Shiny New Languages for AOM

1
Shiny New Languages for AOM?
Jeff Gray
MODELS 2007 Panel The Future of Aspect Modeling
Will MDE Absorb it?
2
Robert Floyd The Paradigms of Programming

• To the designer of modeling languages, I say
• Unless you can support the paradigms I use when
  I model, or at least support my extending your
  language into one that does support my modeling
  methods, I dont need your shiny new languages
• To persuade me of the merit of your language, you
  must show me how to construct models in it.

1978 Turing Award Lecture
Original focus on programming quote modified to
modeling.
3
Case Study Solved with Aspect-Oriented
ModelingCrosscutting Constraints in
Real-Time/Embedded Models
IEEE Computer, February 2006
4
Doesnt it just involve a specific kind of
modeling?

• Elaborationists versus Translationists
• Some view AOM as just a bunch of notations with
  little or no tool support
• AOM as a noun
• Often interested in traceability concerns to code
• Opinion Majority view
• Others view AOM as a means toward improving the
  modeling process itself, with model weavers
  integrated into a modeling tool
• View aspect modeling as a form of model
  transformation
• AOM as a verb
• Opinion Offers best opportunities for AOM
• different meanings to the term

A summary of these views is presented in
(Cottenier et al., MODELS AOM Workshop, 2006).
5
Current AOM Focus Tilted toward General-Purpose
Languages

• After 6 years and 11 AOM Workshops
• Growth (depth) and application (breadth)
  unimpressive
• Most papers describe a general-purpose aspect
  language for a general-purpose modeling language
• Number of papers specific to UML and
  general-purpose descriptions at last four AOM
  workshops
• 2006-AOSD 8 of 8
• 2006-MODELS 7 of 10
• 2007-AOSD 7 of 8
• 2007-MODELS 6 of 7
• See http//www.aspect-modeling.org/

6
Expanding the purview of AOM

• At the code level, DSALs are gaining more
  interest (original focus of AOP)
• Very little investigation into the combinations
  of DSAL, GPALxDSML, GPML at modeling level
• Are there benefits in expanding the
  general-purpose view at both levels?

7
Does it just involve a special kind of model
transformation?

• Consider a view that treats a transformation
  language/engine as machine code to realize
  higher level concern transformation in a separate
  language
• At the code level

FOP
AOP

Higher level Descriptions
Code Refactoring Browser
Concern Translator
Program Transformation Engine (e.g., DMS,
ASFSDF, TXL)
Lower level Mechanisms
8
Does it just involve a special kind of model
transformation?

• At the modeling level

FOM
AOM

Higher level Descriptions
Model Refactoring Browser
Model Compositor
Model Transformation Engine (e.g., ATL, VIATRA2,
C-SAW, GReAT, ATOM3)
Lower level Mechanisms

• Some observations
• AOM is a type of MT, but most end-users would not
  want to use a MT engine directly
• Layering an aspect modeling language on top of an
  MT engine provides an opportunity for improved
  abstraction in the intent of the aspect
  description

9
Will MDE absorb AOSD in the end?

• May depend on which camp you live in
• Elaborationist/Translationist
• Robert Floyds Plea
• Extensions to incorporate desired paradigms
• If you believe Everything is a Model
• No need for aspects and weavers at the code level
  - code generation from the model can generate all
  crosscutting concerns i.e., aspect weaver built
  into codegen