You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by "andylam-db (via GitHub)" <gi...@apache.org> on 2023/10/12 00:45:26 UTC
[PR] [SPARK-45507] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
andylam-db opened a new pull request, #43341:
URL: https://github.com/apache/spark/pull/43341
<!--
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?
<!--
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.
-->
We want to use the count bug handling in `DecorrelateInnerQuery` to detect potential count bugs in scalar subqueries. it It is always safe to use `DecorrelateInnerQuery` to handle count bugs, but for efficiency reasons, like for the common case of COUNT on top of the scalar subquery, we would like to avoid an extra left outer join. This PR therefore introduces a simple check to detect such cases before `decorrelate()` - if true, then don't do count bug handling in `decorrelate()`, otherwise, do count bug handling.
### 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.
-->
This PR fixes correctness issues for correlated scalar subqueries pertaining to the COUNT bug. Examples can be found in the JIRA ticket.
### 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'.
-->
Yes, results will change.
### 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.
-->
Added SQL end-to-end tests in `count.sql`
### Was this patch authored or co-authored using generative AI tooling?
<!--
If generative AI tooling has been used in the process of authoring this patch, please include the
phrase: 'Generated-by: ' followed by the name of the tool and its version.
If no, write 'No'.
Please refer to the [ASF Generative Tooling Guidance](https://www.apache.org/legal/generative-tooling.html) for details.
-->
No.
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for nested correlated scalar subqueries with COUNT aggregates [spark]
Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1361237972
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
Ah I see the confusion. What I meant was move the check for the else case (where mayHaveCountBugOld is non-empty) first.
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
Posted by "andylam-db (via GitHub)" <gi...@apache.org>.
andylam-db commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1357456302
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
If the plan has already been decorrelated and count bugs handled in `decorrelate()` above, then we can't set `mayHaveCountBug` parameter to be true. This is because if we set it as true based on the pre-rewrite plan (which is what's happening now), running `splitSubquery()` in `constructLeftJoins` will fail because it encounters left outer joins.
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1357405604
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
Should we move this check first? Probably doesn't make a difference but in the event the rule runs multiple times it eliminates possible idempotency issues.
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
Posted by "andylam-db (via GitHub)" <gi...@apache.org>.
andylam-db commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1357469232
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
Maybe the issue is the definition of the ScalarSubquery case class parameter `mayHaveCountBug`.
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for nested correlated scalar subqueries with COUNT aggregates [spark]
Posted by "andylam-db (via GitHub)" <gi...@apache.org>.
andylam-db commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1361251051
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
Ah, it makes sense now. Changed!
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for nested correlated scalar subqueries with COUNT aggregates [spark]
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on PR #43341:
URL: https://github.com/apache/spark/pull/43341#issuecomment-1769795285
OK that's definitely unrelated. I'm merging it to master, thanks!
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for nested correlated scalar subqueries with COUNT aggregates [spark]
Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1361237972
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
Ah I see the confusion. What I meant was move the else case (where mayHaveCountBugOld is non-empty) first.
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
Posted by "andylam-db (via GitHub)" <gi...@apache.org>.
andylam-db commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1357456302
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
If the plan has already been decorrelated and count bugs handled in `decorrelate()` above, then we can't set `mayHaveCountBug` parameter to be true. This is because if we set it as true based on the pre-rewrite plan (which is what's happening now), running `splitSubquery()` in `constructLeftJoins` will fail because it encounters left outer joins (invalid operators).
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for nested correlated scalar subqueries with COUNT aggregates [spark]
Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1361237972
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
Ah I see the confusion. What I meant was move the check for the else case (where mayHaveCountBugOld is non-empty) first - in that case we just copy the old mayHaveCountBug value over. I think that shouldn't cause any functional difference?
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for nested correlated scalar subqueries with COUNT aggregates [spark]
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan closed pull request #43341: [SPARK-45507][SQL] Correctness fix for nested correlated scalar subqueries with COUNT aggregates
URL: https://github.com/apache/spark/pull/43341
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
Posted by "andylam-db (via GitHub)" <gi...@apache.org>.
andylam-db commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1357469232
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
Maybe the issue is the definition of the ScalarSubquery case class parameter `mayHaveCountBug`.
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
Posted by "andylam-db (via GitHub)" <gi...@apache.org>.
andylam-db commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1357456302
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
If the plan has already been decorrelated and count bugs handled in `decorrelate()` above, then we can't set `mayHaveCountBug` parameter to be true. This is because if we set it as true, running `splitSubquery()` in `constructLeftJoins` will fail because it encounters left outer joins.
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for nested correlated scalar subqueries with COUNT aggregates [spark]
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on PR #43341:
URL: https://github.com/apache/spark/pull/43341#issuecomment-1767633377
can we retrigger the failed test jobs via GitHub UI?
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
Posted by "andylam-db (via GitHub)" <gi...@apache.org>.
andylam-db commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1357456302
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
If the plan has already been decorrelated and count bugs handled in `decorrelate()` above, then we can't set `mayHaveCountBug` parameter to be true. This is because if we set it as true based on the pre-rewrite plan (which is what's happening now), running `splitSubquery()` in `constructLeftJoins` will fail because it encounters left outer joins (invalid operators).
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for nested correlated scalar subqueries with COUNT aggregates [spark]
Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1361237972
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
Specifically, what I meant is that if mayHaveCountBugOld is non-empty (the else case), then we should be ok to just copy over the old 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
Re: [PR] [SPARK-45507][SQL] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
Posted by "andylam-db (via GitHub)" <gi...@apache.org>.
andylam-db commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1357469232
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
Maybe the issue is the definition of `mayHaveCountBug`.
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1356986664
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,31 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // We want to handle count bug for scalar subqueries, except for the cases where the
+ // subquery is a simple top level Aggregate which can have a count bug (note: the below
+ // logic also takes into account nested COUNTs). This is because for these cases, we don't
+ // want to introduce redundant left outer joins in [[DecorrelateInnerQuery]], when the
+ // necessary left outer join will be added in [[RewriteCorrelatedScalarSubquery]].
+ val shouldHandleCountBug = !(sub match {
Review Comment:
Yes, let's have a legacy behavior flag for safety.
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1357136420
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,31 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // We want to handle count bug for scalar subqueries, except for the cases where the
Review Comment:
Can you edit the comment a bit:
- clarify that this controls handling count bug in DecorrelateInnerQuery, and if we don't handle it here we handle it in constructLeftJoins.
- handling it here is always correct, turning it off to handle it in constructLeftJoins is an optimization
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for nested correlated scalar subqueries with COUNT aggregates [spark]
Posted by "andylam-db (via GitHub)" <gi...@apache.org>.
andylam-db commented on PR #43341:
URL: https://github.com/apache/spark/pull/43341#issuecomment-1768974966
@cloud-fan Retried, looks like only the Docker integration tests are failing 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
Re: [PR] [SPARK-45507][SQL] Correctness fix for nested correlated scalar subqueries with COUNT aggregates [spark]
Posted by "jchen5 (via GitHub)" <gi...@apache.org>.
jchen5 commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1361234951
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
But if mayHaveCountBugOld is set, the code just keeps that value - decorrelation wouldn't actually run again. IIRC this doesn't come up in real queries, but there are unit tests that check that rules are idempotent which is where that came up.
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for nested correlated scalar subqueries with COUNT aggregates [spark]
Posted by "andylam-db (via GitHub)" <gi...@apache.org>.
andylam-db commented on PR #43341:
URL: https://github.com/apache/spark/pull/43341#issuecomment-1766841408
@cloud-fan Failed tests in build. I think the first one is unrelated to the PR, but not sure about the second one
1. Docker integration tests:
```
[error] Error during tests:
[error] org.apache.spark.sql.jdbc.v2.OracleIntegrationSuite
[error] org.apache.spark.sql.jdbc.OracleIntegrationSuite
```
2. pyspark-mllib
```
Error: The operation was canceled.
```
--
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
Re: [PR] [WIP][SPARK-45507][SQL] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
Posted by "andylam-db (via GitHub)" <gi...@apache.org>.
andylam-db commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1355919290
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,31 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // We want to handle count bug for scalar subqueries, except for the cases where the
+ // subquery is a simple top level Aggregate which can have a count bug (note: the below
+ // logic also takes into account nested COUNTs). This is because for these cases, we don't
+ // want to introduce redundant left outer joins in [[DecorrelateInnerQuery]], when the
+ // necessary left outer join will be added in [[RewriteCorrelatedScalarSubquery]].
+ val shouldHandleCountBug = !(sub match {
Review Comment:
Should we flag this with a config?
--
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
Re: [PR] [WIP][SPARK-45507][SQL] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
Posted by "andylam-db (via GitHub)" <gi...@apache.org>.
andylam-db commented on PR #43341:
URL: https://github.com/apache/spark/pull/43341#issuecomment-1758745012
Pinging for first round of reviews :-) @jchen5 @agubichev
--
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
Re: [PR] [SPARK-45507][SQL] Correctness fix for correlated scalar subqueries with COUNT aggregates [spark]
Posted by "andylam-db (via GitHub)" <gi...@apache.org>.
andylam-db commented on code in PR #43341:
URL: https://github.com/apache/spark/pull/43341#discussion_r1357456302
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/subquery.scala:
##########
@@ -360,8 +360,40 @@ object PullupCorrelatedPredicates extends Rule[LogicalPlan] with PredicateHelper
plan.transformExpressionsWithPruning(_.containsPattern(PLAN_EXPRESSION)) {
case ScalarSubquery(sub, children, exprId, conditions, hint, mayHaveCountBugOld)
if children.nonEmpty =>
- val (newPlan, newCond) = decorrelate(sub, plan)
- val mayHaveCountBug = if (mayHaveCountBugOld.isEmpty) {
+
+ def mayHaveCountBugAgg(a: Aggregate): Boolean = {
+ a.groupingExpressions.isEmpty && a.aggregateExpressions.exists(_.exists {
+ case a: AggregateExpression => a.aggregateFunction.defaultResult.isDefined
+ case _ => false
+ })
+ }
+
+ // The below logic controls handling count bug for scalar subqueries in
+ // [[DecorrelateInnerQuery]], and if we don't handle it here, we handle it in
+ // [[RewriteCorrelatedScalarSubquery#constructLeftJoins]]. Note that handling it in
+ // [[DecorrelateInnerQuery]] is always correct, and turning it off to handle it in
+ // constructLeftJoins is an optimization, so that additional, redundant left outer joins are
+ // not introduced.
+ val handleCountBugInDecorrelate = !conf.getConf(
+ SQLConf.LEGACY_SCALAR_SUBQUERY_COUNT_BUG_HANDLING) && !(sub match {
+ // Handle count bug only if there exists lower level Aggs with count bugs. It does not
+ // matter if the top level agg is count bug vulnerable or not, because:
+ // 1. If the top level agg is count bug vulnerable, it can be handled in
+ // constructLeftJoins, unless there are lower aggs that are count bug vulnerable.
+ // E.g. COUNT(COUNT + COUNT)
+ // 2. If the top level agg is not count bug vulnerable, it can be count bug vulnerable if
+ // there are lower aggs that are count bug vulnerable. E.g. SUM(COUNT)
+ case agg: Aggregate => !agg.child.exists {
+ case lowerAgg: Aggregate => mayHaveCountBugAgg(lowerAgg)
+ case _ => false
+ }
+ case _ => false
+ })
+ val (newPlan, newCond) = decorrelate(sub, plan, handleCountBugInDecorrelate)
+ val mayHaveCountBug = if (handleCountBugInDecorrelate) {
+ // Count bug was already handled in the above decorrelate function call.
+ false
+ } else if (mayHaveCountBugOld.isEmpty) {
Review Comment:
If the plan has already been decorrelated and count bugs handled in `decorrelate()` above, then we can't set `mayHaveCountBug` parameter to be true. This is because if we set it as true based on the pre-rewrite plan (which is what's happening now), running `splitSubquery()` in `constructLeftJoins` will fail because it encounters left outer joins (invalid operators).
Edit: but yeah, from the failing build it appears like i'm facing some issues with idempotency.
--
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