Tim Meehan created SPARK-61290:
----------------------------------
Summary: Raise CANNOT_INFER_ARRAY_ELEMENT_TYPE for a list without
an element type in the Column methods of the Python client, as PySpark does
Key: SPARK-61290
URL: https://issues.apache.org/jira/browse/SPARK-61290
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: Tim Meehan
PySpark's Column methods, such as when, otherwise and isin, send a Python list
as an array literal whose element type they infer, and raise
CANNOT_INFER_ARRAY_ELEMENT_TYPE for an empty list or one that starts with None
(pyspark/sql/connect/expressions.py:368). In the Python client of
spark-connect-rust, such a list becomes array(...) of its elements, so these
lists are accepted.
{code:python}
from pyspark.sql import SparkSession, functions as F
spark = SparkSession.builder.remote("sc://localhost:15002").getOrCreate()
spark.range(2).select(F.when(F.col("id") > 0,
F.array(F.lit(1))).otherwise([]).alias("o")).collect()
{code}
Against an OSS Spark 4.3.0 Connect server on loopback:
* pyspark-client 4.3.0: PySparkTypeError: \[CANNOT_INFER_ARRAY_ELEMENT_TYPE\]
Can not infer the element data type, a non-empty list starting with a non-None
value is required. otherwise(\[None, 2\]) raises the same.
* spark-connect-rust master 8768cc4: \[Row(o=\[\]), Row(o=\[1\])\], and with
otherwise(\[None, 2\]), \[Row(o=\[None, 2\]), Row(o=\[1\])\].
>From source: to_column (crates/pyspark-rs/src/functions.rs:224) sends every
>list through pyspark.sql.functions.lit (:239-247), which builds array(...).
>That is right for PySpark's functions: F.when(F.lit(True), \[\]) returns
>\[Row(w=\[\])\] in both clients.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
