bug#24605: Subject: 25.1.50; form-at-point might fail for some THINGS

[npostavs@users.sourceforge.net]

tags 24605 patch
severity 24605 minor
quit

Tino Calancha <tino.calancha@gmail.com> writes:

> Because `thing-at-point' doesn't return always a string, as
> `form-at-point' currently assumes.
[...]
>
> (with-temp-buffer
>   (require 'thingatpt)
>   (insert "1")
>   (let ((res
>          (cond ((and (number-at-point)
>                      (form-at-point 'number)) 0)
>                ((and (number-at-point)
>                      (not (form-at-point 'number))) -1)
>                ((and (not (number-at-point))
>                      (form-at-point 'number)) -2)
>                ((and (not (number-at-point))
>                      (not (form-at-point 'number))) -3))))
>     (pcase res
>       ('0 (message "OK, both 'number-at-point and 'form-at-point successfull"))
>       (-1 (message "'number-at-point' works but (form-at-point 'number) fails"))
>       (-2 (message "(form-at-point 'number) works but 'number-at-point' fails"))
>       (-3 (message "Both 'number-at-point' and (form-at-point 'number) fails")))))
> => "’number-at-point’ works but (form-at-point ’number) fails"

Well, it looks like `form-at-point' was assumed to be used only for
implementing the foo-at-point functions, (form-at-point 'sexp 'numberp)
does work fine for this.  But we may as well let (form-at-point 'number)
work too, since it's the more obvious way to call it.

>  (defun form-at-point (&optional thing pred)
> -  (let ((sexp (ignore-errors
> -		(thing-at-point--read-from-whole-string
> -		 (thing-at-point (or thing 'sexp))))))
> +  (let* ((obj (thing-at-point (or thing 'sexp)))
> +         (sexp (ignore-errors
> +                 (if (stringp obj)
> +                     (thing-at-point--read-from-whole-string obj)
> +                   obj))))

I suggest keeping the `ignore-errors' strictly around the
`thing-at-point--read-from-whole-string' call.

>      (if (or (not pred) (funcall pred sexp)) sexp)))