Customising the cubes that BIRT crosstab use.

Yaytay

Hi,

When we compare BIRT crosstab to other approaches it doesn't fare very well, which seems to be because other approaches can use an OLAP server with a prepared cube, whilst BIRT has to prepare the cube data each time a report is run.

There doesn't seem to be any way to hook BIRT up to an OLAP server (Mondrian or SQL Server).
If we used a SQL query against the OLAP server we'd lose the benefit of having the crosstab there at all (we may as well just make the queries up ourselves and put the results in normal tables).
What I'd like to be able to do it to dynamically reconfigure BIRT crosstabs (using the DE) so that our end-users can do data-mining through BIRT.
And the reason for wanting to do this via BIRT (as opposed to, say, Pentaho) is that we can then use BIRT for all out reporting needs (I find BIRT better for static reports than Pentatho).

So, thinking about hooking BIRT up to an OLAP server, my question is: how ridiculous would it be to try to provide an alternative cube implementation that crosstabs could use for their data, that instead of connecting to a datasource went directly to an OLAP server?
From a brief look at the source it looks feasible, but a lot of work - have I missed something that would make it impossible?

Thanks for any thoughts.

Jim