-
Notifications
You must be signed in to change notification settings - Fork 435
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
[GLUTEN-7267][CH]Support nested column pruning for HiveTableScan json/parquet/orc format
#7268
Open
KevinyhZou
wants to merge
14
commits into
apache:main
Choose a base branch
from
KevinyhZou:support_nested_project_push_down_json
base: main
Could not load branches
Branch not found: {{ refName }}
Loading
Could not load tags
Nothing to show
Loading
Are you sure you want to change the base?
Some commits from the old base branch may be removed from the timeline,
and old review comments may become outdated.
Open
Changes from 11 commits
Commits
Show all changes
14 commits
Select commit
Hold shift + click to select a range
4fa9091
support nested column pruning
bb1def6
rebase and solve conflict
4c202a6
resolve conflict
b29ce7d
use spark shema prunning
b367576
use spark schema pruning
955ad45
support prunning for orc/parquet
KevinyhZou 766e7d4
fix test
KevinyhZou 7f98477
remove useless file
KevinyhZou c865c8c
fix ci test
d45d022
fix ci test
8259217
remove orc test
f3282a1
Merge branch 'main' into support_nested_project_push_down_json
53255e3
add config for nested column pruning
f9ba278
only for ch backend
File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
255 changes: 255 additions & 0 deletions
255
...substrait/src/main/scala/org/apache/spark/sql/hive/HiveTableScanNestedColumnPruning.scala
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,255 @@ | ||
| /* | ||
| * Licensed to the Apache Software Foundation (ASF) under one or more | ||
| * contributor license agreements. See the NOTICE file distributed with | ||
| * this work for additional information regarding copyright ownership. | ||
| * The ASF licenses this file to You under the Apache License, Version 2.0 | ||
| * (the "License"); you may not use this file except in compliance with | ||
| * the License. You may obtain a copy of the License at | ||
| * | ||
| * http://www.apache.org/licenses/LICENSE-2.0 | ||
| * | ||
| * Unless required by applicable law or agreed to in writing, software | ||
| * distributed under the License is distributed on an "AS IS" BASIS, | ||
| * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| * See the License for the specific language governing permissions and | ||
| * limitations under the License. | ||
| */ | ||
| package org.apache.spark.sql.hive | ||
|
|
||
| import org.apache.gluten.backendsapi.BackendsApiManager | ||
| import org.apache.gluten.execution.ProjectExecTransformer | ||
|
|
||
| import org.apache.spark.internal.Logging | ||
| import org.apache.spark.sql.catalyst.catalog.HiveTableRelation | ||
| import org.apache.spark.sql.catalyst.expressions._ | ||
| import org.apache.spark.sql.execution.{FilterExec, LeafExecNode, ProjectExec, SparkPlan} | ||
| import org.apache.spark.sql.hive.HiveTableScanExecTransformer.{ORC_INPUT_FORMAT_CLASS, PARQUET_INPUT_FORMAT_CLASS, TEXT_INPUT_FORMAT_CLASS} | ||
| import org.apache.spark.sql.types.{ArrayType, DataType, MapType, StructType} | ||
| import org.apache.spark.sql.util.SchemaUtils._ | ||
| import org.apache.spark.util.Utils | ||
|
|
||
| object HiveTableScanNestedColumnPruning extends Logging { | ||
| import org.apache.spark.sql.catalyst.expressions.SchemaPruning._ | ||
|
|
||
| def supportNestedColumnPruning(projectExec: ProjectExec): Boolean = { | ||
| if (BackendsApiManager.getSparkPlanExecApiInstance.supportHiveTableScanNestedColumnPruning()) { | ||
| projectExec.child match { | ||
| case HiveTableScanExecTransformer(_, relation, _, _) => | ||
| relation.tableMeta.storage.inputFormat match { | ||
| case Some(inputFormat) | ||
| if TEXT_INPUT_FORMAT_CLASS.isAssignableFrom(Utils.classForName(inputFormat)) => | ||
| relation.tableMeta.storage.serde match { | ||
| case Some("org.openx.data.jsonserde.JsonSerDe") | Some( | ||
| "org.apache.hive.hcatalog.data.JsonSerDe") => | ||
| return true | ||
| case _ => | ||
| } | ||
| case Some(inputFormat) | ||
| if ORC_INPUT_FORMAT_CLASS.isAssignableFrom(Utils.classForName(inputFormat)) => | ||
| return true | ||
| case Some(inputFormat) | ||
| if PARQUET_INPUT_FORMAT_CLASS.isAssignableFrom(Utils.classForName(inputFormat)) => | ||
| return true | ||
| case _ => | ||
| } | ||
| case _ => | ||
| } | ||
| } | ||
| false | ||
| } | ||
|
|
||
| def apply(plan: SparkPlan): SparkPlan = { | ||
| plan match { | ||
| case ProjectExec(projectList, child) => | ||
| child match { | ||
| case h: HiveTableScanExecTransformer => | ||
| val newPlan = prunePhysicalColumns( | ||
| h.relation, | ||
| projectList, | ||
| Seq.empty[Expression], | ||
| (prunedDataSchema, prunedMetadataSchema) => { | ||
| buildNewHiveTableScan(h, prunedDataSchema, prunedMetadataSchema) | ||
| }, | ||
| (schema, requestFields) => { | ||
| h.pruneSchema(schema, requestFields) | ||
| } | ||
| ) | ||
| if (newPlan.nonEmpty) { | ||
| return newPlan.get | ||
| } else { | ||
| return ProjectExecTransformer(projectList, child) | ||
| } | ||
| case _ => | ||
| return ProjectExecTransformer(projectList, child) | ||
| } | ||
| case _ => | ||
| } | ||
| plan | ||
| } | ||
|
|
||
| private def prunePhysicalColumns( | ||
| relation: HiveTableRelation, | ||
| projects: Seq[NamedExpression], | ||
| filters: Seq[Expression], | ||
| leafNodeBuilder: (StructType, StructType) => LeafExecNode, | ||
| pruneSchemaFunc: (StructType, Seq[SchemaPruning.RootField]) => StructType) | ||
| : Option[SparkPlan] = { | ||
| val (normalizedProjects, normalizedFilters) = | ||
| normalizeAttributeRefNames(relation.output, projects, filters) | ||
| val requestedRootFields = identifyRootFields(normalizedProjects, normalizedFilters) | ||
| // If requestedRootFields includes a nested field, continue. Otherwise, | ||
| // return op | ||
| if (requestedRootFields.exists { root: RootField => !root.derivedFromAtt }) { | ||
| val prunedDataSchema = pruneSchemaFunc(relation.tableMeta.dataSchema, requestedRootFields) | ||
| val metaFieldNames = relation.tableMeta.schema.fieldNames | ||
| val metadataSchema = relation.output.collect { | ||
| case attr: AttributeReference if metaFieldNames.contains(attr.name) => attr | ||
| }.toStructType | ||
| val prunedMetadataSchema = if (metadataSchema.nonEmpty) { | ||
| pruneSchemaFunc(metadataSchema, requestedRootFields) | ||
| } else { | ||
| metadataSchema | ||
| } | ||
| // If the data schema is different from the pruned data schema | ||
| // OR | ||
| // the metadata schema is different from the pruned metadata schema, continue. | ||
| // Otherwise, return None. | ||
| if ( | ||
| countLeaves(relation.tableMeta.dataSchema) > countLeaves(prunedDataSchema) || | ||
| countLeaves(metadataSchema) > countLeaves(prunedMetadataSchema) | ||
| ) { | ||
| val leafNode = leafNodeBuilder(prunedDataSchema, prunedMetadataSchema) | ||
| val projectionOverSchema = ProjectionOverSchema( | ||
| prunedDataSchema.merge(prunedMetadataSchema), | ||
| AttributeSet(relation.output)) | ||
| Some( | ||
| buildNewProjection( | ||
| projects, | ||
| normalizedProjects, | ||
| normalizedFilters, | ||
| leafNode, | ||
| projectionOverSchema)) | ||
| } else { | ||
| None | ||
| } | ||
| } else { | ||
| None | ||
| } | ||
| } | ||
|
|
||
| /** | ||
| * Normalizes the names of the attribute references in the given projects and filters to reflect | ||
| * the names in the given logical relation. This makes it possible to compare attributes and | ||
| * fields by name. Returns a tuple with the normalized projects and filters, respectively. | ||
| */ | ||
| private def normalizeAttributeRefNames( | ||
| output: Seq[AttributeReference], | ||
| projects: Seq[NamedExpression], | ||
| filters: Seq[Expression]): (Seq[NamedExpression], Seq[Expression]) = { | ||
| val normalizedAttNameMap = output.map(att => (att.exprId, att.name)).toMap | ||
| val normalizedProjects = projects | ||
| .map(_.transform { | ||
| case att: AttributeReference if normalizedAttNameMap.contains(att.exprId) => | ||
| att.withName(normalizedAttNameMap(att.exprId)) | ||
| }) | ||
| .map { case expr: NamedExpression => expr } | ||
| val normalizedFilters = filters.map(_.transform { | ||
| case att: AttributeReference if normalizedAttNameMap.contains(att.exprId) => | ||
| att.withName(normalizedAttNameMap(att.exprId)) | ||
| }) | ||
| (normalizedProjects, normalizedFilters) | ||
| } | ||
|
|
||
| /** Builds the new output [[Project]] Spark SQL operator that has the `leafNode`. */ | ||
| private def buildNewProjection( | ||
| projects: Seq[NamedExpression], | ||
| normalizedProjects: Seq[NamedExpression], | ||
| filters: Seq[Expression], | ||
| leafNode: LeafExecNode, | ||
| projectionOverSchema: ProjectionOverSchema): ProjectExecTransformer = { | ||
| // Construct a new target for our projection by rewriting and | ||
| // including the original filters where available | ||
| val projectionChild = | ||
| if (filters.nonEmpty) { | ||
| val projectedFilters = filters.map(_.transformDown { | ||
| case projectionOverSchema(expr) => expr | ||
| }) | ||
| val newFilterCondition = projectedFilters.reduce(And) | ||
| FilterExec(newFilterCondition, leafNode) | ||
| } else { | ||
| leafNode | ||
| } | ||
|
|
||
| // Construct the new projections of our Project by | ||
| // rewriting the original projections | ||
| val newProjects = | ||
| normalizedProjects.map(_.transformDown { case projectionOverSchema(expr) => expr }).map { | ||
| case expr: NamedExpression => expr | ||
| } | ||
|
|
||
| if (log.isDebugEnabled) { | ||
| logDebug(s"New projects:\n${newProjects.map(_.treeString).mkString("\n")}") | ||
| } | ||
| ProjectExecTransformer( | ||
| restoreOriginalOutputNames(newProjects, projects.map(_.name)), | ||
| projectionChild) | ||
| } | ||
|
|
||
| private def buildNewHiveTableScan( | ||
| hiveTableScan: HiveTableScanExecTransformer, | ||
| prunedDataSchema: StructType, | ||
| prunedMetadataSchema: StructType): HiveTableScanExecTransformer = { | ||
| val relation = hiveTableScan.relation | ||
| val partitionSchema = relation.tableMeta.partitionSchema | ||
| val prunedBaseSchema = StructType( | ||
| prunedDataSchema.fields.filterNot( | ||
| f => partitionSchema.fieldNames.contains(f.name)) ++ partitionSchema.fields) | ||
| val finalSchema = prunedBaseSchema.merge(prunedMetadataSchema) | ||
| val prunedOutput = getPrunedOutput(relation.output, finalSchema) | ||
| var finalOutput = Seq.empty[Attribute] | ||
| for (p <- hiveTableScan.output) { | ||
| var flag = false | ||
| for (q <- prunedOutput if !flag) { | ||
| if (p.name.equals(q.name)) { | ||
| finalOutput :+= q | ||
| flag = true | ||
| } | ||
| } | ||
| } | ||
| HiveTableScanExecTransformer( | ||
| hiveTableScan.requestedAttributes, | ||
| relation, | ||
| hiveTableScan.partitionPruningPred, | ||
| finalOutput)(hiveTableScan.session) | ||
| } | ||
|
|
||
| // Prune the given output to make it consistent with `requiredSchema`. | ||
| private def getPrunedOutput( | ||
| output: Seq[AttributeReference], | ||
| requiredSchema: StructType): Seq[Attribute] = { | ||
| // We need to update the data type of the output attributes to use the pruned ones. | ||
| // so that references to the original relation's output are not broken | ||
| val nameAttributeMap = output.map(att => (att.name, att)).toMap | ||
| val requiredAttributes = | ||
| requiredSchema.map(f => AttributeReference(f.name, f.dataType, f.nullable, f.metadata)()) | ||
| requiredAttributes.map { | ||
| case att if nameAttributeMap.contains(att.name) => | ||
| nameAttributeMap(att.name).withDataType(att.dataType) | ||
| case att => att | ||
| } | ||
| } | ||
|
|
||
| /** | ||
| * Counts the "leaf" fields of the given dataType. Informally, this is the number of fields of | ||
| * non-complex data type in the tree representation of [[DataType]]. | ||
| */ | ||
| private def countLeaves(dataType: DataType): Int = { | ||
| dataType match { | ||
| case array: ArrayType => countLeaves(array.elementType) | ||
| case map: MapType => countLeaves(map.keyType) + countLeaves(map.valueType) | ||
| case struct: StructType => | ||
| struct.map(field => countLeaves(field.dataType)).sum | ||
| case _ => 1 | ||
| } | ||
| } | ||
| } |
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
这俩orc和parquet的开关在什么使用场景下是false呢
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
当需要使用hive parquet/orc serde 读取 table 时，而不是使用spark内置的parquet/orc reader读取时，这两个配置就需要被设置为false @taiyang-li