ActiveFacts

The Base API (in Ruby) is complete, representing static fact-oriented information models. The Query API is under construction at present; future additions will include events, actions and agents (roles).

The Semantic Model Storage schema (congruent to the Base API), allows any model to be stored, retrieved, and manipulated in a relational database. The Semantic Model Storage module is not yet integrated, however, with the Base API.

Constellation Query Language is available as an export format from the base API, and as a language recogniser (parser). The parser back-end is incomplete at present - complex constraints are not yet handled.

Queries in CQL span tables transparently, crossing conceptual relationships (fact types) without explicit database joins, and return structured data in a nested-relational form (called constellation form). For every user action in an application, a single query can be defined that provides all the data required to respond to that action. There's no need to fetch this, that, and the other. All required results are fetched at once, which improves application structure, performance, and even concurrency.

ActiveFacts can currently read models from NORMA, which are in elementary form, and it can reverse engineer the schemas of existing databases in SQL Server, DB2, Oracle, and open-source databases not yet tested including MySQL, PostgreSQL, etc. Reverse-engineered models are still in composite form - the mapping to elementary form is incomplete. Reverse engineering does not yet analyse actual data.

ActiveFacts relies on the database adapter infrastructure of ActiveRecord, but needs additional fixes in the form of DRYSql (with some further patches from me if they are not yet incorporated into the released version), and the Composite Primary Keys gem (again, with further patches if not already incorporated).

APRIMO allows the definition of vocabulary aspects as well as the addition of entity and value objects and the fact types that link them. Subtyping, data types and units, and complex constraints are still to be implemented.

The Relational Compositor has not been started, and neither has the Semantic Runtime.

NORMA is the latest in a line of products stretching back two decades which implement the Object Role Modeling (now version 2) fact-based modeling notation. NORMA is a design tool that is open-source, but runs as an extension to Microsoft Visual Studio 2005 - Professional Edition (the Express Edition doesn't have the ability to load plug-ins). You can download the latest version of NORMA (and source code, if you wish) from SourceForge.

ActiveFacts is loading all the NORMA models I can find and a few I built myself, with the exception of sample fact data populations.

ActiveFacts also contains a Metamodels directory which contains a diagram and SQL DDL of the relational schema of the Semantic Model Storage schema. The schema also exists in a form modified to be suitable for Ruby on Rails, which is the form APRIMO uses. The Base API can't yet save or load models into APRIMO database.

My meta-model merges mandatory, uniqueness, and frequency constraints into one "Presence Constraint" object. This is because all three constraint types have to do with the number of times a role or role sequence may be played.

For object types (Entity, Value and Objectified Facts), I have separate classes in the Ruby API, but these are absorbed into one ObjectType table in the SQL. This allows proper referential integrity from the Role table. A consequence is that in sample populations, the entity instances and value instances are absorbed into one Instance table as well, with a nullable Value field.

Finally, my API supports sub-models (known as Aspects or as Vocabularies), and more than one named Population, so that you can have a sample population used for model validation, another as a reference population, and even another for the production database itself.

To look at ActiveFacts, install Subversion (either from the home site or one of the bundles like Tortoise SVN, cygwin, etc), and issue the checkout instruction on the RubyForge web-page listed above. You can also browse the files online at the Rubyforge source code repository browser.

My NORMA models are available here. You can read about them in the tutorial.

Most of the models are trivial demonstrations of some interesting feature of fact-based modeling. The exceptions are the Orienteering model and the Oil Supply model. The Orienteering model covers the orienteering system currently hosted at both street.orienteering.com.au and the Victorian Orienteering Association. I compete in these events, and have even built scoring hardware for it using low-power microprocessors (not yet in use). The point being that this is a real life model, which will continue to be my main demonstration model. The OilSupply model is extracted from an actual schema from a major oil company. It shows regional demand predictions against production commitments, taking transport availability into account.

Though the provided models are mostly simple, ActiveFacts has successfully reverse-engineered some large commercial database schemas, including that of ManageSoft, with all sub-products installed (265 tables).

If you have Ruby installed you can load a NORMA model using the command:

      ruby -I lib bin/read_norma.rb -C MyModel.orm

read_norma reads a NORMA file and emits CQL.

With the Polyglot gem and the edge version of Treetop (downloaded using Git from the repository), you can also parse CQL - but not yet into any useful form.

If you also have ActiveRecord, DRYSql (patched with the patches provided), and CPK (composite primary keys), you can reverse engineer your own schema:

      ruby -I lib bin/reflect.rb --dsn=Orienteering \
	  --adapter=sqlserver --mode=ODBC --user=sa \
	  --password=whatever

using whatever ActiveRecord connection options are appropriate to connect to your database.