bug#24627: 24.5; (thing-at-point 'list) may return a non-empty string without a list

["Andreas Röhler" <andreas.roehler@easy-emacs.de>]

On 13.10.2016 10:50, Tino Calancha wrote:
> Andreas Röhler <andreas.roehler@easy-emacs.de> writes:
>
>> beg still needs a check like
>>
>> (not (nth 8 (parse-partial-sexp (point-min) (point))))
>>
>> otherwise it could match inside a string or comment
> I have the feeling that this should return the local list
> at point, even if inside a string or comment.

Yes, but that would be reported by pps. However, when point is at 
opening delimiter, this is not recognised by pps. Then we must be sure 
not being inside a string or comment, where an opening delimiter is 
meaningless, i.e. just a literal.

IMO all needed is  something like

(beg (or (nth 1 (parse-partial-sexp...))

          (and (eq 4 (car (syntax-after (point))))
               (not (nth 8 (parse-partial-sexp...))
               (point)))))
      

Should both fail, there is not list at point.


>   Then, if
> point is inside a comment/string and there is no list there,
> the function might look for a list around (i.e., outside) that
> comment/string region.  See patch below.
>
> Anyway, neither the doc string of `thing-at-point' nor
> `thing-at-point-bounds-of-list-at-point' mention what expect
> when point is inside a comment/string.  That's why i believe it
> might be better to request that in a different bug report.
> Writting additional tests also might be helpful to find a robust implementation.
>
> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
>  From bfc9b7fb739dfeab09c2ffd064a6ebe65a28b686 Mon Sep 17 00:00:00 2001
> From: Tino Calancha <tino.calancha@gmail.com>
> Date: Thu, 13 Oct 2016 16:34:35 +0900
> Subject: [PATCH] (thing-at-point 'list) return nil if no list at point
>
> * lisp/thingatpt.el (thing-at-point-bounds-of-list-at-point):
> Check first if we are at the beginning of a top-level sexp (Bug#24627).
> If found a list inside a comment or string return it.  Otherwise, look
> for a list around the comment/string.
> Escape '[' in doc string.
> ---
>   lisp/thingatpt.el | 32 ++++++++++++++++++--------------
>   1 file changed, 18 insertions(+), 14 deletions(-)
>
> diff --git a/lisp/thingatpt.el b/lisp/thingatpt.el
> index 6d1014b..656d2c7 100644
> --- a/lisp/thingatpt.el
> +++ b/lisp/thingatpt.el
> @@ -219,22 +219,26 @@ 'beginning-of-sexp
>   
>   (defun thing-at-point-bounds-of-list-at-point ()
>     "Return the bounds of the list at point.
> -[Internal function used by `bounds-of-thing-at-point'.]"
> +\[Internal function used by `bounds-of-thing-at-point'.]"
>     (save-excursion
> -    (let ((opoint (point))
> -	  (beg (ignore-errors
> -		 (up-list -1)
> -		 (point))))
> +    (let* ((opoint (point))
> +	   (st (syntax-ppss))
> +	   (find-list-fn (lambda ()
> +                           (ignore-errors
> +                             (up-list -1)
> +                             (point))))
> +	   (beg (if (eq 4 (car (syntax-after (point))))
> +		    (point)
> +		  (funcall find-list-fn))))
> +      ;; If inside a string or comment and there is no list
> +      ;; at point, check for a list surrounding the string/comment region.
> +      (when (and (nth 8 st) (= opoint (point)))
> +	(goto-char (nth 8 st))
> +	(setq beg (funcall find-list-fn)))
>         (ignore-errors
> -	(if beg
> -	    (progn (forward-sexp)
> -		   (cons beg (point)))
> -	  ;; Are we are at the beginning of a top-level sexp?
> -	  (forward-sexp)
> -	  (let ((end (point)))
> -	    (backward-sexp)
> -	    (if (>= opoint (point))
> -		(cons opoint end))))))))
> +        (when beg
> +          (forward-sexp)
> +          (cons beg (point)))))))
>   
>   ;; Defuns
>