EMWDA Working Group on

1

• EMWDA Working Group on
• Lossy Transformations

2
Issue

• (Lossy) Transformations from semantically rich to
  semantically poor languages
• Semantics preserving transformations
• How to give feedback to the users when
  encountering errors / exceptions in
  transformations

3
People Attending

• Ed Willink
• Jim Steel
• Joao Paulo Almeida
• Octavian Patrascoiu
• Dave Akhurst

• Julian Johnson
• Audris Kalnins
• Olivier Le Merdy
• Kevin Dockerill
• Anneke Kleppe

4
Solution

• Discard the term Lossy
• Use two pairs of terms
• Valid versus Invalid
• Partial versus Complete
• Invalid Transformation because
• T is not applicable to all source models
• Gap between spec of T and impl of T

5
Solution (extra)

• Define Equivalence of Systems (and Models before
  Defining what a semantics preserving
  tranformation is.

6
Solution (extra)

• Traceability is not a theoretical issue (large
  machines will be needed in practice)
• Transformations need not be bidirectional, often
  useful tranformations do not have an inverse
  (approx. 90)

7
Solution (extra)

• Pattern Expansion characteristics
• in place source and target model are same,
  same language
• PIM -gt PSM characteristics
• Target model is possibly made up of multiple
  parts, where each part may be written in a
  different language
• Parameterisation of transformations should be
  possible for any type of transformation

8
Advantages of the solution

• Clear terminology

9
Disadvantages of the solution

• Lot of work to be done  

10
To Be Done

• Define Equivalence of Systems (and Models before
  Defining what a semantics preserving
  tranformation is.

11
Roadmap for the solution

• Next workshop!

12
Related Issues

• Taxonomy of transformations
• Different use cases for transformations give rise
  to different characteristics of transformations