I didn't study it in datail but after looking at the first 10-15% of that, I
didn't see anything that should be a show stopper. The thing I saw mostly
boiled down to augmenting the tree with extra data members and the code to
populate them. If a standard device for attaching extra data could be
chosen, all of that code that isn't language agnostic could be separated
out. The logic edits might require adding virtual functions or something.
On Mar 7, 2011 10:38 AM, "Michel Fortin" <michel.fortin at michelf.com> wrote:
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/dmd-internals/attachments/20110307/95a42d94/attachment.html>

Le 2011-03-07 ? 19:04, Benjamin Shropshire a ?crit :

> I didn't study it in datail but after looking at the first 10-15% of that, I didn't see anything that should be a show stopper. The thing I saw mostly boiled down to augmenting the tree with extra data members and the code to populate them. If a standard device for attaching extra data could be chosen, all of that code that isn't language agnostic could be separated out. The logic edits might require adding virtual functions or something

I can see how DMD could be architectured in a way which is easier to extend. If you're interested in improving things on this front I suppose you could give it a try and then propose it as a pull request. Perhaps Walter will accept it.

... or perhaps he won't. Some people tried to convince him to change the .c extension to .cpp at one point, and it never happened. So I have some doubts about proposing a refactoring of the internals.

-- 
Michel Fortin
michel.fortin at michelf.com
http://michelf.com/