[
https://issues.apache.org/jira/browse/SPARK-28067?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16870880#comment-16870880
]
Marco Gaido commented on SPARK-28067:
-------------------------------------
No, it is the same. Are you sure about your configs?
{code}
macmarco:spark mark9$ git log -5 --oneline
5ad1053f3e (HEAD, apache/master) [SPARK-28128][PYTHON][SQL] Pandas Grouped UDFs
skip empty partitions
113f8c8d13 [SPARK-28132][PYTHON] Update document type conversion for Pandas
UDFs (pyarrow 0.13.0, pandas 0.24.2, Python 3.7)
9b9d81b821 [SPARK-28131][PYTHON] Update document type conversion between Python
data and SQL types in normal UDFs (Python 3.7)
54da3bbfb2 [SPARK-28127][SQL] Micro optimization on TreeNode's mapChildren
method
47f54b1ec7 [SPARK-28118][CORE] Add `spark.eventLog.compression.codec`
configuration
macmarco:spark mark9$ ./bin/spark-shell
19/06/24 09:17:58 WARN NativeCodeLoader: Unable to load native-hadoop library
for your platform... using builtin-java classes where applicable
Using Spark's default log4j profile: org/apache/spark/log4j-defaults.properties
Setting default log level to "WARN".
To adjust logging level use sc.setLogLevel(newLevel). For SparkR, use
setLogLevel(newLevel).
Spark context Web UI available at http://.....:4040
Spark context available as 'sc' (master = local[*], app id =
local-1561360686725).
Spark session available as 'spark'.
Welcome to
____ __
/ __/__ ___ _____/ /__
_\ \/ _ \/ _ `/ __/ '_/
/___/ .__/\_,_/_/ /_/\_\ version 3.0.0-SNAPSHOT
/_/
Using Scala version 2.12.8 (Java HotSpot(TM) 64-Bit Server VM, Java 1.8.0_152)
Type in expressions to have them evaluated.
Type :help for more information.
scala> val df = Seq(
| (BigDecimal("10000000000000000000"), 1),
| (BigDecimal("10000000000000000000"), 1),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2)).toDF("decNum", "intNum")
df: org.apache.spark.sql.DataFrame = [decNum: decimal(38,18), intNum: int]
scala> val df2 = df.withColumnRenamed("decNum", "decNum2").join(df,
"intNum").agg(sum("decNum"))
df2: org.apache.spark.sql.DataFrame = [sum(decNum): decimal(38,18)]
scala> df2.explain
== Physical Plan ==
*(2) HashAggregate(keys=[], functions=[sum(decNum#14)])
+- Exchange SinglePartition
+- *(1) HashAggregate(keys=[], functions=[partial_sum(decNum#14)])
+- *(1) Project [decNum#14]
+- *(1) BroadcastHashJoin [intNum#8], [intNum#15], Inner, BuildLeft
:- BroadcastExchange HashedRelationBroadcastMode(List(cast(input[0,
int, false] as bigint)))
: +- LocalTableScan [intNum#8]
+- LocalTableScan [decNum#14, intNum#15]
scala> df2.show(40,false)
+-----------+
|sum(decNum)|
+-----------+
|null |
+-----------+
{code}
> Incorrect results in decimal aggregation with whole-stage code gen enabled
> --------------------------------------------------------------------------
>
> Key: SPARK-28067
> URL: https://issues.apache.org/jira/browse/SPARK-28067
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
> Affects Versions: 2.3.0, 2.4.0
> Environment: Ubuntu LTS 16.04
> Oracle Java 1.8.0_201
> spark-2.4.3-bin-without-hadoop
> spark-shell
> Reporter: Mark Sirek
> Priority: Minor
> Labels: correctness
>
> The following test case involving a join followed by a sum aggregation
> returns the wrong answer for the sum:
>
> {code:java}
> val df = Seq(
> (BigDecimal("10000000000000000000"), 1),
> (BigDecimal("10000000000000000000"), 1),
> (BigDecimal("10000000000000000000"), 2),
> (BigDecimal("10000000000000000000"), 2),
> (BigDecimal("10000000000000000000"), 2),
> (BigDecimal("10000000000000000000"), 2),
> (BigDecimal("10000000000000000000"), 2),
> (BigDecimal("10000000000000000000"), 2),
> (BigDecimal("10000000000000000000"), 2),
> (BigDecimal("10000000000000000000"), 2),
> (BigDecimal("10000000000000000000"), 2),
> (BigDecimal("10000000000000000000"), 2)).toDF("decNum", "intNum")
> val df2 = df.withColumnRenamed("decNum", "decNum2").join(df,
> "intNum").agg(sum("decNum"))
> scala> df2.show(40,false)
> ---------------------------------------
> sum(decNum)
> ---------------------------------------
> 40000000000000000000.000000000000000000
> ---------------------------------------
>
> {code}
>
> The result should be 1040000000000000000000.0000000000000000.
> It appears a partial sum is computed for each join key, as the result
> returned would be the answer for all rows matching intNum === 1.
> If only the rows with intNum === 2 are included, the answer given is null:
>
> {code:java}
> scala> val df3 = df.filter($"intNum" === lit(2))
> df3: org.apache.spark.sql.Dataset[org.apache.spark.sql.Row] = [decNum:
> decimal(38,18), intNum: int]
> scala> val df4 = df3.withColumnRenamed("decNum", "decNum2").join(df3,
> "intNum").agg(sum("decNum"))
> df4: org.apache.spark.sql.DataFrame = [sum(decNum): decimal(38,18)]
> scala> df4.show(40,false)
> -----------
> sum(decNum)
> -----------
> null
> -----------
>
> {code}
>
> The correct answer, 1000000000000000000000.0000000000000000, doesn't fit in
> the DataType picked for the result, decimal(38,18), so an overflow occurs,
> which Spark then converts to null.
> The first example, which doesn't filter out the intNum === 1 values should
> also return null, indicating overflow, but it doesn't. This may mislead the
> user to think a valid sum was computed.
> If whole-stage code gen is turned off:
> spark.conf.set("spark.sql.codegen.wholeStage", false)
> ... incorrect results are not returned because the overflow is caught as an
> exception:
> java.lang.IllegalArgumentException: requirement failed: Decimal precision 39
> exceeds max precision 38
>
>
>
>
>
>
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
