You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by "ulysses-you (via GitHub)" <gi...@apache.org> on 2023/05/08 05:11:48 UTC
[GitHub] [spark] ulysses-you opened a new pull request, #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
ulysses-you opened a new pull request, #41088:
URL: https://github.com/apache/spark/pull/41088
<!--
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.
-->
Scalar subquery can be pushed down as data filter at runtime, since we always execute subquery first. Ideally, we can rewrite `ScalarSubquery` to `Literal` before pushing down filter. The main issue before we do not support that is `ReuseSubquery` is ineffective, see https://github.com/apache/spark/pull/22518. It is not a issue now.
For example:
```sql
SELECT * FROM t1 WHERE c1 > (SELECT min(c2) FROM t2)
```
### 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.
-->
Improve peformance if data filter have scalar subquery.
### 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'.
-->
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.
-->
add 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] cloud-fan commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1281683113
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -189,7 +189,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
Review Comment:
Actually, subquery expressions must be non-correlated when we reach here (they must have been rewritten into joins before). Then we need to deal with two kinds of subqueries:
1. foldable ones like scalar subquery. We can turn them into literals at the beginning and do filtering.
2. non-foldable ones like the DPP subquery. We will evaluate them when we have an initial partition list and need to filter it further.
--
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-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery [spark]
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on PR #41088:
URL: https://github.com/apache/spark/pull/41088#issuecomment-1797129478
@epa095 yes, updated that jira status
--
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 #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1282582380
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -189,7 +189,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
Review Comment:
Can 2 be used in partition filters?
--
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 #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1281629444
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -171,13 +172,12 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
val partitionKeyFilters = DataSourceStrategy.getPushedDownFilters(partitionColumns,
normalizedFilters)
- // subquery expressions are filtered out because they can't be used to prune buckets or pushed
- // down as data filters, yet they would be executed
- val normalizedFiltersWithoutSubqueries =
- normalizedFilters.filterNot(SubqueryExpression.hasSubquery)
-
val bucketSpec: Option[BucketSpec] = fsRelation.bucketSpec
val bucketSet = if (shouldPruneBuckets(bucketSpec)) {
+ // subquery expressions are filtered out because they can't be used to prune buckets
Review Comment:
seems it's fine to not filter out scalar subquery, as the code of `genBucketSet` can skip it already.
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1189335057
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -543,7 +561,7 @@ case class FileSourceScanExec(
dataSchema = relation.dataSchema,
partitionSchema = relation.partitionSchema,
requiredSchema = requiredSchema,
- filters = pushedDownFilters,
+ filters = dynamicallyPushedDownFilters,
Review Comment:
If we first filter out scalar subqueries and then add them back, then the ordering of pushed filter is changed. e.g.,
`c1 > (select min(x) from t) and c2 > 1` -> `c2 > 1 and c1 > (select min(x) from t)`
I'm not sure if it affects the performance in parquet side. So I simply re-translate the whole data filter.
--
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] viirya commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "viirya (via GitHub)" <gi...@apache.org>.
viirya commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1189395690
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -188,7 +188,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
Review Comment:
Is it possible the scalar subquery is correlated?
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1281716224
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -189,7 +189,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
Review Comment:
yes, and list subquery should also be rewritten to join. So we can only match scalar subquery here for data filter. DPP is a partition filter.
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1189269733
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -543,7 +561,7 @@ case class FileSourceScanExec(
dataSchema = relation.dataSchema,
partitionSchema = relation.partitionSchema,
requiredSchema = requiredSchema,
- filters = pushedDownFilters,
+ filters = dynamicallyPushedDownFilters,
Review Comment:
We do not change the timing of trigger for subquery execution. If we call `execute` at `SparkPlan`, then it would always first prepare and call `updateResult` to executing subquery. So it seems there is no change with this pr. This pr only uses the result of subquery and push it down.
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1189452525
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -383,6 +383,24 @@ trait FileSourceScanLike extends DataSourceScanExec {
}).flatMap(DataSourceStrategy.translateFilter(_, supportNestedPredicatePushdown))
}
+ @transient
+ protected lazy val pushedDownFilters: Seq[Filter] = translatePushedDownFilters(dataFilters)
+
+ @transient
+ protected lazy val dynamicallyPushedDownFilters: Seq[Filter] = {
+ if (dataFilters.exists(_.exists(_.isInstanceOf[execution.ScalarSubquery]))) {
+ // Replace scalar subquery to literal so that `DataSourceStrategy.translateFilter` can
+ // support translate it. The subquery must has been materialized since SparkPlan always
+ // execute subquery first.
+ val normalized = dataFilters.map(_.transform {
+ case scalarSubquery: execution.ScalarSubquery => scalarSubquery.toLiteral
+ })
+ translatePushedDownFilters(normalized)
+ } else {
+ pushedDownFilters
+ }
+ }
+
override lazy val metadata: Map[String, String] = {
Review Comment:
I think we are not able to update it. metadata would be used by explain which is invoked before executing and we get the pushed filter with scalar subqueris at runtime.
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1281730562
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -383,6 +383,24 @@ trait FileSourceScanLike extends DataSourceScanExec {
}).flatMap(DataSourceStrategy.translateFilter(_, supportNestedPredicatePushdown))
}
+ @transient
+ protected lazy val pushedDownFilters: Seq[Filter] = translatePushedDownFilters(dataFilters)
+
+ @transient
+ protected lazy val dynamicallyPushedDownFilters: Seq[Filter] = {
+ if (dataFilters.exists(_.exists(_.isInstanceOf[execution.ScalarSubquery]))) {
+ // Replace scalar subquery to literal so that `DataSourceStrategy.translateFilter` can
+ // support translate it. The subquery must has been materialized since SparkPlan always
+ // execute subquery first.
+ val normalized = dataFilters.map(_.transform {
Review Comment:
The `outputPartitioning` and `outputOrdering` use `selectedPartitions` when it is bucket scan which is before do execution. Besides, `selectedPartitions` is a public field, I'm afraid it would break something if we assume it only be called at runtime...
--
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 #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1282626537
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -189,7 +189,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
Review Comment:
ok, so we only need to covert scala subquery to literal because filter pushdown needs v1 filter which doesn't support subquery?
--
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 #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1281653286
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -189,7 +189,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
Review Comment:
thinking about this again. I think all non-correlated subqueries can be allowed? They will be evaluated before the actual plan execution.
--
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] viirya commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "viirya (via GitHub)" <gi...@apache.org>.
viirya commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1189388803
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -383,6 +383,24 @@ trait FileSourceScanLike extends DataSourceScanExec {
}).flatMap(DataSourceStrategy.translateFilter(_, supportNestedPredicatePushdown))
}
+ @transient
+ protected lazy val pushedDownFilters: Seq[Filter] = translatePushedDownFilters(dataFilters)
+
+ @transient
+ protected lazy val dynamicallyPushedDownFilters: Seq[Filter] = {
+ if (dataFilters.exists(_.exists(_.isInstanceOf[execution.ScalarSubquery]))) {
+ // Replace scalar subquery to literal so that `DataSourceStrategy.translateFilter` can
+ // support translate it. The subquery must has been materialized since SparkPlan always
+ // execute subquery first.
+ val normalized = dataFilters.map(_.transform {
+ case scalarSubquery: execution.ScalarSubquery => scalarSubquery.toLiteral
+ })
+ translatePushedDownFilters(normalized)
+ } else {
+ pushedDownFilters
+ }
+ }
+
override lazy val metadata: Map[String, String] = {
Review Comment:
This `metadata` also contains `pushedDownFilters`, do we need to update it? Or, are we able to update it?
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1189315413
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -543,7 +561,7 @@ case class FileSourceScanExec(
dataSchema = relation.dataSchema,
partitionSchema = relation.partitionSchema,
requiredSchema = requiredSchema,
- filters = pushedDownFilters,
+ filters = dynamicallyPushedDownFilters,
Review Comment:
I did not change pushedDownFilters, it should not be a issue with `metadata`. But it's a pity that we can not make metadata aware of scalar subquery filter.
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1190541536
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -188,7 +188,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
+ .filterNot(_.isInstanceOf[BloomFilterMightContain])
+ val dataFilters = normalizedFiltersWithScalarSubqueries.flatMap { f =>
Review Comment:
no, translateFilter does not support scalar subquery. We will first transform scalar subquery to literal then go through translateFilter at runtime. The translateFilter related code in FileSourceStrategy is only used to print log which does not matter.
--
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-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery [spark]
Posted by "epa095 (via GitHub)" <gi...@apache.org>.
epa095 commented on PR #41088:
URL: https://github.com/apache/spark/pull/41088#issuecomment-1795621044
@ulysses-you Does this solve https://issues.apache.org/jira/browse/SPARK-34444 ?
--
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 #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1281647040
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -383,6 +383,24 @@ trait FileSourceScanLike extends DataSourceScanExec {
}).flatMap(DataSourceStrategy.translateFilter(_, supportNestedPredicatePushdown))
}
+ @transient
+ protected lazy val pushedDownFilters: Seq[Filter] = translatePushedDownFilters(dataFilters)
+
+ @transient
+ protected lazy val dynamicallyPushedDownFilters: Seq[Filter] = {
+ if (dataFilters.exists(_.exists(_.isInstanceOf[execution.ScalarSubquery]))) {
+ // Replace scalar subquery to literal so that `DataSourceStrategy.translateFilter` can
+ // support translate it. The subquery must has been materialized since SparkPlan always
+ // execute subquery first.
+ val normalized = dataFilters.map(_.transform {
Review Comment:
should `selectedPartitions` also use the normalized data filters?
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1282632165
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -189,7 +189,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
Review Comment:
yes, v1 pushdown filter seems only support 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] ulysses-you commented on pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on PR #41088:
URL: https://github.com/apache/spark/pull/41088#issuecomment-1537799363
cc @cloud-fan @viirya thank you
--
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] viirya commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "viirya (via GitHub)" <gi...@apache.org>.
viirya commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1190554776
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -188,7 +188,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
+ .filterNot(_.isInstanceOf[BloomFilterMightContain])
+ val dataFilters = normalizedFiltersWithScalarSubqueries.flatMap { f =>
Review Comment:
Oh, yea, I remember this 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] viirya commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "viirya (via GitHub)" <gi...@apache.org>.
viirya commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1190309544
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -188,7 +188,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
+ .filterNot(_.isInstanceOf[BloomFilterMightContain])
+ val dataFilters = normalizedFiltersWithScalarSubqueries.flatMap { f =>
Review Comment:
Hmm, `dataFilters` will go through `translateFilter`. Does `translateFilter` support translate scalar subquery?
--
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] ulysses-you commented on pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on PR #41088:
URL: https://github.com/apache/spark/pull/41088#issuecomment-1657395612
thank you for review, merged to master!
--
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 #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1281629444
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -171,13 +172,12 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
val partitionKeyFilters = DataSourceStrategy.getPushedDownFilters(partitionColumns,
normalizedFilters)
- // subquery expressions are filtered out because they can't be used to prune buckets or pushed
- // down as data filters, yet they would be executed
- val normalizedFiltersWithoutSubqueries =
- normalizedFilters.filterNot(SubqueryExpression.hasSubquery)
-
val bucketSpec: Option[BucketSpec] = fsRelation.bucketSpec
val bucketSet = if (shouldPruneBuckets(bucketSpec)) {
+ // subquery expressions are filtered out because they can't be used to prune buckets
Review Comment:
why scalar subquery can't be used to prune buckets?
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1282550963
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -189,7 +189,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
Review Comment:
it seems 2 and 3 are same but 2 is data filter and 3 is partition filter ?
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1188377686
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -188,7 +188,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
Review Comment:
partition filter has already supported since https://github.com/apache/spark/pull/23802
--
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 #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1282621969
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -189,7 +189,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
Review Comment:
I mean, something like `part_col = (select ... )`
--
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] ulysses-you closed pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you closed pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
URL: https://github.com/apache/spark/pull/41088
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1282596741
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -189,7 +189,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
Review Comment:
Technically it can be. But there is a issue if we want to use data filte to help pruning partition directory https://github.com/apache/spark/pull/41088#discussion_r1281730562. Or, add one more field to do one more list partition directory with normalized data filter.
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1281706092
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -383,6 +383,24 @@ trait FileSourceScanLike extends DataSourceScanExec {
}).flatMap(DataSourceStrategy.translateFilter(_, supportNestedPredicatePushdown))
}
+ @transient
+ protected lazy val pushedDownFilters: Seq[Filter] = translatePushedDownFilters(dataFilters)
+
+ @transient
+ protected lazy val dynamicallyPushedDownFilters: Seq[Filter] = {
+ if (dataFilters.exists(_.exists(_.isInstanceOf[execution.ScalarSubquery]))) {
+ // Replace scalar subquery to literal so that `DataSourceStrategy.translateFilter` can
+ // support translate it. The subquery must has been materialized since SparkPlan always
+ // execute subquery first.
+ val normalized = dataFilters.map(_.transform {
Review Comment:
Seems `selectedPartitions` does not accept subquery expression, it filter out the dpp filter first.
https://github.com/apache/spark/blob/e326b9b6565b1e2b697d41714060e2ff9e424514/sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala#L252-L257
Then in `dynamicallySelectedPartitions`, normalized the partition filter.
https://github.com/apache/spark/blob/e326b9b6565b1e2b697d41714060e2ff9e424514/sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala#L268-L281
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1282624369
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -189,7 +189,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
Review Comment:
yes, it has been worked since https://github.com/apache/spark/pull/23802. So, both `partitionFilters` and `dataFilters` support contain subquery expression.
--
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] ulysses-you commented on pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on PR #41088:
URL: https://github.com/apache/spark/pull/41088#issuecomment-1549359117
@cloud-fan @viirya @wangyum do you have other thought ?
--
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] wangyum commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "wangyum (via GitHub)" <gi...@apache.org>.
wangyum commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1189280141
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -543,7 +561,7 @@ case class FileSourceScanExec(
dataSchema = relation.dataSchema,
partitionSchema = relation.partitionSchema,
requiredSchema = requiredSchema,
- filters = pushedDownFilters,
+ filters = dynamicallyPushedDownFilters,
Review Comment:
Could we split the `dynamicallyPushedDownFilters` to `pushedDownFilters` and `pushedDownRuntimeFilters` and combine them here? Something like this: https://github.com/apache/spark/pull/36128/files#diff-089285f1484c1598cb2839b86b6a9e65b98ab5b30462aedc210fe4bbf44cae78R374-R455
This is because `pushedDownFilters` are also used in [`metadata`](https://github.com/apache/spark/blob/abd841bdcabc64d04e0743b0c16443cf0bb66558/sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala#L416), while `pushedDownRuntimeFilters` can only be used 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 a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1281649535
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -383,6 +383,24 @@ trait FileSourceScanLike extends DataSourceScanExec {
}).flatMap(DataSourceStrategy.translateFilter(_, supportNestedPredicatePushdown))
}
+ @transient
+ protected lazy val pushedDownFilters: Seq[Filter] = translatePushedDownFilters(dataFilters)
+
+ @transient
+ protected lazy val dynamicallyPushedDownFilters: Seq[Filter] = {
+ if (dataFilters.exists(_.exists(_.isInstanceOf[execution.ScalarSubquery]))) {
+ // Replace scalar subquery to literal so that `DataSourceStrategy.translateFilter` can
+ // support translate it. The subquery must has been materialized since SparkPlan always
+ // execute subquery first.
+ val normalized = dataFilters.map(_.transform {
Review Comment:
I feel a bit of inconsistency here. The partition filter can also contain scala subquery, when is it normalized?
--
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 #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1281849044
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -189,7 +189,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
Review Comment:
I think we need to make it clear about the 3 different kinds of predicates
1. simple predicates that can be used to prune partitions, or anything that needs to be accessed during planning
2. foldable subquery expressions that can be turned into literal before further use (is data source filter pushdown the only use?)
3. subquery expressions that can be used to do additional pruning after planning but before execution.
--
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 #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1281713302
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -383,6 +383,24 @@ trait FileSourceScanLike extends DataSourceScanExec {
}).flatMap(DataSourceStrategy.translateFilter(_, supportNestedPredicatePushdown))
}
+ @transient
+ protected lazy val pushedDownFilters: Seq[Filter] = translatePushedDownFilters(dataFilters)
+
+ @transient
+ protected lazy val dynamicallyPushedDownFilters: Seq[Filter] = {
+ if (dataFilters.exists(_.exists(_.isInstanceOf[execution.ScalarSubquery]))) {
+ // Replace scalar subquery to literal so that `DataSourceStrategy.translateFilter` can
+ // support translate it. The subquery must has been materialized since SparkPlan always
+ // execute subquery first.
+ val normalized = dataFilters.map(_.transform {
Review Comment:
For scalar subquery, I think we should turn it into literal so that it can be used to prune partitions?
--
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] ulysses-you commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1189456831
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -188,7 +188,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
+ // execute subquery first.
+ // It has no meaning to push down bloom filter, so skip it.
+ val normalizedFiltersWithScalarSubqueries = normalizedFilters
+ .filterNot(e => e.containsPattern(PLAN_EXPRESSION) && !e.containsPattern(SCALAR_SUBQUERY))
Review Comment:
it seems impossible, we have rewritten the correlated scalar subquery at Optimizer side. Otherwise it would break tons of things, e.g., the partition filter with sclar subquery.
--
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 #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1188178513
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala:
##########
@@ -188,7 +188,13 @@ object FileSourceStrategy extends Strategy with PredicateHelper with Logging {
// Partition keys are not available in the statistics of the files.
// `dataColumns` might have partition columns, we need to filter them out.
val dataColumnsWithoutPartitionCols = dataColumns.filterNot(partitionSet.contains)
- val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
+ // Scalar subquery can be pushed down as data filter at runtime, since we always
Review Comment:
how about partition filters?
--
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 #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1188575493
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -543,7 +561,7 @@ case class FileSourceScanExec(
dataSchema = relation.dataSchema,
partitionSchema = relation.partitionSchema,
requiredSchema = requiredSchema,
- filters = pushedDownFilters,
+ filters = dynamicallyPushedDownFilters,
Review Comment:
One implicit change is, now we will trigger subquery execution when getting the rdd of the file scan node. This worries me a little bit. Can we trigger subquery execution in the AQE framework? e.g. before executing a query stage, transform the scan node and update its `dataFilters` to replace evaluated scalar subqueries with literals.
--
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] wangyum commented on a diff in pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "wangyum (via GitHub)" <gi...@apache.org>.
wangyum commented on code in PR #41088:
URL: https://github.com/apache/spark/pull/41088#discussion_r1189320749
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala:
##########
@@ -543,7 +561,7 @@ case class FileSourceScanExec(
dataSchema = relation.dataSchema,
partitionSchema = relation.partitionSchema,
requiredSchema = requiredSchema,
- filters = pushedDownFilters,
+ filters = dynamicallyPushedDownFilters,
Review Comment:
It is not the issue. Just wanted to make it clearer. `dynamicallyPushedDownFilters` also includes `pushedDownFilters`.
--
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] ulysses-you commented on pull request #41088: [SPARK-43402][SQL] FileSourceScanExec supports push down data filter with scalar subquery
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on PR #41088:
URL: https://github.com/apache/spark/pull/41088#issuecomment-1654874547
I'm going to merge this pr to master(4.0.0) if no any more comments in a few days, cc @viirya @cloud-fan @wangyum
--
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