Skip to content

[SPARK-33673][SQL] Avoid push down partition filters to ParquetScan for DataSourceV2 #30652

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 4 commits into from
Closed

[SPARK-33673][SQL] Avoid push down partition filters to ParquetScan for DataSourceV2 #30652

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
be6cfc8
use dataSchema instead of schema
LuciferYang Dec 7, 2020
fb01c3f
fix ExplainSuite
LuciferYang Dec 8, 2020
b9f8eb2
use readDataSchema instead of dataSchema
LuciferYang Dec 9, 2020
13bfda3
Merge branch 'upmaster' into SPARK-33673
LuciferYang Dec 14, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion ...main/scala/org/apache/spark/sql/execution/datasources/v2/parquet/ParquetScanBuilder.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,7 +50,7 @@ case class ParquetScanBuilder(
val pushDownInFilterThreshold = sqlConf.parquetFilterPushDownInFilterThreshold
val isCaseSensitive = sqlConf.caseSensitiveAnalysis
val parquetSchema =
new SparkToParquetSchemaConverter(sparkSession.sessionState.conf).convert(schema)
Copy link
Member

@viirya viirya Dec 8, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What if the partition column is in file? I guess this might be why we push down partition filter.

Copy link
Contributor Author

@LuciferYang LuciferYang Dec 8, 2020 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is a good question, but it seems that the filter pushed down in DataSource V1 does not contain the filter related to partition columns too.

The dataFilters use to construct FileSourceScanExec and pass to ParquetFileFormat.buildReaderWithPartitionValues to build pushed filters also filtered out partition filters, am I right?

spark/sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategy.scala

Lines 184 to 193 in e4d1c10

// 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(partitionColumns.contains)
val dataFilters = normalizedFiltersWithoutSubqueries.flatMap { f =>
if (f.references.intersect(partitionSet).nonEmpty) {
extractPredicatesWithinOutputSet(f, AttributeSet(dataColumnsWithoutPartitionCols))
} else {
Some(f)
}
}

Copy link
Contributor Author

@LuciferYang LuciferYang Dec 8, 2020 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe another way is refer to parquet FileMetaData#getSchema to determine which filter should be pushed down

Copy link
Member

@wangyum wangyum Dec 8, 2020 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe this is a correct fix. I have a pr to fix partition column is in file.
#30670

Copy link
Contributor Author

@LuciferYang LuciferYang Dec 9, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe we should use readDataSchema() instead of dataSchema

Copy link
Contributor Author

@LuciferYang LuciferYang Dec 9, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Address b9f8eb2 use readDataSchema() instead of dataSchema

new SparkToParquetSchemaConverter(sparkSession.sessionState.conf).convert(readDataSchema())
val parquetFilters = new ParquetFilters(parquetSchema, pushDownDate, pushDownTimestamp,
pushDownDecimal, pushDownStringStartWith, pushDownInFilterThreshold, isCaseSensitive)
parquetFilters.convertibleFilters(this.filters).toArray
Expand Down
2 changes: 1 addition & 1 deletion sql/core/src/test/scala/org/apache/spark/sql/ExplainSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -367,7 +367,7 @@ class ExplainSuite extends ExplainSuiteHelper with DisableAdaptiveExecutionSuite
val basePath = dir.getCanonicalPath + "/" + fmt
val pushFilterMaps = Map (
"parquet" ->
"|PushedFilers: \\[.*\\(id\\), .*\\(value\\), .*\\(id,1\\), .*\\(value,2\\)\\]",
"|PushedFilers: \\[IsNotNull\\(value\\), GreaterThan\\(value,2\\)\\]",
"orc" ->
"|PushedFilers: \\[.*\\(id\\), .*\\(value\\), .*\\(id,1\\), .*\\(value,2\\)\\]",
"csv" ->
Expand Down