Hyukjin Kwon created SPARK-61084:
------------------------------------
Summary: Register progress handlers, and call them with the
progress of each execution, in the Python client as PySpark does
Key: SPARK-61084
URL: https://issues.apache.org/jira/browse/SPARK-61084
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: Hyukjin Kwon
In PySpark Connect, {{spark.registerProgressHandler(handler)}} registers a
callable once (it returns None), {{removeProgressHandler(handler)}} removes
that callable (a {{ValueError}} if it is not registered), and the client calls
each handler with the keyword arguments of {{ProgressHandler}}: {{stages}} (the
{{StageInfo}} of each stage), {{inflight_tasks}}, {{operation_id}} and
{{done}}, for each progress of an execution and once more with {{done=True}}
when it ends. {{pyspark.sql.connect.shell.progress}} has {{ProgressHandler}},
{{StageInfo}} and the {{Progress}} bar of
{{SPARK_CONNECT_PROGRESS_BAR_ENABLED}}.
The Rust-backed Python client returns an id of {{registerProgressHandler}},
removes a handler by that id, and calls a handler with one dict of the
progress, so a handler of PySpark fails. It lacks
{{pyspark.sql.connect.shell.progress}}. Register and call the handlers 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]
