[Mondrian] NullDenominatorProducesNull Property

John V. Sichi jsichi at gmail.com
Tue Dec 9 15:35:41 EST 2008


michael.pflug at thomsonreuters.com wrote:
>   Can I get some clarification about the purpose of the property 
> NullDenominatorProducesNull?  On the Mondrian website and in comments 
> the property is called NullOrZeroDenominatorProducesNull while the 
> actual name of the property is NullDenominatorProducesNull.  The code 
> using this property actually checks both the numerator and demoninator 
> for null, but doesn’t check for zero.  I have a case where I want to 
> catch a divide by zero and return null instead of “infinity” and it 
> seems like this property should be the way to do it (see example 
> below).  Does anyone know why the property was changed to exclude zero 
> to begin with and are there objections to adding the check for zero back 
> in?  

This property was introduced when the discrepancy between Mondrian and 
MSAS behavior was originally noticed.  The property was added as a 
backward-compatibility flag to allow applications which had become 
dependent on the original Mondrian behavior to keep working.

There was some confusion about the property name and semantics, so it 
went through a few iterations before settling on 
NullDenominatorProducesNull.

Rushan, it looks like mondrian/doc/configuration.html never got updated 
to match the final state, leading to the mismatch Michael noticed; can 
you fix that?

Michael, I don't think further changes to the semantics of this property 
would be the best approach, since that can break existing reports; 
instead, perhaps some new UDF's which can be used to explicitly request 
the desired semantics on a case-by-base basis?

JVS




More information about the Mondrian mailing list