Friday, August 31, 2007

mondrian release 2.4 and beyond

I've just released mondrian-2.4 production. In the five months since mondrian-2.3 there have been some significant new features and a host of bug fixes.

The features I'm most pleased with are the ability to aggregate distinct-count measures (much harder to implement than you'd think!) and the ability to generate SQL containing the new GROUPING SETS clause if the database supports it.

No more release lag

In past releases, we've been criticized for a several month lag between a mondrian release and the Pentaho release which contains it - and rightly so. We've now put that right. On Monday, Pentaho just released the 2nd Release Candidate of their 1.6 release, containing mondrian-2.4.1-RC1. Pentaho 1.6 will shortly go production containing the production mondrian-2.4.2.

The future is... olap4j

If you follow this blog, you'll know that I think that the world really needs an open API for accessing OLAP data from Java. I've been working with several other companies and organizations on the olap4j specification.

Since the original draft olap4j specification, version 0.5, released in September 2006, a lot of work has done behind the scenes. There are several new modules in olap4j, including an MDX parser, parse tree model, and a query model, and there are partially complete drivers for mondrian and XMLA.

It's time for a new release of olap4j, and now I have mondrian-2.4 out of the door, I should be able to pull together olap4j release 0.9 by October.

olap4j will also feature prominently in the next mondrian release. Mondrian release 3.0 will include an olap4j driver compatible with the olap4j 0.9 specification. (There will be some other cool features too; see the mondrian roadmap for more details.)

Eventually olap4j will be mondrian's main API. Ever since I started developing mondrian, my mantra has been "don't innovate in the specification: innovate in the implementation". When there are several OLAP servers out there supporting olap4j, and several clients which can talk to any olap4j server, I want people to be choosing mondrian because it is the best implementation. The idea of olap4j is to level the playing field, and may the best OLAP technology win!

Future plans for backwards compatibility

Mondrian's current API will exist unchanged in mondrian-3.0. The API will be deprecated after that, and by mondrian-3.1 we won't guarantee that existing applications will still work. For most applications, moving to olap4j won't be hard: the olap4j API is simpler than mondrian's current API, familiar to anyone who has used JDBC, and better documented too.

There are some other APIs which are deprecated in mondrian-2.4 and will be removed in mondrian-3.0. For details of these, see the roadmap once again.

No comments: