dot-mode.el

;;; dot-mode.el - minor mode to repeat typing or commands
;;; Copyright (C) 1995 James Gillespie
;;; Copyright (C) 2000 Robert Wyrick (rob@wyrick.org)
;;; Copyright (C) 2012 by (fernando@dobladez.com)
;;
;; Purpose of this package: minor mode to repeat typing or commands
;;
;; Installation instructions
;;
;; Install this file somewhere in your load path, byte-compile it and
;; add one of the following to your .emacs file (remove the comment
;; delimiters ;-)
;;
;; If you only want dot-mode to activate when you press "C-.", add the
;; the following to your .emacs:
;;
;;     (autoload 'dot-mode "dot-mode" nil t) ; vi `.' command emulation
;;     (global-set-key [(control ?.)] (lambda () (interactive) (dot-mode 1)
;;                                       (message "Dot mode activated.")))
;;
;; If you want dot-mode all the time (like me), add the following to
;; your .emacs:
;;
;;     (require 'dot-mode)
;;     (add-hook 'find-file-hooks 'dot-mode-on)
;; 
;; You may still want to use the global-set-key above.. especially if you
;; use the *scratch* buffer.
;;
;; To toggle dot mode on or off type `M-x dot-mode'
;;
;; There are only two variables that allow you to modify how dot-mode
;; behaves:
;;           dot-mode-ignore-undo
;;           dot-mode-global-mode
;;
;; dot-mode-ignore-undo - defaults to t.  When nil, it will record keystrokes
;;     that generate an undo just like any other keystroke that changed the
;;     buffer.  I personally find that annoying, but if you want dot-mode to
;;     always remember your undo's:
;;         (setq dot-mode-ignore-undo nil)
;;     Besides, you can always use dot-mode-override to record an undo when
;;     you need to (or even M-x undo).
;;
;; dot-mode-global-mode - defaults to t.  When t, dot-mode only has one
;;     keyboard command buffer.  That means you can make a change in one
;;     buffer, switch buffers, then repeat the change.  When set to nil,
;;     each buffer gets its own command buffer.  That means that after
;;     making a change in a buffer, if you switch buffers, that change
;;     cannot repeated.  If you switch back to the first buffer, your
;;     change can then be repeated again.  This has a nasty side effect
;;     if your change yanks from the kill-ring (You could end up
;;     yanking text you killed in a different buffer).
;;     If you want to set this to nil, you should do so before dot-mode
;;     is activated on any buffers.  Otherwise, you may end up with some
;;     buffers having a local command buffer and others using the global
;;     one.
;;
;; Usage instructions:
;;
;; `C-.'    is bound to dot-mode-execute, which executes the buffer of
;;          stored commands as a keyboard macro.
;;
;; `C-M-.'  is bound to dot-mode-override, which will cause dot-mode
;;          to remember the next keystroke regardless of whether it
;;          changes the buffer and regardless of the value of the
;;          dot-mode-ignore-undo variable.
;;
;; `C-S-.'  is bound to dot-mode-copy-to-last-kbd-macro, which will
;;          copy the current dot mode keyboard macro to the last-kbd-macro
;;          variable.  It can then be executed via call-last-kbd-macro
;;          (normally bound to `C-x-e'), named via name-last-kbd-macro,
;;          and then inserted into your .emacs via insert-kbd-macro.
;;
;; Known bugs:
;;
;; none
;;

;;; This program is free software; you can redistribute it and/or modify
;;; it under the terms of the GNU General Public License as published by
;;; the Free Software Foundation; either version 1, or (at your option)
;;; any later version.

;;; This program is distributed in the hope that it will be useful,
;;; but WITHOUT ANY WARRANTY; without even the implied warranty of
;;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
;;; GNU General Public License for more details.

;;; A copy of the GNU General Public License can be obtained from 
;;; the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA
;;; 02139, USA.

;;; COMMENTARY
;;;
;;; This mode is written to address one argument in the emacs vs. vi
;;; jihad :-)  It emulates the vi `redo' command, repeating the
;;; immediately preceding sequence of commands.  This is done by
;;; recording input commands which change the buffer, i.e. not motion
;;; commands.

;;; DESIGN
;;;
;;; The heart of this minor mode is a state machine.  The function
;;; dot-mode-after-change is called from after-change-functions and
;;; sets a variable (is there one already?  I couldn't find it) which
;;; is examined by dot-mode-loop, called from from post-command-hook.
;;; This variable, dot-mode-changed, is used in conjunction with
;;; dot-mode-state to move to the next state in the state machine.
;;; The state machine is hard coded into dot-mode-loop in the
;;; interests of speed; it uses two normal states (idle and store)
;;; and two corresponding override states which allow the user to
;;; forcibly store commands which do not change the buffer.

(defconst dot-mode-version "1.12"
  "Report bugs to: Robert Wyrick <rob@wyrick.org>")

;;; CHANGE HISTORY
;;;
;;; 1.1
;;; Wrote dot-mode.el
;;;
;;; 1.2
;;; At the suggestion of Scott Evans <gse@ocsystems.com>, added
;;; 'dot-mode-override' to allow the user to force dot mode to store a
;;; motion command
;;;
;;; 1.3
;;; Changed dot-mode-loop to use a state machine instead of several
;;; booleans
;;;
;;; 1.4
;;; Hard coded the state machine into dot-mode-loop in the hope of
;;; speeding it up
;;;
;;; 1.5
;;; Ported to GNU Emacs - nearly: the keymap doesn't seem to install
;;; correctly.
;;;
;;; 1.6
;;; Rob Wyrick (that's me) took over maintenance of the package from
;;; Jim Gillespie.
;;;
;;; In some versions of Emacs, (this-command-keys) returns a empty
;;; vector by the time it is called from the 'post-command-hook.
;;; So, I split the functionality... now dot-mode-command-keys
;;; stores (this-command-keys) output in a temp variable to be used
;;; by dot-mode-loop and dot-mode-command-keys is called from the
;;; pre-command-hook.  Also re/ported to XEmacs/GNU Emacs.  It works
;;; on both now.  dot-mode-command-keys could have been put on the
;;; after-change-functions hook, but I've begun preliminary work to
;;; capture what's going on in the minibuffer and I'm certain I need
;;; it where it is.
;;;
;;; 1.7
;;; Added my first attempt to capture what the user is doing with
;;; execute-extended-command (M-x).  It even works if the executed
;;; command prompts the user.
;;; Also added some error recovery if the user interrupts or there is
;;; an error during execution of the stored macro.
;;;
;;; 1.8
;;; Second attempt to capture what the user is doing with 
;;; execute-extended-command (M-x).  The previous version didn't work
;;; in XEmacs.  This version works in both XEmacs and GNUEmacs.
;;;
;;; 1.9
;;; Third attempt to capture what the user is doing with 
;;; execute-extended-command (M-x).  Wow was I making things hard.
;;; It's cost me a lot of version numbers in a short amount of time,
;;; so we won't discuss my previous attempts. *grin*  My second attempt
;;; worked just fine, but it was more complicated and maybe not as
;;; portable to older version of X/GNU Emacs.
;;; Other things:
;;;   - Yet another restructuring of the code.  By doing so, 
;;;     quoted-insert (C-q) is properly stored by dot-mode.
;;;     (quoted-insert has been broken since ver 1.6)
;;;   - Deleted an extraneous state and the "extended-state" added
;;;     in ver 1.8.  We're down to just two normal states and two
;;;     override states.
;;;   - Added dot-mode-ignore-undo and dot-mode-global-mode variables
;;;     as well as the new function dot-mode-copy-to-last-kbd-macro.
;;;
;;; 1.10
;;; Fixed a bug where the META key wasn't properly recorded on GNU
;;; Emacs.  Actually, if you used ESC for META (like me), everything
;;; worked fine.  But using ALT for META was broken.
;;; Now I'm using this-command-keys-vector when I can.
;;; I also added the dot-mode-event-to-string function to make the
;;; output a little prettier.
;;; Thanks to Scott Evans <gse@antisleep.com> for reporting the bug!
;;;
;;; 1.11
;;; Fixed a bug where dot-mode would give an error if you used
;;; dot-mode-override to record a <right> and then tried to call
;;; dot-mode-execute.  The bug was in dot-mode-event-to-string
;;; Thanks to Scott Evans <gse@antisleep.com> for reporting the bug!
;;;
;;; 1.12

;;; Made compatible with Emacs 24.
;;; Change of behavior: pointer now stays were it was when "dot-mode-execute" started
;;; (this is *not* how VI does it, but I like it much better)
;;; Change some default bindings

(defvar dot-mode nil
  "Whether dot mode is on or not")
(make-variable-buffer-local 'dot-mode)

(defvar dot-mode-map 
  (let ((map (make-sparse-keymap)))
    (if (fboundp 'read-kbd-macro)
        (progn
          (define-key map (read-kbd-macro "C-.")   'dot-mode-execute)
          (define-key map (read-kbd-macro "C-M-.") 'dot-mode-override)
          (define-key map (read-kbd-macro "C->") 'dot-mode-copy-to-last-kbd-macro)
        )
      ;; ELSE - try this way...
      (define-key map [(control ?.)]         'dot-mode-execute)
      (define-key map [(control meta ?.)]    'dot-mode-override)
      (define-key map [(control ?>)]     'dot-mode-copy-to-last-kbd-macro)
    )
    map)
  "Keymap used in dot mode buffers")

;; Make sure add-minor-mode exists
(if (not (fboundp 'add-minor-mode))
    (defun add-minor-mode (mode name map)
      (or (assoc mode minor-mode-alist)
          (setq minor-mode-alist
                (cons (list mode name) minor-mode-alist)))
      (or (assoc mode minor-mode-map-alist)
          (setq minor-mode-map-alist
                (cons (cons mode map) minor-mode-map-alist)))))

(add-minor-mode 'dot-mode " Dot" dot-mode-map) ;; depends on add-minor-mode

(defvar dot-mode-global-mode t
  "Should dot-mode share its command buffer between buffers?")

(defvar dot-mode-ignore-undo t
  "Should dot-mode ignore undo?")

(defvar dot-mode-changed nil
  "Did last command change buffer?")

(defvar dot-mode-cmd-buffer nil
  "Saved commands.")

(defvar dot-mode-cmd-keys nil
  "Saved keys.")

(defvar dot-mode-state 0
  "Current state of dot mode.
0 - Initial (no changes)
1 - Recording buffer changes
2 - Override from state 0
3 - Override from state 1")

(defvar dot-mode-minibuffer-input nil
  "Global buffer to capture minibuffer input")

;; The below statements are another possible definition of dot-mode-command-keys
;; that I *think* will work.  It hasn't been well tested so I'm leaving it
;; out for now.  This was written at the same time I found this-command-keys-vector.
;; Since I don't know what version of emacs has that func, I was working on
;; the below as an alternate work-around.
;; ------------------------------------------------------------------------
;;  ((and (boundp 'current-prefix-arg)
;;        (boundp 'meta-prefix-char)
;;        (fboundp 'this-single-command-keys)) ;; also see this-single-command-raw-keys
;;   (defun dot-mode-command-keys ()
;;     (if (null current-prefix-arg)
;;         (this-single-command-keys)
;;       (vconcat (char-to-string meta-prefix-char) (number-to-string current-prefix-arg) (this-single-command-keys)))))

(cond 
 ((fboundp 'this-command-keys-vector)
  (fset 'dot-mode-command-keys (symbol-function 'this-command-keys-vector)))
 (t
  (defun dot-mode-command-keys ()
    (let ((tmp (this-command-keys)))
      (cond ((vectorp tmp)
             tmp)
            ((stringp tmp)
             (string-to-vector tmp))
            ((fboundp 'character-to-event) ;; xemacs
             (character-to-event tmp))
            (t ;; probably never get here
             (vconcat tmp))))))
)

(defun dot-mode-copy-to-last-kbd-macro ()
  "Copy the current dot-mode command buffer to the last-kbd-macro variable.
Then it can be called with call-last-kbd-macro, named with name-last-kbd-macro,
or even saved for later use with name-last-kbd-macro"
  (interactive)
  (if (null dot-mode-cmd-buffer)
      (message "Nothing to copy.")
    (setq last-kbd-macro dot-mode-cmd-buffer)
    (message "Copied."))
)

(cond ((and (fboundp 'event-modifiers) (fboundp 'event-basic-type))
       (defun dot-mode-event-to-string (ev)
         "Return the event as a string."
       (let
         ((em (event-modifiers ev))
          (eb (event-basic-type ev)))
         (if (and (not (symbolp eb)) (equal em '(control)))
           (char-to-string ev)
           (concat
             (mapconcat (lambda(x) (concat "<" (symbol-name x) ">")) em "")
             (if (symbolp eb) (concat "<" (symbol-name eb) ">") (char-to-string eb))
      )))))
      (t
       (defun dot-mode-event-to-string (ev)
         "Return the event as a string."
         (char-to-string ev)))
)

(defun dot-mode-buffer-to-string ()
       "Return the macro buffer as a string."
  (let ((str dot-mode-cmd-buffer))
    (if (fboundp 'character-to-event) ; we're on X-Emacs
        (progn
          (setq str (prin1-to-string str))
          (setq str (replace-in-string str " *#<keypress-event +" "<"))
          (setq str (replace-in-string str " *<\\(.\\)> *" "\\1"))
          (setq str (replace-in-string str "^\\[\\(.*\\)\\]$" "\\1"))
        )
      ;; ELSE - we're on GNU Emacs
      (setq str (mapconcat (lambda (arg)
                             (cond ((and (fboundp 'eventp) (eventp arg))
                                    (dot-mode-event-to-string arg))
                                   ((symbolp arg)
                                    (concat "<" (symbol-name arg) ">"))
                                   (t
                                    (char-to-string arg)))) str ""))
    )
    str)
)

(defun dot-mode-minibuffer-exit ()
  "Catch minibuffer exit"
  ;; Just store it as a string buffer... 
  ;;     On X Emacs, we'll call character-to-event later
  ;;     On GNU Emacs, vconcat will handle strings
  (setq dot-mode-minibuffer-input 
        (concat dot-mode-minibuffer-input (buffer-string) "\r"))

  ;; I'd really like to check this-command to see if it's exit-minibuffer
  ;; and remove this function from the minibuffer-exit-hook if it is.
  ;; Unfortunately, if an extended command asks for 2 or more arguments,
  ;; the first arg would be the only one to get recorded since exit-minibuffer
  ;; is called between each argument.
)

(defun dot-mode-execute ()
  "Execute stored commands."
  (interactive)
  ;; Don't want execution to kick off infinite recursion
  (if (null dot-mode-cmd-buffer)
      (message "Nothing to repeat")
    (remove-hook 'pre-command-hook 'dot-mode-pre-hook t)
    (remove-hook 'post-command-hook 'dot-mode-loop t)
    (remove-hook 'after-change-functions 'dot-mode-after-change t)

    ;; Do the business
    (if dot-mode-keep-cursor-position
        (save-excursion (dot-mode-execute--do-it))
      (dot-mode-execute--do-it))

    ;; Put the hooks back
    (add-hook 'pre-command-hook 'dot-mode-pre-hook nil t)
    (add-hook 'post-command-hook 'dot-mode-loop nil t)
    (add-hook 'after-change-functions 'dot-mode-after-change nil t)
  )
)

(defun dot-mode-execute--do-it ()
  (message "Repeating \"%s\"" (dot-mode-buffer-to-string))
  (condition-case nil
      (execute-kbd-macro dot-mode-cmd-buffer)
    ((error quit exit)
     (setq dot-mode-cmd-buffer nil
           dot-mode-state      0)
     (message "Dot mode reset")))
  (if (not (null dot-mode-cmd-buffer))
      ;; I message before AND after a macro execution.
      ;; On XEmacs, I never saw the Repeating message above...
      ;; Besides, this way you'll know if your macro somehow
      ;; hangs during execution (on GNU Emacs, anyway).
      (message "Repeated \"%s\"" (dot-mode-buffer-to-string))))

(defun dot-mode-override ()
  "Override standard behaviour and store next keystroke no matter what."
  (interactive)
  (setq dot-mode-state (+ dot-mode-state 2))
  (message "dot-mode will remember the next keystroke..."))

(defun dot-mode-after-change (start end prevlen)
  "Dot mode's after-change-functions hook"
  ;; By the time we get here, dot-mode-pre-hook has already setup
  ;; dot-mode-cmd-keys.  It'll be a vector, t, or nil.

  (cond ((vectorp dot-mode-cmd-keys) ;; we just did an execute-extended-command
                                     ;; or an override

         (if (not dot-mode-changed)     ;; if dot-mode-changed is t, we're in override
             (progn
               ;; remove hook
               (remove-hook 'minibuffer-exit-hook 'dot-mode-minibuffer-exit)
               (if (not (null dot-mode-minibuffer-input))
                   (progn
                     (if (fboundp 'character-to-event) ;; we're on X-Emacs
                         (setq dot-mode-minibuffer-input 
                               (mapcar 'character-to-event dot-mode-minibuffer-input)))
                     (setq dot-mode-cmd-keys (vconcat dot-mode-cmd-keys 
                                                      dot-mode-minibuffer-input))
                   )
               )
             )
           ;; ELSE - we're in override and the keys have already been read
         )
        )
        ;; Normal mode
        (dot-mode-cmd-keys
         (setq dot-mode-cmd-keys (dot-mode-command-keys))
        )
        ;; Else, do nothing dot-mode-cmd-keys will remain nil. (Only happens on ignore-undo)
  )
  (if dot-mode-cmd-keys
      (setq dot-mode-changed t))
)

(defun dot-mode-pre-hook ()
  "Dot mode's pre-command-hook"

  ;; remove hook (should already be removed... but double check)
  ;; The only time this will ever do any good is if you did a
  ;; quit out of the minibuffer.  In that case, the hook will
  ;; still be there.  It won't really hurt anything, it will just
  ;; continue to record everything you do in the minibuffer 
  ;; regardless of whether or not it is an execute-extended-command.
  ;; And the dot-mode-minibuffer-input buffer could get quite large.
  (remove-hook 'minibuffer-exit-hook 'dot-mode-minibuffer-exit)

  (cond
   ;; Is this an execute-extended-command?
   ((eq this-command 'execute-extended-command)
    (setq dot-mode-minibuffer-input nil
          ;; Must get this (M-x) now!  It's gone later.
          dot-mode-cmd-keys         (dot-mode-command-keys)
          dot-mode-changed          nil ;; ignore an override
    )
    ;; Must be a global hook
    (add-hook 'minibuffer-exit-hook 'dot-mode-minibuffer-exit)
   )
   (dot-mode-changed            ;; on override, dot-mode-changed is t
    ;; Always read the keys here on override _UNLESS_ it's a quoted-insert.
    ;; This is to make sure we capture keys that don't change the buffer.
    ;; On quoted-insert, all we get here is  , but in dot-mode-after-change,
    ;; we get   plus the following key (and we're guaranteed to change the
    ;; buffer)
    (setq dot-mode-cmd-keys (or (eq this-command 'quoted-insert)
                                (dot-mode-command-keys)))
   )
   ;; Should we ignore this key sequence? (is it an undo?)
   ((and dot-mode-ignore-undo
         (or (eq this-command 'advertised-undo)
             (eq this-command 'undo)))
    (setq dot-mode-cmd-keys nil)
   )
   (t
    (setq dot-mode-cmd-keys t)  ;; signal to read later (in dot-mode-after-change)
   )
  )
)

(defun dot-mode-loop ()
  "The heart of dot mode."
;;  (message "in: state is %d" dot-mode-state)
;;  (message "in: cmd-buffer is '%s'" (dot-mode-buffer-to-string))
  (cond ((= dot-mode-state 0)           ; idle
         (if dot-mode-changed
             (setq dot-mode-state       1
                   dot-mode-changed     nil
                   dot-mode-cmd-buffer  dot-mode-cmd-keys))
        )
        ((= dot-mode-state 1)           ; recording
         (if dot-mode-changed
             (setq dot-mode-changed     nil
                   dot-mode-cmd-buffer  (vconcat dot-mode-cmd-buffer dot-mode-cmd-keys))
           (setq dot-mode-state 0))
        )
        (t ; = 2 or 3                   ; override
         (setq dot-mode-state       (- dot-mode-state 2)
               dot-mode-changed     t)
        )
  )
;;  (message "out: state is %d" dot-mode-state)
;;  (message "out: cmd-buffer is '%s'" (dot-mode-buffer-to-string))
)

(defun dot-mode (arg)
  "Toggle dot mode.
With arg, turn dot mode on iff arg is positive.

Dot mode mimics the `.' function in vi, repeating sequences of
commands and/or typing delimited by motion events.  Use `C-.' rather
than just `.'."
  (interactive "P")
  (setq dot-mode
        (if (null arg)
            (not dot-mode)
          (> (prefix-numeric-value arg) 0)))
  (if (not dot-mode)
      (progn
        (remove-hook 'pre-command-hook 'dot-mode-pre-hook t)
        (remove-hook 'post-command-hook 'dot-mode-loop t)
        (remove-hook 'after-change-functions 'dot-mode-after-change t)
      )
    ;; ELSE
    ;; The hooks are _ALWAYS_ local since dot-mode may not be on in every buffer
    (add-hook 'pre-command-hook 'dot-mode-pre-hook nil t)
    (add-hook 'post-command-hook 'dot-mode-loop nil t)
    (add-hook 'after-change-functions 'dot-mode-after-change nil t)
    (if dot-mode-global-mode
        (progn
          (kill-local-variable 'dot-mode-cmd-buffer)
          (kill-local-variable 'dot-mode-cmd-keys)
          (kill-local-variable 'dot-mode-state)
          (kill-local-variable 'dot-mode-changed)
        )
      ;; ELSE
      (make-local-variable 'dot-mode-cmd-buffer)
      (make-local-variable 'dot-mode-cmd-keys)
      (make-local-variable 'dot-mode-state)
      (make-local-variable 'dot-mode-changed)
      (setq dot-mode-state        0
            dot-mode-changed      nil
            dot-mode-cmd-buffer   nil
            dot-mode-cmd-keys     nil
      )
    )
  )
  (cond ((fboundp 'force-mode-line-update)
         (force-mode-line-update))
        ((fboundp 'redraw-modeline)
         (redraw-modeline)))
;;  (set-buffer-modified-p (buffer-modified-p)) ;; Why was I doing this?
)

(defun dot-mode-on ()
  "Turn on dot-mode."
  (interactive)
  (dot-mode 1))

(defalias 'turn-on-dot-mode 'dot-mode-on)




;;; customization
(defgroup dot-mode nil
  "Mode for repeating the last typing or commands (similar to VI's . (dot) command"
  :group 'convenience
  :version dot-mode-version)

(defcustom dot-mode-keep-cursor-position nil
  "Non-nil (true) means that after repeating a command the cursor is put back where it was
before repeating the command.
When nil (false), the cursor stays where the repeated command left it. This is VI's behavior"
  :type 'boolean
  :group 'dot-mode)



(provide 'dot-mode)

;;; dot-mode.el ends here