Re: mykie.el

[Ted Zlatanov <tzz@lifelogs.com>]

On Mon, 06 Jan 2014 19:37:01 -0500 Stefan Monnier <monnier@iro.umontreal.ca> wrote: 

SM> Is that really much better than [vanilla Emacs lisp]?
>> Yes.  The difference is huge for a beginner.

SM> I'm far from convinced.  The beginner will have to learn the
SM> mykie-specific sublanguage.  Maybe it's slightly simpler, but then this
SM> effort won't help him understand other code.

That's a problem with anything that hides code complexity, and yet we do
it all the time.  I doubt beginners can handle `pcase' and the
subtleties of context-sensitive key definitions, so I think for this
specific case the simplification is necessary.

SM> Making it easier to customize Emacs is great, but it's better if it also
SM> gets you some way towards hacking on Emacs so as to become a contributor.

I completely disagree with that line of thinking.

SM> Of course, there are cosmetic issues in the mykie.el syntax that rub me
SM> the wrong way.  Most obvious is the need to quote code.  I already spend
SM> enough time telling people not to quote their lambdas and trying to fix
SM> code that abuses `eval' in all kinds of useless ways.

That's a good point.

...
SM>    (global-set-key "C-j"
SM>       (sm-combined-command
SM>        (:C-u&eolp (fill-region (point-at-bol) (point-at-eol)))
SM>        (:region   'query-replace-regexp)
SM>        (t
SM>         (delete-trailing-whitespace)
SM>         (case major-mode
SM>           (org-mode (org-return-indent))
SM>           (t        (newline-and-indent))))))

SM> And sm-combined-command is a very simple macro:

SM>   (defconst sm-combined-command-predicates
SM>     '((:region . (use-region-p))
SM>       (:C-u . current-prefix-arg)
SM>       (:C-u&eolp . (and current-prefix-arg (eolp)))
SM>       ...))

SM>   (defmacro sm-combined-command (branches)
SM>     `(lambda ()
SM>        (interactive)
SM>        (cond
SM>         ,@(mapcar (lambda (branch)
SM>                     `(,(or (cdr (assq (car branch)
SM>                                       sm-combined-command-predicates))
SM>                            (car branch))
SM>                       ,@(pcase (cdr branch)
SM>                           (`(',cmd) `((call-interactively ',cmd)))
SM>                           (,cmds cmds))))
SM>                   branches))))

That's very nice (you and your pcase :).  I think it's fairly difficult
to explain *how* it works to a beginner and would need lots of
documentation, but from the user's standpoint it's exactly what I was
hoping to bring out of mykie.el.  The developers can then contribute
more predicates and the users can happily use them.

The important question for you and Yamada-san is, would you be
interested in something like `sm-combined-command' in the Emacs core,
regardless of how mykie.el evolves?  Then mykie.el can remain a place
for experimentation and perhaps it can use `sm-combined-command'
internally, enriching it with special extensions that maybe don't make
sense in the core.

The other way I imagine is to put mykie.el in the GNU ELPA, but I would
rather enable this facility by default if there are no objections.

Ted