You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by GitBox <gi...@apache.org> on 2021/02/08 03:29:44 UTC
[GitHub] [spark] wangyum commented on pull request #31485: [SPARK-34137][SQL] Update suquery's stats when build LogicalPlan's stats
wangyum commented on pull request #31485:
URL: https://github.com/apache/spark/pull/31485#issuecomment-774841202
It may just not be explained correctly, for example:
```sql
create table t1 using parquet as select id as a, id as b, id as c from range(1000000);
create table t2 using parquet as select id as a, id as b, id as c from range(1000000);
create table t3 using parquet as select id as a, id as b, id as c from range(1000000);
analyze table t1 compute statistics for all columns;
analyze table t2 compute statistics for all columns;
analyze table t3 compute statistics for all columns;
set spark.sql.cbo.enabled=true;
explain cost
select * from t3 where c > (select max(t1.c) as tc from t1 join t2 on t1.a = t2.a and t2.b < 10);
== Optimized Logical Plan ==
Filter (isnotnull(c#35782L) AND (c#35782L > scalar-subquery#35774 [])), Statistics(sizeInBytes=30.5 MiB, rowCount=1.00E+6)
: +- Aggregate [max(c#35785L) AS tc#35773L]
: +- Project [c#35785L]
: +- Join Inner, (a#35783L = a#35786L)
: :- Project [a#35783L, c#35785L]
: : +- Filter isnotnull(a#35783L)
: : +- Relation[a#35783L,b#35784L,c#35785L] parquet
: +- Project [a#35786L]
: +- Filter ((isnotnull(b#35787L) AND (b#35787L < 10)) AND isnotnull(a#35786L))
: +- Relation[a#35786L,b#35787L,c#35788L] parquet
+- Relation[a#35780L,b#35781L,c#35782L] parquet, Statistics(sizeInBytes=30.5 MiB, rowCount=1.00E+6)
== Physical Plan ==
AdaptiveSparkPlan isFinalPlan=false
+- Filter (isnotnull(c#35782L) AND (c#35782L > Subquery subquery#35774, [id=#201]))
: +- Subquery subquery#35774, [id=#201]
: +- AdaptiveSparkPlan isFinalPlan=false
: +- HashAggregate(keys=[], functions=[max(c#35785L)], output=[tc#35773L])
: +- Exchange SinglePartition, ENSURE_REQUIREMENTS, [id=#199]
: +- HashAggregate(keys=[], functions=[partial_max(c#35785L)], output=[max#35791L])
: +- Project [c#35785L]
: +- BroadcastHashJoin [a#35783L], [a#35786L], Inner, BuildRight, false
: :- Filter isnotnull(a#35783L)
: : +- FileScan parquet default.t1[a#35783L,c#35785L] Batched: true, DataFilters: [isnotnull(a#35783L)], Format: Parquet, Location: InMemoryFileIndex(1 paths)[file:/root/opensource/apache-spark/spark-warehouse/t1], PartitionFilters: [], PushedFilters: [IsNotNull(a)], ReadSchema: struct<a:bigint,c:bigint>
: +- BroadcastExchange HashedRelationBroadcastMode(List(input[0, bigint, true]),false), [id=#194]
: +- Project [a#35786L]
: +- Filter ((isnotnull(b#35787L) AND (b#35787L < 10)) AND isnotnull(a#35786L))
: +- FileScan parquet default.t2[a#35786L,b#35787L] Batched: true, DataFilters: [isnotnull(b#35787L), (b#35787L < 10), isnotnull(a#35786L)], Format: Parquet, Location: InMemoryFileIndex(1 paths)[file:/root/opensource/apache-spark/spark-warehouse/t2], PartitionFilters: [], PushedFilters: [IsNotNull(b), LessThan(b,10), IsNotNull(a)], ReadSchema: struct<a:bigint,b:bigint>
+- FileScan parquet default.t3[a#35780L,b#35781L,c#35782L] Batched: true, DataFilters: [isnotnull(c#35782L)], Format: Parquet, Location: InMemoryFileIndex(1 paths)[file:/root/opensource/apache-spark/spark-warehouse/t3], PartitionFilters: [], PushedFilters: [IsNotNull(c)], ReadSchema: struct<a:bigint,b:bigint,c:bigint>
```
It should be SMJ if stats incorrect.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org