View Full Version : How to use PME output to generate documentation?

07-13-2009, 03:58 PM
Hi all!

I am in charge of finding a way to use the metadata generated with PME to create data documentation reports. Simply put, as I already have the metadata telling from where does came every field in my database, why not run a report where I can see it?

Like, report "Source Data" would give a list comprised of columns telling the database, the column, the field name and type.

I have browsed the data stored in a db-persisted CWM repository (Postgres, using the pimper repository configuration to save to data base), but I couldn't make sense out of the <Binary> field... I have browsed the web for quite a couple of hours and not found anything on how to navigate those structures and see the metadata, nor found a mention to anything in Pentaho fora.

Thanks and best regards,


07-13-2009, 04:04 PM
The CWM repository is not meant to be read by humans and has no sensible, explainable structure (Sort of). The database structure will change when the contents of your metadata model changes.

It might be a lot saner to either
* use the metadata runtime model for this excercise.
* wait for Citrus and PMD 2.0, which drops the CWM madness and uses a more simplified repository.

If you have to have a solution today, I would fire up a Java-IDE and implement a set of classes that produce TableModels. Then you can use the "Static-DataFactory" to call these classes to use the produced tablemodels in the reporting engine. I would rather try to save the world, solve world-hunger and build a perpetuum-mobile before I would contemplate to dig into the SQL generated by the CWM-repository implementation.

07-13-2009, 04:17 PM

Thanks a lot for your prompt answer!

Beeing still a lit bit noob, two extra questions:

1. Is the metadata.xmi the runtime model, or do you refer to a sort of in-memory expansion of the CWM used by BI Server to generate the SQL?

2. How long until PMD 2.0 (did you drop the "Editor" from the name?)?

Best regards,


07-13-2009, 05:48 PM
The MetaData runtime I'm talking about is the API contained in the "pentaho-metadata" project ( http://source.pentaho.org/svnroot/pentaho-metadata/ ). It allows you to access the business models and business-tables etc without having to deal with either insane databases or insane xml files.

The *.xmi files are an XML representation of the CWM model. Therefore they are as ugly and complex as the repository. (Well, not that bad. Its hard to top the SQL-model of the Netbeans-CWM-engine :D)

PMD 2.0 is in the forge, will be part of Citrus and should be integrated into the reporting engine in the next weeks. Pentaho-Metadata is the backend system - the stuff that does the work - while the Pentaho-Metadata-Editor is just an UI to define models. Forcing users to manually edit XMI files is considered torture in some countries. You dont need the UI to work with models from within a Java-Application.