Xiangrui Meng created SPARK-7903:
------------------------------------
Summary: PythonUDT shouldn't get serialized on the Scala side
Key: SPARK-7903
URL: https://issues.apache.org/jira/browse/SPARK-7903
Project: Spark
Issue Type: Bug
Components: PySpark, SQL
Affects Versions: 1.4.0
Reporter: Xiangrui Meng
Assignee: Xiangrui Meng
A round trip for a pure Python UDT should be: Python UDT -> Python SQL internal
types -> Scala/Java SQL internal types -> transformation -> Scala/Java SQL
internal types -> Python SQL internal types -> Python UDT. So the serialization
shouldn't be invoked on the Scala side if no Scala code is applied to the UDT.
Code (from [~rams]) to reproduce this bug:
{code}
from pyspark.mllib.linalg import SparseVector
from pyspark.sql.functions import udf
from pyspark.sql.types import IntegerType
df = sqlContext.createDataFrame([(SparseVector(2, {0: 0.0}),)], ["features"])
sz = udf(lambda s: s.size, IntegerType())
df.select(sz(df.features).alias("sz")).collect()
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
