Under the gracious invitation of the Pentaho folks I'm attending a week's training in Orlando. Now being in Orlando without my kids is a bit painful but is offset by the knowledge that I'll be back in a week's time for vacation, oh I can't wait for that.

But I digress, back to the point of the post.

Pentaho is a group of folks with years in the BI space, all are from one of the big 3 players and most have spent time together in other companies that were acquired by one of the big three. Its a close nit group and you can tell, makes for a very fun environment where special things can happen and often do.

So far in the class time we have covered the over all architecture of the product and reporting. Docs can be found here, I know I've seen an architecture document on their site too that I can't find right now. I'll update if/when I do find it.

Architecture: So far I'm impressed with the architecture. They have a solid product that looks like its easy to extend. You can for example provide your own integration with your favorite reporting solution (as an example) with a few hundred lines of java. Its very similar to the SPI idea you find in JNDI and many other tool sets. Pentaho defines a set of api's and you implement then and bada-boom you have integration. They of course provide support out of the box for good reporting but the point is extensibility, they have it in spades. There are lots of other interesting bits to the architecture. Just as another example, you can burst reports. So from a simple bit of XML you can specify that 1000 people are to get a report and farm that out to 10 servers to build the reports (this is accomplished through posting messages to JMS and each of the 10 machines picking up the work). Anyway enough geeking out, its cool.

Reporting: Looks very cool architecturally, they are integrated with JFreeReports (whats with the ape on the top of their page anyway, was it written with the help of Primate Programming?), BIRT and another one I don't recall at the moment. Anyway the way that these various engines are integrated is again with a couple of hundred lines of java code and its done. I think in the long haul they should focus on one and continue to make it possible to do others. I my opinion its just too much work to maintain all the integrations, but I digress. So with the solution they currently have you define a report in the respective reporting language/tool set (i.e. XML for the JFreeReports report). This definition is then matched with the data in a Pentaho Action Definition and executed and you have a report. The Actions in the action sequence are essentially an implementation of the Command Pattern. So they are easy to string together into the sequences. Its cool stuff.

So anyway if you are doing a Java EE project then you should check out this stuff. There are some rough edges but what I've tried works.