Attention: Here be dragons

This is the latest (unstable) version of this documentation, which may document features not available in or compatible with released stable versions of Godot.

Checking the stable version of the documentation...

C++ usage guidelines

Rationale

Since Godot 4.0, the C++ standard used throughout the codebase is a subset of C++17. While modern C++ brings a lot of opportunities to write faster, more readable code, we chose to restrict our usage of C++ to a subset for a few reasons:

It makes it easier to review code in online editors. This is because engine contributors don't always have access to a full-featured IDE while reviewing code.
It makes the code easier to grasp for beginner contributors (who may not be professional C++ programmers). Godot's codebase is known to be easy to learn from, and we'd like to keep it that way.

To get your pull request merged, it needs to follow the C++ usage guidelines outlined here. Of course, you can use features not allowed here in your own C++ modules or GDExtensions.

Note

Prior to Godot 4.0, the C++ standard used throughout the codebase was C++03, with a handful of C++14 extensions. If you are contributing a pull request to the 3.x branch rather than master, your code can't use C++17 features. Instead, your code must be able to be built with a C++14 compiler.

The guidelines below don't apply to third-party dependencies, although we generally favor small libraries instead of larger solutions. See also Best practices for engine contributors.

Disallowed features

Any feature not listed below is allowed. Using features like constexpr variables and nullptr is encouraged when possible. Still, try to keep your use of modern C++ features conservative. Their use needs to serve a real purpose, such as improving code readability or performance.

Standard Template Library

We don't allow using the STL as Godot provides its own data types (among other things). See Why does Godot not use STL (Standard Template Library)? for more information.

This means that pull requests should not use std::string, std::vector and the like. Instead, use Godot's datatypes as described in the Core types documentation.

A 📜 icon denotes the type is part of Variant. This means it can be used as a parameter or return value of a method exposed to the scripting API.

`auto` keyword

Please don't use the auto keyword for type inference. While it can avoid repetition, it can also lead to confusing code:

// Not so confusing...
auto button = memnew(Button);

// ...but what about this?
auto result = EditorNode::get_singleton()->get_complex_result();

Keep in mind hover documentation often isn't readily available for pull request reviewers. Most of the time, reviewers will use GitHub's online viewer to review pull requests.

The auto keyword can be used in some special cases, like C++ lambda or Objective-C block definitions and C++ templates. Please ask before using templates with auto in a pull request.

// Full type definitions.
void (*mult64to128)(uint64_t, uint64_t, uint64_t &, uint64_t &) = [](uint64_t u, uint64_t v, uint64_t &h, uint64_t &l) { ... }
void (^JOYSTICK_LEFT)(GCControllerDirectionPad *__strong, float, float) = ^(GCControllerDirectionPad *dpad, float xValue, float yValue) { ... }

// Less clutter with auto.
auto mult64to128 = [](uint64_t u, uint64_t v, uint64_t &h, uint64_t &l) { ... }
auto JOYSTICK_LEFT = ^(GCControllerDirectionPad *dpad, float xValue, float yValue) { ... }

// Compare function for different types.
template <typename T1, typename T2>
constexpr auto MIN(const T1 m_a, const T2 m_b) {
    return m_a < m_b ? m_a : m_b;
}

We chose to forbid auto in all other cases. Thank you for your understanding.

Lambdas

Lambdas should be used conservatively when they make code effectively faster or simpler, and do not impede readability. Please ask before using lambdas in a pull request.

`#ifdef`-based include guards

Starting with 4.5, all files now use the #pragma once directive, as they improve readability and declutter macros. Use of #ifdef-based include guards are now actively discouraged.

`try`-`catch` blocks

C++ style exception handling using try and catch blocks is forbidden. This restriction is in place for several reasons, including performance, binary size and code complexity. Use Error macros instead.