Hyukjin Kwon created SPARK-62078:
------------------------------------
Summary: The methods of the UDF and UDTF registrations and of the
streaming classes lack the docstrings that PySpark Connect copies of PySpark's
classic classes, in the Python client
Key: SPARK-62078
URL: https://issues.apache.org/jira/browse/SPARK-62078
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: 5.0.0
Reporter: Hyukjin Kwon
PySpark Connect's modules give the methods of their classes the docstrings of
PySpark's classic classes:
- python/pyspark/sql/connect/udf.py:331-346: `register.__doc__ =
PySparkUDFRegistration.register.__doc__`, and the same of registerJavaFunction
and registerJavaUDAF;
- python/pyspark/sql/connect/udtf.py:253: `register.__doc__ =
PySparkUDTFRegistration.register.__doc__`;
- python/pyspark/sql/connect/streaming/query.py:59-253: 12 members of
StreamingQuery (id, runId, name, isActive, awaitTermination, status,
recentProgress, lastProgress, processAllAvailable, stop, explain, exception)
and 6 of StreamingQueryManager (active, get, awaitAnyTermination,
resetTerminated, addListener, removeListener);
- python/pyspark/sql/connect/streaming/readwriter.py:72-765: 14 methods of
DataStreamReader (format, schema, option, options, name, load, json, orc,
parquet, text, csv, xml, table, changes) and the class itself
(`DataStreamReader.__doc__ = PySparkDataStreamReader.__doc__`), and 12 of
DataStreamWriter (outputMode, format, option, options, partitionBy, clusterBy,
queryName, trigger, foreach, foreachBatch, start, toTable).
The drop-in's members have docstrings of their own, or none, for all 49 of
them. Repro: `help(spark.udf.register)`, `spark.readStream.format.__doc__`,
`type(spark.streams).active.__doc__`.
- pyspark-client: PySpark's documentation, of 4536, 1161 and 793 characters,
with its parameters, notes and examples.
- The drop-in: "Register a Python UDF with the given name. ..." of 789
characters, "Set the source format. As PySpark's, None unsets it, ..." of 97,
and None.
The registrations of the drop-in are Python classes, which can assign the
docstrings as PySpark Connect does, as SPARK-62071 vendored the classic
pyspark.sql.udf and pyspark.sql.udtf. The streaming classes are native, whose
docstrings are those that the extension is built with.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]