With the keybinding patch you can bind a key to run an external
command.  There are two types of bindings:

1. When the wmx-modifier is pressed [the one defined to CONFIG_ALT_KEY
in Config.h].

2. When no modifiers are pressed.

Actually, in both of the above cases the state of caps-lock and
num-lock is ignored.

When wmx starts up it examines the ~/.wmx directory, which
traditionally consisted of two things.  Executable files (or usually
symlinks to them) are the commands that appear on the Command Menu
(the middle button in the root, and the options file is parsed at
startup (by Config.C).

With the keybinding patch wmx checks for 2 additional directories:
.keys and .mkeys.

When wmx starts up it examines these directories for executable files
whose filename is a valid keymap name.  For example "F1" and
"KP_Enter" are valid.  So is "1" but "f1" is probably not.

Files in the .keys directory are activated when the key is pressed
without any modifier (such as Shift, Control or Meta).

Files in the .mkeys directory are activated when the key is pressed
and the wmx-modifier key is pressed.  (The wmx-modifier is defined in
Config.h;  the default is Alt, but I reccomend changing it.  Super_L
is usually a good choice.)

(For convenience the num-lock and caps-lock is ignored, so "no
modifiers" and "only wmx modifier" really isn't quite correct.)


For example,  suppose ~/.wmx/.keys/F1 contains

------------------------------------------------------------------------
#! /bin/sh
xlogo -title F1
------------------------------------------------------------------------

then when you press F1 an xlogo pops up with F1 in the title bar.  If
the file were moved to ~/.wmx/.mkeys/F1 then you would have to have
the wmx-modifier key held down when you pressed F1.

Ok, that's not an exciting example.


A really useful program is xwit:
	http://www.x.org/contrib/utilities/xwit-3.4.tar.gz
This allows you to do "window manager type things" from a shell script.
For example, you can move a window to an absolute location or relative
to its current position.

When wmx calls a file because of a keybinding it passes it with one
argument,  the window id.  This is how we can tell xwit which window
to work with.

For example, I have these two:

File ~/.wmx/.mkeys/comma
------------------------------------------------------------------------
#! /bin/sh

# move current window left 50 pixels
xwit -id $1 -rmove -50 0
------------------------------------------------------------------------

File ~/.wmx/.mkeys/period
------------------------------------------------------------------------
#! /bin/sh

# move current window right 50 pixels
xwit -id $1 -rmove 50 0
------------------------------------------------------------------------


Now, by pressing the wmx-modifier and the comma key the current window
is moved left a bit,  and with the period it goes right a bit.
I used comma and period because they are above the "<" and ">" keys.
You can't use filenames "less" and "greter" because the keysym "less"
is bound to the the "period" key when the shift modifier is prespent.


With the remote patch you can do fancier things as explained there.

The standard wmx client menu shows the label.  If you're looking for a
specific window it can be hard to find if there are many.

With the "class" option turned on the client window will display
for example

[emacs] README.menu

instead of the default:

README.menu

If you also turn on the "sort" option the names are sorted,  so all
the emacs ones are connsecutive.


To turn on class and/or sort you can

1. Change Config.h to have
#define CLASS_IN_MENU True

2. With the default setting of
#define CLASS_IN_MENU	(dConfig.classInMenu())
you can turn it on by adding "class:on" to the options string.

E.g.
cd ~/.wmx
ln -sf class:on/sort:on options

Or, of course, add to your existing options such as
ln -sf right:toggleheight/class:on/sort:on options


3. If you have the remote option patch then in the telnet window
you can use the options command:
options class:on/sort:on

With the remote control patch if you invoke wmx with either the "-r"
or "-p" option then wmx opens a socket.  With the -r option the
default port of 6999 is used,  the -p option allows you to specific
the port number.  Wmx listens on this port for commands and executes
them.


If you telnet into this port you will see:

Welcome to wmx remote control.  Type help for commands

Each reply (including error messages) always ends with an empty line
to allow a program to parse it.  For example:

input:	channel 2
output: OK
output:

With that example wmx will have switched to channel two,  so it would
have been a good idea to make the above telnet window a sticky one!

By itself the remote control is fun to play with, but what good is it?

Well, with the keybinding patch you can run a command that can send a
command to the socket.  The easiest way to do this from a shell script
is using the socket program.  If you don't have this you should!

	http://www.jnickelsen.de/socket/

This is a very handy program which can do many things.  The nice
feature here is that you can invoke it from a shell script to send
commands to a socket.

Here is an example file that hides the curent window:

------------------------------------------------------------------------
#! /bin/sh

# Hide current window

PORT=6999
echo hide $1 | socket -q localhost $PORT
------------------------------------------------------------------------

If this file is copied to ~/.wmx/.keys/F1 with execute permission
(and wmx restarted) then when you press F1 the window under the cursor
disappears.  Ok, so you can do the same thing by moving the pointer to
the box and clicking.  Sometmes, though, having a key bound can be
faster.

Here's a more interesting example:

------------------------------------------------------------------------
#! /bin/sh

# Move this window to the next channel

# go to next channel
echo channel | socket -q localhost 6999  | \
	gawk  '/Current channel is/{ n = $4; n++ ; print "wchannel",'"$1"', n}' |
	socket -q localhost 6999
------------------------------------------------------------------------

As the comment says,  the current window is moved to the next channel
but the current channel is left as is.

To do the same thing with the mouse you would have to move the pointer
to the frame, middle click (or use the mouse wheel),  then move back
to the previous channel by moving the pointer away from the window
and going back to the previous channel.

Cycling amoungst windows often gets ones you're not interested in.
This file allows cycling between just the same client classes.

------------------------------------------------------------------------
#! /bin/sh

# warp to the next client of the same class on the same channel
# ignoring hidden and withdrawn clients

(echo clients ; echo clients ; echo quit) | socket localhost 6999 | \

gawk -v W=$1 '

BEGIN { CLASS = ""; }

(CLASS != "" && $3 !~ /.*[HW].*/ && $4 == CLASS && $2 == CHANNEL) {
	print "warp", $1
	print "quit"
	exit;
}

($1 == W) {
	CLASS = $4
	CHANNEL = $2
}
' | \
  tee -a $HOME/Log | \
socket -q localhost 6999
------------------------------------------------------------------------