MetaModelling

MDA

• Model driven architecture effectively defines a philosophy and loose process for developing systems.
• It is model-centric, everything is a model: system, requirements, platform, the transformations that are used to move from one model to another.
• System models are separated into two parts:
  • Platform independent.
  • Platform dependent.
• Mappings and transformations are needed to include platform specific details - they should be modelled to.

MDA Models

PIM

• Platform Independent Model.
  • Formal specifications of a system (business and service model) that abstracts away technical detail.
  • e.g. a Billing System in UML.
• "Model once, generate everywhere."

PSM

• Platform Specific Model.
  • Business and Service Model expressed in terms of the specification model of the platform.
  • e.g. a Billing System expressed in UML Profile for CORBA.

Platform Model

• Model of platform component constructs e.g. of Eiffel, C#, EJB etc.

MDA Technologies

MOF

• All languages used in MDA will be based on MOF.

QVT

• Query-View-Transformation: a standard for modelling transformations between languages.
• Expresses relationships between model elements and constraints on when the transformations can be fired.

XMI

• Tool interchange.

Metamodel

• A metamodel is used to define the characteristics of a valid model.
• A model conforms to its metamodel.
• A model is an instance of its metamodel.
• e.g in UML, a Class is an instanceof Classifier.

Uses

• To define the syntax and semantics of a language.
  • UML is a language. So the document that defines UML is the UML Metamodel.
  • e.g. a Class IS-A Classifier, an Association IS-A Relationship.
• To explain the language.
• To compare languages rigorously.
• To specify requirements for a tool for the language.
• To enable interchange between tools.

Tools to build Metamodels

UML/MOF

• Advantages: Promotes understandability, deals with reasonably large structures, CASE tools exist.
• Disadvantages: Meta circularity, semantic checking.

Formal Specification Language

• Advantages: Avoids meta circularity, proof system for semantic checking : possible to construct proofs of consistency/soundness.
• Disadvantages: Difficult to understand, requires expertise.

Executable Language

• e.g. Ocaml, Eiffel, JML.
• Advantages: understandable, works for large and small model, pre-existing tools for compilation, metamodels can be tested and simulated easily.
• Disadvantages: Risk of losing abstraction level, transformation from modeling language to implementation needs to be transparent, misconception that programs and model must be in different languages.

Concrete Syntax

• Textual (code) or Visual.
• Text good for complex expressions. Beyond a certain size, difficult to comprehend and manage.
• Visual good for expressing large amounts of detail in intuitive form.However works only till a certain level of detail.

Abstract Syntax

• Describes a vocabulary of concepts in a language and how they may be combined to create models.
• Concerned strictly with form/structure. Omits presentation, semantics.
• e.g. State Machines.
• Key concepts: states, transitions, events.
• A State Machine Language must be able to describe these concepts.
• We can then model form and structure, transition associated with 2 states etc.
• However nothing about presentation or semantics.

OMG Metamodelling Stack

• 4) M3 - Meta-Meta Model - MOF
• 3) M2 - Meta Model - UML MetaModels and other metamodels
• 2) M1 - Model - UML Model
• 1) M0 - Real world - A particular use of a UML model.