Re: [Fwd: Re: beginning-of-defun]

[Andreas Roehler]

Richard Stallman schrieb:
> Thanks for starting to explore this issue.
>
> Because it only finds `defun' calls, it fails to find other constructs
> that define functions or macros.  

This would be the task of `beginning-of-form' - a more
abstract utility.

> Also, it would get rather confused
> when dealing with top-level forms that don't define functions at all:
> it just skips them.
>
>   
Imagine to make a `beginning-of-form' working the
way `thing-at-point' does: (beginning-of-form 'defvar)
etc.
> The example suggests that this code avoids finding a match inside a
> string, but in fact there is no code to check that.  It just checks
> that the (defun doesn't follow any non-indentation and that it includes
> a function name and argument list.
>   
Indeed. Will proceed here. Will it be possible to write a reg-exp which
matches a functions definition reliable? Just couldn't
get it until now. Probably a more prosperous way is to
write genuine mode-functions, in Emacs Lisp using the
list- and sexp-moving commands.

> I think it would be more useful to specially recognize certain
> top-level combining constructs, such as eval-when-compile and `when',
> and treat their subforms as 

Don't understand completely what this part means. Will
reflect it further.

> defuns.
>   
IMO different meanings of
`defun' in Emacs are the reason of a major difficulty
for beginners (at least for non-programmers).

Altogether I see no use and no reason to have it,
beside of historical ones.

To keep backward compatibility while replacing all
`defun'-namings where `function' or just top-level-form
is meant, it could be done while defining aliases at all
cases where `function' is the right name.

Of course the manuals must be corrected afterwards, but
that's seems not a too complicated task, as quite
often only a word is to change.

>     ;; The mode should set this, if possible
>     (defcustom beginning-of-function-return-value t
>       "Specify the return value if function succeeds: either t or the value 
>     of point"
>       :group 'lisp
>         :type '(choice (const t)
>                (sexp :tag "Return function" (point))))
>
> I think that feature is unhelpful complexity.  It would be better to
> have just one return value convention.  If the caller of
> beginning-of-function wants the value of point, it can do (if
> (beginning-of-function) (point))
>   


Would not consider it a important question. Just need the
point as return value and would save writing with it.

`beginning-of-defun-raw' at the moment closes:

and (re-search-backward ...)
    (progn (goto-char (1- (match-end 0)))) t))

Why the `t' instead of the more useful (point)? Isn't
the value of point true anyway?

As there are probably reasons to put it that way which
I ignore, I didn't abolish it, just turned to
(point). With the default as `t' the
user won't remark a changing here.


__
Andreas Roehler