Making code more Emacs Lisp-like

Making code more Emacs Lisp-like

[Sarir Khamsi]

I wrote this function

(defun sk-convert-to-ctor-init-list ()
  "Convert a C++ member variable declaration to a ctor init list line.
To use this, copy all member variable declarations into the constructor's member initialization list area and execute this command on each line. At the end of this command, it moves you to the next line setting it up for a key binding."
  (interactive)
  (save-excursion
    (end-of-line)
    (delete-horizontal-space)           ; delete whitespace at end of line
    (backward-delete-char-untabify 1)   ; delete the ";"
    (insert "(),")
    (search-backward-regexp "[\t *]")     ; look for whitespace or "*"
    (set-mark (+ (point) 1))              ; save the current position
    (beginning-of-line)
    (kill-region (point) (mark))
    (c-indent-command))
  (next-line 1))

to take a region that containers C++ code for member variable
declarations and convert it to a constructor's member init list. Is
there a better (or a more ELisp) way than this approach? Thanks.

Sarir

-- 
Sarir Khamsi
software guy
sarir.khamsi@raytheon.com

Re: Making code more Emacs Lisp-like

[Joseph Brenner]

Sarir Khamsi <sarir.khamsi@raytheon.com> writes:

> I wrote this function
>
> (defun sk-convert-to-ctor-init-list ()
>   "Convert a C++ member variable declaration to a ctor init list line.
> To use this, copy all member variable declarations into the constructor's member initialization list area and execute this command on each line. At the end of this command, it moves you to the next line setting it up for a key binding."
>   (interactive)
>   (save-excursion
>     (end-of-line)
>     (delete-horizontal-space)           ; delete whitespace at end of line
>     (backward-delete-char-untabify 1)   ; delete the ";"
>     (insert "(),")
>     (search-backward-regexp "[\t *]")     ; look for whitespace or "*"
>     (set-mark (+ (point) 1))              ; save the current position
>     (beginning-of-line)
>     (kill-region (point) (mark))
>     (c-indent-command))
>   (next-line 1))
>
> to take a region that containers C++ code for member variable
> declarations and convert it to a constructor's member init list. Is
> there a better (or a more ELisp) way than this approach? Thanks.

If you look up the Help screen for each function (using the default
bindings: cursor to the function, and do a "f1 f"), you'll often see
hints on better ways of doing things from inside of elisp.

For example, if you look up "set-mark", there's an example that's
easily adapted to what you're doing here:

>     (set-mark (+ (point) 1))              ; save the current position
>     (beginning-of-line)
>     (kill-region (point) (mark))

Would become:

   (let ((beg (point)))
      (beginning-of-line)
      (delete-region beg (point)))

The goal there is to avoid messing with the kill-ring or the mark
history if that's not necessary.  Otherwise you may surprise the
user with the unwanted side-effects.

Re: Making code more Emacs Lisp-like

[Sarir Khamsi]

Joseph Brenner <doom@kzsu.stanford.edu> writes:

> For example, if you look up "set-mark", there's an example that's
> easily adapted to what you're doing here:
>
>>     (set-mark (+ (point) 1))              ; save the current position
>>     (beginning-of-line)
>>     (kill-region (point) (mark))
>
> Would become:
>
>    (let ((beg (point)))
>       (beginning-of-line)
>       (delete-region beg (point)))

Thanks, this looks better.

> The goal there is to avoid messing with the kill-ring or the mark
> history if that's not necessary.  Otherwise you may surprise the
> user with the unwanted side-effects.

save-excursion is supposed to "Save point, mark, and current buffer;
execute BODY; restore those things" but I understand that you want to
avoid modifying the kill ring. Thanks.

Sarir

-- 
Sarir Khamsi
sarir.khamsi@raytheon.com

Re: Making code more Emacs Lisp-like

[Joseph Brenner]

Sarir Khamsi <sarir.khamsi@raytheon.com> writes:
> Joseph Brenner <doom@kzsu.stanford.edu> writes:

>> The goal there is to avoid messing with the kill-ring or the mark
>> history if that's not necessary.  Otherwise you may surprise the
>> user with the unwanted side-effects.
>
> save-excursion is supposed to "Save point, mark, and current buffer;
> execute BODY; restore those things" but I understand that you want to
> avoid modifying the kill ring. Thanks.

Good point... but don't always assume that save-excursion is really
going to save what you think it will.  Every now and then I find
a case where it doesn't seem to want to restore the current buffer,
and I need to do that manually.

(No, I haven't yet filed a bug report.  Maybe next time I see the
problem.)