[Mondrian] Custom aggregate Functions

mondrian at pentaho.org mondrian at pentaho.org
Wed Dec 13 07:10:49 EST 2006

>> P.S.: How can I help implement custom aggregate functions,
>> I will need that!
> Let's discuss custom aggregate functions in a different thread, say in the
> Open Discussion forum.
> One thing I'd like to understand is (a) how your functions roll up, and (b)
> whether the function can be evaluated using a SQL expression built up from
> standard aggregate SQL aggregate functions like SUM. If it is not possible
> to translate the aggregate function into a SQL expression, Mondrian will
> need to read back all of the fact table data, and that will be inefficient.
> So, I'd like to see a few use-cases before we go off and build the wrong
> feature!
The use case is for "Measures" that are not aggregatable.
(for example a comment to each row in the fact table)
At the time of writing I was thinking about having a user defined
aggregate function that would show a placeholder if aggregated
and the comment if only one row would be shown.
But I realized, the proper solution would be to use drill-through,
and, e.g. count as an aggregate function.

Alas, the current drill through on JPivot seems to show all fact rows for
all members of the hierarchy on an axis (the view uses a parent-child
hierarchy on one axis and measures on the other), which is impracticable
for even small or medium
size fact tables.

I figured out that JPivot uses the MondrianCell#getDrillThroughSQL(true);
And I currently am trying to switch it to false.

Here is another issue: It is hard to know which source version of JPivot
with which source version of mondrian, especially because JPivot and
Mondrian both
don't use version numbers on their jars respectively.
(Do you happen to know, which JPivot version matches with Mondrian 2.1.1?)

I currently use maven 2 as a build system, and it makes managing
versions and dependencies a lot easier.....

Thank you for your fast response, I hope my feedback is helpful in some

Kind regards,

More information about the Mondrian mailing list