Re: parser improvement suggestion

[asmwarrior]

On 2011-6-8 15:52, Jean-Marc Saffroy wrote:
> > Mostly, I think you need to use the "recursive decent parser", so you create a
> > lot of functions like:
> > HandleClass();
> > HandleNamespace();
> > HandleWhile();
> With the approach you suggest, is it still possible to parse things like
> this?
>
> #ifdef foo
> class foo {
> #else
> class bar {
> #endif
> ...
Yes, currently, codeblocks' parser can handle this kind of preprocessor 
directive. But there are some limitations.
because
#if Const Expression
It need to do Const expression calculation and Macro expansion. Mostly, 
it need a Macro symbol table look-up. and note that each translation 
unit should have its own symbol table, so this is a time consuming work.

Current way, we just simplified this, if a symbol is not found, we just 
set if to "0", so we can always take a conditional preprocessor branch, 
although it is not 100% correct branch.

ctags has the same way to handle conditional preprocessor directives.
> Global felt quite mature to me the last 10 years I used it. :)
> Cheers,
> Jean-Marc
Well, I'd admit that if your code is very complex (contains a lot of c++ 
template and template partial specialization), It is hard to parse them. 
This time, a full compiler front end can solve the issue, but a full 
compiler is always running slowly.:-)
Clang compiler has an index feature, but gcc currently doesn't have such 
feature.

asmwarrior