Hyukjin Kwon created SPARK-60967:
------------------------------------
Summary: Deserialize collected values of user-defined types, and
give their dtypes, in Spark Connect Rust as PySpark does
Key: SPARK-60967
URL: https://issues.apache.org/jira/browse/SPARK-60967
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: Hyukjin Kwon
In PySpark Connect, a value of a user-defined type that a DataFrame collects is
deserialized by its UDT, and {{dtypes}} gives the UDT's {{simpleString()}}.
With {{df = spark.sql("SELECT array(1.5D, 2.0D) AS
a").select(array_to_vector("a").alias("vec"))}} of {{pyspark.ml.functions}},
{{df.first().vec}} is {{DenseVector([1.5, 2.0])}} and {{df.dtypes}} is
{{[('vec', 'vector')]}} in PySpark 4.3.0.
The Rust-backed Python client gives the value of the UDT's SQL type, a {{Row}},
and {{[('vec', 'udt')]}}, though the field type of {{df.schema}} is the
{{VectorUDT}}. Deserialize the collected values of a user-defined type, and
give the dtypes of their UDTs, 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]
