Hyukjin Kwon created SPARK-61151:
------------------------------------
Summary: Vendor pyspark.sql.streaming.listener, of the listener
and its event and progress classes, in the Python client as PySpark has it
Key: SPARK-61151
URL: https://issues.apache.org/jira/browse/SPARK-61151
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: Hyukjin Kwon
The Python client's pyspark.sql.streaming.listener is its own rewrite of
PySpark's, which differs from PySpark 4.3.0's:
- StreamingQueryProgress.numInputRows, inputRowsPerSecond and
processedRowsPerSecond are None for the progress of the server, whose JSON
leaves them out. PySpark's sums those of the sources then. So
q.lastProgress.numInputRows and q.recentProgress[i].numInputRows are None.
- StreamingQueryListener has no spark property and no _set_spark_session.
- __all__ lists the events and progress classes, where PySpark's has only
StreamingQueryListener.
The module is pure Python and needs nothing of the client: it should be
PySpark's, vendored verbatim and checked by the drift check. Apache Spark's
test_batch_dataframe_in_batch_session_sql_is_supported and
test_batch_dataframe_join_in_batch_session_is_supported of
test_parity_foreach_batch.py fail on this client ('>' not supported between
instances of 'NoneType' and 'int').
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
