Hyukjin Kwon created SPARK-61539:
------------------------------------

             Summary: Raise PySpark's TypeError for a jdbc property whose key 
is not a str, in the Python client
                 Key: SPARK-61539
                 URL: https://issues.apache.org/jira/browse/SPARK-61539
             Project: Spark
          Issue Type: Bug
          Components: Connect
    Affects Versions: connect-rust-4.3.0
            Reporter: Hyukjin Kwon


PySpark's DataFrameReader.jdbc and DataFrameWriter.jdbc pass the properties as 
`options(**properties)`. A key that is not a str raises the TypeError of that 
unpacking, "keywords must be strings", before any property is set. The format, 
the options of a column (reader) and the mode (writer) are already set on the 
reader or writer by then.

The Python client sent the str() of each key as an option name instead: 
properties={b"user": "u"} set an option "b'user'", so the user property was 
silently lost, and {1: "v"} set an option "1". The jdbc methods now raise 
PySpark's TypeError, and keep the reader's and the writer's settings as 
PySpark's do. A str key is set as protobuf takes it (crate::str_argument), not 
through str().




--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to