⌨️

xremap is a key remapper for Linux. Unlike xmodmap, it supports app-specific remapping and Wayland.

Concept

• Fast - Xremap is written in Rust, which is faster than JIT-less interpreters like Python.

• Cross-platform - Xremap uses evdev and uinput, which works whether you use X11 or Wayland.

• Language-agnostic - The config is JSON-compatible. Generate it from any language, e.g. Ruby, Python.

Features

• Remap any keys, e.g. Ctrl or CapsLock.
• Remap any key combination to another, even to a key sequence.
• Remap a key sequence as well. You could do something like Emacs's C-x C-c.
• Remap a key to two different keys depending on whether it's pressed alone or held.
• Application-specific remapping. Even if it's not supported by your application, xremap can.
• Device-specific remapping.
• Automatically remap newly connected devices by starting xremap with --watch.
• Support Emacs-like key remapping, including the mark mode.
• Trigger commands on key press/release events.
• Use a non-modifier key as a virtual modifier key.

Installation

Download a binary from Releases.

If it doesn't work, please install Rust and run one of the following commands:

cargo install xremap --features x11     # X11
cargo install xremap --features gnome   # GNOME Wayland
cargo install xremap --features kde     # KDE-Plasma Wayland
cargo install xremap --features wlroots # Sway, Hyprland, etc.
cargo install xremap                    # Others

You may also need to install libx11-dev to run the xremap binary for X11.

Arch Linux

If you are on Arch Linux and X11, you can install xremap-x11-bin from AUR.

NixOS

If you are using NixOS, xremap can be installed and configured through a flake.

Usage

Write a config file directly, or generate it with xremap-ruby or xremap-python.

Then start the xremap daemon by running:

sudo xremap config.yml

(You will need to leave it running for your mappings to take effect.)

If you want to run xremap without sudo, click here.

Running xremap without sudo

To do so, your normal user should be able to use evdev and uinput without sudo. In Ubuntu, this can be configured by running the following commands and rebooting your machine.

sudo gpasswd -a YOUR_USER input
echo 'KERNEL=="uinput", GROUP="input", TAG+="uaccess"' | sudo tee /etc/udev/rules.d/input.rules

Arch Linux

The following can be used on Arch.

lsmod | grep uinput

If this module is not loaded, add to /etc/modules-load.d/uinput.conf:

uinput

Then add udev rule.

echo 'KERNEL=="uinput", GROUP="input", TAG+="uaccess"' | sudo tee /etc/udev/rules.d/99-input.rules

Then reboot the machine.

Debian

Make sure uinput is loaded same as in Arch:

lsmod | grep uinput

If it shows up empty:

echo uinput | sudo tee /etc/modules-load.d/uinput.conf

Add your user to the input group and add the same udev rule as in Ubuntu:

sudo gpasswd -a YOUR_USER input
echo 'KERNEL=="uinput", GROUP="input", TAG+="uaccess"' | sudo tee /etc/udev/rules.d/input.rules

Reboot the machine afterwards or try:

sudo modprobe uinput
sudo udevadm control --reload-rules && sudo udevadm trigger

Other platforms

In other platforms, you might need to create an input group first and run echo 'KERNEL=="event*", NAME="input/%k", MODE="660", GROUP="input"' | sudo tee /etc/udev/rules.d/input.rules as well.

If you do this, in some environments, --watch may fail to recognize new devices due to temporary permission issues. Using sudo might be more useful in such cases.

---

See the following instructions for your environment to make application-specific remapping work.

X11

If you use sudo to run xremap, you may need to run xhost +SI:localuser:root if you see No protocol specified.

GNOME Wayland

Install xremap's GNOME Shell extension from this link, switching OFF to ON.

If you use sudo to run xremap, also click here.

Update /usr/share/dbus-1/session.conf as follows, and reboot your machine.

   <policy context="default">
+    <allow user="root"/>
     <!-- Allow everything to be sent -->
     <allow send_destination="*" eavesdrop="true"/>
     <!-- Allow everything to be received -->

KDE-Plasma Wayland

Xremap cannot be run as root. Follow the instructions above to run xremap without sudo.

Configuration

Your config.yml should look like this:

modmap:
  - name: Except Chrome
    application:
      not: Google-chrome
    remap:
      CapsLock: Esc
keymap:
  - name: Emacs binding
    application:
      only: Slack
    remap:
      C-b: left
      C-f: right
      C-p: up
      C-n: down

See also: example/config.yml and example/emacs.yml

modmap

modmap is for key-to-key remapping like xmodmap. Note that remapping a key to a modifier key, e.g. CapsLock to Control_L, is supported only in modmap since keymap handles modifier keys differently.

modmap:
  - name: Name # Optional
    exact_match: false # Optional, defaults to false
    remap: # Required
      # Replace a key with another
      KEY_XXX: KEY_YYY # Required
      # Dispatch different keys depending on whether you hold it or press it alone
      KEY_XXX:
        held: [KEY_YYY] # Required
        alone: [KEY_ZZZ] # Required
        alone_timeout_millis: 1000 # Optional
      # Hook `keymap` action on key press/release events.
      KEY_XXX:
        press: { launch: ["xdotool", "mousemove", "0", "7200"] } # Required
        release: { launch: ["xdotool", "mousemove", "0", "0"] } # Required
    application: # Optional
      not: [Application, ...]
      # or
      only: [Application, ...]
    device: # Optional
      not: [Device, ...]
      # or
      only: [Device, ...]

For KEY_XXX and KEY_YYY, use these names. You can skip KEY_ and the name is case-insensitive. So KEY_CAPSLOCK, CAPSLOCK, and CapsLock are the same thing. Some custom aliases like SHIFT_R, CONTROL_L, etc. are provided.

If you specify a map containing held and alone, you can use the key for two purposes. The key is considered alone if it's pressed and released within alone_timeout_millis (default: 1000) before any other key is pressed. Otherwise it's considered held.

keymap

keymap is for remapping a sequence of key combinations to another sequence of key combinations or other actions.

keymap:
  - name: Name # Optional
    remap: # Required
      # Key press -> Key press
      MOD1-KEY_XXX: MOD2-KEY_YYY
      # Sequence (MOD1-KEY_XXX, MOD2-KEY_YYY) -> Key press (MOD3-KEY_ZZZ)
      MOD1-KEY_XXX:
        remap:
          MOD2-KEY_YYY: MOD3-KEY_ZZZ
        timeout_millis: 200 # Optional. No timeout by default.
      # Key press (MOD1-KEY_XXX) -> Sequence (MOD2-KEY_YYY, MOD3-KEY_ZZZ)
      MOD1-KEY_XXX: [MOD2-KEY_YYY, MOD3-KEY_ZZZ]
      # Execute a command
      MOD1-KEY_XXX:
        launch: ["bash", "-c", "echo hello > /tmp/test"]
      # Let `with_mark` also press a Shift key (useful for Emacs emulation)
      MOD1-KEY_XXX: { set_mark: true } # use { set_mark: false } to disable it
      # Also press Shift only when { set_mark: true } is used before
      MOD1-KEY_XXX: { with_mark: MOD2-KEY_YYY }
      # The next key press will ignore keymap
      MOD1-KEY_XXX: { escape_next_key: true }
      # Set mode to configure Vim-like modal remapping
      MOD1-KEY_XXX: { set_mode: default }
    application: # Optional
      not: [Application, ...]
      # or
      only: [Application, ...]
    device: # Optional
      not: [Device, ...]
      # or
      only: [Device, ...]
    mode: default # Optional
default_mode: default # Optional

For KEY_XXX, use these names. You can skip KEY_ and the name is case-insensitive. So KEY_CAPSLOCK, CAPSLOCK, and CapsLock are the same thing.

For the MOD1- part, the following prefixes can be used (also case-insensitive):

• Shift: SHIFT-
• Control: C-, CTRL-, CONTROL-
• Alt: M-, ALT-
• Windows: SUPER-, WIN-, WINDOWS-

You can use multiple prefixes like C-M-Shift-a. You may also suffix them with _L or _R (case-insensitive) so that remapping is triggered only on a left or right modifier, e.g. Ctrl_L-a.

If you use virtual_modifiers explained below, you can use it in the MOD1- part too.

exact_match defines whether to use exact match when matching key presses. For example, given a mapping of C-n: down and exact_match: false (default), and you pressed C-Shift-n, it will automatically be remapped to Shift-down, without you having to define a mapping for C-Shift-n, which you would have to do if you use exact_match: true.

application

application can be used for both modmap and keymap, which allows you to specify application-specific remapping.

application:
  not: Application
  # or
  not: [Application, ...]
  # or
  only: Application
  # or
  only: [Application, ...]

The application name can be specified as a normal string to exactly match the name, or a regex surrounded by /s like /application/.

To check the application names, you can use the following commands:

X11

$ wmctrl -x -l
0x02800003  0 slack.Slack           ubuntu-jammy Slack | general | ruby-jp
0x05400003  0 code.Code             ubuntu-jammy application.rs - xremap - Visual Studio Code

You may use the entire string of the third column (slack.Slack, code.Code), or just the last segment after . (Slack, Code).

GNOME Wayland

Use the following command or check windows' WMClass by pressing Alt+F2 and running lg command in LookingGlass:

busctl --user call org.gnome.Shell /com/k0kubun/Xremap com.k0kubun.Xremap WMClasses

KDE-Plasma Wayland

Xremap prints the active window to the console. However, it will only start printing, once a mapping has been triggered that uses an application filter. So you have to create a mapping with a filter using a dummy application name and trigger it. Then each time you switch to a new window xremap will print its caption, class, and name in the following style: active window: caption: '<caption>', class: '<class>', name: '<name>' You want to use the class for the filter.

If you use a systemd-daemon to manage xremap, the prints will be visible in the system-logs (Can be opened with journalctl -f)

Sway

swaymsg -t get_tree

Locate app_id in the output.

application-specific key overrides

Sometimes you want to define a generic key map that is available in all applications, but give specific keys in that map their own definition in specific applications. You can do this by putting the generic map at the bottom of the config, after any specific overrides, as follows.

# Emacs-style word-forward and word-back
keymap:
  - name: override to make libreoffice-writer go to end of word but before final space like emacs
    application:
      only: libreoffice-writter
    remap:
      Alt-f: [right, C-right, left]
  - name: generic for all apps
    remap:
      Alt-f: C-right
      Alt-b: C-left

Note how Alt-f and Alt-b work in all apps, but the definition of Alt-f is slightly different in LibreOffice Writer. When that app is active, the first definition overrides the second definition; but for any other app, only the second definition is found. This is because xremap uses the first matching definition that it finds.

device

Much like application, you may specify {keymap,modmap}.device.{not,only} in your configuration for device-specific remapping. Consistent with the global --device flag, device-matching strings may be any of:

• the full path of the device
• the filename of the device
• the device name
• a substring of the device name

To determine the names and paths of your devices, examine xremap's log output at startup.

device:
  not: '/dev/input/event0'
  # or
  not: ['event0', ...]
  # or
  only: 'Some Cool Device Name'
  # or
  only: ['Cool Device', ...]
  # etc...

Unlike for application, regexs are not supported for device.

virtual_modifiers

You can declare keys that should act like a modifier.

virtual_modifiers:
  - CapsLock
keymap:
  - remap:
      CapsLock-i: Up
      CapsLock-j: Left
      CapsLock-k: Down
      CapsLock-l: Right

keypress_delay_ms

Some applications have trouble understanding synthesized key events, especially on Wayland. keypress_delay_ms can be used to workaround the issue. See #179 for the detail.

License

xremap is available as open source under the terms of the MIT License.