[Mondrian] Segment cache and multi-valued dimensions

[Ricardo Fradinho]

Hi,

I have a question around many to many / multi-valued dimensions support in
Mondrian:

Can I use Mondrian on top of a schema with 1:N relation on the fact =>
dimension without having the segment cache ?

Here's a simplified version of the model I'm looking at:

Fact books:

isbn        prints
----------- ------------
123-XYZ     1

Dimension book_authors:

isbn        author
----------- ------------
123-XYZ     Mr 001
123-XYZ     Mr 002
...
123-XYZ     Mr 010

I have a fact table that joins with a dimension that has several authors
for the same book.
I want Mondrian to return sum(prints)=10 when I join the fact with the
dimension table, but
I want Mondrian to return sum(prints)=1 when don't query over the authors
dimension, ie, just looking at the fact or another dimensions with a 1:1
relation.

If execute a query first with a authors breakdown and the information is
stored in the segment cache, does Mondrian use that segment if I query a
second time without the authors breakdown ?
If so, Mondrian will aggregate over the cached segment and sum(prints)=10.

Put differently, does Mondrian expect a 1:1 relation on the fact foreign
key to dimension primary key ?

I'm not asking about the many to many feature support or semi/non additive
measures, rather I'm looking at the consistency requirements when dealing
with segment aggregates.


Thanks,
Ricardo Fradinho.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.pentaho.org/pipermail/mondrian/attachments/20150504/90787693/attachment.html