[Geany-devel] Geany FTW - better autocompletion

[Colomban Wendling]

Le 17/11/2010 23:45, Enrico Tröger a écrit :
> On Thu, 18 Nov 2010 09:34:03 +1100, Lex wrote:
> 
>> On 18 November 2010 04:05, Colomban Wendling
>> <lists.ban at herbesfolles.org> wrote:
>>> Le 17/11/2010 15:54, Nick Treleaven a écrit :
>>>> On Tue, 16 Nov 2010 17:32:01 +0100
>>>> Simone Pellegrini <spellegrini at dps.uibk.ac.at> wrote:
>>>>
>>>>> I really like geany as it uses scintilla, the only feature I feel
>>>>> is missing is the intellisense like code competition for C++. I
>>>>> would like to ask if anyone is working on a better code
>>>>> competition for C++. I am not talking about ctags but something
>>>>> more advanced that can give me code hints on the fly without the
>>>>> need of rebuilding the index.
>>>>
>>>> I don't think anyone's working on that ATM. There was some work to
>>>> make Geany able to parse tags from memory instead of from disk, but
>>>> it wasn't finished.
>>> Well, I had an attempt a while ago to achieve this [1], but I didn't
>>> finished it already. I used a small library I written for the
>>> occasion, called MIO [2], that replicates C's file IO functions, but
>>> allowing to choose between memory or file IO. This makes the
>>> tagmanager update quite easy.
>>>
>>> BTW almost everything is done, but it needs someone to check whether
>>> it actually work in memory and don't do some FILE I/O (if Geany
>>> correctly demands memory IO, and if the tagmanager doesn't do extra
>>> file IO). Unfortunately, I wasn't sure how to check for this...
>>> (just thinking I may add a debug statement in my IO object creation
>>> functions)
>>
>> You could try using strace to check what file IO is being done.
Yes, may be a good idea :) Hope there is not toooooooo much open calls
in other places... (I'm a bit scared though :-')

> Yup. Or simply use stat(1) to check the atime (provided the filesystem
> in use supports atime) of the opened file.
IIRC, the problem was that the tag manager seems to use some tempfiles
in some cases. And such tempfiles are not acceptable for an in-memory
parsing, from which we expect real-time parsing.
But it's probably easy to at least move these temp files to memory
chunks. Need to be investigated deeper.

> Though, Lex' suggestion is more secure and can be easily grepped :).
> 
> 
>>> But I can try to bring this upper in my todo-list, and somebody with
>>> better knowledge of the tagmanager (Enrico? Nick? :D) may want to
>>> help :)
> 
> knowledge? me? no no.
> 
> Seriously, this looks interesting!
> Great idea to create MIO!
Thanks, and happy to see you like it :) (but remember, we spoke of this
a while ago ;))

> Since MIO seems quite small (code-size), we maybe even could distribute
> as static lib with Geany at the beginning and then a bit later once it
> used and got stable, separate it as a standalone library.
I actually wrote it partially thinking about using it as a static
library because Geany's policy is not to have much dependencies. But I'm
of the "share everything" school, so I feel even better with it as a
shared library :)
About the stability, even though it would be completely blind to think
it is stable regarding the few real-world tests I made, I think the test
suite should help a lot for this -- and since all tests passed (and
still passes I guess!), it should work quite correctly :)

> If I only would have more time, I'd like to dig into this...
Almost the same here... but I'll try to move it up in my todo-list.

Regards,
Colomban