[SPARK-20688][SQL] correctly check analysis for scalar sub-queries

In `CheckAnalysis`, we should call `checkAnalysis` for `ScalarSubquery` at the beginning, as later we will call `plan.output` which is invalid if `plan` is not resolved.

new regression test

Author: Wenchen Fan <[email protected]>

Closes #17930 from cloud-fan/tmp.

(cherry picked from commit 789bdbe)
Signed-off-by: Wenchen Fan <[email protected]>

Author: cloud-fan

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

@@ -128,6 +128,8 @@ trait CheckAnalysis extends PredicateHelper {
             }
 
           case s @ ScalarSubquery(query, conditions, _) =>
+            checkAnalysis(query)
+
             // If no correlation, the output must be exactly one column
             if (conditions.isEmpty && query.output.size != 1) {
               failAnalysis(
@@ -186,7 +188,6 @@ trait CheckAnalysis extends PredicateHelper {
                 case fail => failAnalysis(s"Correlated scalar subqueries must be Aggregated: $fail")
               }
             }
-            checkAnalysis(query)
             s
 
           case s: SubqueryExpression =>

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

@@ -72,7 +72,7 @@ class SubquerySuite extends QueryTest with SharedSQLContext {
     }
   }
 
-  test("rdd deserialization does not crash [SPARK-15791]") {
+  test("SPARK-15791: rdd deserialization does not crash") {
     sql("select (select 1 as b) as b").rdd.count()
   }
 
@@ -839,4 +839,12 @@ class SubquerySuite extends QueryTest with SharedSQLContext {
         Row(0) :: Row(1) :: Nil)
     }
   }
+
+  test("SPARK-20688: correctly check analysis for scalar sub-queries") {
+    withTempView("t") {
+      Seq(1 -> "a").toDF("i", "j").createTempView("t")
+      val e = intercept[AnalysisException](sql("SELECT (SELECT count(*) FROM t WHERE a = 1)"))
+      assert(e.message.contains("cannot resolve '`a`' given input columns: [i, j]"))
+    }
+  }
 }