Model Driven Generative Programming

1
Model Driven Generative Programming
ECE1770 Trends in Middleware Systems

• Reza Azimi
• February 6, 2003

2
Motivations

• Easy to use
• Platform independent

• Too hard to program and debug
• Too machine- specific

• Still hard to program
• Still platform-specific
• Hard to reuse

3
Model-Driven Programming

• Goal
• To create an abstract (platform-independent),
  complete, precise, and executable model of the
  system.
• Advantages
• Reduced required development effort
• Early system debugging
• Challenges
• Is there such a model?
• How do we verify a model?
• How do we translate a model to code?

4
What should be modeled?

• Data objects
• contents, relations, constraints
• Objects lifetime
• states, and transitions
• Objects interactions
• message exchange, calls
• Objects actions
• computable algorithms
• Concurrency and synchronization
• timing constraints (e.g. Real-time systems)

5
Unified Modelling Language (UML)

• Abstract
• Platform independent
• Readable by non-technical people
• Complete (?)
• Formal action semantics
• Precise model for concurrency and synchronization
• Used widely
• Capable enough for modeling large systems

6
Modeling Data Objects

• Class Diagram

7
Modeling Behaviour

• State Charts

Collaboration Diagrams
8
Modeling Actions

• Action Semantics
• Object and attribute actions
• Selection expressions
• Link actions
• Control structures (loops and if-then-else)

9
Model Checking

• Automatic Verification
• Completeness
• Consistency
• Debugging
• Generating test cases
• Executing the model

10
Model Compilers

• Design Patterns
• Translation Rules
• Pre-built Libraries

11
Design Patterns

• An abstract template for solving a well-known
  problem
• Creational Patterns
• Abstract Factory, Builder, etc.
• Structural Patterns
• Adapter, Bridge, Proxy, etc.
• Behavioural Patterns
• Interpreter, Iterator, Mediator, etc.

12
Translation Rules

• .for each object in O_OBJ
• public class obj.name extends SateMachine
• private SateMachineSate currentState
• .select many attributes related by
  object-gtO_ATTRR105
• .for each attribute in attributes
• private attribute.implType attribute.name
• .end for
• .
• .
• .select many signals related by
  object-gtSM_EVTR303
• .for each signal in signals
• protected void signal.name() throws
  ooaException
• .end for
• .emit to file obj.name.java
• .end for

Highlights of a translation rule that creates a
Java Class
13
Existing Tools

• BridgePoint
• By Project Technologies
• http//www.projtech.com
• Kabira
• http//www.kabira.com

14
Conclusions

• Potentials
• Faster software development
• Early defect removal
• Faster technology adoption
• Limitations
• Model compilers
• Effectiveness
• Complexity