On 3 déc. 2009, at 18:57, Stéphane Ducasse wrote:


On Dec 3, 2009, at 10:46 PM, Tudor Girba wrote:

Hi Cyrille,

The MSE Import Wizard is a good step forward, but it has a problem: The entity types are hardcoded. This is a big problem because we should be able to load any entity.

It would be better if that list would be gathered from the types from MooseModel meta.

ImportingContext should not be hardcoded either, but it should be computed from the same repository by traversing the properties.

For example, in FAMIXMethod>>parentType says that FAMIXClass is required for FAMIXMethod. In the same time, FAMIXMethod is not required for a FAMIXClass because the property FAMIXClass>>methods isDerived. Basically in this way, we should be able to create the graph dynamically and thus we can use the mechanisms for any meta-model.

What do you think?

yes
the design of ImportContext predates the one of FAME and indeed we should be able to have any metamodel and their dependency
Now it may happen that we will found some hidden (manually expressed in the importingcontext and importer) dependency
that are not expressed in the metamodel: famix comment, literal.

A related issue is to encode the importing context in the mse. So that we know what is not in the mse.
See comment #2 in the issue

--
 Simon