Repository: spark Updated Branches: refs/heads/branch-2.0 47fc0b9f4 -> 0a593db36
[SPARK-17616][SQL] Support a single distinct aggregate combined with a non-partial aggregate We currently cannot execute an aggregate that contains a single distinct aggregate function and an one or more non-partially plannable aggregate functions, for example: ```sql select grp, collect_list(col1), count(distinct col2) from tbl_a group by 1 ``` This is a regression from Spark 1.6. This is caused by the fact that the single distinct aggregation code path assumes that all aggregates can be planned in two phases (is partially aggregatable). This PR works around this issue by triggering the `RewriteDistinctAggregates` in such cases (this is similar to the approach taken in 1.6). Created `RewriteDistinctAggregatesSuite` which checks if the aggregates with distinct aggregate functions get rewritten into two `Aggregates` and an `Expand`. Added a regression test to `DataFrameAggregateSuite`. Author: Herman van Hovell <hvanhov...@databricks.com> Closes #15187 from hvanhovell/SPARK-17616. (cherry picked from commit 0d634875026ccf1eaf984996e9460d7673561f80) Signed-off-by: Herman van Hovell <hvanhov...@databricks.com> Project: http://git-wip-us.apache.org/repos/asf/spark/repo Commit: http://git-wip-us.apache.org/repos/asf/spark/commit/0a593db3 Tree: http://git-wip-us.apache.org/repos/asf/spark/tree/0a593db3 Diff: http://git-wip-us.apache.org/repos/asf/spark/diff/0a593db3 Branch: refs/heads/branch-2.0 Commit: 0a593db360b3b7771f45f482cf45e8500f0faa76 Parents: 47fc0b9 Author: Herman van Hovell <hvanhov...@databricks.com> Authored: Thu Sep 22 14:29:27 2016 -0700 Committer: Herman van Hovell <hvanhov...@databricks.com> Committed: Thu Sep 22 16:22:31 2016 -0700 ---------------------------------------------------------------------- .../optimizer/RewriteDistinctAggregates.scala | 18 ++-- .../RewriteDistinctAggregatesSuite.scala | 94 ++++++++++++++++++++ .../spark/sql/DataFrameAggregateSuite.scala | 8 ++ 3 files changed, 111 insertions(+), 9 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/spark/blob/0a593db3/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/RewriteDistinctAggregates.scala ---------------------------------------------------------------------- diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/RewriteDistinctAggregates.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/RewriteDistinctAggregates.scala index 0f43e7b..d6a39ec 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/RewriteDistinctAggregates.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/RewriteDistinctAggregates.scala @@ -119,14 +119,16 @@ object RewriteDistinctAggregates extends Rule[LogicalPlan] { .filter(_.isDistinct) .groupBy(_.aggregateFunction.children.toSet) - // Aggregation strategy can handle the query with single distinct - if (distinctAggGroups.size > 1) { + // Check if the aggregates contains functions that do not support partial aggregation. + val existsNonPartial = aggExpressions.exists(!_.aggregateFunction.supportsPartial) + + // Aggregation strategy can handle queries with a single distinct group and partial aggregates. + if (distinctAggGroups.size > 1 || (distinctAggGroups.size == 1 && existsNonPartial)) { // Create the attributes for the grouping id and the group by clause. - val gid = - new AttributeReference("gid", IntegerType, false)(isGenerated = true) + val gid = AttributeReference("gid", IntegerType, nullable = false)(isGenerated = true) val groupByMap = a.groupingExpressions.collect { case ne: NamedExpression => ne -> ne.toAttribute - case e => e -> new AttributeReference(e.sql, e.dataType, e.nullable)() + case e => e -> AttributeReference(e.sql, e.dataType, e.nullable)() } val groupByAttrs = groupByMap.map(_._2) @@ -135,9 +137,7 @@ object RewriteDistinctAggregates extends Rule[LogicalPlan] { def patchAggregateFunctionChildren( af: AggregateFunction)( attrs: Expression => Expression): AggregateFunction = { - af.withNewChildren(af.children.map { - case afc => attrs(afc) - }).asInstanceOf[AggregateFunction] + af.withNewChildren(af.children.map(attrs)).asInstanceOf[AggregateFunction] } // Setup unique distinct aggregate children. @@ -265,5 +265,5 @@ object RewriteDistinctAggregates extends Rule[LogicalPlan] { // NamedExpression. This is done to prevent collisions between distinct and regular aggregate // children, in this case attribute reuse causes the input of the regular aggregate to bound to // the (nulled out) input of the distinct aggregate. - e -> new AttributeReference(e.sql, e.dataType, true)() + e -> AttributeReference(e.sql, e.dataType, nullable = true)() } http://git-wip-us.apache.org/repos/asf/spark/blob/0a593db3/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/optimizer/RewriteDistinctAggregatesSuite.scala ---------------------------------------------------------------------- diff --git a/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/optimizer/RewriteDistinctAggregatesSuite.scala b/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/optimizer/RewriteDistinctAggregatesSuite.scala new file mode 100644 index 0000000..0b973c3 --- /dev/null +++ b/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/optimizer/RewriteDistinctAggregatesSuite.scala @@ -0,0 +1,94 @@ +/* + * 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.catalyst.optimizer + +import org.apache.spark.sql.catalyst.SimpleCatalystConf +import org.apache.spark.sql.catalyst.analysis.{Analyzer, EmptyFunctionRegistry} +import org.apache.spark.sql.catalyst.catalog.{InMemoryCatalog, SessionCatalog} +import org.apache.spark.sql.catalyst.dsl.expressions._ +import org.apache.spark.sql.catalyst.dsl.plans._ +import org.apache.spark.sql.catalyst.expressions.{If, Literal} +import org.apache.spark.sql.catalyst.expressions.aggregate.{CollectSet, Count} +import org.apache.spark.sql.catalyst.plans.PlanTest +import org.apache.spark.sql.catalyst.plans.logical.{Aggregate, Expand, LocalRelation, LogicalPlan} +import org.apache.spark.sql.types.{IntegerType, StringType} + +class RewriteDistinctAggregatesSuite extends PlanTest { + val conf = SimpleCatalystConf(caseSensitiveAnalysis = false, groupByOrdinal = false) + val catalog = new SessionCatalog(new InMemoryCatalog, EmptyFunctionRegistry, conf) + val analyzer = new Analyzer(catalog, conf) + + val nullInt = Literal(null, IntegerType) + val nullString = Literal(null, StringType) + val testRelation = LocalRelation('a.string, 'b.string, 'c.string, 'd.string, 'e.int) + + private def checkRewrite(rewrite: LogicalPlan): Unit = rewrite match { + case Aggregate(_, _, Aggregate(_, _, _: Expand)) => + case _ => fail(s"Plan is not rewritten:\n$rewrite") + } + + test("single distinct group") { + val input = testRelation + .groupBy('a)(countDistinct('e)) + .analyze + val rewrite = RewriteDistinctAggregates(input) + comparePlans(input, rewrite) + } + + test("single distinct group with partial aggregates") { + val input = testRelation + .groupBy('a, 'd)( + countDistinct('e, 'c).as('agg1), + max('b).as('agg2)) + .analyze + val rewrite = RewriteDistinctAggregates(input) + comparePlans(input, rewrite) + } + + test("single distinct group with non-partial aggregates") { + val input = testRelation + .groupBy('a, 'd)( + countDistinct('e, 'c).as('agg1), + CollectSet('b).toAggregateExpression().as('agg2)) + .analyze + checkRewrite(RewriteDistinctAggregates(input)) + } + + test("multiple distinct groups") { + val input = testRelation + .groupBy('a)(countDistinct('b, 'c), countDistinct('d)) + .analyze + checkRewrite(RewriteDistinctAggregates(input)) + } + + test("multiple distinct groups with partial aggregates") { + val input = testRelation + .groupBy('a)(countDistinct('b, 'c), countDistinct('d), sum('e)) + .analyze + checkRewrite(RewriteDistinctAggregates(input)) + } + + test("multiple distinct groups with non-partial aggregates") { + val input = testRelation + .groupBy('a)( + countDistinct('b, 'c), + countDistinct('d), + CollectSet('b).toAggregateExpression()) + .analyze + checkRewrite(RewriteDistinctAggregates(input)) + } +} http://git-wip-us.apache.org/repos/asf/spark/blob/0a593db3/sql/core/src/test/scala/org/apache/spark/sql/DataFrameAggregateSuite.scala ---------------------------------------------------------------------- diff --git a/sql/core/src/test/scala/org/apache/spark/sql/DataFrameAggregateSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/DataFrameAggregateSuite.scala index 69a3b5f..cb505ac 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/DataFrameAggregateSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/DataFrameAggregateSuite.scala @@ -485,4 +485,12 @@ class DataFrameAggregateSuite extends QueryTest with SharedSQLContext { spark.sql("select avg(a) over () from values 1.0, 2.0, 3.0 T(a)"), Row(2.0) :: Row(2.0) :: Row(2.0) :: Nil) } + + test("SPARK-17616: distinct aggregate combined with a non-partial aggregate") { + val df = Seq((1, 3, "a"), (1, 2, "b"), (3, 4, "c"), (3, 4, "c"), (3, 5, "d")) + .toDF("x", "y", "z") + checkAnswer( + df.groupBy($"x").agg(countDistinct($"y"), sort_array(collect_list($"z"))), + Seq(Row(1, 2, Seq("a", "b")), Row(3, 2, Seq("c", "c", "d")))) + } } --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@spark.apache.org For additional commands, e-mail: commits-h...@spark.apache.org