Will Gorman’s Blog

During the development of 5.1, Pentaho has taken steps to integrate Mondrian 4 into our business analytics platform.  This article goes over what we have accomplished so far, where we are headed, and also instructions for getting Mondrian 4 working with Pentaho 5.1 Community Edition.

Pentaho Enterprise Edition has Mondrian 4 bundled for a specific reason - we’ve now introduced native MongoDB support as a plugin to Mondrian 4.  This use case allows customers to slice and dice data from MongoDB collections in Pentaho Analyzer.  You can learn more about the capability here:  http://www.pentaho.com/request-analyzer-mongodb

As we continue to evolve the Pentaho Platform, we need a more flexible plugin architecture for driving innovation.  To allow both Mondrian 3 and Mondrian 4 runtime environments, we’ve introduced OSGi as a core part of the platform.  Mondrian 4 is our first use case, but we’ll be introducing many others in future versions.

Once Mondrian 4 is installed as an OSGi bundle, it is available as an OLAP4J resource to the platform via Pentaho’s proxy system Driver aptly named “PentahoSystemDriver”.  The following steps below walk you through getting Pentaho Mondrian up and running within Pentaho CE 5.1.  Note that these instructions won’t work against previous versions of Pentaho, and these instructions are not necessary in Pentaho EE 5.1, as Mondrian 4 is already configured and installed.

Download Pentaho 5.1 CE

You can download Pentaho 5.1 from sourceforge here: http://sourceforge.net/projects/pentaho/files/Business%20Intelligence%20Server/5.1/

Make sure it is working normally before continuing with these instructions.

Deploy Mondrian 4’s required OSGi bundles

We first need to add Mondrian 4 and its dependencies as OSGi Bundles.  Copy the following JARs, which are now OSGi compatible, to pentaho-solutions/system/osgi/bundles:

• mondrian-4.0.0-SNAPSHOT.jar http://nexus.pentaho.org/content/groups/omni/pentaho/mondrian/4.0.0-SNAPSHOT/
• olap4j-xmlaserver-1.2.0.jar  http://nexus.pentaho.org/content/groups/omni/org/olap4j/olap4j-xmlaserver/1.2.0/olap4j-xmlaserver-1.2.0.jar
• eigenbase-properties-1.1.4.jar  http://nexus.pentaho.org/content/groups/omni/eigenbase/eigenbase-properties/1.1.4/eigenbase-properties-1.1.4.jar
• eigenbase-resgen-1.3.5.jar  http://nexus.pentaho.org/content/groups/omni/eigenbase/eigenbase-resgen/1.3.5/eigenbase-resgen-1.3.5.jar
• eigenbase-xom-1.3.4.jar  http://nexus.pentaho.org/content/groups/omni/eigenbase/eigenbase-xom/1.3.4/eigenbase-xom-1.3.4.jar
• commons-io-1.4.jar  http://repo1.maven.org/maven2/commons-io/commons-io/1.4/commons-io-1.4.jar
• commons-lang-2.4.jar  http://repo1.maven.org/maven2/commons-lang/commons-lang/2.4/commons-lang-2.4.jar

Also, you may copy the Mondrian properties file to “mondrian.cfg” in the same folder to customize various Mondrian 4 properties.

Install the CDA Plugin via the Marketplace

Go to the Marketplace perspective and install Community Data Access, this plugin allows you to query various data sources including Mondrian 4 via OLAP4J.

Setup a Mondrian 4 Database and Schema

I already had Foodmart installed in a local MySQL instance, I also deployed the Mondrian 4 FoodMart.mondrian.xml schema (https://github.com/pentaho/mondrian/blob/lagunitas/demo/FoodMart.mondrian.xml) to the BA Server by uploading the file in the /public/Foodmart folder.

Create a new CDA file that queries Foodmart

I copied the olap4j example that comes with CDA.  Here are the important parts:

<DataSources>
<Connection id=”1″ type=”olap4j”>
<Driver>mondrian.olap4j.MondrianOlap4jDriver</Driver>
<Url>jdbc:mondrian4:</Url>
<Property name=”JdbcUser”>foodmart</Property>
<Property name=”JdbcPassword”>foodmart</Property>
<Property name=”Jdbc”>jdbc:mysql://localhost:3306/foodmart_mondrian_4</Property>
<Property name=”JdbcDrivers”>com.mysql.jdbc.Driver</Property>
<Property name=”Catalog”>solution:/public/Foodmart/FoodMart.mondrian.xml</Property>
</Connection>
</DataSources>
…
<Query>
select {[Measures].[Unit Sales]} ON COLUMNS,
NON EMPTY [Time].[Time].[1997].Children ON ROWS
from [Sales]
where ([Product].[${productfamily}])
</Query>

You can download the full CDA file I used here: mondrian4.cda

Upload the CDA file in the /public/Foodmart folder.

You can now run the CDA file and see the results come back from Mondrian 4!

Additional Info

For developers wanting to access Mondrian 4 metadata like they can access Mondrian 3 schemas by schema name in the Pentaho Platform, we’ve done some low level plumbing to get folks started.  If you define a Mondrian 4 connection in pentaho-solutions/system/olap4j.properties, you can gain access to that programmatically through code like the following:

import org.pentaho.platform.plugin.services.connections.mondrian.MDXOlap4jConnection;
import org.pentaho.platform.plugin.services.importexport.legacy.MondrianCatalogRepositoryHelper;
import org.pentaho.platform.plugin.services.importexport.legacy.MondrianCatalogRepositoryHelper.Olap4jServerInfo;
...
final MondrianCatalogRepositoryHelper helper = new MondrianCatalogRepositoryHelper( repo );
if ( helper.getOlap4jServers().contains( catalogName ) ) {
final Olap4jServerInfo serverInfo = helper.getOlap4jServerInfo( catalogName );
properties.setProperty( “url”, serverInfo.URL );
properties.setProperty( “driver”, serverInfo.className );
if ( serverInfo.user != null ) {
properties.setProperty( “user”, serverInfo.user );
}
if ( serverInfo.password != null ) {
properties.setProperty( “password”, serverInfo.password );
}
MDXOlap4jConnection connection =
(MDXOlap4jConnection) PentahoConnectionFactory.getConnection( IPentahoConnection.MDX_OLAP4J_DATASOURCE,
properties, PentahoSessionHolder.getSession(), null );
}

Using this utility code is nice because the MDXOlap4jConnection will manage mapping Pentaho’s roles to Mondrian’s.

So how does all of this work?

Pentaho has bundled Apache Felix into the Pentaho Platform.  Felix is an OSGi container which now manages Mondrian 4 and its dependencies.  The core bundles that make up Pentaho’s OSGi container can be found in pentaho-solutions/system/osgi/core_bundles, here you’ll find a number of utility OSGi jars including Gemini’s Blueprint, which we use for wiring OSGi components.  Blueprint is similar to the Spring Framework.  Also, the Mondrian 4 jar contains some metadata that registers it with the Pentaho platform as an available OLAP4J driver with the JDBC prefix name of “mondrian4″.  You can check out the metadata file OSGI-INF/blueprint/beans.xml to see the specific XML to declare the driver.  To see how the internal wiring is done, and how PentahoSystemDriver is involved, you can check out the pentaho-platform package org.pentaho.platform.osgi.

There is still a lot of work to do!

Here are some of the areas we will need to complete in future versions to make this a seamless experience:

• Ship the Mondrian 4 binaries with the release!
• Update Pentaho’s Datasource Manager to easily manage Mondrian 4 connections instead of editing olap4j.properties
• Enable Mondrian 4 to access Pentaho Database Connections
• Create an easy to use Mondrian 4 Schema Editor - Maybe Ivy Information System’s work will move in that direction?
• Have projects such as Pivot4J and Saiku support Mondrian 4 connections via OSGi

If any of this work is of interest to you, folks in the Mondrian mailing list and at Pentaho would be happy to help point the way!

Thanks!

Will

There has been a lot of buzz recently about Pentaho’s acquisition of LucidEra’s ClearView technology, now known as Pentaho Analyzer Enterprise Edition, including this detailed writeup by Julian Hyde. But one
important aspect of this acquisition that has gone unmentioned, which I am most excited about, is the addition of Benny Chow to the Pentaho engineering team.

At LucidEra, Benny was the Director of Engineering and was responsible for the development of ClearView. In his short time here at Pentaho, he’s already had a great impact on our engineering efforts especially with the integration work necessary to make Analyzer an integral part of Pentaho’s User Console. Let me be the first to publicly welcome Benny to the Pentaho team!

I just got back from spending the week in Santa Clara, California at the MySQL Conference. I really enjoyed spreading the word about Open Source Business Intelligence. Many folks attending the conference were new to Business Intelligence and Pentaho, so I enjoyed demoing our products, showing off Adhoc Querying with Metadata and building Transformations with Kettle.

I was very impressed with Marten Mickos and Jonathan Schwartz’s keynotes. Both focused on the value and benefits of Open Source technology. Marten celebrated the MySQL community by recognizing some of the top community contributors of MySQL, and Jonathan discussed the big picture of how Sun powered by Open Source can change the world.

Other great highlights included Lance Walter giving a talk on Operational Business Intelligence, and Julian Hyde discussing Interactive OLAP, which included a demo of Halogen. To cap off the week, I visited my good friend Dan Morrill at the Googleplex and got a tour of their wonderfully creative environment, including pin ball machines and space ships!

Over the past few weeks, we’ve been working hard to get Mondrian 3.0 feature complete and relatively bug free. This is a major release, with large architectural and functional changes. I spent the last part of 2007 implementing the new shared dimensions architecture for Mondrian. This change cleaned up a lot of code in Mondrian, along with enabling new features like sharing the same dimension multiple times in a single cube, and also better management of shared dimensions within Virtual Cubes.

In January, I worked with Julian on squashing bugs. Earlier in the month we triaged all the bugs that have been logged since 2006. Now that we’ve assigned and prioritized them, tracking down bugs has been a great way for me to become more familiar with the Mondrian code base.

Some of the other major features that Julian has been working on this round include better role support, olap4j API support, and a much richer set of VBA and Excel function support.

Read about all the changes in the change log, and of course download this milestone release, kick the tires, and let us know what you think!

Bill Seyler has been working on an open source web 2.0 GWT pivot viewer based on the new olap4j API.  He recently posted the first cut over on this google code project.  Julian Hyde has a few words to say about the release over on his blog.  I’ve been working with Bill and Julian on the olap4j query implementation behind Halogen.  By separating the UI from concepts like pivoting, etc, it makes it much easier for other UI developers to incorporate multi-dimensional queries into their own applications.  Hopefully olap4j will encourage new OLAP UI development in the open source community!

Last week, the architects of Pentaho’s products convened in Orlando to sync up and discuss plans for 2008.  Matt Casters of Kettle, Julian Hyde of Mondrian, Thomas Morgner of JFreeReport, and Mark Hall of Weka attended the festivities.  Some of the highlights included roadmap discussions on Pentaho Metadata, a great Weka brown bag, and lots of discussion on improving communication and process as Pentaho scales its engineering department.

One aspect of the discussion included having Mondrian, our OLAP server, work more closely with Pentaho Metadata.  Over a year ago, Matt Casters wrote a prototype of a Mondrian Schema Generator based on Pentaho Metadata.  This functionality currently exists as an easter egg within Pentaho Metadata Editor, by clicking “CTRL-ALT-o”.  While brain storming, we’ve thought of many different ways that Metadata, Mondrian, and the Pentaho Platform can work together to make a better user experience.