Max Härtwig created SPARK-30473:
-----------------------------------
Summary: PySpark enum subclass crashes when used inside UDF
Key: SPARK-30473
URL: https://issues.apache.org/jira/browse/SPARK-30473
Project: Spark
Issue Type: Bug
Components: PySpark
Affects Versions: 2.4.4
Environment: Databricks Runtime 6.2 (includes Apache Spark 2.4.4,
Scala 2.11)
Reporter: Max Härtwig
PySpark enum subclass crashes when used inside a UDF.
Example:
{code:java}
from enum import Enum
class Direction(Enum):
NORTH = 0
SOUTH = 1
{code}
Working:
{code:java}
Direction.NORTH{code}
Crashing:
{code:java}
@udf
def fn(a):
Direction.NORTH
return ""
df.withColumn("test", fn("a")){code}
Stacktrace:
{noformat}
SparkException: Job aborted due to stage failure: Task 0 in stage 9.0 failed 4
times, most recent failure: Lost task 0.3 in stage 9.0 (TID 235, 10.139.64.21,
executor 0): org.apache.spark.api.python.PythonException: Traceback (most
recent call last): File "/databricks/spark/python/pyspark/serializers.py", line
182, in _read_with_length return self.loads(obj) File
"/databricks/spark/python/pyspark/serializers.py", line 695, in loads return
pickle.loads(obj, encoding=encoding) File
"/databricks/python/lib/python3.7/enum.py", line 152, in __new__ enum_members =
{k: classdict[k] for k in classdict._member_names} AttributeError: 'dict'
object has no attribute '_member_names'{noformat}
I suspect the problem is in `python/pyspark/cloudpickle.py`. On line 586 in the
function `_save_dynamic_enum`, the attribute `_member_names` is removed from
the enum. Yet, this attribute is required by the `Enum` class and Enum
subclasses will crash.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
