Technical Stuff
Keyboard Switch is written using C# and .NET 6. Following is the list of technologies used in this app and some other technical aspects.
UI
The Keyboard Switch Settings app is written using Avalonia, a cross-platform UI framework for .NET. I chose it because unlike other UI frameworks for .NET, it's cross-platform, and I had to do literally nothing for it to work on Linux as well. The style is provided by FluentAvalonia.
Previous versions of the app used .NET Framework and WPF. Back then I didn't expect to have it working on different platforms.
Previous versions of the app also included a tray icon so you could see that the app is running. I removed it in version 3.0 as I don't see much value in keeping it and cluttering your tray.
Core Logic
Windows
The service app is just an app which runs in the background. I could have made it into a Windows service, but decided against it when I read somewhere that Windows services cannot set up keyboard hooks. I didn't actually verify this info, so I'm not sure it's correct, but the background app approach works just fine, so it's likely to stay this way.
The service app calls various native functions from the Windows' user32.dll. It uses Vanara to call them.
In the previous versions (up to 3.0) the service and the settings app were not separated. It was just an app running with a hidden window which showed up when you opened it. Having a hidden window always loaded is not that great of an idea, even though it used very little RAM.
Linux
At first I decided to run the app as a systemd service, but it proved not to provide much value, so I've since reverted this decision.
The service app uses X11, especially the X Keyboard Extension, and the X Test Extension. Currently it doesn't work on Wayland, even through XWayland. It uses the X11 package to call various native Xlib functions.
The app also uses Bash and xsel, but those are not used directly - rather they are used by the TextCopy library which the app uses for copying and pasting text.
The Settings App
The core logic of the settings app is implemented using ReactiveUI as the MVVM framework. I must say that it’s a real pleasure to use this framework. Also, Splat is used for service location.
The Global Keyboard Hook
The app uses libuiohook to create a global keyboard hook. The library is cross-platform and supports almost all popular OSes and architectures. Since this library is native and it's non-trivial to build, I've developed a .NET wrapper for it, SharpHook, so that I don't have to deal with a native library in this project.
Previously the app used a Windows keyboard hook directly, but that doesn't work on other systems.
Preferences
Preferences are stored in the user's local app data folder (under the KeyboardSwitch directory) on Windows. I didn't put it into the normal app data folder because that one is shared if you use the same Windows account on multiple machines, and that kind of defeats the purpose of this app being configured for each machine individually.
On Linux the settings are stored in the ~/.keyboard-switch directory.
This app uses Akavache as the engine for storing settings.
Logging
This app uses a plain-text log file to log the stuff it’s doing and errors it encounters along the way. Serilog is used as the logging library. The log file is stored in the same folder as the preferences and is rolled over after reaching 10 MB. If you really want to (and know how), you can change the logging configuration in the appsettings.json file (which is located in the same folder as the app itself).
Tests
There are none. Honestly, I don't even know how to approach testing, because the core logic of the app is essentially creating system-wide side effects, so don't think it's really testable. As for the settings app, well, I could test it, but that would take a lot of time, and the app is pretty simple, so I don't plan on doing it in the nearest future.
Docs
The docs are built and hosted on GitBook.
Previously these articles were built using Jekyll and hosted on GitHub Pages, and used the Minimal Mistakes theme.
Building from Source
You can build Keyboard Switch from source if you like. All projects (except the Windows installer) require .NET 6.
The Projects
The app consists of 7 projects:
KeyboardSwitch: The Keyboard Switch service itself
KeyboardSwitch.Core: Core functionality and interfaces for the projects
KeyboardSwitch.Settings: The Keyboard Switch Settings app
KeyboardSwitch.Settings.Core: The core logic of the settings app
KeyboardSwitch.Windows: The implementation of the core functionality for the Windows platform
KeyboardSwitch.Linux: The implementation of the core functionality for the Linux platform
KeyboardSwitch.Windows.Setup: The setup project
Building the App Itself
Building the service app and the settings app is quite straightforward.
On Windows you can simply use the Build-Portable.ps1 script located in the build/windows directory. Bear in mind that you have to call this script from the solution root directory: .\build\windows\Build-Portable.ps1
. It will create the KeyboardSwitch-Portable.zip file in the bin directory which you can then unpack into wherever you like.
On Linux you should use the build-tar.sh script to build the tar.gz file, the build-deb.sh script to build the deb package, and the build-rpm.sh script to build the RPM package. They are located in the build/tar, build/deb, and build/rpm directories respectively. As with the Windows script, you have to execute these scripts from the solution root: ./build/tar/build-tar.sh
, ./build/deb/build-deb.sh
, and ./build/rpm/build-rpm.sh
. The scripts' results will be in the bin directory.
Alternatively, you can build the projects using Visual Studio 2019 or later, or with the dotnet
tool. Note however that you have to build the whole solution. The startup project is KeyboardSwitch.Settings
, and if you build it, it won't actually build the KeyboardSwitch
project, because the settings app doesn't directly depend on the service app.
All projects (except the installer) are built into a shared bin folder located in the solution root. Again, this is because the settings app needs the service app to be in the same folder, but the projects don't depend on each other.
It's better to use dotnet publish
than simply using the raw build results. You can look into how the build scripts call dotnet publish
.
The installer project is excluded from the solution build sequence as it's not always needed.
Building the Windows Installer
If you want to properly install the app on Windows, you can build the installer. Unlike the other projects, this one requires .NET Framework 4.8. This is because it's built with WixSharp, which in turn is based on the WiX toolset, and as of version 3.11.2 WiX doesn't support .NET 6.
Simply run the build, and it will generate the MSI installer in the project's bin folder. Before the build, it calls dotnet publish
to use its output.
Last updated