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

[Massimo Cora' <maxcvs email it>]

On 07/30/2013 09:07 PM, Sébastien Granjoux wrote:
> Hi Massimo,
>
>
> Le 28/07/2013 18:54, Massimo Cora' a écrit :
> > I remember that I wrote a patch against ctags to get the return type of
> > a function, and that I implemented a way to get the pointer type of a
> > variable in the cxx-parser, using some code from CodeLite - and a lexer.
> > You can look at parser-cxx plugin.
> >
> > Anyway please note that currently symbol-db won't give you the pointer
> > info because its ctags-backend doesn't provide it.
> >
> > The best choice would be to replace totally ctags in view of a clang
> > based parser.
>
> I'm interested to improve this part.

Yeah I had also to help you doing this but unfortunately I'm flooded 
with other inputs at the moment, sorry.

I have thought about moving the
> code reading the output of ctags in its own plugin allowing to easily
> use a different backend. It will make the code of the symbol-db smaller
> and it will be easier to choose a different backend. But I'm not sure it
> worths it. Else as we use anjuta-ctags anyway, we can just improve
> anjuta-ctags. What do you think?

This is a delicate part.
IMHO I would suggest to drop ctags at all. It's too weak to have some 
serious symbol parsing. Secondly, it appears to be a dead project for a 
long time now - so no help from its devs.

I would point all on clang, but there are a couple of things to verify 
before analysing a porting:

1. Does it support static symbol listing in a ctags fashion? If yes 
we'll be safe.

2. The point 1 can be done fast? So fast that we can consider it nearly 
real-time? Ok, if yes we could drop the db thing, at least for 
small-medium projects. For the big ones there can be some scalability 
problems and we have to think about some cache system for the scan - but 
maybe clang has it already?

3. The completion. I'm sure I've seen a clang api that list the 
suggestions given a text-file and a column-row parameters - we could 
trash cxx-parser and swap it with the clang beast.

4. Other languages than C/C++. Well, is it worth keeping the old 
fashioned ctags to have a really tiny support for some symbols here and 
there in the source files? IMHO we have to take a decision here. We 
should have some numbers on how many people uses Anjuta and for what 
languages. Python and Javascript for instance have their compilers 
installed into Anjuta. Why don't we let them provide the symbols given a 
common-interface?

Regards,
Massimo