[Mondrian] UDF and result type

[Anton Nikitin]

Done.
https://sourceforge.net/tracker/index.php?func=detail&aid=1755569&group_id=3
5302&atid=414613

Hope my test case will work...

Anton

-----Original Message-----
From: mondrian-bounces at pentaho.org [mailto:mondrian-bounces at pentaho.org] On
Behalf Of Julian Hyde
Sent: Tuesday, July 17, 2007 7:54 PM
To: 'Mondrian developer mailing list'
Subject: RE: [Mondrian] UDF and result type

I'm guessing that the code is instantiating the type based on category (the
approximate type: integer, string, member, dimension etc.) rather than the
type returned from getReturnType. 

Please log a bug and provide a testcase method to add to UdfTest.

Julian

> -----Original Message-----
> From: mondrian-bounces at pentaho.org 
> [mailto:mondrian-bounces at pentaho.org] On Behalf Of Anton Nikitin
> Sent: Tuesday, July 17, 2007 4:48 AM
> To: 'Mondrian developer mailing list'
> Subject: RE: [Mondrian] UDF and result type
> 
> My case is more complex. Function accepts two arguments: 
> member and dimension (not the one for first argument). Return 
> type should be member from dimension, specified in 2nd parameter.
> 
> I have the following definition of getReturnType() :
> 
>     public Type getReturnType(Type[] parameterTypes) {
>         DimensionType dimType = (DimensionType) parameterTypes[1];
> 
>         if (dimType.getDimension()==null) {
>           memberType =  MemberType.Unknown;
>         } else {
>           memberType =
> MemberType.forHierarchy(dimType.getDimension().getHierarchy());
>         }
> 
>         // This one provides the same error!
>         //MemberType memberType =
> MemberType.forDimension(dimType.getDimension());
> 
>         return memberType;
>     }
> 
> After that I can see that resolver considers return type as a 
> member from the same dimension as first argument, not the 
> second one! Then (because I use this member as parameter in 
> tuple expression I get the following):
> 
> Caused by: mondrian.olap.MondrianException: Mondrian 
> Error:Tuple contains more than one member of dimension '[2_6]'.
> 	at
> mondrian.resource.MondrianResource$_Def0.ex(MondrianResource.java:752)
> 	at
> mondrian.olap.fun.TupleFunDef.checkDimensions(TupleFunDef.java:95)
> 	at 
> mondrian.olap.fun.TupleFunDef.getResultType(TupleFunDef.java:73)
> 	at mondrian.olap.fun.FunDefBase.createCall(FunDefBase.java:194)
> 	at 
> mondrian.mdx.UnresolvedFunCall.accept(UnresolvedFunCall.java:93)
> 	...
> 
> Anton
> 
> -----Original Message-----
> From: mondrian-bounces at pentaho.org 
> [mailto:mondrian-bounces at pentaho.org] On Behalf Of Julian Hyde
> Sent: Monday, July 16, 2007 10:22 PM
> To: 'Mondrian developer mailing list'
> Subject: RE: [Mondrian] UDF and result type
> 
> I just wrote a test with a function 'F(<Integer>, <String>) 
> returns <String>'. FunDefBase.guessResultType would guess 
> that this returns an integer, but that didn't happen. The 
> test worked fine, because I implemented
> 
> 
>         public Type getReturnType(Type[] parameterTypes) {
>             return new StringType();
>         }
> 
> in my UDF.
> 
> Julian
> 
> > -----Original Message-----
> > From: mondrian-bounces at pentaho.org
> > [mailto:mondrian-bounces at pentaho.org] On Behalf Of Anton Nikitin
> > Sent: Monday, July 16, 2007 3:33 AM
> > To: 'Mondrian developer mailing list'
> > Subject: [Mondrian] UDF and result type
> > 
> > 
> > Hello.
> > 
> > I have a question about User Defined Functions.
> > 
> > AFAIK, UDF is resolved into UdfResolver.UdfFunDef instance which 
> > derives from FunDefBase.
> > It forces using getResultType() from FunDefBase which doesn't work 
> > correctly in our case (FunDefBase.guessResultType() 
> provides incorrect 
> > result). In turn, it is impossible to implement anything in 
> UDF class 
> > to correct its behavior (in contrast to FunDef-based 
> approach, where I 
> > could simply override getResultType()).
> > 
> > Any thoughts, workarounds?
> > 
> > Thanks,
> > 
> > Anton
> > 
> > 
> > _______________________________________________
> > Mondrian mailing list
> > Mondrian at pentaho.org
> > http://lists.pentaho.org/mailman/listinfo/mondrian
> > 
> 
> _______________________________________________
> Mondrian mailing list
> Mondrian at pentaho.org
> http://lists.pentaho.org/mailman/listinfo/mondrian
> 
> _______________________________________________
> Mondrian mailing list
> Mondrian at pentaho.org
> http://lists.pentaho.org/mailman/listinfo/mondrian
> 

_______________________________________________
Mondrian mailing list
Mondrian at pentaho.org
http://lists.pentaho.org/mailman/listinfo/mondrian