Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1173210161


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +600,32 @@ object RewriteCorrelatedScalarSubquery extends 
Rule[LogicalPlan] with AliasHelpe
         if (Utils.isTesting) {
           assert(mayHaveCountBug.isDefined)
         }
+
+        def queryOutputFoldable(list: Seq[NamedExpression]): Boolean = {
+          trimAliases(list.filter(p => 
p.exprId.equals(query.output.head.exprId)).head).foldable
+        }
+
+        lazy val resultFoldable = {
+          query match {
+            case Project(expressions, _) =>
+              queryOutputFoldable(expressions)
+            case Aggregate(_, expressions, _) =>
+              queryOutputFoldable(expressions)
+            case _ =>
+              false
+          }
+        }
+
         if (resultWithZeroTups.isEmpty) {
           // CASE 1: Subquery guaranteed not to have the COUNT bug because it 
evaluates to NULL
           // with zero tuples.
           planWithoutCountBug
+        } else if (resultFoldable) {

Review Comment:
   Already merged before create this PR. 
   
   > if the subquery is something like `select false from ... group by c` then 
it will still actually return null on empty inputs.
   
   That's why we should judge whether the column returned by subquery is 
foldable. If the result depends on input, the `resultFoldable` will be `false`. 
 It will skip the logic I created.
   
   > I think your fix will fix the COUNT(*) is null case, but break the false 
case, e.g. in a query like select *, (select any_value(false) as result from t1 
where t0.a = t1.c) from t0) Can you confirm this?
   
   Is this answer right?
   <img width="952" alt="image" 
src="https://user-images.githubusercontent.com/32387433/233515508-8a7055f6-7366-413c-b137-2bdcbbabb0a1.png";>
   



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org

Reply via email to