Hyukjin Kwon created SPARK-61085:
------------------------------------
Summary: Collect the results of the UDF profilers, and show them
with spark.profile, in the Python client as PySpark does
Key: SPARK-61085
URL: https://issues.apache.org/jira/browse/SPARK-61085
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: Hyukjin Kwon
In PySpark Connect, {{spark.profile}} is a {{pyspark.sql.profiler.Profile}} of
a {{ConnectProfilerCollector}}, which collects the results of the UDF profilers
({{spark.sql.pyspark.udf.profiler}} of "perf" or "memory"): the server sends
them in the {{__python_accumulator__}} observed metrics of an execution, and
{{spark.profile.show()}}, {{dump}}, {{render}} and {{clear}} show, write and
clear them.
The Rust-backed Python client returns a {{ProfilerCollector}} of its own from
{{spark.profile}}, which collects nothing, and whose {{show}} returns "No
profile data collected" and {{dump(path)}} raises a {{TypeError}}. Collect and
show the profiles as PySpark does.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
