[
https://issues.apache.org/jira/browse/SPARK-24034?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Emilio Dorigatti updated SPARK-24034:
-------------------------------------
Description:
Consider the following code
{noformat}
def mapper(xx):
if xx % 2 == 0:
raise StopIteration()
else:
return xx
sc.parallelize(range(100)).map(mapper)collect()
{noformat}
The result I get is `[57, 71, 85]`
I think it happens because {{map }}is implemented in terms of
{{mapPartitionsWithIndex}} using a custom iterator, so the {{StopIteration}}
raised by the mapper is handled by that iterator. I think this should be raised
to the user instead
NB: this may be the underlying cause of
https://issues.apache.org/jira/browse/SPARK-23754
was:
Consider the following code
def mapper(xx):
if xx % 2 == 0:
raise StopIteration()
else:
return xx
sc.parallelize(range(100)).map(mapper)collect()
The result I get is `[57, 71, 85]`
I think it happens because `map` is implemented in terms of
`mapPartitionsWithIndex` using a custom iterator, so the StopIteration raised
by the mapper is handled by that iterator. I think this should be raised to the
user instead
NB: this may be the underlying cause of
https://issues.apache.org/jira/browse/SPARK-23754
> StopIteration in pyspark mapper results in partial results
> ----------------------------------------------------------
>
> Key: SPARK-24034
> URL: https://issues.apache.org/jira/browse/SPARK-24034
> Project: Spark
> Issue Type: Bug
> Components: PySpark
> Affects Versions: 2.3.0
> Reporter: Emilio Dorigatti
> Priority: Major
>
> Consider the following code
> {noformat}
> def mapper(xx):
> if xx % 2 == 0:
> raise StopIteration()
> else:
> return xx
> sc.parallelize(range(100)).map(mapper)collect()
> {noformat}
> The result I get is `[57, 71, 85]`
> I think it happens because {{map }}is implemented in terms of
> {{mapPartitionsWithIndex}} using a custom iterator, so the {{StopIteration}}
> raised by the mapper is handled by that iterator. I think this should be
> raised to the user instead
> NB: this may be the underlying cause of
> https://issues.apache.org/jira/browse/SPARK-23754
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
