With Autoexist I thought that AS 2005 would finally work as a relational
database "inner join" between dimension tables and fact-tables. My thought
was that if I put a customer dimension, with 3000 members, together with a
product dimension, with 2000 members, on the row axis, I would only see the
valid combinations of Customers and Products, actually bought, in a sales
AS-2005 cube. Not a relational crossjoin between Customers and Products.

After reading comments from Chris Webb I realize that this is not correct,
or?

What You actually mean(MS) is that this is only valid for a single dimension
and the "attribute hierarchies" and "user hierarchies" within that
dimension?

If this is correct I cannot get this to fit with the talk of the UDM as a
source for all types of reporting and analysis. I will still have to use
TSQL and reporting services to get only the real combination of Customers,
Products and Facts that have occurred in Sales?

Regards
Thomas Ivarsson

All you need is to say you want the NON EMPTY crossjoin of the three
dimensions and it will return you only the rows that have non empty measure
values...

Thanks,
Akshai
--
This posting is provided "AS IS" with no warranties, and confers no rights
Please do not send email directly to this alias. This alias is for newsgroup
purposes only.

"Thomas Ivarsson" <TI (AT) nospam (DOT) com> wrote

Yes I know. This is not the question. So this model have not changed and it
is still relational crossjoin between dimensions and dimensions-facts that
is the model in AS 2005? This means that I cannot use the UDM to get a
relational inner join? I must use TSQL and RS 2005 to solve this?

Regards
/Thomas

"Akshai Mirchandani [MS]" <akshaim (AT) online (DOT) microsoft.com> wrote

Yes, Auto-exists is automatic only within a dimension. Dimensions are
special in OLAP and so yes, it is by default a full crossjoin between
dimensions and dimensions and facts.

But I was also answering this comment:

Just because the "default" mechanism of crossjoining dimensions doesn't give
you the inner join doesn't mean you can't get it and "still have to use TSQL
and reporting services".

You can also use the function Exists( <set>, [<filter set>], "measure group
name" ) to obtain the inner join with respect to a measure group. This will
return the tuples in the set that have data associated with them in the
measure group. So essentially you can now do:

SELECT Exists( [Customers].members * [Product].members, ,"Sales" ) ON 0
FROM [Sales]
would return you the customers and products that have sales.

Or:

SELECT Exists( [Customers].members, { [Product].[Food] }, "Sales" ) ON 0
FROM [Sales]
which would return you the customers that bought the product Food in the
Sales fact table.


But yes, you still cannot do cross-dimensional natural joins very
efficiently. E.g. something like:
SELECT [BillToCustomers].members * [ShipToCustomers].members
where BillToCustomers.[EmailAlias] = ShipToCustomers.[EmailAlias]
isn't a very natural OLAP construct and requires you to use a Filter over
the entire crossjoin...

Thanks,
Akshai
--
This posting is provided "AS IS" with no warranties, and confers no rights
Please do not send email directly to this alias. This alias is for newsgroup
purposes only.

"Thomas Ivarsson" <TI (AT) nospam (DOT) com> wrote

Thank's for your comment.

Regards
/Thomas
"Akshai Mirchandani [MS]" <akshaim (AT) online (DOT) microsoft.com> wrote