Re: dabbrev.sl

[Günter Milde <g.milde@xxxxxx>]

Hi John,

excuse me for cc-ing to the list as well, I have 2 reasons for this:
 
 - other users can bring in their opinion and wishes in a prerelease state

 - I am afraid, previous mail from me to you got silently eaten up by
   spam-preventing software.

On Sun, Mar 14, 2004 at 01:54:35AM -0500, John E. Davis wrote:
> Hi,
> 
> I am currently testing your version 2.2 of dabbrev.sl for inclusion
> into the next release. I have a few questions:
> 
> The function get_buflist takes a scope argument:
> %     0 = current_buffer,
> %     1 = visible_buffers,  % including the current
> %     2 = open_buffers_of_same_mode,
> %     3 = open_buffers
> %     4 = other_visible_buffers,  % excluding the current
> %     5 = other_open_buffers_of_same_mode,
> %     6 = other_open_buffers
> 
> I do not understand some of these values, in particular 3, and 6.
> What is an "open_buffer", and why are there others?

In my wording, "open_buffers" means simply all buffers. You are right to
nag this, as saying "all buffers" is definitely more clear. (I suppose I
came to this term thinking of open files).

> My feeling is that scope should be a bitmapped value, but I have not
> thought too much about it.

I always find bitmapped flags harder to understand and handle than a simple
list.

I think, bitmapping might be advantageous for "orthogonal" (i.e.
independend) features. This could be used with dabbrev (but whether it is
really an advantage is not clear to me

   bit 1    current buffer / all buffers

with bit 1 to TRUE there are 2 independent filters

   bit 2    only visible buffers
   bit 3    only buffers of same mode
 
> Also, instead of assigning get_buflist to a key, perhaps it would be
> better to have dabbrev take a prefix argument, e.g.,
> 
>    ESC 4 Ctrl-A

IMHO, this should not be instead but rather as a separate option.
I use the prefix argument rather seldom but can imagine it to be usefull
in some circumstances or for other users ==> implemented (in v 2.4)

> which would be equivalent to expansion with scope=4.  It might even be
> useful to use a buffer-local variable for the scope, and default to using
> Dabbrev_Default_Buflist if such a variable does not work.

This is prepared in v 2.4 but not activated as it would need the function
get_blocal(name, default) from sl_utils (or need to emulate it). Simply
uncomment the lines
	% buflist_scope = get_blocal("Dabbrev_Default_Buflist", 
	% 		  	      Dabbrev_Default_Buflist;
when get_blocal is available by default...

> Finally, is there a way I can define a buffer that consists of a list
> of words and always have that buffer added to the list of completion
> buffers?

Yes, of course. One of the design-goals of the new dabbrev is
extensibility. 

If you only want expansion from "*wordlist*", it is as simple as
  % setkey("dabbrev(\"*wordlist*\")", "^A");

But your question was to add the wordlist. So you need to define and bind
a new function. The following example assumes the wordlist to be in a
buffer "*wordlist*" (of course it could also be an optional argument or
whatever):

define dabbrev_with_wordlist()
{
   variable buflist = "*wordlist* ";
   variable buflist_scope = prefix_argument(-1);
   if (buflist_scope == -1)
    buflist_scope = get_blocal("Dabbrev_Default_Buflist", 
    		                Dabbrev_Default_Buflist;
   buflist += get_buflist(buflist_scope);
   dabbrev(buflist);
}

The buflist-computation is done for the first expansion only, as
the function name starts with "dabbrev".
(c.f. the keypress check loop in dabbrev():

   do
     {
        ...  % expand
        
	% Check next keypress: 
	(type, fun) = get_key_binding();
     }
   while (andelse {type >= 0}                    % key bound ...
	{is_substr(fun, _function_name) == 1});  % ... to "dabbrev.*"

)


I just did an update of the dabbrev mode at 
  http://jedmodes.sf.net/mode/dabbrev
The most recent version of dabbrev is now 2.4, with substantial enhancements

   2.2.1 bugfix: invalid mark when buffer of last
                 expansion killed (P. Boekholt)
   2.3 Dec 2003 prevent flooding the undo-buffer (using getkey for
                subsequent invocations)
   2.3.1 2003-12-05 replaced unget_keystring() with buffer_keystring()
   2.4   2004-03-15 dabbrev() takes a prefix argument for the 
                    buflist-scope (this is checked in dab_reset)
                    clearer documentation (open_buffers -> all buffers)
                    (hints by J. E. Davis)



Günter


-- 
G.Milde at web.de