Albertus Kelvin created SPARK-28562:
---------------------------------------
Summary: PySpark profiling is not understandable
Key: SPARK-28562
URL: https://issues.apache.org/jira/browse/SPARK-28562
Project: Spark
Issue Type: Question
Components: Optimizer
Affects Versions: 2.4.0
Reporter: Albertus Kelvin
I was profiling code in PySpark. What I did was set the "spark.python.profile"
in the config to "true". I also made a simple method consisting of several
dataframe operations, such as "withColumn" and "join". Here's the code sample:
{code:python}
def join_df(df, df1):
df = df.withColumn('rowa', F.lit(100))
df = df.withColumn('rowb', df['rowa'] * F.lit(100))
joined_df = df.join(df1,'rowid',how='left')
return joined_df
{code}
However, after the driver exits, the output of the profiler was not
understandable because there were no my filename and the corresponding methods.
All exists was Spark's built-in files and methods, such as "rdd.py",
"worker.py", and "serializers.py".
The question is, how to show all of my methods that become the bottlenecks? For
example, using the above code sample, I'd like to know the time needed for
"withColumn" and "join" operation.
Thanks.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
