Re: Inject some eshell features into shell?

[Kevin Rodgers]

Kai Grossjohann wrote:

> Hey, cool!  It's not quite what I meant, but you're showing me the
> interesting bits.  I need to set comint-input-sender to my function,
> and my function needs to fall back using comint-simple-send.


Right.  It would also be nice to keep the shell prompt in sync when you
run an Emacs command:

        (progn
          (comint-simple-send process (concat "# " command "\n"))
          (apply command-symbol (cdr command-list)))

But it looks like that might not get into the input history (see 
comint-input-history-ignore).


> What I meant was something less automatic: Have an alist which says
> which Lisp to invoke depending on the first word.
>
> But maybe it's also good to use the eshell way of doing things: just
> look for a kshell/foo function if the user entered foo.
>
> When using eshell, I noticed that many Lisp functions are not suitable
> to be called from the shell prompt.  For example, cvs-update is a good
> candiate, but it requires me to pass FLAGS (and DIRECTORY).  It just
> fees unnatural to type "cvs-update . nil" at the shell prompt, when
> "cvs-update" ought to do.

I think you could use the kshell/foo approach to provide required arguments
that can be inferred from the context (current directory, etc.).

--
Kevin Rodgers