Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[BUG] When+sequence could trigger "Illegal sequence boundaries" error #5023

Closed
viadea opened this issue Mar 23, 2022 · 2 comments · Fixed by #5042
Closed

[BUG] When+sequence could trigger "Illegal sequence boundaries" error #5023

viadea opened this issue Mar 23, 2022 · 2 comments · Fixed by #5042
Assignees
@andygrove
Labels
bug Something isn't working P0 Must have for release
Milestone
Mar 21 - Apr 1

Comments

@viadea
Copy link
Collaborator

viadea commented Mar 23, 2022 •
edited
Loading

Below is a minimum repro in pyspark:

Env:
22.02GA jars
or
22.04Snapshot jars(20220322)

Reproduced on Databricks 9.1ML GPU and also local workstation.

Repro

data = [
 ("James,,Smith",["Java","Scala","C++"],["Spark","Java"],"OH","CA"),
 ("Michael,Rose,",["Spark"],["Spark","Java"],"NY","NJ"),
 ("Robert,,Williams",["CSharp","VB"],[],"UT","NV")
]

from pyspark.sql.types import StringType, ArrayType,StructType,StructField
schema = StructType([ 
    StructField("name",StringType(),True), 
    StructField("languagesAtSchool",ArrayType(StringType()),True), 
    StructField("languagesAtWork",ArrayType(StringType()),True), 
    StructField("currentState", StringType(), True), 
    StructField("previousState", StringType(), True)
  ])

df = spark.createDataFrame(data=data,schema=schema)
df.write.format("parquet").mode("overwrite").save("/tmp/testparquet")
df2 = spark.read.parquet("/tmp/testparquet")
df2.show()

spark.conf.set("spark.rapids.sql.enabled",True)
df2.withColumn("new_column", when (  (size(col("languagesAtWork")) > 0 ) , sequence(lit(1), size(col("languagesAtWork")),lit(1))) ).show()

Error:

java.lang.IllegalArgumentException: requirement failed: Illegal sequence boundaries: step > 0 but start > stop

Full stacktrace:

Caused by: java.lang.IllegalArgumentException: requirement failed: Illegal sequence boundaries: step > 0 but start > stop
	at scala.Predef$.require(Predef.scala:281)
	at org.apache.spark.sql.rapids.GpuSequenceUtil$.$anonfun$checkSequenceInputs$5(collectionOperations.scala:558)
	at org.apache.spark.sql.rapids.GpuSequenceUtil$.$anonfun$checkSequenceInputs$5$adapted(collectionOperations.scala:557)
	at com.nvidia.spark.rapids.Arm.withResource(Arm.scala:28)
	at com.nvidia.spark.rapids.Arm.withResource$(Arm.scala:26)
	at org.apache.spark.sql.rapids.GpuSequenceUtil$.withResource(collectionOperations.scala:533)
	at org.apache.spark.sql.rapids.GpuSequenceUtil$.$anonfun$checkSequenceInputs$2(collectionOperations.scala:557)
	at org.apache.spark.sql.rapids.GpuSequenceUtil$.$anonfun$checkSequenceInputs$2$adapted(collectionOperations.scala:549)
	at com.nvidia.spark.rapids.Arm.withResource(Arm.scala:28)
	at com.nvidia.spark.rapids.Arm.withResource$(Arm.scala:26)
	at org.apache.spark.sql.rapids.GpuSequenceUtil$.withResource(collectionOperations.scala:533)
	at org.apache.spark.sql.rapids.GpuSequenceUtil$.$anonfun$checkSequenceInputs$1(collectionOperations.scala:549)
	at org.apache.spark.sql.rapids.GpuSequenceUtil$.$anonfun$checkSequenceInputs$1$adapted(collectionOperations.scala:541)
	at com.nvidia.spark.rapids.Arm.withResource(Arm.scala:28)
	at com.nvidia.spark.rapids.Arm.withResource$(Arm.scala:26)
	at org.apache.spark.sql.rapids.GpuSequenceUtil$.withResource(collectionOperations.scala:533)
	at org.apache.spark.sql.rapids.GpuSequenceUtil$.checkSequenceInputs(collectionOperations.scala:541)
	at org.apache.spark.sql.rapids.GpuSequenceUtil$.computeSequenceSizes(collectionOperations.scala:599)
	at org.apache.spark.sql.rapids.GpuSequence.$anonfun$columnarEval$7(collectionOperations.scala:681)
	at com.nvidia.spark.rapids.Arm.closeOnExcept(Arm.scala:87)
	at com.nvidia.spark.rapids.Arm.closeOnExcept$(Arm.scala:85)
	at org.apache.spark.sql.rapids.GpuSequence.closeOnExcept(collectionOperations.scala:653)
	at org.apache.spark.sql.rapids.GpuSequence.$anonfun$columnarEval$4(collectionOperations.scala:680)
	at com.nvidia.spark.rapids.Arm.withResource(Arm.scala:28)
	at com.nvidia.spark.rapids.Arm.withResource$(Arm.scala:26)
	at org.apache.spark.sql.rapids.GpuSequence.withResource(collectionOperations.scala:653)
	at org.apache.spark.sql.rapids.GpuSequence.$anonfun$columnarEval$3(collectionOperations.scala:676)
	at com.nvidia.spark.rapids.Arm.withResource(Arm.scala:37)
	at com.nvidia.spark.rapids.Arm.withResource$(Arm.scala:35)
	at org.apache.spark.sql.rapids.GpuSequence.withResource(collectionOperations.scala:653)
	at org.apache.spark.sql.rapids.GpuSequence.$anonfun$columnarEval$1(collectionOperations.scala:672)
	at com.nvidia.spark.rapids.Arm.withResource(Arm.scala:28)
	at com.nvidia.spark.rapids.Arm.withResource$(Arm.scala:26)
	at org.apache.spark.sql.rapids.GpuSequence.withResource(collectionOperations.scala:653)
	at org.apache.spark.sql.rapids.GpuSequence.columnarEval(collectionOperations.scala:671)
	at com.nvidia.spark.rapids.RapidsPluginImplicits$ReallyAGpuExpression.columnarEval(implicits.scala:34)
	at com.nvidia.spark.rapids.GpuConditionalExpression.$anonfun$computeIfElse$2(conditionalExpressions.scala:159)
	at com.nvidia.spark.rapids.Arm.withResource(Arm.scala:28)
	at com.nvidia.spark.rapids.Arm.withResource$(Arm.scala:26)
	at com.nvidia.spark.rapids.GpuCaseWhen.withResource(conditionalExpressions.scala:309)
	at com.nvidia.spark.rapids.GpuConditionalExpression.$anonfun$computeIfElse$1(conditionalExpressions.scala:158)
	at com.nvidia.spark.rapids.Arm.withResourceIfAllowed(Arm.scala:73)
	at com.nvidia.spark.rapids.Arm.withResourceIfAllowed$(Arm.scala:71)
	at com.nvidia.spark.rapids.GpuCaseWhen.withResourceIfAllowed(conditionalExpressions.scala:309)
	at com.nvidia.spark.rapids.GpuConditionalExpression.computeIfElse(conditionalExpressions.scala:157)
	at com.nvidia.spark.rapids.GpuConditionalExpression.computeIfElse$(conditionalExpressions.scala:152)
	at com.nvidia.spark.rapids.GpuCaseWhen.computeIfElse(conditionalExpressions.scala:309)
	at com.nvidia.spark.rapids.GpuCaseWhen.$anonfun$columnarEval$6(conditionalExpressions.scala:362)
	at scala.collection.IndexedSeqOptimized.foldRight(IndexedSeqOptimized.scala:65)
	at scala.collection.IndexedSeqOptimized.foldRight$(IndexedSeqOptimized.scala:72)
	at scala.collection.mutable.ArrayBuffer.foldRight(ArrayBuffer.scala:49)
	at com.nvidia.spark.rapids.GpuCaseWhen.columnarEval(conditionalExpressions.scala:360)
	at com.nvidia.spark.rapids.RapidsPluginImplicits$ReallyAGpuExpression.columnarEval(implicits.scala:34)
	at com.nvidia.spark.rapids.GpuExpressionsUtils$.columnarEvalToColumn(GpuExpressions.scala:93)
	at com.nvidia.spark.rapids.GpuUnaryExpression.columnarEval(GpuExpressions.scala:202)
	at com.nvidia.spark.rapids.RapidsPluginImplicits$ReallyAGpuExpression.columnarEval(implicits.scala:34)
	at com.nvidia.spark.rapids.GpuAlias.columnarEval(namedExpressions.scala:109)
	at com.nvidia.spark.rapids.RapidsPluginImplicits$ReallyAGpuExpression.columnarEval(implicits.scala:34)
	at com.nvidia.spark.rapids.GpuExpressionsUtils$.columnarEvalToColumn(GpuExpressions.scala:93)
	at com.nvidia.spark.rapids.GpuProjectExec$.projectSingle(basicPhysicalOperators.scala:102)
	at com.nvidia.spark.rapids.GpuProjectExec$.$anonfun$project$1(basicPhysicalOperators.scala:109)
	at com.nvidia.spark.rapids.RapidsPluginImplicits$MapsSafely.$anonfun$safeMap$1(implicits.scala:162)
	at com.nvidia.spark.rapids.RapidsPluginImplicits$MapsSafely.$anonfun$safeMap$1$adapted(implicits.scala:159)
	at scala.collection.immutable.List.foreach(List.scala:392)
	at com.nvidia.spark.rapids.RapidsPluginImplicits$MapsSafely.safeMap(implicits.scala:159)
	at com.nvidia.spark.rapids.RapidsPluginImplicits$AutoCloseableProducingSeq.safeMap(implicits.scala:194)
	at com.nvidia.spark.rapids.GpuProjectExec$.project(basicPhysicalOperators.scala:109)
	at com.nvidia.spark.rapids.GpuProjectExec$.projectAndClose(basicPhysicalOperators.scala:73)
	at com.nvidia.spark.rapids.GpuProjectExec.$anonfun$doExecuteColumnar$1(basicPhysicalOperators.scala:149)
	at scala.collection.Iterator$$anon$10.next(Iterator.scala:459)
	at com.nvidia.spark.rapids.GpuBaseLimitExec$$anon$1.next(limit.scala:71)
	at com.nvidia.spark.rapids.GpuBaseLimitExec$$anon$1.next(limit.scala:65)
	at org.apache.spark.sql.rapids.execution.GpuShuffleExchangeExecBase$$anon$1.partNextBatch(GpuShuffleExchangeExecBase.scala:288)
	at org.apache.spark.sql.rapids.execution.GpuShuffleExchangeExecBase$$anon$1.hasNext(GpuShuffleExchangeExecBase.scala:304)
	at org.apache.spark.shuffle.sort.BypassMergeSortShuffleWriter.write(BypassMergeSortShuffleWriter.java:132)
	at org.apache.spark.shuffle.ShuffleWriteProcessor.write(ShuffleWriteProcessor.scala:59)
	at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:99)
	at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:52)
	at org.apache.spark.scheduler.Task.run(Task.scala:131)
	at org.apache.spark.executor.Executor$TaskRunner.$anonfun$run$3(Executor.scala:497)
	at org.apache.spark.util.Utils$.tryWithSafeFinally(Utils.scala:1439)
	at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:500)
	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
	... 1 more

CPU Spark works fine.

Basically the column "languagesAtWork" has some empty array value, seems even if "when clause" is FALSE, we are evaluating the expression(which contains sequence).
@viadea viadea added bug Something isn't working ? - Needs Triage Need team to review and classify labels Mar 23, 2022
@viadea
Copy link
Collaborator Author

viadea commented Mar 23, 2022

The input dataframe is :

+----------------+------------------+---------------+------------+-------------+
|            name| languagesAtSchool|languagesAtWork|currentState|previousState|
+----------------+------------------+---------------+------------+-------------+
|    James,,Smith|[Java, Scala, C++]|  [Spark, Java]|          OH|           CA|
|   Michael,Rose,|           [Spark]|  [Spark, Java]|          NY|           NJ|
|Robert,,Williams|      [CSharp, VB]|             []|          UT|           NV|
+----------------+------------------+---------------+------------+-------------+

And the expected output is:

+----------------+------------------+---------------+------------+-------------+----------+
|            name| languagesAtSchool|languagesAtWork|currentState|previousState|new_column|
+----------------+------------------+---------------+------------+-------------+----------+
|    James,,Smith|[Java, Scala, C++]|  [Spark, Java]|          OH|           CA|    [1, 2]|
|   Michael,Rose,|           [Spark]|  [Spark, Java]|          NY|           NJ|    [1, 2]|
|Robert,,Williams|      [CSharp, VB]|             []|          UT|           NV|      null|
+----------------+------------------+---------------+------------+-------------+----------+

@viadea
Copy link
Collaborator Author

viadea commented Mar 23, 2022

Is it related to #3849 regarding case/when evaluation?

@sameerz sameerz added P0 Must have for release and removed ? - Needs Triage Need team to review and classify labels Mar 23, 2022
@sameerz sameerz added this to the Mar 21 - Apr 1 milestone Mar 23, 2022
This was referenced Mar 24, 2022
Closed
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@andygrove andygrove

Labels
bug Something isn't working P0 Must have for release
Projects
None yet
Milestone
Mar 21 - Apr 1
Development

Successfully merging a pull request may close this issue.

Implement hasSideEffects for GpuSequence andygrove/spark-rapids
3 participants
@andygrove @sameerz @viadea