On 17.03.2017 14:05, Noam Postavsky wrote:
> On Fri, Mar 17, 2017 at 4:16 AM, Andreas Röhler
> <andreas.roehler@online.de> wrote:
>>>> --8<---------------cut here---------------start------------->8---
>>>> (defun my-command (beg end)
>>>>    (interactive "r")
>>>>    (if (use-region-p)
>>>>        (my-command-region beg end)
>>>>      (my-command-non-region)))
>>>> --8<---------------cut here---------------end--------------->8---
>>> AFAIK, you cannot use the (interactive "r") form for this "dwim-ish"
>>> kind of behavior.
>> Why that? If a region is set, why should (use-region-p) fail?
> (use-region-p) doesn't fail, the "r" in interactive fails, when there
> is no mark (i.e., when a region is not and has never been set). It's
> basically equivalent to this:
>
> (defun my-command (beg end)
>    (interactive (sort (list (point) (mark)) #'<))
>    (if (use-region-p)
>        (my-command-region beg end)
>      (my-command-non-region)))

Okay, thanks. So that's no occasion to question the use-region-p, 
region-active-p design.
For the case given, think the error is reasonable, as a command 
explicitly working on region might expect an existing one.