Homepage: https://www.gnu.org/software/emacs
Author: Peter Stephenson
Manipulate an alternative ("virtual") cursor
Latest changes
==============
- *IMPORTANT* `vcursor-key-bindings' is now nil by default, to avoid
side-effects when the package is loaded. This means no keys are
bound by default. Use customize to change it to t to restore
the old behavior. (If you do it by hand in .emacs, it
must come before vcursor is loaded.)
- You can alter the main variables and the vcursor face via
M-x customize: go to the Editing group and find Vcursor.
- `vcursor-auto-disable' can now be 'copy (actually any value not nil
or t), which means that copying from the vcursor will be turned
off after any operation not involving the vcursor, but the
vcursor itself will be left alone.
- works on dumb terminals
- new keymap `vcursor-map' for binding to a prefix key
- `vcursor-compare-windows' substantially improved
- `vcursor-execute-{key,command}' much better about using the
right keymaps and arranging for the correct windows to be used
- `vcursor-window-funcall' can call functions interactively
- `vcursor-interpret-input' for special effects
Introduction
============
Virtual cursor commands. I got this idea from the old BBC micro.
You need a window system for the best effects.
This is much easier to use than the instructions are to read.
First, you need to let vcursor define some keys: setting
vcursor-key-bindings to t before loading, or by customize, will
define various keys with the prefix C-S. You'll have to read
further if you don't want this. Then I suggest you simply load it
and play around with holding down Ctrl and Shift and pressing up,
down, left, right, tab, return, and see what happens. (Find a
scratch buffer before using C-S-tab: that toggles copying.)
Most of the functions described in this documentation are in
parentheses so that if you have the package loaded you can type C-h
f on top of them for help.
Using the cursor keys with both control and shift held down moves
around a virtual cursor, which is initially at point. When active,
it appears with an underline through it to distinguish it from the
normal cursor. You can then use one of the other commands to copy
characters from the location of the virtual cursor to point. This
is very useful, for example, when copying some previous text while
making changes to it at the same time, since you never have to move
the "real" cursor away from where you are inserting.
The remaining default key bindings are based around the PC-type
cluster found above the cursor keys on a lot of keyboards, the
function keys which my limited knowledge of X terminals expects to
find at the top. Some functions are duplicated in more obvious
places for the X version.
All the keybindings require you to hold down control and shift at
once. I assumed this combination wouldn't be heavily bound by most
people and that it would be easy to type with the left hand.
Inevitably it will clash with some other packages, but I can't help
that: an intuitive binding is a prerequisite here. See below for
other alternatives (search for "Oemacs"). There is also a keymap
which you can bind to a prefix key, which may give some more
intuitive alternatives in some cases, see `The vcursor keymap' below.
Holding down control and shift and pressing insert (vcursor-copy)
copies one character from wherever the virtual cursor is to point;
point and the virtual cursor advance in the separate and equal
station to which... (etc.). M-C-S-return (vcursor-copy-line)
copies to the end of the line instead of just one character,
C-S-delete or C-S-remove (vcursor-copy-word) copies a word.
A more general way of copying is to use C-S-tab, which is a toggle.
In the "on" state, moving the virtual cursor will copy the
moved-over text to the normal cursor position (including when going
backwards, though each piece of text moved over is copied forwards:
compare the behavior of C-S-up and C-S-left).
However, that's just a small part of the magic. If the virtual
cursor goes off the display, it will be redisplayed in some other
window. (See the function (vcursor-find-window) for details of how
this window is chosen.) This gives you fingertip control over two
windows at once.
C-S-return (vcursor-disable) disables the virtual cursor, removing
it so that it starts from point whenever you move it again --- note
that simply moving the cursor and virtual cursor on top of one
another does not have this effect.
If you give C-S-return a positive prefix arg, it will also delete the
window (unless it's the current one). Whenever the virtual cursor
goes off-screen in its own window, point in that window is moved as
well to restore it to view. (It's easier that way, that's why.
However, point doesn't move unless the view in the window does, so
it's not tied to the virtual cursor location.)
You can also use C-S-return with a negative prefix argument which
forces the vcursor to appear at point. This is particularly useful if
you actually want to edit in another window but would like to
remember the current cursor location for examining or copying from
that buffer. (I just hit C-S-right C-S-left, but I'm a hopeless
lowbrow.)
There is also C-S-f6 (vcursor-other-window) which behaves like
C-x o on the virtual rather than the real cursor, except that it
will create another window if necessary.
The keys C-S-prior (vcursor-scroll-down) and C-S-next
(vcursor-scroll-up) (i.e., PageUp and PageDown) will scroll the
virtual cursor window, appropriately chosen. They will always
create a new window or take over an old one if necessary.
Likewise, M-C-S-left and M-C-S-right move you to the
beginning or end of a line, C-S-home and C-S-end the
beginning or end of a buffer (these are also on M-C-S-up and
M-C-S-down for those of us stuck with DEC keyboards).
C-S-f7 (vcursor-goto) will take you to the vcursor position
(swapping windows if it seems sensible) and (unless you give it a
prefix argument) delete the virtual cursor, so this is useful for
you to take over editing at the virtual cursor position. It is not
an error if the virtual cursor is not active; it simply leaves you
at point, because that is where the virtual cursor would start
from.
In a similar vein, M-C-S-tab (hope your left hand's flexible;
C-S-select on DEC keyboards) (vcursor-swap-point) will take you to
the virtual cursor position but simultaneously put the virtual
cursor at the old cursor position. It is also supposed to ensure
that both are visible.
C-S-f8 (C-S-find on DEC keyboards) (vcursor-isearch-forward)
allows you to do an isearch in another window. It works a bit like
vcursor-scroll-*; it moves into another window, calls isearch
there, and sets the virtual cursor position to the point found. In
other words, it works just like isearch but with the virtual cursor
instead of the real one (that's why it's called a "virtual
cursor"). While you are isearching, you are editing in the virtual
cursor window, but when you have finished you return to where you
started. Note that once you are in isearch all the keys are normal
--- use C-s, not C-S-f8, to search for the next occurrence.
If you set the variable vcursor-auto-disable, then any command
which does not involve moving or copying from the virtual cursor
causes the virtual cursor to be disabled. If you set it to non-nil
but not t, then the vcursor itself will remain active, but copying
will be turned off, so that the next time the vcursor is moved no
text is copied over. Experience shows that this setting is
particularly useful. If you don't intend to use this, you can
comment out the `add-hook' line at the bottom of this file. (This
feature partially emulates the way the "copy" key on the BBC micro
worked; actually, the copy cursor was homed when you hit return.
This was in keeping with the line-by-line way of entering BASIC,
but is less appropriate here.)
vcursor-compare-windows is now a reliable adaption of
compare-windows, which compares between point in the current buffer
and the vcursor location in the other one. It is an error if
vcursor is not set, however it will be brought up in another window
if it is not currently visible. The prefix argument acts just like
compare-windows, ignoring whitespace if set. (In versions before
1.6, this simply called compare-windows, which was much less likely
to pick the two windows you wanted.)
There is a way of moving the virtual cursor using ordinary
commands: C-S-f9 (vcursor-execute-key) reads a key string,
moves to the virtual cursor position, executes the command bound to
the string, then returns to the original point. Thus C-S-f9 M-m
moves the virtual cursor back to the first non-whitespace character
on its line. As the command is called interactively all the usual
ways of passing information to the command called, such as by a
prefix argument, are available. This has many uses not necessarily
related to moving the vcursor itself; it can do essentially
everything that the \C-x 4 series of commands can do and a lot
more. Note, however, that a new window is not used if the vcursor
is visible in the current one: this can lead to some strange effects,
but it is preferable to making a new window every time the vcursor
is moved in this may.
C-S-f10 (C-S-x) (vcursor-execute-command) behaves the same way but
you enter the name of the command. To do anything really
complicated, you are better off using M-C-S-tab
(vcursor-swap-point), doing whatever it is, then calling M-C-S-tab
again.
If you want to add your own moving or copying functions you should
be able to do this fairly easily with (vcursor-relative-move) and
(vcursor-copy) together with (vcursor-get-char-count). If you want to
do something in a different window, use (vcursor-window-funcall).
Key bindings
============
There is an alternative set of key bindings named "Oemacs" (for
historical reasons). This set uses separate control, shift and
meta keys with function keys 1 to 10. In particular, movement keys
are concentrated on f5 to f8 with (in increasing order of distance
traveled) C-, M- and S- as prefixes. See the actual bindings below
(search for C-f1). This is because the C-S- prefix is represented
by weird key sequences and the set is incomplete; if you don't mind
that, some hints are given in comments below.
You can specify which set of key bindings to use by customizing the
user option `vcursor-key-bindings'.
Usage on dumb terminals
=======================
If Emacs has set the variable window-system to nil, vcursor will
assume that overlays cannot be displayed in a different face, and
will instead use a string (the variable vcursor-string, by default "**>")
to show its position. Unlike the old-fashioned overlay arrow (as
used by debuggers), this appears between existing text, which can
make it hard to read if you're not used to it. (This seemed the
better option here.) This means moving the vcursor up and down is
a very efficient way of locating it!
Everything else should function as expected, but there is no way to
get an easy key binding for the vcursor keys on a generic terminal.
Consequently a special keymap is defined for you to use traditional
methods: the keymap, however, is available on any terminal type.
The vcursor keymap
==================
In addition to any other bindings, vcursor-map contains key definitions
for handling the vcursor. You should assign this to a prefix key
in the usual way, e.g.
(global-set-key [f14] vcursor-map)
and also as usual \C-h in this map will list the key definitions, which
are designed to be easy to remember.
A special feature is provided by (vcursor-use-vcursor-map), bound
to t in that keymap. With this in effect, the main keymap
is overridden by the vcursor map, so keys like \C-p and so on
move the vcursor instead. Remember how to turn it off (type t),
or you are in serious trouble! Note that the cursor keys are not
bound by default in this keymap and will continue to move the
ordinary cursor.
Interpreted input
=================
Just occasionally, you may want to pretend the strings copied from
the vcursor position are to be interpreted as if you had typed them
from the keyboard. Normally, they will just insert themselves anyway,
but in some modes (Info and calc for example) typing ordinary characters
does something else. To get this effect, set
vcursor-interpret-input to t. This is normally not a good idea as
interpreting input is very much slower than copying text.
Un-features
===========
- The vcursor will not move to point-max, since otherwise it would
disappear. However, no error is flagged as point-max is a valid
point in the buffer. Thus cursor right or down at the second
last point in the file does not flag an error, which is inconsistent,
and if copying is on the last character (typically newline) will
be repeatedly copied. (I've tried making it flag an error
instead and that's worse since often the vcursor is sent to
point in some other window, which may be point-max.)
- The vcursor widens when over a tab character or right at the
end of the line. You're welcome to consider this a feature;
it's just a part of how overlays work.
- The vcursor obscures the real cursor. Creative use of overlays
could cure this.
- The vcursor does not remember its own previous positions. If
you cycle it back into a window it was in before, it will be at
point in that window. Often, that is where a previous recenter
left point, not where the vcursor was before.
(Note, however, that the vcursor does remember where it *is*,
even if it's off-screen. This can also lead to surprises, but I
don't think it's a bug.)
- vcursor-window-funcall could perhaps be smarter about restoring
the previous window state on failure.
- The logic in vcursor-find-window is rather complicated and
therefore bug-prone, though in practice it seems to work OK.
Possible enhancements:
It would be easy to implement vcursor-push (save vcursor position
as mark and deactivate) and vcursor-pop (deactivate vcursor and
move to last pushed position) functions.