Writing a command that assumes control of the keyboard, like isearch does

Writing a command that assumes control of the keyboard, like isearch does

[Sean McAfee]

I want to write a command that assumes control of the keyboard until it
returns, highlighting and modifying the current buffer in the meantime,
like the isearch family of commands do.

Specifically, I have some text that is the output from an OCR process,
in which the spaces between words have sometimes been lost; for example:

  GNUEmacsisanextensible,customizabletexteditor—andmore.

I want to write a command that will wait for me to type a letter, then
advance the cursor to that letter and provisionally insert a space after
it, highlighting the passed-over text (the provisional word, that is).
If I then type a space, I confirm the provisional word; if I type
another letter, the provisional space is removed, and the cursor skips
to the newer letter I typed and inserts another provisional space.
Therefore, I could break the words in the sample text above with this
key sequence:

  u SPC s SPC s SPC n SPC e e SPC e SPC t t SPC r SPC d SPC

Some kind of special handling would probably be needed for punctuation,
but I can resolve those details later.  And I suppose RET would probably
terminate the command.

Since the behavior I want to implement broadly resembles what isearch
does, I went and read the source code, but found it unsuitable for
casual reading.  I did note this command near the beginning:

;; For programmed use of isearch-mode, e.g. calling (isearch-forward),
;; isearch-mode behaves modally and does not return until the search
;; is completed.  It uses a recursive-edit to behave this way.

So, does that mean it *doesn't* use a recursive edit when called
interactively?  What does it do in that case?  Not very illuminating.

Browsing the manual, it seems that I might be able to write my own mini
event loop using read-char, but I can't help but feel that there must be
an easier way.  As best I can express it right now, I want to
temporarily install a keymap where every key that normally calls
self-insert-command calls a routine of my choice instead; SPC and RET
have the special meanings described above; and any other key sequence
exits my command and has its normal effect (again, like isearch).

Any pointers to get me on the right track, or even just references to
the appropriate places in the manual that I might have missed, would be
greatly appreciated.

Re: Writing a command that assumes control of the keyboard, like isearch does

[harven]

Sean McAfee <eefacm@gmail.com> writes:

> I want to write a command that assumes control of the keyboard until it
> returns, highlighting and modifying the current buffer in the meantime,
> like the isearch family of commands do.
>
> Specifically, I have some text that is the output from an OCR process,
> in which the spaces between words have sometimes been lost; for example:
>
>   GNUEmacsisanextensible,customizabletexteditor—andmore.
>
> I want to write a command that will wait for me to type a letter, then
> advance the cursor to that letter and provisionally insert a space after
> it, highlighting the passed-over text (the provisional word, that is).
> If I then type a space, I confirm the provisional word; if I type
> another letter, the provisional space is removed, and the cursor skips
> to the newer letter I typed and inserts another provisional space.

> Therefore, I could break the words in the sample text above with this
> key sequence:
>
>   u SPC s SPC s SPC n SPC e e SPC e SPC t t SPC r SPC d SPC
>
> Some kind of special handling would probably be needed for punctuation,
> but I can resolve those details later.  And I suppose RET would probably
> terminate the command.

> Any pointers to get me on the right track, or even just references to
> the appropriate places in the manual that I might have missed, would be
> greatly appreciated.

A simple while loop (with no body !) together with read-event may be sufficient.
Something like

(defun essai ()
   (interactive)
   (let (key 
        (echo-keystrokes 0))
     (setq key (read-event))
     (while 
         (when (characterp key)
            (search-forward (char-to-string key) nil t)
            (insert " ")
            (setq key (read-event))
            (if (equal key ?\ )
                (setq key (read-event))
              (delete-char -1)
              t)))))

BTW it seems there is a missing e
   u SPC s SPC s SPC n SPC e e e SPC e SPC t t SPC r SPC d SPC
                               ^

Re: Writing a command that assumes control of the keyboard, like isearch does

[Alan Mackenzie]

Sean McAfee <eefacm@gmail.com> wrote:
> I want to write a command that assumes control of the keyboard until it
> returns, highlighting and modifying the current buffer in the meantime,
> like the isearch family of commands do.

> Specifically, I have some text that is the output from an OCR process,
> in which the spaces between words have sometimes been lost; for example:

>  GNUEmacsisanextensible,customizabletexteditor?andmore.

> I want to write a command that will wait for me to type a letter, then
> advance the cursor to that letter and provisionally insert a space
> after it, highlighting the passed-over text (the provisional word, that
> is).  If I then type a space, I confirm the provisional word; if I type
> another letter, the provisional space is removed, and the cursor skips
> to the newer letter I typed and inserts another provisional space.
> Therefore, I could break the words in the sample text above with this
> key sequence:

>  u SPC s SPC s SPC n SPC e e SPC e SPC t t SPC r SPC d SPC

> Some kind of special handling would probably be needed for punctuation,
> but I can resolve those details later.  And I suppose RET would
> probably terminate the command.

> Since the behavior I want to implement broadly resembles what isearch
> does, I went and read the source code, but found it unsuitable for
> casual reading.  I did note this command near the beginning:

isearch.el is worth studying at a deeper level.

> ;; For programmed use of isearch-mode, e.g. calling (isearch-forward),
> ;; isearch-mode behaves modally and does not return until the search
> ;; is completed.  It uses a recursive-edit to behave this way.

> So, does that mean it *doesn't* use a recursive edit when called
> interactively?  What does it do in that case?  Not very illuminating.

Interactively, isearch mode switches on a minor mode, and most
importantly, sets the `overriding-terminal-local-map' to the keymap
isearch-mode-map.  isearch-mode-map has the keybinding
isearch-printing-char (which increments the search string) for all
printing characters, and suitable commands are bound to C-s, M-r, ....,
and so on.  Random control sequences are bound to
isearch-other-control-char, which typically exits isearch mode by
clearing `overriding-terminal-local-map'.

> Browsing the manual, it seems that I might be able to write my own mini
> event loop using read-char, but I can't help but feel that there must be
> an easier way.  As best I can express it right now, I want to
> temporarily install a keymap where every key that normally calls
> self-insert-command calls a routine of my choice instead; SPC and RET
> have the special meanings described above; and any other key sequence
> exits my command and has its normal effect (again, like isearch).

I think that is indeed what you want to do.  Don't forget C-g!

> Any pointers to get me on the right track, or even just references to
> the appropriate places in the manual that I might have missed, would be
> greatly appreciated.

Study isearch.el more thoroughly.  The stuff you want to emulate is
indeed there.

-- 
Alan Mackenzie (Nuremberg, Germany).