CEDET

Hi Ville

>>> Ville Vainio <vainio24@...> seems to think that:
>I need autocompletion, and following the guidelines (intellisense page) on
>the cedet webpage doesn't do what I need. What I need to do is:
>
>1. Create the database from 1e6+1 C++ header files in an SDK directory
>(i.e. parse all the .h files under one directory, recursively)

In semantic 1.4.x (the version currently in XEmacs) the only way to
do this is to pull up every file individually, and wait for Emacs to
parse the file.

In the beta version (recently released) there is a shell script that
will do it for you.  (semanticdb.sh)

>2. Use this database for autocompletion in my own projects, i.e. when I
>type:
>
>LibClass c;
>c.
>
>And execute some semantic-show-possible-completions:ish command, I would
>get the list of methods and their signatures in speedbar or a new buffer.
>I don't really care about actually completing the method, typing is not a
>problem, remembering what the methods are is.
>
>If I type one letter, autocompletion tries to complete it. But I really
>have no use for this, since I possible don't even know what letter to
>type. If I don't type any letters after the period I get a zillion
>possible completions, most of which are not the methods of the class.
  [ ... ]

There is no simple way to do this with semantic 1.4 without hacking
some lisp code.

In the beta version it is possible to create a "system" database
using M-x semanticdb-create-system-database.  After creation (which
can be a bit slow) it will auto-load the database every time you restart
Emacs, and it will be included for all searches in files of that type
(such as C or C++ files.)

I do not know the current state of autocompletion stuff in the beta.
It should work, but I haven't evaluated it lately.

Eric

-- 
          Eric Ludlam:                 zappo@..., eric@...
   Home: http://www.ludlam.net            Siege: http://www.siege-engine.com
Emacs: http://cedet.sourceforge.net               GNU: http://www.gnu.org

On Sat, 29 Nov 2003, Eric M. Ludlam wrote:

> >(i.e. parse all the .h files under one directory, recursively)
>
> In semantic 1.4.x (the version currently in XEmacs) the only way to
> do this is to pull up every file individually, and wait for Emacs to
> parse the file.
>
> In the beta version (recently released) there is a shell script that
> will do it for you.  (semanticdb.sh)

Does the script work w/ Cygwin on windows? Additionally, is it easy to
install the beta on windows? It seemed to use makefiles and in general be
more unix-oriented.

> I do not know the current state of autocompletion stuff in the beta.
> It should work, but I haven't evaluated it lately.

What method do you use to know what member functions are implemented,
then?

-- 
Ville Vainio   http://www.students.tut.fi/~vainio24

>>> Ville Vainio <vainio24@...> seems to think that:
>On Sat, 29 Nov 2003, Eric M. Ludlam wrote:
>
>> >(i.e. parse all the .h files under one directory, recursively)
>>
>> In semantic 1.4.x (the version currently in XEmacs) the only way to
>> do this is to pull up every file individually, and wait for Emacs to
>> parse the file.
>>
>> In the beta version (recently released) there is a shell script that
>> will do it for you.  (semanticdb.sh)
>
>Does the script work w/ Cygwin on windows? Additionally, is it easy to
>install the beta on windows? It seemed to use makefiles and in general be
>more unix-oriented.

I do not have a windows box to test on, so I do not know.  It doesn't
do anything other than things supported by bash or Emacs so it should
work w/ Cygwin.  The build process also works with Cygwin and XEmacs,
that has been tested by others.

>> I do not know the current state of autocompletion stuff in the beta.
>> It should work, but I haven't evaluated it lately.
>
>What method do you use to know what member functions are implemented,
>then?

The process captures the current symbol, finds it's definition, looks
up the type for that symbol, and assumes that fields within that
structure are what you are completing against.  Of those elements, it
also tries to match the type needed in that location (function arg,
or variable/field assignment) and provide only those entries that
satisfy that requirement.

By not work, I meant it may throw signals and errors.  The logic
should still be good.

Eric

-- 
          Eric Ludlam:                 zappo@..., eric@...
   Home: http://www.ludlam.net            Siege: http://www.siege-engine.com
Emacs: http://cedet.sourceforge.net               GNU: http://www.gnu.org