The keyboard remapping software you never knew you wanted.
Inspired by the open source keyboard firmware project,
QMK, the goal of qsk
is enable similar
features on arbitrary keyboards connected to a host system. For example, the
built-in keyboard on a laptop or that crummy old generic keyboard that came
with your 1990s era department store desktop computer.
- standard keyboard remapping, eg remap
F
->U
- composable layers of keymappings activated by keys with special functionality
- "tap toggle", which causes a given key to send its usual keystroke when tapped within a given time limit and to activate a specified layer while held
This feature set is still fairly small relative to QMK's quite prolific feature set. Features are implemented on an as-needed basis -- contributions welcome!
Install:
cargo install qsk
Get a list of available devices:
qsk list-devices
After identifying the device you want to use, run the remapper:
sudo qsk remap /path/to/device-file
Note: sudo
is necessary above because by default your linux login
user won't have the permissions necessary to grab your chosen keyboard input
device nor to create new virtual keyboard device through which your remapped
key strokes will be emitted.
The previous section describes how to use the binary shipped via crates.io,
which for now can't have its keymaps customized. In the future it will be
possible to pass it a path to a file with a keymapping DSL/script. For now,
keyboard remapping definitions must be compiled in. To make this easier, qsk
provides a cargo-generate
template
that helps you get started quickly to create a qsk
project of your own:
cargo generate --git https://github.com/waynr/qsk.git qsk-template
cargo-generate
will prompt you for values to fill in the qsk
template
project, one of which will be "Project Name". The value you pass to this will
be the name of the directory of your new qsk
project. To build it:
cd $PROJECT_NAME
cargo build
Get a list of available devices:
./target/debug/$PROJECT_NAME list-devices
After identifying the device you want to use, run the remapper:
sudo ./target/debug/$PROJECT_NAME remap /path/to/device-file
The abovementioned template produces a main.rs
that looks like the following:
use std::error;
use qsk_macros;
use qsk::entrypoint;
fn main() -> Result<(), Box<dyn error::Error>> {
let layer_composer = qsk_macros::remap!(
ModLayer[Active]: {
Y -> HOME,
F -> TapToggle(Navigation, F),
},
Navigation: {
END -> Exit(),
Y -> HOME,
U -> PAGEDOWN,
I -> PAGEUP,
O -> END,
H -> LEFT,
J -> DOWN,
K -> UP,
SEMICOLON -> RIGHT,
},
)?;
entrypoint(layer_composer)?;
Ok(())
}
This demonstrates the qsk_macros::remap!
macro which takes as input a
mini-DSL that simplifies definition of layered keyboard remapping. The only
alternative to this currently would be defining a qsk_types::LayerComposer
directly in Rust. All you have to do in your generated project is to update the
key mappings, save, and build.
There are several categories of identifier to be concerned with when defining keyboard remapping layers:
Layer Name
are identifiers likeNavigation
andModLayer
shown above. These names precede a colon with an optional set of square brackets and are used by Key Functions. For example,Navigation
inTapToggle
indicates that theNavigation
layer should be activated when the key on the left side of the->
is held.Layer Option
are identifiers likeActive
in square brackets above. These are used to configure individual layers.Key Codes
are identifiers likeK
,END
, andUP
shown above. On the left side of a->
the key code indicates the "input" key that will be remapped. On the right side of a->
this indicates what key code will be output given the key code on the left.Key Functions
are identifies likeExit
andTapToggle
shown above. These can only appear on the right side of a->
and are used to bestow special properties on the corresponding key indicated on the left side of the->
.
TapToggle(<layer_ref>, <tap_key>)
When the key on the left side of the->
is pressed and held, the layer named<layer_ref>
is activated. When it is tapped within the default tap toggle timeout (180 milliseconds).Exit()
When the key on the left side of the->
is pressed, the program will exit gracefully.
Active
indicates that the layer should be set to "active" state on program initialization.
Assuming you are familiar with QMK, you might be interested to know how this project differs from it.
If you're not familiar with QMK, then the TL;DR is that it is software that you can use to customize the behavior of supported keyboards to dynamically alter the behavior of keys according. If you would like to know more, please check out the QMK documentation site but be warned that it is a somewhat deep rabbithole.
QMK compiles to firmware that must be loaded onto a given target keyboard that it supports. As such, it imposes no resource consumption burden on the host system and minimizes latency due to the (presumably) dedicated nature of its microcontroller.
qsk
, on the otherhand, compiles to a binary that necessarily runs on the host
system receiving the original hardware input events and sending the same or
different events as determined by its configuration.
It requires permissions on the host system necessary to:
- Grab the input of an existing input device to receive its input events.
- Create a new virtual input device to which it sends keystrokes that it either generates or passes through from the source input device.
Additionally, you must tell qsk
what source input device to grab when
executing the binary.
As you can imagine, there is potential for a tool like qsk
to inject
non-trivial between the time it receives a keyboard event and the time that it
sends corresponding potentially altered keyboard events.
The intent in choosing Rust for this tool, aside from indulging a personal preference, is to safely minimize latency while providing opportunities to extend its features along a number of axes. That said, there has not yet been an effort to characterize the latency involved here but I (waynr) can attest that it doesn't seem to be noticeable for everyday use.
In qsk
we don't map desired keyboard events/behaviors to specific hardware
keys but to other keyboard or input events. Because of this you have to be
conscious of what input events your desired physical device and target host OS
map to in order to effectively remap it.
It is possible that we could in the future do something fancy like inspect details of a given input device and allow the user to configure it using a GUI and a presumed default layout presented to us by the input event interface. Contributions in this area are encouraged!
Because of the nature of qsk
it's most likely that support needs to be
kernel/OS specific since that is the most natural API boundary where an
interface might be made available for such things.
For Linux we implement device support through
evdev
, a generic input event
interface provided by the kernel. This can be illustrated with the following
rough chain of relationships:
brain -> fingers -> keyboard -> CPU interrupts -> interrupt handlers (kernel) ->
evdev subsystem (kernel/userspace) -> qsk -> X11 input drivers OR libinput for
Wayland -> your program
Beware that while qsk
is attached to a given input source it will "grab" that
input so that it has the exclusive right to read events from it.
I don't have any mac computers so it's not practical for me to implement support for them. I am somewhat skeptical that it is as easy as for Linux, but happy to be proven wrong by someone with a mac. Please feel encouraged to implement support, I would be happy to provide whatever guidance you want or need!
I don't have any windows computers so it's not practical for me to implement support for them. I am somwhat skeptical that it is as easy as for Linux, but am happy to be proven wrong by someone with a windows. Please feel encouraged to implement support, I would be happy to provide whatever guidance you want or need!
Got an operating system or computing paradigm I don't know about? Let me know!
Wayne Warren is an everyday normal guy who likes to write software in Rust and
was driven to write qsk
out of annoyance at the lack of muscle memory
compatibility between his highly-customized mechanical keyboard firmware and
his various laptops' super uncustomizable built-in keyboards.