| [modeling-pmc] MDT/OCL Project Scope |
HiThe MDT/OCL project was originally set up to support the OMG OCL meta-model and expression evaluation. The MDT/OCL has also provided an 'example' OCL console supporting interactive evaluation for at least three years.
An MDT/OCL Tools project was proposed to provide editors and code generation, but failed to reach the creation review.
An OCL editor has bounced from GMT/UMLX (SWT-based) via M2M/QVTd (IMP-based) to MDT/OCL examples in Helios, where IP issues required a late redevelopment to be Xtext-based.
Xtext provides a dramatically simpler and arguably much better parser, which merits revisiting many of the internal models so that the official OCL specification validation and well-formedness constraints can be used and debugged rather than paraphrased. However use of interpreted OCL is liable to give very disappointing performance, so Java code generation from OCL becomes important to MDT/OCL itself. Java code generation is also important to capitalize on the support for OCL delegates that EMF added in Helios.
Xtext is an excellent example of the new generation of modeling tools that really add model-driven value. With the aid of QVTo and Acceleo, Java code generation should be much more disciplined and tractable.
It therefore makes sense to add generation of code from OCL and OCL editing within the scope of the MDT/OCL project rather than rejuvenating the failed MDT/OCL Tools project. With the advent of tooling, it would seem appropriate for MDT/OCL to emulate MWE with multiple update site zips. An OCL-Core build at +1 would continue to offer the basic meta-model and evaluation support. A second OCL-Tools build at +3 would provide the functionality currently lurking as MDT/OCL examples without the dilemma of releasing examples at +1 with +3 dependencies.
In summary, is it alright for MDT/OCL to expand its scope to code generation from OCL, to OCL editing and to interactive OCL evaluation?
Regards
Ed Willink