Lecture 02

MDSD — Lecture 2. (18/02/2015)

MDSD Principles – Languages and Models

3 pillars: Language, Model, X

Model

Representation of a system.
Only a selection of features/properties of the modelled system.

• reduced complexity
• selected the important features/properties

Concepts

• code: coding language
• model: modelling language, frameworks, etc.
• metamodelling language: a common language to describe/capture/specify models of a domain
• code -> bytecode: When you use a compiler to generate bytecode (use a higher level of abstraction), you don't want to modify the bytecode itself. The more mature a compiler is, the more it could optimize the code. You may loose optimization by using an abstract layer. But (!) your productivity may increase!

---

Modeling Languages

Domain Specific Languages (DSLs): languages that are designed specifically for a certain domain or context. DSLs have been largely used in computer science. Examples: HTML, Logo, VHDL, Mathematica, SQL

*Often you knows the domain of your model, so you could crate less expensive language. But you will gain analysability. *

General Purpose Modeling Languages (GPMLs, GMLs, or GPLs): languages that can be applied to any sector or domain for (software) modeling purposes. The typical examples are: UML, Petri-nets, or state machines

Domain Specific Modeling Language

• Abstract Model: "The Model" -- the core for most of the tools we develop. The tooling is based on this representation.
• Concrete Model: Graphical/Textual view of the instance model. You don't need to present every information every time.

Types of Models

1. Modelled properties
   • static models: structural shape, connection, etc. class diagram, ER-diagram.
   • dynamic models: the dynamic behaviour -- sequence diagram, BPMN, compiler event processing
2. Purpose/usage
   • traceability models: follow the chain of elements and bind the requirements to a part of the system meaning that it is responsible for a specific requirement. And the other way around as well.
   • execution trace models: -- sequence models. The specific run of the system
   • analysis models
   • simulation models

Metamodeling

To represent the models itself -- means of abstraction, highlighting the properties of the models. MOF -- 4 levels -> there can be more, only the concept of metamodeling is important.

Model Transformation

E.g. transforming COBOL legacy code to newer languages.
The concept of code transformation, queries exists in the world of models. And between them! Model-to-Model (M2M), M2T (Text), T2T, T2M

---

---

First, languages are represented by their metamodel, a precise specification of the language. The concrete model is formalized with the language/metamodel. Then the translation rules between the languages/models is described. This expresses, how any of the instance model elements are to be transformed. Last, these rules are passed by auto transformation engine and the transformation is executed.

---

---

Domain modeling

• metamodels: Precise specification of domain language. Language for describing the abstract syntax. Aggregation is providing an important concept: container hierarchy.

Why generalisation instead of inheritance?

• in programing, we use top-down concept
• in modelling we use the opposite, bottom-up

The evolution of the language describes this.

##Type conformance

Each model element is an instance of a metamodel element.

• Direct type: there is no other type exists lower in the hierarchy
• Indirect type: superclass, there is a lower type.

"is a" could mean two things:

• classification (instance of): not transitive.
• generalization (super type of relation): transitive.

In modelling multiple inheritance is possible.
Restriction (usually): each model element only has single direct type. But multiple classification, from multiple domains is allowed. (In class, there are some exceptions.)

Type conformance of references

Type conformance (formally): A link in a model is type conformant, if type(src(link)) is subtype of src(type(link)), type(trg(link)) is subtype of trg(type(link)).

Informally:
The type of the source object is a subtype of the source class of the link’s type. The type of the target object is a subtype of the target class of the link’s type.

Container hierarchy

• single root element
• if an element is not contained it won't be saved/serialized
  • no circular containment
  • nodes themselves can be in 2 circle in the metamodel.

Multiplicity restrictions

upper and lower bounds:

• 0..1
• 0..*
• 1 (usually containment)