Re: command-line-args

[Miles Bader]

On Fri, Sep 19, 2003 at 12:15:54PM -0400, Stefan Monnier wrote:
> > Because of this behavior, I can't use the following:
> 
> > emacs -f w3m http://some.news.site/daily-news.html&;

The only way this bug report makes sense to me is if w3m was already special
casing for command-line arguments, i.e., if it had code like this:

   (defun w3m (&optional url ...)
     (interactive ...)
     (when (null url)
       ...try to get command-line args...))

Right?

> That was not intended.
> I'm not sure how best to get both behaviors.
> 
> What do people think of a patch like the one below ?
...
> +              (condition-case nil
> +                  (funcall tem)
> +                ((invalid-function wrong-number-of-arguments)
> +                 (command-execute tem))))
>  
>                  ((member argi '("-eval" "-execute"))
>                   (eval (read (or argval (pop command-line-args-left)))))

That's sort of convoluted, and likely to cause confusion (because different
functions will be called in different ways for not-very-obvious reasons).  It
seems much preferable for command-line functions to be called one consistent
way, either interactively or non-interactively -- and certainly in the
majority of cases, interactive would make sense (for the same reason that you
wanted it for gdb).

A function like w3m that (if I presume correctly) special-cases command
line arguments, if it use code in its interactive spec (instead of a string),
it could just add the special case to the interactive spec code.  If it uses
a string interactive spec, I'm not sure; it could change to using code, or I
suppose `interactive' could add a special directive that says `if called from
the command line, try to read arguments from there too.'  Since I suspect
hardly any functions try to do this, though, probably it's OK to just convert
to using a code-form interactive spec.

-Miles
-- 
Saa, shall we dance?  (from a dance-class advertisement)