Re: [anjuta-devel] How to get the type of member that is a pointer using symbol-db

[Massimo Cora' <maxcvs email it>]

Hi,

On 07/29/2013 01:08 PM, Marco Diego Aurélio Mesquita wrote:
> On Sun, Jul 28, 2013 at 1:54 PM, Massimo Cora' <maxcvs email it> wrote:
> > The best choice would be to replace totally ctags in view of a clang based
> > parser.
>
>
> That would be a too much work for the feature I want to implement. For
> now, I'm thinking on a workaround. Suppose that I want to get type of
> "help_window" that is declared as:
>
> GtkWindow *main_window, *help_window;
>
> Currently, it is easy to correctly get the offsets of GtkWindow and
> help_window. So, now we'll store the piece of the string between
> GtkWindow and help_window in a buffer. Such buffer would then contain:
> " *main_window, *". To get the correct type, we'll scan this buffer
> backwards and stop scanning once we find a "," or we reach the
> beginning of the buffer and store each non-commented char in another
> buffer. This other buffer would then contain: " *". This other buffer
> can then be concatenated with the first buffer to correctly get the
> type of "help_window": "GtkWidget *".
>
> Do you think this approach is acceptable?


in short, no.
The best approach is indeed the lexer one (where this method is already 
implemented).
The top-down method - i.e. proceeding by simple regular expressions - 
will fail once the complexity of the expression raise.
That's why cxx-parser isn't working good.

Regards,
Massimo