[Mondrian] DRILLTHROUGH behavior change in lagunitas

Paul Stoellberger p.stoellberger at gmail.com
Wed Mar 6 05:37:51 EST 2013


I would expect the raw data (fact table records) to be returned
There are some use cases where people want to use the DRILLTHROUGH to get the exact fact table rows that make up the cell 

This would match the description you posted from the SSAS specs

-Paul

On Mar 6, 2013, at 12:55 AM, Julian Hyde <jhyde at pentaho.com> wrote:

> Luc changed the behavior of DRILLTHROUGH on lagunitas branch in August (see https://github.com/pentaho/mondrian/commit/cb9854af5a654641bc6f524ed9f452a283f04459), and I'm only now reviewing the change. The effect is that DRILLTHROUGH now returns the cells at the "bottom of the cube" (i.e. at the lowest level of every hierarchy in the cube) as opposed to the rows in the fact table. 
> 
> Before we release this change in mondrian-4, I want to check two things:
> 1. Is the new behavior consistent with the spec?
> 2. Which behavior do our users want, the old or the new?
> 
> Luc is half right when he says in his check-in comment that the new behavior is "per SSAS specs". The latest specs (http://technet.microsoft.com/en-us/library/ms145964.aspx) say
> 
>> By default, a drillthrough result set is derived from the table rows that were evaluated
>> to calculate the value of the selected Analysis Services, the results are retrieved
>> directly from MOLAP storage, unless ROLAP partitions or dimensions are queried.
> 
> In other words, an engine should return the lowest level cell data that it has. If SSAS is running in ROLAP mode, it will return table rows.  Since Mondrian is a ROLAP engine, it should return table rows too. They basically cut SSAS some slack when it is running in MOLAP mode.
> 
> Now, the more important question: What do our users want? Users, please chime in. If I hear nothing, we will probably keep the new behavior.
> 
> Julian
> _______________________________________________
> Mondrian mailing list
> Mondrian at pentaho.org
> http://lists.pentaho.org/mailman/listinfo/mondrian



More information about the Mondrian mailing list