Re: Convert an existing keyboard macro to elisp code?

[Raffaele Ricciardi <rfflrccrd@gmail.com>]

On 08/03/2012 05:33 PM, PJ Weisberg wrote:> On Fri, Aug 3, 2012 at 8:10 
AM, Thorsten Jolitz <tjolitz@googlemail.com> wrote:
 >> Unfortunately, the answer was more a less a 'NO'(or 'YOU HAVE TO WRITE
 >> IT YOURSELF').
 >>
 >> Since Emacs is a pretty dynamic project, I thought I give it another try
 >> on the mailing list - maybe this feature request has been implemented in
 >> the meantime by somebody?
 >
 > If it were just a matter of reading the key sequence and mapping it to
 > functions it wouldn't be too bad, but the more I think about this, the
 > more it seems like a rat hole.  How would you translate
 >
 > (fset 'my-macro
 >     [?\M-x ?m ?a ?g tab ?i ?t tab ?s ?t tab ?t tab return ?m ?a tab 
return])
 >
 > into
 >
 > (defun my-macro ()
 >    (interactive)
 >    (magit-status "c:/Users/PJ/Documents/magit/"))
 >
 > ?
 >
 > You would need to know what auto-completion happened when the user
 > pressed tab, and then you would need to know how the function
 > translated its user input into interactive arguments.  I don't think
 > there's a way to do it without running the macro and (somehow)
 > watching what happens.  And that could have unpleasant side effects.
 >
 > And this particular example wouldn't even be portable, since when I
 > type "M-x magit-status RET magit RET", Magit actually does some magic
 > to translate that into a full path.  So on my other machine, the
 > proper definition of that macro would be
 >
 > (defun my-macro ()
 >    (interactive)
 >    (magit-status "/home/pj/source-code/magit"))
 >
 >
 > And that doesn't even take into account different keymaps.
 >
 > -PJ
 >
 > Gehm's Corollary to Clark's Law: Any technology distinguishable from
 > magic is insufficiently advanced.
 >

Be stumped:

C-h f command-history
C-h v command-history

Emacs already keeps track of commands that read input from terminal, I 
don't see
the reason it couldn't keep track of all commands, as the third-party 
library
mwe-log-commands.el does it, while discarding the arguments.  Maybe 
Emacs does
it already in some other variable I don't know.

Regarding commands requiring further input and possibly performing 
conversions
on their own, Emacs could feed them the same input you did.  Of course, the
resulting translation will be context-sensitive just like the original macro
was.  Adjustments to make the code portable would be up to the user.