[
https://issues.apache.org/jira/browse/SPARK-42346?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17684992#comment-17684992
]
Ritika Maheshwari commented on SPARK-42346:
-------------------------------------------
I have Spark 3.3.0 and I do not have 39887 fix . I am not able to reproduce
this issue. Am I missing something?
scala> val df = Seq(("a","b")).toDF("surname","first_name")
*df*: *org.apache.spark.sql.DataFrame* = [surname: string, first_name: string]
scala> df.createOrReplaceTempView("input_table")
scala> spark.sql("select(Select Count(Distinct first_name) from input_table) As
distinct_value_count from input_table Union all select (select count(Distinct
surname) from input_table) as distinct_value_count from input_table").show()
+--------------------+
|distinct_value_count|
+--------------------+
| 1|
| 1|
+--------------------+
= Physical Plan ==
AdaptiveSparkPlan isFinalPlan=false
+- Union
:- Project [cast(Subquery subquery#46, [id=#114] as string) AS
distinct_value_count#62]
: : +- Subquery subquery#46, [id=#114]
: : +- AdaptiveSparkPlan isFinalPlan=false
: : +- HashAggregate(keys=[], functions=[count(first_name#12)],
output=[count(DISTINCT first_name)#53L])
: : +- Exchange SinglePartition, ENSURE_REQUIREMENTS, [id=#112]
: : +- HashAggregate(keys=[],
functions=[partial_count(first_name#12)], output=[count#67L])
: : +- LocalTableScan [first_name#12]
: +- LocalTableScan [_1#6, _2#7]
+- Project [cast(Subquery subquery#48, [id=#125] as string) AS
distinct_value_count#64]
: +- Subquery subquery#48, [id=#125]
: +- AdaptiveSparkPlan isFinalPlan=false
: +- HashAggregate(keys=[], functions=[count(surname#11)],
output=[count(DISTINCT surname)#55L])
: +- Exchange SinglePartition, ENSURE_REQUIREMENTS, [id=#123]
: +- HashAggregate(keys=[],
functions=[partial_count(surname#11)], output=[count#68L])
: +- LocalTableScan [surname#11]
+- LocalTableScan [_1#50, _2#51|#50, _2#51]
This is what I have in my SparkOptimizer.scala
override def defaultBatches: Seq[Batch] = (preOptimizationBatches ++
super.defaultBatches :+
Batch("Optimize Metadata Only Query", Once, OptimizeMetadataOnlyQuery(catalog))
:+
Batch("PartitionPruning", Once,
PartitionPruning) :+
Batch("InjectRuntimeFilter", FixedPoint(1),
InjectRuntimeFilter,
RewritePredicateSubquery) :+
Batch("MergeScalarSubqueries", Once,
MergeScalarSubqueries) :+
Batch("Pushdown Filters from PartitionPruning", fixedPoint,
PushDownPredicates) :+
Batch
> distinct(count colname) with UNION ALL causes query analyzer bug
> ----------------------------------------------------------------
>
> Key: SPARK-42346
> URL: https://issues.apache.org/jira/browse/SPARK-42346
> Project: Spark
> Issue Type: Bug
> Components: SQL
> Affects Versions: 3.3.0, 3.4.0, 3.5.0
> Reporter: Robin
> Assignee: Peter Toth
> Priority: Major
> Fix For: 3.3.2, 3.4.0, 3.5.0
>
>
> If you combine a UNION ALL with a count(distinct colname) you get a query
> analyzer bug.
>
> This behaviour is introduced in 3.3.0. The bug was not present in 3.2.1.
>
> Here is a reprex in PySpark:
> {{df_pd = pd.DataFrame([}}
> {{ \{'surname': 'a', 'first_name': 'b'}}}
> {{])}}
> {{df_spark = spark.createDataFrame(df_pd)}}
> {{df_spark.createOrReplaceTempView("input_table")}}
> {{sql = """}}
> {{SELECT }}
> {{ (SELECT Count(DISTINCT first_name) FROM input_table) }}
> {{ AS distinct_value_count}}
> {{FROM input_table}}
> {{UNION ALL}}
> {{SELECT }}
> {{ (SELECT Count(DISTINCT surname) FROM input_table) }}
> {{ AS distinct_value_count}}
> {{FROM input_table """}}
> {{spark.sql(sql).toPandas()}}
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
