On 20-Mar-14 14:35, gautier_niouzes@hotmail.com wrote:
> Good luck with this ambitious project.

Thank you.

> A challenge is that at some point the AST knows too much about the program.

Ah, you mean things like visibility and private/public access to things, 
right?

> I've developed a similar system 25 years ago (in Turbo Pascal!) for Ada 83.

Heh! so then this is in a lot of ways a re-hash of your work. :D
If you'd like to share tips/, info, or even stories about it I'd love to 
hear.

> The pros: no more parsing, compilation is straightforward, links are readily available,
> the system is able to manage the layout itself, to indent etc.

Much the same that I was thinking.

> The cons: formatting rigidity, and if the programer needs to make a big change,  use
> another package, types and so on, plus all depending changes, he needs to have his
> program "incorrect" for a while. In my system, it meant export to text, rework the
> text and then reimport the changed text. Too cumbersome - and tools like GPS offer
> nowadays navigation tools and autocomplete even in the middle of a such a rework,
> thanks to smart guessing, and all that at text level...

Ah, I see. The export/edit/import is certainly a way to handle that -- I 
was leaning towards a sort of buffer/workspace methodology, keeping the 
inconsistent things separated out of the [main/shared] system.

One of the references "Workspaces and Experimental Databases: Automated 
Support for Software Maintenance and Evolution" (URI: 
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.26.2533&rep=rep1&type=pdf 
) shows an interesting way to use a hierarchical structuring of 
databases to handle consistency/correctness for a project.