Lightweight DomainSpecific Modeling and ModelDriven Development

1
Lightweight Domain-Specific Modeling and
Model-Driven Development

• Risto Pitkänen and Tommi Mikkonen
• Tampere University of Technology
• Institute of Software Systems
• Finland
• risto.pitkanen_at_tut.fi tommi.mikkonen_at_tut.fi

2
Outline of the Talk

• Full-blown domain-specific modeling (DSM)
  approach is always not feasible

Lightweight Hybrid DSM/MDD
Why apply such an approach and when?
Conclusions
3
Full-Blown DSM is Always Not Feasible

• DSM languages and tools provide significant
  increase of productivity (5 to 10 times e.g. the
  foreword of the proceedings of this workshop)

• but
• they require a high initial investment
• are usually targeted at a certain implementation
  platform
• they pay off only if many software products with
  enough similarities are produced

4
The Generic to Specific Continuum

• Languages are not necessarily either fully
  generic or fully domain specific

generic
domain specific
Language for describing systems GENERIC
Language for describing reactive
systems SOMEWHAT DOMAIN SPECIFIC
Language for describing real-time mobile
robot control systems DOMAIN SPECIFIC
5
Lightweight Hybrid DSM/MDD Compared withDSM
6
Abstract specification of a mobile robotin a
somewhat domain-specific language

• layer abstract_robot is
• constant min_stopped time 0.5
• class Robot(1) is
• mode (STOPPED, MOVING)
• allowed_to_move_at time 0.0
• end
• action move(r Robot) is
• when r.modeSTOPPED and now gt
  r.allowed_to_move_at do
• r.mode MOVING()
• end
• action stop(r Robot) is
• r.mode STOPPED r.allowed_to_move_at
  now MIN_STOPPED
• end
• end abstract_robot

7
Why Apply Such an Approach and When?

• If only one or a small number of similar projects
  are planned, and the high initial investment
  required by DSM cannot be justified
• If precise enough estimates about performance
  issues cannot be done without first writing some
  sort of a specification, and if the results of
  this evaluation affect the choice of archicture
  and/or implementation technologies
• If several different implementation platforms are
  targeted

8
Conclusions

• Lightweight Hybrid DSM/MDD can be applied when
  the high initial investment of full-blown DSM
  cannot be justified
• It can pave the way to a full-blown DSM approach
• any implemented transformations can later be
  utilized in a code generator
• properties required by a DSM language can be
  explored
• On the other hand, it does not provide as high a
  productivity increase as full-blown DSM, because
• modeling language is not as domain-specific
• the transformation process towards an
  implementation requires more developer
  intervention and guidance
• Our example approach is based on the DisCo
  language and targeted towards reactive systems.
  The paper describes a real-time control system of
  a simple robot as an example.