It might be interesting in some cases to have very detailed moose models, at the level of single instruction (there is a loop, its test compares variable x to 5, ...)
That would generates incredibly huge models, of course, but for localized tasks it could be useful.
Was there some work along these lines at any point in time?
Or are we the only one to think it could be useful?
nicolas
On Thu, Sep 6, 2012 at 8:56 AM, Nicolas Anquetil Nicolas.Anquetil@inria.fr wrote:
It might be interesting in some cases to have very detailed moose models, at the level of single instruction (there is a loop, its test compares variable x to 5, ...)
That would generates incredibly huge models, of course, but for localized tasks it could be useful.
Was there some work along these lines at any point in time?
Or are we the only one to think it could be useful?
nicolas _______________________________________________ Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Personally, I would like to be able to track the movement of data through the code - where it was first entered, what methods/classes/interfaces/whatever it passed through, where it finally ended up. I've been meaning to ask if the moose model (or MSE?) was conceived for such analysis, or is it targetted at a larger scale view (which is what I think I've seen).
So, thanks for asking. That's roughly what I'd like to know, too.
-Chris
Hi Nicolas,
Of course it can be interesting.
Are you hinting towards AST information? Or do you have in mind a more relaxed, but more general model?
If it is the former, in Smalltalk the AST is already available :). In other languages you would have to implement it.
In projects where PetitParser is being used to import sources, the AST is most likely provided (because for less than it is needed for import). For example, I guess that the Delphi implementation will offer it (right, Stephan? :)). Also, I built a couple of such projects for custom languages and there I have a full AST that is available on demand from the FAMIX model.
For other languages, like Java, we would need to serialize this AST from the external parser. It would definitely be a very welcome addition.
As for the size of the model, the AST could also be loaded on demand for a restricted amount of entities. For example, suppose that each source file (or source entity) would have an associated AST file. In that case, we could load the AST on demand just for the interesting parts.
Cheers, Doru
On Thu, Sep 6, 2012 at 5:56 PM, Nicolas Anquetil Nicolas.Anquetil@inria.fr wrote:
It might be interesting in some cases to have very detailed moose models, at the level of single instruction (there is a loop, its test compares variable x to 5, ...)
That would generates incredibly huge models, of course, but for localized tasks it could be useful.
Was there some work along these lines at any point in time?
Or are we the only one to think it could be useful?
nicolas _______________________________________________ Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
-
Chris Cunningham -
Nicolas Anquetil -
Tudor Girba