Github user wzhfy commented on a diff in the pull request:
https://github.com/apache/spark/pull/16431#discussion_r94953739
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/estimation/AggregateEstimation.scala
---
@@ -0,0 +1,59 @@
+/*
+ * 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.plans.logical.estimation
+
+import org.apache.spark.sql.catalyst.expressions.Attribute
+import org.apache.spark.sql.catalyst.plans.logical.{Aggregate, Statistics}
+
+
+object AggregateEstimation {
+ import EstimationUtils._
+
+ def estimate(agg: Aggregate): Option[Statistics] = {
+ val childStats = agg.child.statistics
+ // Check if we have column stats for all group-by columns.
+ val colStatsExist = agg.groupingExpressions.forall { e =>
+ e.isInstanceOf[Attribute] &&
childStats.attributeStats.contains(e.asInstanceOf[Attribute])
+ }
+ if (rowCountsExist(agg.child) && colStatsExist) {
+ // Initial value for agg without group expressions
+ var outputRows: BigInt = 1
+ agg.groupingExpressions.map(_.asInstanceOf[Attribute]).foreach {
attr =>
+ val colStat = childStats.attributeStats(attr)
+ // Multiply distinct counts of group by columns. This is an upper
bound, which assumes
+ // the data contains all combinations of distinct values of group
by columns.
+ outputRows *= colStat.distinctCount
+ }
+
+ // The number of output rows must not be larger than child's number
of rows.
+ // Note that this also covers the case of uniqueness of column. If
one of the group-by columns
--- End diff --
If the aggregate has three group-by columns, e.g. group by a, b, c, the
number of output rows is estimated by `ndv(a) * ndv(b) * ndv(c)`. It's an upper
bound by assuming the data has every combination of values of a, b and c. But
this product can become very large. So previously, I had two methods to set
tighter bounds.
1. #row of the aggregate must be <= #row of child.
2. if one of the group-by columns is a primary key (e.g. column a), each
distinct value of a can appear only once in records, then the number of
possible combinations of a, b, c is equal to ndv(a), thus #row of the
aggregate with group by a, b, c is equal to ndv(a).
But later, I noticed that since a is a primary key, ndv(a) is actually
equal to #row of child. So case 2 is covered by case 1.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
