Re: cc-mode Objective C method names

[Alan Mackenzie <acm@muc.de>]

Hello, Alan.

On Wed, Jan 01, 2020 at 11:27:57 +0000, Alan Third wrote:
> One small annoyance I’ve had with developing Emacs is that the
> helpers for filling in the changelog entries don’t work with Objective
> C methods. I’ve tried writing a patch to make it work.

> For reference, an Objective C class looks something like:

> @implementation ClassName

> - (int)doSomething
> {
>   /* do that something */
>   return 1;
> }

> - (void)doSomethingTo: (SomeClass *)object with: (int)someParam
> {
>   return;
> }

> @end

> And I think the methods’ names should be written something like:

> -[ClassName doSomething]
> -[ClassName doSomethingTo:with:]

> The ‘-’ means it’s an instance method and a ‘+’ would mean it was a
> class method.

If either ClassName or doSomethingTo is long, you might be taking up too
much space on, for example, the first line of a commit message.  But
you've probably already thought this through.  How long are these
extended names in practice?

> It appears to work for me, but I’m not great at Emacs lisp so I
> thought it best to run this by the mailing list in case I’ve made any
> boneheaded errors.

I've got just a few comments to add to Eli's and Stefan's, so ....

> Thanks!
> -- 
> Alan Third

> >>From 2945f1c6c57eeabdbeb8e7c058070587a9bf4c0a Mon Sep 17 00:00:00 2001
> From: Alan Third <alan@idiocy.org>
> Date: Mon, 30 Dec 2019 16:38:47 +0000
> Subject: [PATCH] Add ability to find ObjC method names

> * lisp/progmodes/cc-cmds.el (c-defun-name-1): Add Objective-C method
> name ability.
> ---
>  lisp/progmodes/cc-cmds.el | 30 ++++++++++++++++++++++++++++++
>  1 file changed, 30 insertions(+)

> diff --git a/lisp/progmodes/cc-cmds.el b/lisp/progmodes/cc-cmds.el
> index 0343f9df32..9165398132 100644
> --- a/lisp/progmodes/cc-cmds.el
> +++ b/lisp/progmodes/cc-cmds.el
> @@ -2024,6 +2024,36 @@ c-defun-name-1
>  	     (c-backward-syntactic-ws)
>  	     (point))))
 
> +	 ((looking-at "[-+]\\s-*(")     ; Objective-C method

I'd be happier here if you could also check we're really in Objc with
(c-major-mode-is 'objc-mode).

> +	  (let ((class
> +		 (save-excursion
> +		   (re-search-backward "@\\(implementation\\|class\\|interface\\)")

Here, you might want to give a NOERROR argument to re-search-backward,
so that if implementation etc., isn't found, you can handle the error
gracefully.  Something like

                   (if (re-search-backward "@\\(....\\)" nil t)
		       (progn
		         (c-forward-token-2)
			 .....)
	             "")

> +		   (c-forward-token-2)
> +		   (thing-at-point 'word t)))
> +		(type (buffer-substring-no-properties (point) (+ (point) 1)))
> +		(name
> +		 (save-excursion
> +		   (c-forward-token-2 2 t)

c-forward-token-2 might not be the best function, here.  It moves to the
next token, but if there isn't a next token it doesn't move at all.
c-forward-over-token-and-ws might be better.  Alternatively, you could
check the result of c-forward-token-2 (say, with zerop) and take special
action if the call has failed.  This comment also applies to the later
uses of c-forward-token-2.

> +		   (let ((name ""))

I feel there ought to be a standard CC Mode function which would do the
following scanning to the {, but I can't find it at the moment.

> +		     (while (not (looking-at "[{;]"))

Or, (not (memq (char-after) '(?{ ?\;))), which would be minutely faster.
It would also not change the match-data if you cared about that (which
you don't, here).

> +		       (let ((start (point))
> +			     (end
> +			      (progn
> +				(c-forward-syntactic-ws)
> +				(forward-word)
> +				(if (looking-at ":")
> +				    (+ (point) 1)
> +				  (point)))))
> +			 (when (looking-at ":")
> +			   (c-forward-token-2)
> +			   (if (looking-at "(")
> +			       (c-forward-token-2 2 t)
> +			     (c-forward-token-2 1 t)))
> +			 (c-forward-syntactic-ws)
> +			 (setq name (concat name (buffer-substring-no-properties start end)))))

Just a small point: isn't name always the empty string before this form?
If so, you could get rid of the concat and just use
buffer-substring-no-properties.

> +		     name))))
> +	    (format "%s[%s %s]" type class name)))

As an alternative to format, you could use (concat type "[" class " "
                                            name "]")
, which is more direct, but probably doesn't matter here.

> +
>  	 (t				; Normal function or initializer.
>  	  (when (looking-at c-defun-type-name-decl-key) ; struct, etc.
>  	    (goto-char (match-end 0))
> -- 
> 2.24.0

-- 
Alan Mackenzie (Nuremberg, Germany).