Attention: Here be dragons (unstable version)
This is the latest
(unstable) version of this documentation, which may document features
not available in or compatible with released stable versions of Redot.
Checking the stable version of the documentation...
What is GDExtension?
Introduction
GDExtension is a Redot-specific technology that lets the engine interact with native shared libraries at runtime. You can use it to run native code without compiling it with the engine.
Note
GDExtension is not a scripting language and has no relation to GDScript.
Differences between GDExtension and C++ modules
You can use both GDExtension and C++ modules to run C or C++ code in a Redot project.
They also both allow you to integrate third-party libraries into Redot. The one you should choose depends on your needs.
Warning
Our long-term goal is that GDExtensions targeting an earlier version of Redot will work in later minor versions, but not vice-versa. For example, a GDExtension targeting Redot 4.2 should work just fine in Redot 4.3, but one targeting Redot 4.3 won't work in Redot 4.2.
However, GDExtension is currently experimental, which means that we may break compatibility in order to fix major bugs or include critical features. For example, GDExtensions created for Redot 4.0 aren't compatible with Redot 4.1 (see Updating your GDExtension for 4.1).
Advantages of GDExtension
Unlike modules, GDExtension doesn't require compiling the engine's source code, making it easier to distribute your work. It gives you access to most of the API available to GDScript and C#, allowing you to code game logic with full control regarding performance. It's ideal if you need high-performance code you'd like to distribute as an add-on in the asset library.
Also:
GDExtension is not limited to C and C++. Thanks to third-party bindings, you can use it with many other languages.
You can use the same compiled GDExtension library in the editor and exported project. With C++ modules, you have to recompile all the export templates you plan to use if you require its functionality at runtime.
GDExtension only requires you to compile your library, not the whole engine. That's unlike C++ modules, which are statically compiled into the engine. Every time you change a module, you need to recompile the engine. Even with incremental builds, this process is slower than using GDExtension.
Advantages of C++ modules
We recommend C++ modules in cases where GDExtension isn't enough:
C++ modules provide deeper integration into the engine. GDExtension's access is not as deep as static modules.
You can use C++ modules to provide additional features in a project without carrying native library files around. This extends to exported projects.
Note
If you notice that specific systems are not accessible via GDExtension but are via custom modules, feel free to open an issue on the godot-cpp repository to discuss implementation options for exposing the missing functionality.
Supported languages
The Redot developers officially support the following language bindings for GDExtension:
C++ (tutorial)
Note
There are no plans to support additional languages with GDExtension officially. That said, the community offers several bindings for other languages (see below).
The bindings below are developed and maintained by the community:
Note
Not all bindings mentioned here may be production-ready. Make sure to research options thoroughly before starting a project with one of those. Also, double-check whether the binding is compatible with the Redot version you're using.
Version compatibility
GDExtension add-ons compiled for a given Redot version are only guaranteed to work with the same minor release series. For example, a GDExtension add-on compiled for Redot 4.0 will only work with Redot 4.0, 4.0.1, 4.0.2. In addition, GDExtension is not compatible with Redot 3.x.
GDExtension add-ons are also only compatible with engine builds that use the level of floating-point precision the extension was compiled for. This means that if you use an engine build with double-precision floats, the extension must also be compiled for double-precision floats. See Large world coordinates for details.