Skip to content
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.

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

[VL] Make bloom_filter_agg fall back when might_contain is not transformable #3917

Merged
merged 5 commits into from
Dec 6, 2023
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
31 changes: 3 additions & 28 deletions docs/velox-backend-limitations.md
Original file line number Diff line number Diff line change
Expand Up @@ -11,34 +11,6 @@ Gluten avoids to modify Spark's existing code and use Spark APIs if possible. Ho

So you need to ensure preferentially load the Gluten jar to overwrite the jar of vanilla spark. Refer to [How to prioritize loading Gluten jars in Spark](https://github.com/oap-project/gluten/blob/main/docs/velox-backend-troubleshooting.md#incompatible-class-error-when-using-native-writer).


### Runtime BloomFilter

Velox BloomFilter's implementation is different from Spark's. So if `might_contain` falls back, but `bloom_filter_agg` is offloaded to velox, an exception will be thrown.

#### example

```sql
SELECT might_contain(null, null) both_null,
might_contain(null, 1L) null_bf,
might_contain((SELECT bloom_filter_agg(cast(id as long)) from range(1, 10000)),
null) null_value
```

The below exception will be thrown.

```
Unexpected Bloom filter version number (512)
java.io.IOException: Unexpected Bloom filter version number (512)
at org.apache.spark.util.sketch.BloomFilterImpl.readFrom0(BloomFilterImpl.java:256)
at org.apache.spark.util.sketch.BloomFilterImpl.readFrom(BloomFilterImpl.java:265)
at org.apache.spark.util.sketch.BloomFilter.readFrom(BloomFilter.java:178)
```

#### Solution

Set the gluten config `spark.gluten.sql.native.bloomFilter=false` to fall back to vanilla bloom filter, you can also disable runtime filter by setting spark config `spark.sql.optimizer.runtime.bloomFilter.enabled=false`.

### Fallbacks
Except the unsupported operators, functions, file formats, data sources listed in , there are some known cases also fall back to Vanilla Spark.

Expand All @@ -52,6 +24,9 @@ Gluten only supports spark default case-insensitive mode. If case-sensitive mode
In velox, lookaround (lookahead/lookbehind) pattern is not supported in RE2-based implementations for Spark functions,
such as `rlike`, `regexp_extract`, etc.

#### Runtime BloomFilter
Velox BloomFilter's serialization format is different from Spark's. BloomFilter binary generated by Velox can't be deserialized by vanilla spark. So if `might_contain` falls back, we fall back `bloom_filter_agg` to vanilla spark also.

#### FileSource format
Currently, Gluten only fully supports parquet file format and partially support ORC. If other format is used, scan operator falls back to vanilla spark.

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -815,6 +815,7 @@ case class ColumnarOverrideRules(session: SparkSession)
(spark: SparkSession) => PlanOneRowRelation(spark),
(_: SparkSession) => FallbackEmptySchemaRelation(),
(_: SparkSession) => AddTransformHintRule(),
(_: SparkSession) => FallbackBloomFilterAggIfNeeded(),
(_: SparkSession) => TransformPreOverrides(isAdaptiveContext),
(spark: SparkSession) => RewriteTransformer(spark),
(_: SparkSession) => EnsureLocalSortRequirements
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,7 @@ import io.glutenproject.utils.PhysicalPlanSelector
import org.apache.spark.api.python.EvalPythonExecTransformer
import org.apache.spark.sql.SparkSession
import org.apache.spark.sql.catalyst.InternalRow
import org.apache.spark.sql.catalyst.expressions.{AttributeReference, SortOrder}
import org.apache.spark.sql.catalyst.expressions.{AttributeReference, Expression, SortOrder}
import org.apache.spark.sql.catalyst.optimizer.{BuildLeft, BuildRight}
import org.apache.spark.sql.catalyst.plans.FullOuter
import org.apache.spark.sql.catalyst.rules.Rule
Expand Down Expand Up @@ -270,6 +270,49 @@ case class FallbackEmptySchemaRelation() extends Rule[SparkPlan] {
}
}

/**
* Velox BloomFilter's implementation is different from Spark's. So if might_contain falls back, we
* need fall back related bloom filter agg.
*/
case class FallbackBloomFilterAggIfNeeded() extends Rule[SparkPlan] {
override def apply(plan: SparkPlan): SparkPlan =
if (GlutenConfig.getConf.enableNativeBloomFilter) {
plan.transformDown {
case p if TransformHints.isAlreadyTagged(p) && TransformHints.isNotTransformable(p) =>
handleBloomFilterFallback(p)
p
}
} else {
plan
}

object SubPlanFromBloomFilterMightContain {
def unapply(expr: Expression): Option[SparkPlan] =
SparkShimLoader.getSparkShims.extractSubPlanFromMightContain(expr)
}

private def handleBloomFilterFallback(plan: SparkPlan): Unit = {
def tagNotTransformableRecursive(p: SparkPlan): Unit = {
p match {
case agg: org.apache.spark.sql.execution.aggregate.ObjectHashAggregateExec
if SparkShimLoader.getSparkShims.hasBloomFilterAggregate(agg) =>
TransformHints.tagNotTransformable(agg, "related BloomFilterMightContain falls back")
tagNotTransformableRecursive(agg.child)
case a: org.apache.spark.sql.execution.adaptive.AdaptiveSparkPlanExec =>
tagNotTransformableRecursive(a.executedPlan)
case _ =>
p.children.map(tagNotTransformableRecursive)
}
}

plan.transformAllExpressions {
case expr @ SubPlanFromBloomFilterMightContain(p: SparkPlan) =>
tagNotTransformableRecursive(p)
expr
}
}
}

// This rule will try to convert a plan into plan transformer.
// The doValidate function will be called to check if the conversion is supported.
// If false is returned or any unsupported exception is thrown, a row guard will
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -51,8 +51,6 @@ class VeloxTestSettings extends BackendTestSettings {
.exclude("string split function with positive limit")
.exclude("string split function with negative limit")
enableSuite[GlutenBloomFilterAggregateQuerySuite]
// fallback might_contain, the input argument binary is not same with vanilla spark
.exclude("Test NULL inputs for might_contain")
enableSuite[GlutenDataSourceV2DataFrameSessionCatalogSuite]
enableSuite[GlutenDataSourceV2DataFrameSuite]
enableSuite[GlutenDataSourceV2FunctionSuite]
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -51,8 +51,6 @@ class VeloxTestSettings extends BackendTestSettings {
.exclude("string split function with positive limit")
.exclude("string split function with negative limit")
enableSuite[GlutenBloomFilterAggregateQuerySuite]
// fallback might_contain, the input argument binary is not same with vanilla spark
.exclude("Test NULL inputs for might_contain")
enableSuite[GlutenDataSourceV2DataFrameSessionCatalogSuite]
enableSuite[GlutenDataSourceV2DataFrameSuite]
enableSuite[GlutenDataSourceV2FunctionSuite]
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@ import org.apache.spark.sql.catalyst.expressions.{Expression, PlanExpression}
import org.apache.spark.sql.catalyst.plans.physical.Distribution
import org.apache.spark.sql.connector.catalog.Table
import org.apache.spark.sql.connector.expressions.Transform
import org.apache.spark.sql.execution.FileSourceScanExec
import org.apache.spark.sql.execution.{FileSourceScanExec, SparkPlan}
import org.apache.spark.sql.execution.datasources.{FilePartition, FileScanRDD, PartitionDirectory, PartitionedFile, PartitioningAwareFileIndex}
import org.apache.spark.sql.execution.datasources.v2.BatchScanExec
import org.apache.spark.sql.execution.datasources.v2.text.TextScan
Expand Down Expand Up @@ -81,4 +81,9 @@ trait SparkShims {
start: Long,
length: Long,
@transient locations: Array[String] = Array.empty): PartitionedFile

def hasBloomFilterAggregate(
agg: org.apache.spark.sql.execution.aggregate.ObjectHashAggregateExec): Boolean

def extractSubPlanFromMightContain(expr: Expression): Option[SparkPlan]
}
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,7 @@ import org.apache.spark.sql.catalyst.expressions.Expression
import org.apache.spark.sql.catalyst.plans.physical.{Distribution, HashClusteredDistribution}
import org.apache.spark.sql.connector.catalog.Table
import org.apache.spark.sql.connector.expressions.Transform
import org.apache.spark.sql.execution.{FileSourceScanExec, PartitionedFileUtil}
import org.apache.spark.sql.execution.{FileSourceScanExec, PartitionedFileUtil, SparkPlan}
import org.apache.spark.sql.execution.datasources.{BucketingUtils, FilePartition, FileScanRDD, PartitionDirectory, PartitionedFile, PartitioningAwareFileIndex}
import org.apache.spark.sql.execution.datasources.FileFormatWriter.Empty2Null
import org.apache.spark.sql.execution.datasources.v2.BatchScanExec
Expand Down Expand Up @@ -101,4 +101,9 @@ class Spark32Shims extends SparkShims {
length: Long,
@transient locations: Array[String] = Array.empty): PartitionedFile =
PartitionedFile(partitionValues, filePath, start, length, locations)

override def hasBloomFilterAggregate(
agg: org.apache.spark.sql.execution.aggregate.ObjectHashAggregateExec): Boolean = false

override def extractSubPlanFromMightContain(expr: Expression): Option[SparkPlan] = None
}
Original file line number Diff line number Diff line change
Expand Up @@ -127,6 +127,24 @@ class Spark33Shims extends SparkShims {
@transient locations: Array[String] = Array.empty): PartitionedFile =
PartitionedFile(partitionValues, filePath, start, length, locations)

override def hasBloomFilterAggregate(
agg: org.apache.spark.sql.execution.aggregate.ObjectHashAggregateExec): Boolean = {
agg.aggregateExpressions.exists(
expr => expr.aggregateFunction.isInstanceOf[BloomFilterAggregate])
}

override def extractSubPlanFromMightContain(expr: Expression): Option[SparkPlan] = {
expr match {
case mc @ BloomFilterMightContain(sub: org.apache.spark.sql.execution.ScalarSubquery, _) =>
Some(sub.plan)
case mc @ BloomFilterMightContain(
g @ GetStructField(sub: org.apache.spark.sql.execution.ScalarSubquery, _, _),
_) =>
Some(sub.plan)
case _ => None
}
}

private def invalidBucketFile(path: String): Throwable = {
new SparkException(
errorClass = "INVALID_BUCKET_FILE",
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -130,6 +130,24 @@ class Spark34Shims extends SparkShims {
@transient locations: Array[String] = Array.empty): PartitionedFile =
PartitionedFile(partitionValues, SparkPath.fromPathString(filePath), start, length, locations)

override def hasBloomFilterAggregate(
agg: org.apache.spark.sql.execution.aggregate.ObjectHashAggregateExec): Boolean = {
agg.aggregateExpressions.exists(
expr => expr.aggregateFunction.isInstanceOf[BloomFilterAggregate])
}

override def extractSubPlanFromMightContain(expr: Expression): Option[SparkPlan] = {
expr match {
case mc @ BloomFilterMightContain(sub: org.apache.spark.sql.execution.ScalarSubquery, _) =>
Some(sub.plan)
case mc @ BloomFilterMightContain(
g @ GetStructField(sub: org.apache.spark.sql.execution.ScalarSubquery, _, _),
_) =>
Some(sub.plan)
case _ => None
}
}

private def invalidBucketFile(path: String): Throwable = {
new SparkException(
errorClass = "INVALID_BUCKET_FILE",
Expand Down
Loading