS query 32

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

@@ -124,6 +124,10 @@ trait CheckAnalysis extends PredicateHelper {
                 s"Scalar subquery must return only one column, but got ${query.output.size}")
 
           case s @ ScalarSubquery(query, conditions, _) if conditions.nonEmpty =>
+
+            // Collect the columns from the subquery for further checking.
+            var subqueryColumns = conditions.flatMap(_.references).filter(query.output.contains)
+
             def checkAggregate(agg: Aggregate): Unit = {
               // Make sure correlated scalar subqueries contain one row for every outer row by
               // enforcing that they are aggregates which contain exactly one aggregate expressions.
@@ -136,24 +140,35 @@ trait CheckAnalysis extends PredicateHelper {
                 failAnalysis("The output of a correlated scalar subquery must be aggregated")
               }
 
-              // SPARK-18504: block cases where GROUP BY columns
-              // are not part of the correlated columns
-              val groupByCols = ExpressionSet.apply(agg.groupingExpressions.flatMap(_.references))
-              val predicateCols = ExpressionSet.apply(conditions.flatMap(_.references))
-              val invalidCols = groupByCols.diff(predicateCols)
+              // SPARK-18504/SPARK-18814: Block cases where GROUP BY columns
+              // are not part of the correlated columns.
+              val groupByCols = AttributeSet(agg.groupingExpressions.flatMap(_.references))
+              val correlatedCols = AttributeSet(subqueryColumns)
+              val invalidCols = groupByCols -- correlatedCols
               // GROUP BY columns must be a subset of columns in the predicates
               if (invalidCols.nonEmpty) {
                 failAnalysis(
-                  "a GROUP BY clause in a scalar correlated subquery " +
+                  "A GROUP BY clause in a scalar correlated subquery " +
                     "cannot contain non-correlated columns: " +
                     invalidCols.mkString(","))
               }
             }
 
-            // Skip projects and subquery aliases added by the Analyzer and the SQLBuilder.
+            // Skip subquery aliases added by the Analyzer and the SQLBuilder.
+            // For projects, do the necessary mapping and skip to its child.
             def cleanQuery(p: LogicalPlan): LogicalPlan = p match {
               case s: SubqueryAlias => cleanQuery(s.child)
-              case p: Project => cleanQuery(p.child)
+              case p: Project =>
+                // SPARK-18814: Map any aliases to their AttributeReference children
+                // for the checking in the Aggregate operators below this Project.
+                subqueryColumns = subqueryColumns.map {
+                  xs => p.projectList.collectFirst {
+                    case e @ Alias(child : AttributeReference, _) if e.exprId == xs.exprId =>
+                      child
+                  }.getOrElse(xs)
+                }
+
+                cleanQuery(p.child)
               case child => child
             }
 

sql/core/src/test/resources/sql-tests/inputs/scalar-subquery.sql

@@ -0,0 +1,20 @@
+CREATE OR REPLACE TEMPORARY VIEW p AS VALUES (1, 1) AS T(pk, pv);
+CREATE OR REPLACE TEMPORARY VIEW c AS VALUES (1, 1) AS T(ck, cv);
+
+-- SPARK-18814.1: Simplified version of TPCDS-Q32
+SELECT pk, cv
+FROM   p, c
+WHERE  p.pk = c.ck
+AND    c.cv = (SELECT avg(c1.cv)
+               FROM   c c1
+               WHERE  c1.ck = p.pk);
+
+-- SPARK-18814.2: Adding stack of aggregates
+SELECT pk, cv
+FROM   p, c
+WHERE  p.pk = c.ck
+AND    c.cv = (SELECT max(avg)
+	       FROM   (SELECT   c1.cv, avg(c1.cv) avg
+		       FROM     c c1
+		       WHERE    c1.ck = p.pk
+                       GROUP BY c1.cv));

sql/core/src/test/resources/sql-tests/results/scalar-subquery.sql.out

@@ -0,0 +1,46 @@
+-- Automatically generated by SQLQueryTestSuite
+-- Number of queries: 4
+
+
+-- !query 0
+CREATE OR REPLACE TEMPORARY VIEW p AS VALUES (1, 1) AS T(pk, pv)
+-- !query 0 schema
+struct<>
+-- !query 0 output
+
+
+
+-- !query 1
+CREATE OR REPLACE TEMPORARY VIEW c AS VALUES (1, 1) AS T(ck, cv)
+-- !query 1 schema
+struct<>
+-- !query 1 output
+
+
+
+-- !query 2
+SELECT pk, cv
+FROM   p, c
+WHERE  p.pk = c.ck
+AND    c.cv = (SELECT avg(c1.cv)
+               FROM   c c1
+               WHERE  c1.ck = p.pk)
+-- !query 2 schema
+struct<pk:int,cv:int>
+-- !query 2 output
+1	1
+
+
+-- !query 3
+SELECT pk, cv
+FROM   p, c
+WHERE  p.pk = c.ck
+AND    c.cv = (SELECT max(avg)
+	       FROM   (SELECT   c1.cv, avg(c1.cv) avg
+		       FROM     c c1
+		       WHERE    c1.ck = p.pk
+                       GROUP BY c1.cv))
+-- !query 3 schema
+struct<pk:int,cv:int>
+-- !query 3 output
+1	1

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

@@ -491,7 +491,7 @@ class SubquerySuite extends QueryTest with SharedSQLContext {
         sql("select (select sum(-1) from t t2 where t1.c2 = t2.c1 group by t2.c2) sum from t t1")
       }
       assert(errMsg.getMessage.contains(
-        "a GROUP BY clause in a scalar correlated subquery cannot contain non-correlated columns:"))
+        "A GROUP BY clause in a scalar correlated subquery cannot contain non-correlated columns:"))
     }
   }