spark-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Gregor Moehler (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (SPARK-19311) UDFs disregard UDT type hierarchy
Date Fri, 20 Jan 2017 14:16:26 GMT

     [ https://issues.apache.org/jira/browse/SPARK-19311?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Gregor Moehler updated SPARK-19311:
-----------------------------------
    Description: 
When you define UDTs based on hierarchical traits UDFs disregard the type hierarchy:

E.g. I have 2 UDTs based on 2 hierarchical traits. I then define 2 UDFs: The first one returns
the derived type, the second takes the base type. This results in an error, although i believe
it should be feasible:

{quote}
(...)cannot resolve 'UDF(UDF(22))' due to data type mismatch: argument 1 requires exampleBaseType
type, however, 'UDF(22)' is of exampleFirstSubType type.
{quote}

The reason is that DataType defines
{quote}
override private[sql] def acceptsType(dataType: DataType) =
    this.getClass == dataType.getClass
{quote}

However I believe it should be:

{quote}
override private[sql] def acceptsType(dataType: DataType) = dataType match {
    case other: UserDefinedType[_] =>
      this.getClass == other.getClass || this.userClass.isAssignableFrom(other.userClass)
    case _ => false
  }
{quote}

  was:
When you define UDTs based on hierarchical traits UDFs disregard the type hierarchy:

E.g. I have 2 UDTs based on 2 hierarchical traits. I then define 2 UDFs: The first one returns
the derived type, the second takes the base type. This results in an error, although i believe
it should be feasible:

(...)cannot resolve 'UDF(UDF(22))' due to data type mismatch: argument 1 requires exampleBaseType
type, however, 'UDF(22)' is of exampleFirstSubType type.

The reason is that DataType defines
{quote}
override private[sql] def acceptsType(dataType: DataType) =
    this.getClass == dataType.getClass
{quote}

However I believe it should be:

{quote}
override private[sql] def acceptsType(dataType: DataType) = dataType match {
    case other: UserDefinedType[_] =>
      this.getClass == other.getClass || this.userClass.isAssignableFrom(other.userClass)
    case _ => false
  }
{quote}


> UDFs disregard UDT type hierarchy
> ---------------------------------
>
>                 Key: SPARK-19311
>                 URL: https://issues.apache.org/jira/browse/SPARK-19311
>             Project: Spark
>          Issue Type: Bug
>          Components: SQL
>    Affects Versions: 2.1.0
>            Reporter: Gregor Moehler
>   Original Estimate: 1h
>  Remaining Estimate: 1h
>
> When you define UDTs based on hierarchical traits UDFs disregard the type hierarchy:
> E.g. I have 2 UDTs based on 2 hierarchical traits. I then define 2 UDFs: The first one
returns the derived type, the second takes the base type. This results in an error, although
i believe it should be feasible:
> {quote}
> (...)cannot resolve 'UDF(UDF(22))' due to data type mismatch: argument 1 requires exampleBaseType
type, however, 'UDF(22)' is of exampleFirstSubType type.
> {quote}
> The reason is that DataType defines
> {quote}
> override private[sql] def acceptsType(dataType: DataType) =
>     this.getClass == dataType.getClass
> {quote}
> However I believe it should be:
> {quote}
> override private[sql] def acceptsType(dataType: DataType) = dataType match {
>     case other: UserDefinedType[_] =>
>       this.getClass == other.getClass || this.userClass.isAssignableFrom(other.userClass)
>     case _ => false
>   }
> {quote}



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org


Mime
View raw message