dtenedor commented on code in PR #36527:
URL: https://github.com/apache/spark/pull/36527#discussion_r871913208
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala:
##########
@@ -196,7 +196,8 @@ abstract class Optimizer(catalogManager: CatalogManager)
ReplaceDeduplicateWithAggregate) ::
Batch("Aggregate", fixedPoint,
RemoveLiteralFromGroupExpressions,
- RemoveRepetitionFromGroupExpressions) :: Nil ++
+ RemoveRepetitionFromGroupExpressions,
+ RewriteNonAggregateFirst) :: Nil ++
Review Comment:
Just curious, is there a reason we elect to put the new rule here vs. some
other location in the optimizer rule batches?
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/RewriteNonAggregateFirst.scala:
##########
@@ -0,0 +1,72 @@
+/*
+ * 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.expressions.{Alias, Literal}
+import
org.apache.spark.sql.catalyst.expressions.aggregate.{AggregateExpression, First}
+import org.apache.spark.sql.catalyst.plans.logical._
+import org.apache.spark.sql.catalyst.rules.Rule
+import org.apache.spark.sql.catalyst.trees.TreePattern.FIRST
+import org.apache.spark.sql.types.IntegerType
+
+/**
+ * Rewrite aggregate plan with a single [[First]] function when grouping is
absent. In such a case
+ * the query is equivalent to simple projection with limit 1.
+ *
+ * Input Pseudo-Query:
+ * {{{
+ * SELECT FIRST(col) FROM table
+ * }}}
+ *
+ * Rewritten Query:
+ * {{{
+ * SELECT col FROM table LIMIT 1
+ * }}}
+ *
+ * Note that using IGNORE NULLS with [[First]] blocks rewrite logic since
projection with NOT NULL
+ * filter might return different result than [[First]] if all values are NULL.
+ */
+object RewriteNonAggregateFirst extends Rule[LogicalPlan] {
+ override def apply(plan: LogicalPlan): LogicalPlan = {
+ plan.transformWithPruning(_.containsPattern(FIRST), ruleId) {
+ case agg: Aggregate if isNonAggregateFirst(agg) =>
+ rewriteNonAggregateFirstToProject(agg)
+ }
+ }
+
+ private def isNonAggregateFirst(agg: Aggregate): Boolean = {
Review Comment:
We only call this method once and the implementation is relatively
simple...maybe we just inline it at the call site instead?
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/RewriteNonAggregateFirst.scala:
##########
@@ -0,0 +1,72 @@
+/*
+ * 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.expressions.{Alias, Literal}
+import
org.apache.spark.sql.catalyst.expressions.aggregate.{AggregateExpression, First}
+import org.apache.spark.sql.catalyst.plans.logical._
+import org.apache.spark.sql.catalyst.rules.Rule
+import org.apache.spark.sql.catalyst.trees.TreePattern.FIRST
+import org.apache.spark.sql.types.IntegerType
+
+/**
+ * Rewrite aggregate plan with a single [[First]] function when grouping is
absent. In such a case
+ * the query is equivalent to simple projection with limit 1.
+ *
+ * Input Pseudo-Query:
+ * {{{
+ * SELECT FIRST(col) FROM table
+ * }}}
+ *
+ * Rewritten Query:
+ * {{{
+ * SELECT col FROM table LIMIT 1
+ * }}}
+ *
+ * Note that using IGNORE NULLS with [[First]] blocks rewrite logic since
projection with NOT NULL
+ * filter might return different result than [[First]] if all values are NULL.
+ */
+object RewriteNonAggregateFirst extends Rule[LogicalPlan] {
+ override def apply(plan: LogicalPlan): LogicalPlan = {
+ plan.transformWithPruning(_.containsPattern(FIRST), ruleId) {
+ case agg: Aggregate if isNonAggregateFirst(agg) =>
+ rewriteNonAggregateFirstToProject(agg)
+ }
+ }
+
+ private def isNonAggregateFirst(agg: Aggregate): Boolean = {
+ agg.aggregateExpressions.length == 1 && agg.groupingExpressions.isEmpty
+ }
+
+ private def rewriteNonAggregateFirstToProject(agg: Aggregate): LogicalPlan =
{
+ var skipRewrite = false
+ val projectList = agg.aggregateExpressions.map {
+ case alias: Alias =>
Review Comment:
I think we could simplify this by just matching against the Alias <- First
sequence in one go, e.g. `case a@Alias(f: First, _) if !f.ignoreNulls =>
Some(a.copy(child = f.child))` and then catch all other cases with `case _ =>
None`. Then at the end you can check `if (projectList.forall(_.isDefined) { ...
projectList.map(_.get) } else { agg }`.
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/RewriteNonAggregateFirst.scala:
##########
@@ -0,0 +1,72 @@
+/*
+ * 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.expressions.{Alias, Literal}
+import
org.apache.spark.sql.catalyst.expressions.aggregate.{AggregateExpression, First}
+import org.apache.spark.sql.catalyst.plans.logical._
+import org.apache.spark.sql.catalyst.rules.Rule
+import org.apache.spark.sql.catalyst.trees.TreePattern.FIRST
+import org.apache.spark.sql.types.IntegerType
+
+/**
+ * Rewrite aggregate plan with a single [[First]] function when grouping is
absent. In such a case
+ * the query is equivalent to simple projection with limit 1.
+ *
+ * Input Pseudo-Query:
+ * {{{
+ * SELECT FIRST(col) FROM table
+ * }}}
+ *
+ * Rewritten Query:
+ * {{{
+ * SELECT col FROM table LIMIT 1
+ * }}}
+ *
+ * Note that using IGNORE NULLS with [[First]] blocks rewrite logic since
projection with NOT NULL
+ * filter might return different result than [[First]] if all values are NULL.
+ */
+object RewriteNonAggregateFirst extends Rule[LogicalPlan] {
+ override def apply(plan: LogicalPlan): LogicalPlan = {
+ plan.transformWithPruning(_.containsPattern(FIRST), ruleId) {
+ case agg: Aggregate if isNonAggregateFirst(agg) =>
+ rewriteNonAggregateFirstToProject(agg)
+ }
+ }
+
+ private def isNonAggregateFirst(agg: Aggregate): Boolean = {
+ agg.aggregateExpressions.length == 1 && agg.groupingExpressions.isEmpty
+ }
+
+ private def rewriteNonAggregateFirstToProject(agg: Aggregate): LogicalPlan =
{
Review Comment:
same here, we can probably just put all the code in the `apply` method.
##########
sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/optimizer/RewriteNonAggregateFirstSuite.scala:
##########
@@ -0,0 +1,72 @@
+/*
+ * 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.dsl.expressions._
+import org.apache.spark.sql.catalyst.dsl.plans._
+import org.apache.spark.sql.catalyst.expressions.aggregate.First
+import org.apache.spark.sql.catalyst.plans.PlanTest
+import org.apache.spark.sql.catalyst.plans.logical._
+
+class RewriteNonAggregateFirstSuite extends PlanTest {
+ val testRelation: LocalRelation = LocalRelation('a.string, 'b.string)
+
+ private def checkRewrite(rewrite: LogicalPlan): Unit = rewrite match {
+ case GlobalLimit(_, LocalLimit(_, Project(_, _))) =>
+ case _ => fail(s"Plan is not rewritten:\n$rewrite")
+ }
+
+ test("no group by and single first aggregate") {
+ val input = testRelation.select(first('a)).analyze
+ val rewrite = RewriteNonAggregateFirst(input)
+ checkRewrite(rewrite)
+ }
+
+ test("single first aggregate with group by") {
+ val input = testRelation
+ .groupBy('a)(first('b))
+ .analyze
+ val rewrite = RewriteNonAggregateFirst(input)
+ comparePlans(input, rewrite)
+ }
+
+ test("multiple aggregates with group by") {
+ val input = testRelation
+ .groupBy('a)(
+ first('a').as('agg1),
+ max('b).as('agg2))
+ .analyze
+ val rewrite = RewriteNonAggregateFirst(input)
+ comparePlans(input, rewrite)
+ }
+
+ test("multiple aggregates without group by") {
+ val input = testRelation
+ .select(first('a).as('agg1), max('b).as('agg2))
+ .analyze
+ val rewrite = RewriteNonAggregateFirst(input)
+ comparePlans(input, rewrite)
+ }
+
+ test("no group by and single first aggregate with ignore nulls") {
+ val input = testRelation
+ .select(First('a, ignoreNulls = true).toAggregateExpression())
Review Comment:
add another test with multiple First aggregates (to show the optimization
applies), plus one with multiple First aggregates plus one non-First aggregate
(to show the optimization does not apply)?
##########
sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/optimizer/RewriteNonAggregateFirstSuite.scala:
##########
@@ -0,0 +1,72 @@
+/*
+ * 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.dsl.expressions._
+import org.apache.spark.sql.catalyst.dsl.plans._
+import org.apache.spark.sql.catalyst.expressions.aggregate.First
+import org.apache.spark.sql.catalyst.plans.PlanTest
+import org.apache.spark.sql.catalyst.plans.logical._
+
+class RewriteNonAggregateFirstSuite extends PlanTest {
+ val testRelation: LocalRelation = LocalRelation('a.string, 'b.string)
Review Comment:
Symbol literals are banned by the databricks style guide:
https://github.com/databricks/scala-style-guide#symbol-literals
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/RewriteNonAggregateFirst.scala:
##########
@@ -0,0 +1,72 @@
+/*
+ * 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.expressions.{Alias, Literal}
+import
org.apache.spark.sql.catalyst.expressions.aggregate.{AggregateExpression, First}
+import org.apache.spark.sql.catalyst.plans.logical._
+import org.apache.spark.sql.catalyst.rules.Rule
+import org.apache.spark.sql.catalyst.trees.TreePattern.FIRST
+import org.apache.spark.sql.types.IntegerType
+
+/**
+ * Rewrite aggregate plan with a single [[First]] function when grouping is
absent. In such a case
Review Comment:
We can apply this optimization if there are more than one FIRST aggregate
function, correct? We just can't have any other aggregate functions. Maybe
update the comment and example to mention this, and update the rule
implementation to allow this case as well, plus a couple unit tests with
multiple FIRST aggregate functions?
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
