Hyukjin Kwon created SPARK-60974:
------------------------------------
Summary: Accept a DDL string schema in mapInPandas, applyInPandas
and the other map functions in Spark Connect Rust as PySpark does
Key: SPARK-60974
URL: https://issues.apache.org/jira/browse/SPARK-60974
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: Hyukjin Kwon
In PySpark, the schema of {{DataFrame.mapInPandas}}, of
{{GroupedData.applyInPandas}} and {{applyInArrow}}, and of the cogrouped
{{applyInPandas}} and {{applyInArrow}} can be a DDL string, as can the output
type of {{applyInPandasWithState}} and {{transformWithStateInPandas}}. The
client parses the string into a {{StructType}} and pickles the function with
that type. The Rust-backed Python client pickles the string itself, so the
Python worker fails. For example:
{code:python}
df.groupBy("k").applyInPandas(lambda pdf: pdf, "k long, v double").collect()
{code}
This works in PySpark 4.3.0. The client raises {{PythonException}} with
{{AssertionError: dataType k long, v double should be an instance of <class
'pyspark.sql.types.DataType'>}}. With a DDL string, five of the six batch
functions fail, and {{applyInArrow}} raises
{{UNSUPPORTED_DATA_TYPE_FOR_ARROW_CONVERSION}}. Pickle the parsed type 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]
