Victor Alor created SPARK-25371:
-----------------------------------
Summary: Vector Assembler with no input columns throws an exception
Key: SPARK-25371
URL: https://issues.apache.org/jira/browse/SPARK-25371
Project: Spark
Issue Type: Bug
Components: ML, MLlib
Affects Versions: 2.3.1, 2.3.0
Reporter: Victor Alor
When `VectorAssembler ` is given an empty array as its inputColumns it throws
an opaque error. In versions less than 2.3 `VectorAssembler` it simply appends
a column containing empty vectors.
{code:java}
val inputCols = Array()
val outputCols = Array("A")
val vectorAssembler = new VectorAssembler()
.setInputCols(inputCols)
.setOutputCol(outputCols)
vectorAssmbler.fit(data).transform(df)
{code}
In versions 2.3 > this throws the exception below
{code:java}
org.apache.spark.sql.AnalysisException: cannot resolve 'named_struct()' due to
data type mismatch: input to function named_struct requires at least one
argument;;
{code}
Whereas in versions less than 2.3 it just adds a column containing an empty
vector.
I'm not certain if this is an intentional choice or an actual bug. If this is a
bug, the `VectorAssembler` should be modified to append an empty vector column
if it detects no inputCols. If it is a design decision it would be nice to
throw a human readable exception explicitly stating inputColumns must not be
empty. The current error is somewhat opaque.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
