Tim Meehan created SPARK-61435:
----------------------------------

             Summary: Send the bytes() of a bytes or bytearray subclass in a 
binary column of createDataFrame in the Python client, as PySpark does
                 Key: SPARK-61435
                 URL: https://issues.apache.org/jira/browse/SPARK-61435
             Project: Spark
          Issue Type: Bug
          Components: Connect
    Affects Versions: connect-rust-4.3.0
            Reporter: Tim Meehan


createDataFrame of the Python client of spark-connect-rust, for a column of 
BinaryType, reads the buffer of a bytes or bytearray subclass. PySpark's 
LocalDataToArrowConversion calls bytes() on it, which takes the __bytes__ of 
the subclass. A subclass with its own __bytes__ is sent with other bytes, and 
no error is raised.

{code:python}
from pyspark.sql import SparkSession

class B(bytes):
    def __bytes__(self):
        return b"other"

spark = SparkSession.builder.remote("sc://localhost:15002").getOrCreate()
print(spark.createDataFrame([(B(b"orig"),)], "b binary").collect())
{code}

Against an OSS Spark 4.3.0 Connect server on loopback:
* pyspark-client 4.3.0: [Row(b=b'other')].
* spark-connect-rust master 125654b: [Row(b=b'orig')].

The same holds for a bytearray subclass with a __bytes__ (b'other' against 
b'orig'). A subclass without a __bytes__, and bytes and bytearray themselves, 
give the same bytes in both clients.

>From source, at 125654b: the BinaryType branch of convert 
>(crates/pyspark-rs/src/infer.rs:1217-1224) casts the value to PyBytes or 
>PyByteArray, which also succeeds for a subclass, and copies the buffer. 
>PySpark's convert_binary (pyspark/sql/conversion.py) returns an exact bytes as 
>it is and calls bytes(value) on any other bytes or bytearray.




--
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