You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by "Hisoka-X (via GitHub)" <gi...@apache.org> on 2023/04/20 04:54:12 UTC

[GitHub] [spark] Hisoka-X opened a new pull request, #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Hisoka-X opened a new pull request, #40865:
URL: https://github.com/apache/spark/pull/40865

   <!--
   Thanks for sending a pull request!  Here are some tips for you:
     1. If this is your first time, please read our contributor guidelines: https://spark.apache.org/contributing.html
     2. Ensure you have added or run the appropriate tests for your PR: https://spark.apache.org/developer-tools.html
     3. If the PR is unfinished, add '[WIP]' in your PR title, e.g., '[WIP][SPARK-XXXX] Your PR title ...'.
     4. Be sure to keep the PR description updated to reflect all changes.
     5. Please write your PR title to summarize what this PR proposes.
     6. If possible, provide a concise example to reproduce the issue for a faster review.
     7. If you want to add a new configuration, please read the guideline first for naming configurations in
        'core/src/main/scala/org/apache/spark/internal/config/ConfigEntry.scala'.
     8. If you want to add or modify an error type or message, please read the guideline first in
        'core/src/main/resources/error/README.md'.
   -->
   
   ### What changes were proposed in this pull request?
   Example query:
   ```sql
   
   create or replace temp view t0 (a, b)
   as values
       (1, 1.0),
       (2, 2.0);
   create or replace temp view t1 (c, d)
   as values
       (2, 3.0); 
   
   spark.sql("select *, (select (count(1)) is null from t1 where t0.a = t1.c) from t0").collect()
   res6: Array[org.apache.spark.sql.Row] = Array([1,1.0,null], [2,2.0,false])  
   ```
   In this subquery, count(1) always evaluates to a non-null integer value, so count(1) is null is always false. The correct evaluation of the subquery is always false.
   
   We incorrectly evaluate it to null for empty groups. The reason is that NullPropagation rewrites `Aggregate [c] [isnull(count(1))]` to `Aggregate [c] [false]`, this rewrite would be correct normally, but in the context of a scalar subquery it breaks our count bug handling in RewriteCorrelatedScalarSubquery.constructLeftJoins . By the time we get there, the query appears to not have the count bug - it looks the same as if the original query had a subquery with `select any_value(false) from r...`, and that case is not subject to the count bug.
   
   Postgres comparison show correct always-false result: http://sqlfiddle.com/#!17/67822/5
   
   Solution:
   
   In scalar subquery must return only one column, and this return column is the result of aggregation. We can judge whether the column returned by subquery is foldable. If it is, it means that the result of this value has no substantial relationship with the data, and the presence or absence of data will not affect this column. So in this case, this column can be extracted from the JOIN to ensure that this value can be obtained regardless of whether the data JOIN is successful or not.
   
   <!--
   Please clarify what changes you are proposing. The purpose of this section is to outline the changes and how this PR fixes the issue. 
   If possible, please consider writing useful notes for better and faster reviews in your PR. See the examples below.
     1. If you refactor some codes with changing classes, showing the class hierarchy will help reviewers.
     2. If you fix some SQL features, you can provide some references of other DBMSes.
     3. If there is design documentation, please add the link.
     4. If there is a discussion in the mailing list, please add the link.
   -->
   
   
   ### Why are the changes needed?
   <!--
   Please clarify why the changes are needed. For instance,
     1. If you propose a new API, clarify the use case for a new API.
     2. If you fix a bug, you can clarify why it is a bug.
   -->
   
   
   ### Does this PR introduce _any_ user-facing change?
   <!--
   Note that it means *any* user-facing change including all aspects such as the documentation fix.
   If yes, please clarify the previous behavior and the change this PR proposes - provide the console output, description and/or an example to show the behavior difference if possible.
   If possible, please also clarify if this is a user-facing change compared to the released Spark versions or within the unreleased branches such as master.
   If no, write 'No'.
   -->
   
   
   ### How was this patch tested?
   <!--
   If tests were added, say they were added here. Please make sure to add some test cases that check the changes thoroughly including negative and positive cases if possible.
   If it was tested in a way different from regular unit tests, please clarify how you tested step by step, ideally copy and paste-able, so that other reviewers can test and check, and descendants can verify in the future.
   If tests were not added, please describe why they were not added and/or why it was difficult to add.
   If benchmark tests were added, please run the benchmarks in GitHub Actions for the consistent environment, and the instructions could accord to: https://spark.apache.org/developer-tools.html#github-workflow-benchmarks.
   -->
   


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] allisonwang-db commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "allisonwang-db (via GitHub)" <gi...@apache.org>.
allisonwang-db commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177746579


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,7 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))

Review Comment:
   @Hisoka-X Can we add a comment here to explain why we want to keep this literal value and how it is related to the COUNT bug?



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177426150


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,7 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))
+        (alias.exprId, l)

Review Comment:
   I think this one-line change is all we need. Count bug does not apply to the count function only, but all expressions in Aggregate that return non-null value for empty input. This includes all constant values.
   
   Assume there are two table `t1(c1 int)`, `t2(c2 int)`. `t1` has two rows (1, 2), `t2` has two rows (1,3)
   
   ```
   scala> sql("select (select f from (select false as f, max(c1) from t1 where c1 = c2)) from t2").show
   +------------------+
   |scalarsubquery(c2)|
   +------------------+
   |             false|
   |              NULL|
   +------------------+
   ```
   The result is wrong and it should be false, false
   
   `select f from (select false as f, max(c1) from t1 where c1 = c2)` is to build a global aggregate that has no aggregate function. If SQL syntax supports group by no column, then we can simply write `select false as f from t1 group by () where c1 = c2`



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] jchen5 commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1518139762

   I checked the case of `any_value(false)` in a debugger and it works because resultWithZeroTups is NULL there, so that explains why it works - because there's an aggregation value around false, and not just constant false.
   
   This fix seems to work with the cases I've thought of (results match SQL standard semantics and postgres).
   
   I think this PR is mergeable. I'll plan to add some more test coverage for related cases in a future PR.


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177477109


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +599,39 @@ 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
+        }
+
+        // SPARK-43156: We can judge whether the column returned by subquery is
+        // foldable (already handle by [[NullPropagation]]). If it is, it means that
+        // the result of this value has no substantial relationship with the data,
+        // and the presence or absence of data will not affect this column. So in
+        // this case, this column can be extracted from the JOIN to ensure that this
+        // value can be obtained regardless of whether the data JOIN is successful or not.
+        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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&

Review Comment:
   > can you show me which test will fail if we don't have the changes here?
   
   if without this, the logic will use `Literal(0)` as result to replace `count(1)`.



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1517170830

   > I think your fix will fix the COUNT(*) is null case, but break the false case
   
   Hi @jchen5 , can you help add a test for it in a new PR? The current PR passes all tests which means our test coverage is not good.


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1517543820

   > Before we make a decision, let's try the query with postgres, oracle, db2, etc.
   
   Ok. By the way, I check with postgres, it return `null`
   ![image](https://user-images.githubusercontent.com/32387433/233600204-e254ca5c-daec-43cb-9490-0bf406d5d844.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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1517521661

   > @Hisoka-X do you mean the bug is already fixed now?
   
   Depends on what is correct results of `select *, (select any_value(false) as result from t1 where t0.a = t1.c) from t0)` ?
   
   > if the subquery is something like select false from ... group by c then it will still actually return null on empty inputs.
   
   The current state is as described by @jchen5  . `select false from ... group by c` will result null on empty inputs, not return `false`. I think this is what @jchen5 worry about. You can check by https://github.com/apache/spark/pull/40865#discussion_r1173210161 .
   
   If you think `select false from ... group by c` in subquery should return `false` on empty inputs, this PR have bug now. If you think it should return `null` on empty inputs, this PR are correct.
   
   


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177465919


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +599,39 @@ 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
+        }
+
+        // SPARK-43156: We can judge whether the column returned by subquery is
+        // foldable (already handle by [[NullPropagation]]). If it is, it means that
+        // the result of this value has no substantial relationship with the data,
+        // and the presence or absence of data will not affect this column. So in
+        // this case, this column can be extracted from the JOIN to ensure that this
+        // value can be obtained regardless of whether the data JOIN is successful or not.
+        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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&

Review Comment:
   can you show me which test will fail if we don't have the changes here?



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1172544078


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,8 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))
+        // SPARK-43156: copy value to get the correct value when extracting in constructLeftJoins
+        (alias.exprId, l.copy(value = l.value))

Review Comment:
   My fault, just want to create new `Literal` instance with same value.



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1525132518

   It's better to have this fix in 3.4, @Hisoka-X can you open a new PR for 3.4 to backport https://github.com/apache/spark/pull/40946 and this PR together?


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177536071


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +599,39 @@ 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
+        }
+
+        // SPARK-43156: We can judge whether the column returned by subquery is
+        // foldable (already handle by [[NullPropagation]]). If it is, it means that
+        // the result of this value has no substantial relationship with the data,
+        // and the presence or absence of data will not affect this column. So in
+        // this case, this column can be extracted from the JOIN to ensure that this
+        // value can be obtained regardless of whether the data JOIN is successful or not.
+        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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&

Review Comment:
   There are some change after I create this PR. Seem like `resultFoldable` are useful when I create PR but not now.



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1176203863


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +600,33 @@ 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 = {

Review Comment:
   We need more comments here. Why is foldable result a special case that we need to take care here?



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1523077396

   @Hisoka-X I think it's clearer to use my example to explain the bug. `NullPropagation` triggers the bug for `(count(1)) is null`, but the root cause is the wrong handling of literals when dealing with the count bug. The bug can be triggered when the scalar subquery has a global aggregate with a constant, which can be produced by `NullPropagation` or by the user query directly.


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] jchen5 commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1173128023


##########
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:
   I just changed this code in https://github.com/apache/spark/pull/40811 so you'll need to merge with that - your check should go after my new one, because if the subquery is something like `select false from ... group by c` then it will still actually return null on empty inputs. I have a test case for it added in my PR.



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] jchen5 commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1173668725


##########
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:
   That answer does look right, let me debug this code to figure out why it's right and if I can find a counterexample.
   
   Your check needs to go after the check immediately below to be correct (that's the cause of the test results change Allison commented).



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1517492697

   @Hisoka-X do you mean the bug is already fixed now?


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177427622


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,7 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))
+        (alias.exprId, l)

Review Comment:
   We can optimize the code a bit, to build simpler plan if `resultWithZeroTups.get` is a `Literal`



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177508036


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +599,39 @@ 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
+        }
+
+        // SPARK-43156: We can judge whether the column returned by subquery is
+        // foldable (already handle by [[NullPropagation]]). If it is, it means that
+        // the result of this value has no substantial relationship with the data,
+        // and the presence or absence of data will not affect this column. So in
+        // this case, this column can be extracted from the JOIN to ensure that this
+        // value can be obtained regardless of whether the data JOIN is successful or not.
+        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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&

Review Comment:
   Yes, after I merge the code. The test will passed. But the plan will be
   ```
   == Optimized Logical Plan ==
   Project [_1#220 AS a#225, (count(1) IS NULL)#266 AS scalarsubquery(a)#267]
   +- Join LeftOuter, (_1#220 = c#236)
      :- Project [_1#220]
      :  +- Filter (isnotnull(_1#220) AND (_1#220 < 4))
      :     +- LocalRelation [_1#220, _2#221]
      +- Aggregate [c#236], [false AS (count(1) IS NULL)#266, c#236]
         +- Project [_1#231 AS c#236]
            +- Filter ((_1#231 < 4) AND isnotnull(_1#231))
               +- LocalRelation [_1#231, _2#232]
   ```
   not
   ```
   == Optimized Logical Plan ==
   Project [false AS scalarsubquery(c2)#252]
   +- LocalRelation [col1#247]
   ```
   on `select a, (select (count(1)) is null from r where a = c group by c) from l where a < 4`



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177534269


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +599,39 @@ 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
+        }
+
+        // SPARK-43156: We can judge whether the column returned by subquery is
+        // foldable (already handle by [[NullPropagation]]). If it is, it means that
+        // the result of this value has no substantial relationship with the data,
+        // and the presence or absence of data will not affect this column. So in
+        // this case, this column can be extracted from the JOIN to ensure that this
+        // value can be obtained regardless of whether the data JOIN is successful or not.
+        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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&

Review Comment:
   Should I split it in two PR? Or just chagne title and description?



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1516131796

   cc @cloud-fan 


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1518501790

   kindly ping @cloud-fan . All CI passed.


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1517538486

   Before we make a decision, let's try the query with postgres, oracle, db2, etc.


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1173671566


##########
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:
   Yes, I add `mayHaveCountBug.getOrElse(false)` into check part.



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] jchen5 commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1176630244


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +600,33 @@ 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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&
+          !conf.getConf(SQLConf.DECORRELATE_SUBQUERY_LEGACY_INCORRECT_COUNT_HANDLING_ENABLED)) {
+          val alias = Alias(resultWithZeroTups.get, origOutput.name)()
+          subqueryAttrMapping += (origOutput -> alias.toAttribute)
+          Project(
+            currentChild.output :+ alias,
+            Join(currentChild, query, LeftOuter, conditions.reduceOption(And), joinHint))

Review Comment:
   Quick PR to add some test coverage for it: https://github.com/apache/spark/pull/40946



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177437984


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,7 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))
+        (alias.exprId, l)

Review Comment:
   > We can optimize the code a bit, to build simpler plan if resultWithZeroTups.get is a Literal
   
   Can't, in `select *, (select count(1) from t2 where t1.c1 = t2.c1) from t1`. The resultWithZeroTups.get will return `Literal(0)`. That's why we need use `resultFoldable`.



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177944444


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,7 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))

Review Comment:
   Done



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1172523177


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,8 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))
+        // SPARK-43156: copy value to get the correct value when extracting in constructLeftJoins
+        (alias.exprId, l.copy(value = l.value))

Review Comment:
   what does `l.copy(value = l.value)` do? seems nothing?



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1517200147

   > > I think your fix will fix the COUNT(*) is null case, but break the false case
   > 
   > Hi @jchen5 , can you help add a test for it in a new PR? The current PR passes all tests which means our test coverage is not good.
   
   If you don't mind, I can also add the required test cases to my PR. However, after I merged #40811, it seems that the current use case has been able to prove the correctness of this PR.😂


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] jchen5 commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1173756531


##########
sql/core/src/test/resources/sql-tests/results/subquery/scalar-subquery/scalar-subquery-count-bug.sql.out:
##########
@@ -86,14 +86,14 @@ from l
 -- !query schema
 struct<a:int,b:decimal(2,1),scalarsubquery(a):boolean>
 -- !query output
-1	2.0	NULL
-1	2.0	NULL
+1	2.0	false
+1	2.0	false

Review Comment:
   Yes, old result here was wrong, new result is correct.
   
   @Hisoka-X There's a comment about it in the input file, please update it to remove the "but this case is wrong due to bug SPARK-43156" part



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1525125052

   thanks, merging to master/3.4!


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1176211624


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,8 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))
+        // SPARK-43156: copy value to get the correct value when extracting in constructLeftJoins

Review Comment:
   My mistake. Fixed.



##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +600,33 @@ 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 = {

Review Comment:
   Done



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1173730596


##########
sql/core/src/test/resources/sql-tests/results/subquery/scalar-subquery/scalar-subquery-count-bug.sql.out:
##########
@@ -86,14 +86,14 @@ from l
 -- !query schema
 struct<a:int,b:decimal(2,1),scalarsubquery(a):boolean>
 -- !query output
-1	2.0	NULL
-1	2.0	NULL
+1	2.0	false
+1	2.0	false

Review Comment:
   @jchen5 was the old result wrong?



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177550428


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +599,39 @@ 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
+        }
+
+        // SPARK-43156: We can judge whether the column returned by subquery is
+        // foldable (already handle by [[NullPropagation]]). If it is, it means that
+        // the result of this value has no substantial relationship with the data,
+        // and the presence or absence of data will not affect this column. So in
+        // this case, this column can be extracted from the JOIN to ensure that this
+        // value can be obtained regardless of whether the data JOIN is successful or not.
+        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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&

Review Comment:
   please split the PR. We should have a PR with the one-line change and tests, also with detailed PR description to explain how the bug happens.



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177426150


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,7 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))
+        (alias.exprId, l)

Review Comment:
   I think this one-line change is all we need. Count bug does not apply to the count function only, but all expressions in Aggregate that return non-null value for empty input. This includes all constant values.
   
   Assume there are two table `t1(c1 int)`, `t2(c2 int)`. `t2` has two rows (1, 2), `t3` has two rows (1,3)
   
   ```
   scala> sql("select (select f from (select false as f, max(c1) from t1 where c1 = c2)) from t2").show
   +------------------+
   |scalarsubquery(c2)|
   +------------------+
   |             false|
   |              NULL|
   +------------------+
   ```
   The result wrong and it should be false, false



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] allisonwang-db commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "allisonwang-db (via GitHub)" <gi...@apache.org>.
allisonwang-db commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1173566280


##########
sql/core/src/test/resources/sql-tests/results/subquery/scalar-subquery/scalar-subquery-count-bug.sql.out:
##########
@@ -106,14 +106,14 @@ from l
 -- !query schema
 struct<a:int,b:decimal(2,1),scalarsubquery(a):boolean>
 -- !query output
-1	2.0	NULL

Review Comment:
   These results are incorrect. They should be `NULL` instead of `false`. 



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] jchen5 commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1517695970

   > Depends on what is correct results of select *, (select any_value(false) as result from t1 where t0.a = t1.c) from t0) ?
   
   Yes, this should return null on empty data.
   
   I will try to find another test case to check the potential issue I mentioned. Maybe something like `select *, (select false as result from t1 where t0.a = t1.c) from t0 limit 1)`(correct answer for that is also null).


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177426150


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,7 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))
+        (alias.exprId, l)

Review Comment:
   I think this one-line change is all we need. Count bug does not apply to the count function only, but all expressions in Aggregate that return non-null value for empty input. This includes all constant values.
   
   Assume there are two table `t1(c1 int)`, `t2(c2 int)`. `t2` has two rows (1, 2), `t3` has two rows (1,3)
   
   ```
   scala> sql("select (select f from (select false as f, max(c1) from t1 where c1 = c2)) from t2").show
   +------------------+
   |scalarsubquery(c2)|
   +------------------+
   |             false|
   |              NULL|
   +------------------+
   ```
   The result is wrong and it should be false, false



##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,7 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))
+        (alias.exprId, l)

Review Comment:
   cc @jchen5 



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177439615


##########
sql/core/src/test/scala/org/apache/spark/sql/SubquerySuite.scala:
##########
@@ -1928,6 +1928,15 @@ class SubquerySuite extends QueryTest
     checkAnswer(df, Nil)
   }
 
+  test("SPARK-43156: scalar subquery with Literal result like `COUNT(1) is null`") {

Review Comment:
   Removed



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan closed pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan closed pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery
URL: https://github.com/apache/spark/pull/40865


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1525570134

   > It's better to have this fix in 3.4, @Hisoka-X can you open a new PR for 3.4 to backport #40946 and this PR together?
   
   Please check #40977 


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] jchen5 commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1176379731


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +600,33 @@ 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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&
+          !conf.getConf(SQLConf.DECORRELATE_SUBQUERY_LEGACY_INCORRECT_COUNT_HANDLING_ENABLED)) {
+          val alias = Alias(resultWithZeroTups.get, origOutput.name)()
+          subqueryAttrMapping += (origOutput -> alias.toAttribute)
+          Project(
+            currentChild.output :+ alias,
+            Join(currentChild, query, LeftOuter, conditions.reduceOption(And), joinHint))

Review Comment:
   This change works for the new decorrelation framework (and the old one too). This code happens after / on top of the DecorrelateInnerQuery framework.



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1173667469


##########
sql/core/src/test/resources/sql-tests/results/subquery/scalar-subquery/scalar-subquery-count-bug.sql.out:
##########
@@ -106,14 +106,14 @@ from l
 -- !query schema
 struct<a:int,b:decimal(2,1),scalarsubquery(a):boolean>
 -- !query output
-1	2.0	NULL

Review Comment:
   Thanks for remind. I fixed this bug and add test to cover it. I will update `sql.out` later.



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1173768209


##########
sql/core/src/test/resources/sql-tests/results/subquery/scalar-subquery/scalar-subquery-count-bug.sql.out:
##########
@@ -86,14 +86,14 @@ from l
 -- !query schema
 struct<a:int,b:decimal(2,1),scalarsubquery(a):boolean>
 -- !query output
-1	2.0	NULL
-1	2.0	NULL
+1	2.0	false
+1	2.0	false

Review Comment:
   Done



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1172544078


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,8 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))
+        // SPARK-43156: copy value to get the correct value when extracting in constructLeftJoins
+        (alias.exprId, l.copy(value = l.value))

Review Comment:
   My fault, just want to create new `Literal` instance with seem value.



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177426150


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,7 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))
+        (alias.exprId, l)

Review Comment:
   I think this one-line change is all we need. Count bug does not apply to the count function only, but all expressions in Aggregate that return non-null value for empty input. This includes all constant values.
   
   Assume there are two table `t1(c1 int)`, `t2(c2 int)`. `t1` has two rows (1, 2), `t2` has two rows (1,3)
   
   ```
   scala> sql("select (select f from (select false as f, max(c1) from t1 where c1 = c2)) from t2").show
   +------------------+
   |scalarsubquery(c2)|
   +------------------+
   |             false|
   |              NULL|
   +------------------+
   ```
   The result is wrong and it should be false, false



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177475934


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +599,39 @@ 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
+        }
+
+        // SPARK-43156: We can judge whether the column returned by subquery is
+        // foldable (already handle by [[NullPropagation]]). If it is, it means that
+        // the result of this value has no substantial relationship with the data,
+        // and the presence or absence of data will not affect this column. So in
+        // this case, this column can be extracted from the JOIN to ensure that this
+        // value can be obtained regardless of whether the data JOIN is successful or not.
+        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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&

Review Comment:
   Test case: 
   <img width="625" alt="image" src="https://user-images.githubusercontent.com/32387433/234505362-b17db89d-7f27-4c62-979f-b39238f1ed6b.png">
   Remove `resultFoldable`: 
   <img width="399" alt="image" src="https://user-images.githubusercontent.com/32387433/234505527-fc0f6d2e-85d6-4dfb-82d4-2821aa92339c.png">
   Should be:
   <img width="398" alt="image" src="https://user-images.githubusercontent.com/32387433/234505644-a9b6de27-cdf3-4ec1-90ce-1c087efcb2f6.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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177508036


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +599,39 @@ 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
+        }
+
+        // SPARK-43156: We can judge whether the column returned by subquery is
+        // foldable (already handle by [[NullPropagation]]). If it is, it means that
+        // the result of this value has no substantial relationship with the data,
+        // and the presence or absence of data will not affect this column. So in
+        // this case, this column can be extracted from the JOIN to ensure that this
+        // value can be obtained regardless of whether the data JOIN is successful or not.
+        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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&

Review Comment:
   Yes, after I merge the code. The test will passed. But the plan will be
   ```
   == Optimized Logical Plan ==
   Project [_1#220 AS a#225, (count(1) IS NULL)#266 AS scalarsubquery(a)#267]
   +- Join LeftOuter, (_1#220 = c#236)
      :- Project [_1#220]
      :  +- Filter (isnotnull(_1#220) AND (_1#220 < 4))
      :     +- LocalRelation [_1#220, _2#221]
      +- Aggregate [c#236], [false AS (count(1) IS NULL)#266, c#236]
         +- Project [_1#231 AS c#236]
            +- Filter ((_1#231 < 4) AND isnotnull(_1#231))
               +- LocalRelation [_1#231, _2#232]
   ```
   not
   ```
   == Optimized Logical Plan ==
   Project [false AS scalarsubquery(c2)#252]
   +- LocalRelation [col1#247]
   ```



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177531457


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +599,39 @@ 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
+        }
+
+        // SPARK-43156: We can judge whether the column returned by subquery is
+        // foldable (already handle by [[NullPropagation]]). If it is, it means that
+        // the result of this value has no substantial relationship with the data,
+        // and the presence or absence of data will not affect this column. So in
+        // this case, this column can be extracted from the JOIN to ensure that this
+        // value can be obtained regardless of whether the data JOIN is successful or not.
+        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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&

Review Comment:
   making plan simpler is an optimization, not a correctness issue.



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1176204595


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +600,33 @@ 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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&
+          !conf.getConf(SQLConf.DECORRELATE_SUBQUERY_LEGACY_INCORRECT_COUNT_HANDLING_ENABLED)) {
+          val alias = Alias(resultWithZeroTups.get, origOutput.name)()
+          subqueryAttrMapping += (origOutput -> alias.toAttribute)
+          Project(
+            currentChild.output :+ alias,
+            Join(currentChild, query, LeftOuter, conditions.reduceOption(And), joinHint))

Review Comment:
   @allisonwang-db does the new decorrelation framework need the same fix?



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177586693


##########
sql/core/src/test/resources/sql-tests/inputs/subquery/scalar-subquery/scalar-subquery-select.sql:
##########
@@ -235,3 +235,6 @@ SELECT c, (
     FROM (VALUES (0, 6), (1, 5), (2, 4), (3, 3)) t1(a, b)
     WHERE a + b = c
 ) FROM (VALUES (6)) t2(c);
+
+-- SPARK-43156: scalar subquery with Literal result like `COUNT(1) is null`
+SELECT *, (SELECT count(1) is null FROM t2 WHERE t1.c1 = t2.c1) FROM t1

Review Comment:
   can you also test the query I mentioned before that has no count function?



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on PR #40865:
URL: https://github.com/apache/spark/pull/40865#issuecomment-1516248434

   also cc @jchen5 @allisonwang-db 


-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1176200981


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,8 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))
+        // SPARK-43156: copy value to get the correct value when extracting in constructLeftJoins

Review Comment:
   `Literal` is immutable, why a copy is needed?



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177434949


##########
sql/core/src/test/scala/org/apache/spark/sql/SubquerySuite.scala:
##########
@@ -1928,6 +1928,15 @@ class SubquerySuite extends QueryTest
     checkAnswer(df, Nil)
   }
 
+  test("SPARK-43156: scalar subquery with Literal result like `COUNT(1) is null`") {

Review Comment:
   this is no longer needed as we have golden file test



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177458137


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -459,7 +459,7 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
       case alias @ Alias(_: AttributeReference, _) =>
         (alias.exprId, Literal.create(null, alias.dataType))
       case alias @ Alias(l: Literal, _) =>
-        (alias.exprId, l.copy(value = null))
+        (alias.exprId, l)

Review Comment:
   > I think this one-line change is all we need. Count bug does not apply to the count function only, but all expressions in Aggregate that return non-null value for empty input. This includes all constant values.
   > 
   > Assume there are two table `t1(c1 int)`, `t2(c2 int)`. `t1` has two rows (1, 2), `t2` has two rows (1,3)
   > 
   > ```
   > scala> sql("select (select f from (select false as f, max(c1) from t1 where c1 = c2)) from t2").show
   > +------------------+
   > |scalarsubquery(c2)|
   > +------------------+
   > |             false|
   > |              NULL|
   > +------------------+
   > ```
   > 
   > The result is wrong and it should be false, false
   
   After this PR, the result will be false, false
   
   > 
   > `select f from (select false as f, max(c1) from t1 where c1 = c2)` is to build a global aggregate that has no aggregate function. If SQL syntax supports group by no column, then we can simply write `select false as f from t1 group by () where c1 = c2`
   
   After this PR, already is.
   
   ```
   == Parsed Logical Plan ==
   'Project [unresolvedalias(scalar-subquery#245 [], None)]
   :  +- 'Project ['f]
   :     +- 'SubqueryAlias __auto_generated_subquery_name
   :        +- 'Project [false AS f#244, unresolvedalias('max('c1), None)]
   :           +- 'Filter ('c1 = 'c2)
   :              +- 'UnresolvedRelation [t1], [], false
   +- 'UnresolvedRelation [t2], [], false
   
   == Analyzed Logical Plan ==
   scalarsubquery(c2): boolean
   Project [scalar-subquery#245 [c2#246] AS scalarsubquery(c2)#252]
   :  +- Project [f#244]
   :     +- SubqueryAlias __auto_generated_subquery_name
   :        +- Aggregate [false AS f#244, max(c1#248) AS max(c1)#251]
   :           +- Filter (c1#248 = outer(c2#246))
   :              +- SubqueryAlias spark_catalog.default.t1
   :                 +- View (`spark_catalog`.`default`.`t1`, [c1#248])
   :                    +- Project [cast(col1#249 as int) AS c1#248]
   :                       +- LocalRelation [col1#249]
   +- SubqueryAlias spark_catalog.default.t2
      +- View (`spark_catalog`.`default`.`t2`, [c2#246])
         +- Project [cast(col1#247 as int) AS c2#246]
            +- LocalRelation [col1#247]
   
   == Optimized Logical Plan ==
   Project [false AS scalarsubquery(c2)#252]
   +- LocalRelation [col1#247]
   
   == Physical Plan ==
   *(1) Project [false AS scalarsubquery(c2)#252]
   +- *(1) LocalTableScan [col1#247]
   ```
   
   



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177491250


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +599,39 @@ 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
+        }
+
+        // SPARK-43156: We can judge whether the column returned by subquery is
+        // foldable (already handle by [[NullPropagation]]). If it is, it means that
+        // the result of this value has no substantial relationship with the data,
+        // and the presence or absence of data will not affect this column. So in
+        // this case, this column can be extracted from the JOIN to ensure that this
+        // value can be obtained regardless of whether the data JOIN is successful or not.
+        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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&

Review Comment:
   are you using the latest Spark version? It works fine if I only fix `l.copy(value = null)`
   ```
   scala> sql("SELECT (SELECT count(1) FROM t2 WHERE t1.c1 = t2.c2) FROM t1").show
   +------------------+
   |scalarsubquery(c1)|
   +------------------+
   |                 1|
   |                 0|
   +------------------+
   ```



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177508036


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +599,39 @@ 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
+        }
+
+        // SPARK-43156: We can judge whether the column returned by subquery is
+        // foldable (already handle by [[NullPropagation]]). If it is, it means that
+        // the result of this value has no substantial relationship with the data,
+        // and the presence or absence of data will not affect this column. So in
+        // this case, this column can be extracted from the JOIN to ensure that this
+        // value can be obtained regardless of whether the data JOIN is successful or not.
+        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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&

Review Comment:
   Yes, after I merge the code. The test will passed. But the plan will be
   ```
   == Optimized Logical Plan ==
   Project [if (isnull(alwaysTrue#253)) 0 else count(1)#250L AS scalarsubquery(c2)#251L]
   +- Join LeftOuter, (c1#247 = c2#245)
      :- Project [col1#246 AS c2#245]
      :  +- LocalRelation [col1#246]
      +- Aggregate [c1#247], [count(1) AS count(1)#250L, c1#247, true AS alwaysTrue#253]
         +- Project [col1#248 AS c1#247]
            +- LocalRelation [col1#248]
   ```
   not
   ```
   == Optimized Logical Plan ==
   Project [false AS scalarsubquery(c2)#252]
   +- LocalRelation [col1#247]
   ```



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177621724


##########
sql/core/src/test/resources/sql-tests/inputs/subquery/scalar-subquery/scalar-subquery-select.sql:
##########
@@ -235,3 +235,6 @@ SELECT c, (
     FROM (VALUES (0, 6), (1, 5), (2, 4), (3, 3)) t1(a, b)
     WHERE a + b = c
 ) FROM (VALUES (6)) t2(c);
+
+-- SPARK-43156: scalar subquery with Literal result like `COUNT(1) is null`
+SELECT *, (SELECT count(1) is null FROM t2 WHERE t1.c1 = t2.c1) FROM t1

Review Comment:
   Done



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] cloud-fan commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177585044


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,6 +599,7 @@ object RewriteCorrelatedScalarSubquery extends Rule[LogicalPlan] with AliasHelpe
         if (Utils.isTesting) {
           assert(mayHaveCountBug.isDefined)
         }
+

Review Comment:
   unnecessary change



-- 
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-unsubscribe@spark.apache.org

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


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

[GitHub] [spark] Hisoka-X commented on a diff in pull request #40865: [SPARK-43156][SQL] Fix `COUNT(*) is null` bug in correlated scalar subquery

Posted by "Hisoka-X (via GitHub)" <gi...@apache.org>.
Hisoka-X commented on code in PR #40865:
URL: https://github.com/apache/spark/pull/40865#discussion_r1177567056


##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -599,10 +599,39 @@ 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
+        }
+
+        // SPARK-43156: We can judge whether the column returned by subquery is
+        // foldable (already handle by [[NullPropagation]]). If it is, it means that
+        // the result of this value has no substantial relationship with the data,
+        // and the presence or absence of data will not affect this column. So in
+        // this case, this column can be extracted from the JOIN to ensure that this
+        // value can be obtained regardless of whether the data JOIN is successful or not.
+        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 (mayHaveCountBug.getOrElse(false) && resultFoldable &&

Review Comment:
   Done.



-- 
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-unsubscribe@spark.apache.org

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


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