I made the first release candidate of mondrian-2.2 a few days ago.

The most exciting feature is Pentaho’s brand new Cube Designer:

The Pentaho Cube Designer simplifies and expedites the analysis process by eliminating the need for administrators to use XML to create new OLAP schemas. Instead, Cube Designer’s wizard-driven graphical user interface walks administrators through the steps of connecting to relational data sources, defining analytical dimensions, identifying measures for analysis, and publishing the schema to the Pentaho server.

At last, a serious UI for building Mondrian schemas! I’ve been helping the Pentaho folks develop this for the last few months, and was delighted when they decided to release it open-source.

It’s been almost four months since Mondrian release 2.1. There are no earth-shatteringly huge new features in the core Mondrian 2.2, because I’ve been busy with the Cube Designer and olap4j. But if you peruse the change log, you’ll see there have been lots of development activity, resulting in numerous fixes in areas such as performance and native SQL generation.

A useful addition is a set of functions for accessing the current date, contributed by Zelaine Fong and Benny Chow. If MDX had a date datatype, we would have implemented these functions long ago. But MDX doesn’t have a date datatype, because everyone who uses dates wants to access them as a dimension (say, [Time].[2006].[October].[15]), not scalar values. To make things worse, we noticed that everyone’s Time dimension is structured differently: some have quarters while some don’t; some use month names while others use month numbers; and some time dimensions are structured on weeks rather than months.

So, Benny and Zelaine created a function which formats the current date as a string, then looks up a member of the time dimension. For example,

SELECT { [Measures].[Unit Sales] } ON COLUMNS,
  { CurrentDateMember([Time],
  : CurrentDateMember([Time],
      '[\"Time\"]\\.[yyyy]\\.[\"Q\"q]\\.[m]\\.[d]') } ON ROWS
FROM [Sales]

prints the sales total for today and the previous three days.

The trick is with the format string, ‘[\"Time\"]\\.[yyyy]\\.[\"Q\"q]\\.[m]\\.[d]’. This is substituted with the current date, to give [Time].[2006].[Q4].[10].[15]. Then Mondrian looks in the [Time] dimension to find a member with that unique name.

The quotes around "Time" and the backslashes before the dots (\\.) are needed to prevent the formatting facility from replacing things that it shouldn’t. It makes the code a little difficult to read, but it allowed us to use the existing formatting facility, which is well-tested and well-documented, rather than inventing a new one just for this function.

Now suppose that the granularity of the time dimension is less than daily. In a monthly time dimension, should 15th October fall resolve to [2006].[10] or [2006].[11]? The BEFORE, AFTER and EXACT keywords help Mondrian decide. So,


would evaluate to the greatest member which is less than the current date, namely, [Time].[2006].[10].

This function fits well with another feature in mondrian-2.2, which I intend to blog about in a few days, is the extension of Parameters to system, schema, and session scope. Parameters are now much, much more than bind variables, and are a great way of sharing constants and expressions across your whole application.

So, while you’re waiting for that blog entry… go and download the new mondrian release candidate, and try it out. Hopefully, the full release will be out this week or next.