Try #71:

Author: bors[bot]

examples/dodge-the-creeps/rust/src/player.rs

@@ -62,7 +62,7 @@ impl GodotExt for Player {
             .base
             .get_node_as::<AnimatedSprite2D>("AnimatedSprite2D");
 
-        let mut velocity = Vector2::new(0.0, 0.0).inner();
+        let mut velocity = Vector2::new(0.0, 0.0);
 
         // Note: exact=false by default, in Rust we have to provide it explicitly
         let input = Input::singleton();
@@ -80,7 +80,7 @@ impl GodotExt for Player {
         }
 
         if velocity.length() > 0.0 {
-            velocity = velocity.normalize() * self.speed;
+            velocity = velocity.normalized() * self.speed;
 
             let animation;
 
@@ -101,10 +101,10 @@ impl GodotExt for Player {
         }
 
         let change = velocity * delta as f32;
-        let position = self.base.get_global_position().inner() + change;
+        let position = self.base.get_global_position() + change;
         let position = Vector2::new(
-            position.x.max(0.0).min(self.screen_size.inner().x),
-            position.y.max(0.0).min(self.screen_size.inner().y),
+            position.x.max(0.0).min(self.screen_size.x),
+            position.y.max(0.0).min(self.screen_size.y),
         );
         self.base.set_global_position(position);
     }

godot-core/src/builtin/mod.rs

@@ -4,9 +4,36 @@
  * file, You can obtain one at https://mozilla.org/MPL/2.0/.
  */
 
-//! Built-in types like `Vector2`, `GodotString` or `Variant`.
+//! Built-in types like `Vector2`, `GodotString` and `Variant`.
+//!
+//! # Background on the design of vector algebra types
+//!
+//! The basic vector algebra types like `Vector2`, `Matrix4` and `Quaternion` are re-implemented
+//! here, with an API similar to that in the Godot engine itself. There are other approaches, but
+//! they all have their disadvantages:
+//!
+//! - We could invoke API methods from the engine. The implementations could be generated, but it
+//!   is slower and prevents inlining.
+//!
+//! - We could re-export types from an existing vector algebra crate, like `glam`. This removes the
+//!   duplication, but it would create a strong dependency on a volatile API outside our control.
+//!   The `gdnative` crate started out this way, using types from `euclid`, but [found it
+//!   impractical](https://github.com/godot-rust/gdnative/issues/594#issue-705061720). Moreover,
+//!   the API would not match Godot's own, which would make porting from GDScript (slightly)
+//!   harder.
+//!
+//! - We could opaquely wrap types from an existing vector algebra crate. This protects users of
+//!   `gdextension` from changes in the wrapped crate. However, direct field access using `.x`,
+//!   `.y`, `.z` is no longer possible. Instead of `v.y += a;` you would have to write
+//!   `v.set_y(v.get_y() + a);`. (A `union` could be used to add these fields in the public API,
+//!   but would make every field access unsafe, which is also not great.)
+//!
+//! - We could re-export types from the [`mint`](https://crates.io/crates/mint) crate, which was
+//!   explicitly designed to solve this problem. However, it falls short because [operator
+//!   overloading would become impossible](https://github.com/kvark/mint/issues/75).
 
 mod macros;
+mod vector_macros;
 
 mod arrays;
 mod color;
@@ -16,8 +43,11 @@ mod string;
 mod string_name;
 mod variant;
 mod vector2;
+mod vector2i;
 mod vector3;
+mod vector3i;
 mod vector4;
+mod vector4i;
 
 pub mod meta;
 
@@ -29,5 +59,8 @@ pub use string::*;
 pub use string_name::*;
 pub use variant::*;
 pub use vector2::*;
+pub use vector2i::*;
 pub use vector3::*;
+pub use vector3i::*;
 pub use vector4::*;
+pub use vector4i::*;

godot-core/src/builtin/vector2.rs

@@ -4,80 +4,111 @@
  * file, You can obtain one at https://mozilla.org/MPL/2.0/.
  */
 
+use std::fmt;
+
 use godot_ffi as sys;
 use sys::{ffi_methods, GodotFfi};
 
-type Inner = glam::f32::Vec2;
-//type Inner = glam::f64::DVec2;
+use crate::builtin::Vector2i;
 
-#[derive(Default, Copy, Clone, Debug, PartialEq)]
+/// Vector used for 2D math using floating point coordinates.
+///
+/// 2-element structure that can be used to represent positions in 2D space or any other pair of
+/// numeric values.
+///
+/// It uses floating-point coordinates of 32-bit precision, unlike the engine's `float` type which
+/// is always 64-bit. The engine can be compiled with the option `precision=double` to use 64-bit
+/// vectors, but this is not yet supported in the `gdextension` crate.
+/// 
+/// See [`Vector2i`] for its integer counterpart.
+#[derive(Debug, Default, Clone, Copy, PartialEq)]
 #[repr(C)]
 pub struct Vector2 {
-    inner: Inner,
+    /// The vector's X component.
+    pub x: f32,
+    /// The vector's Y component.
+    pub y: f32,
 }
 
+impl_vector_operators!(Vector2, f32, (x, y));
+impl_vector_index!(Vector2, f32, (x, y), Vector2Axis, (X, Y));
+impl_common_vector_fns!(Vector2, f32);
+impl_float_vector_fns!(Vector2, f32);
+
 impl Vector2 {
-    pub fn new(x: f32, y: f32) -> Self {
-        Self {
-            inner: Inner::new(x, y),
-        }
+    /// Constructs a new `Vector2` from the given `x` and `y`.
+    pub const fn new(x: f32, y: f32) -> Self {
+        Self { x, y }
     }
 
-    pub fn from_inner(inner: Inner) -> Self {
-        Self { inner }
+    /// Constructs a new `Vector2` with all components set to `v`.
+    pub const fn splat(v: f32) -> Self {
+        Self { x: v, y: v }
     }
 
-    /// only for testing
-    pub fn inner(self) -> Inner {
-        self.inner
+    /// Constructs a new `Vector2` from a [`Vector2i`].
+    pub const fn from_vector2i(v: Vector2i) -> Self {
+        Self { x: v.x as f32, y: v.y as f32 }
     }
 
-    // Hacks for example
-    // pub fn length(self) -> f32 {
-    //     self.inner.length()
-    // }
-    // pub fn normalized(self) -> Vector2 {
-    //     Self::from_inner(self.inner.normalize())
-    // }
-    pub fn rotated(self, angle: f32) -> Self {
-        Self::from_inner(glam::Affine2::from_angle(angle).transform_vector2(self.inner))
-    }
-}
+    /// Zero vector, a vector with all components set to `0.0`.
+    pub const ZERO: Self = Self::splat(0.0);
 
-impl GodotFfi for Vector2 {
-    ffi_methods! { type sys::GDExtensionTypePtr = *mut Self; .. }
-}
+    /// One vector, a vector with all components set to `1.0`.
+    pub const ONE: Self = Self::splat(1.0);
 
-impl std::fmt::Display for Vector2 {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        self.inner.fmt(f)
-    }
-}
+    /// Infinity vector, a vector with all components set to `INFIINTY`.
+    pub const INF: Self = Self::splat(f32::INFINITY);
 
-// ----------------------------------------------------------------------------------------------------------------------------------------------
+    /// Left unit vector. Represents the direction of left.
+    pub const LEFT: Self = Self::new(-1.0, 0.0);
 
-type IInner = glam::IVec2;
+    /// Right unit vector. Represents the direction of right.
+    pub const RIGHT: Self = Self::new(1.0, 0.0);
 
-#[derive(Default, Copy, Clone, Debug, Eq, PartialEq)]
-#[repr(C)]
-pub struct Vector2i {
-    inner: IInner,
+    /// Up unit vector. Y is down in 2D, so this vector points -Y.
+    pub const UP: Self = Self::new(0.0, -1.0);
+
+    /// Down unit vector. Y is down in 2D, so this vector points +Y.
+    pub const DOWN: Self = Self::new(0.0, 1.0);
+
+    /// Returns the result of rotating this vector by `angle` (in radians).
+    pub fn rotated(self, angle: f32) -> Self {
+        Self::from_glam(glam::Affine2::from_angle(angle).transform_vector2(self.to_glam()))
+    }
+
+    /// Converts the corresponding `glam` type to `Self`.
+    fn from_glam(v: glam::Vec2) -> Self {
+        Self::new(v.x, v.y)
+    }
+
+    /// Converts `self` to the corresponding `glam` type.
+    fn to_glam(self) -> glam::Vec2 {
+        glam::Vec2::new(self.x, self.y)
+    }
 }
 
-impl Vector2i {
-    pub fn new(x: i32, y: i32) -> Self {
-        Self {
-            inner: IInner::new(x, y),
-        }
+/// Formats this vector in the same way the Godot engine would.
+impl fmt::Display for Vector2 {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "({}, {})", self.x, self.y)
     }
 }
 
-impl GodotFfi for Vector2i {
+impl GodotFfi for Vector2 {
     ffi_methods! { type sys::GDExtensionTypePtr = *mut Self; .. }
 }
 
-impl std::fmt::Display for Vector2i {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        self.inner.fmt(f)
-    }
+/// Enumerates the axes in a [`Vector2`].
+#[derive(Copy, Clone, Debug, Eq, Ord, PartialEq, PartialOrd)]
+#[repr(i32)]
+pub enum Vector2Axis {
+    /// The X axis.
+    X,
+    /// The Y axis.
+    Y,
+}
+
+impl GodotFfi for Vector2Axis {
+    ffi_methods! { type sys::GDExtensionTypePtr = *mut Self; .. }
 }

godot-core/src/builtin/vector2i.rs

@@ -0,0 +1,105 @@
+/*
+ * This Source Code Form is subject to the terms of the Mozilla Public
+ * License, v. 2.0. If a copy of the MPL was not distributed with this
+ * file, You can obtain one at https://mozilla.org/MPL/2.0/.
+ */
+
+use std::fmt;
+
+use godot_ffi as sys;
+use sys::{ffi_methods, GodotFfi};
+
+use crate::builtin::Vector2;
+
+/// Vector used for 2D math using integer coordinates.
+///
+/// 2-element structure that can be used to represent positions in 2D space or any other pair of
+/// numeric values.
+/// 
+/// It uses integer coordinates and is therefore preferable to [`Vector2`] when exact precision is
+/// required. Note that the values are limited to 32 bits, and unlike [`Vector2`] this cannot be
+/// configured with an engine build option. Use `i64` or [`PackedInt64Array`] if 64-bit values are
+/// needed.
+#[derive(Debug, Default, Clone, Copy, Eq, Ord, PartialEq, PartialOrd)]
+#[repr(C)]
+pub struct Vector2i {
+    /// The vector's X component.
+    pub x: i32,
+    /// The vector's Y component.
+    pub y: i32,
+}
+
+impl_vector_operators!(Vector2i, i32, (x, y));
+impl_vector_index!(Vector2i, i32, (x, y), Vector2iAxis, (X, Y));
+impl_common_vector_fns!(Vector2i, i32);
+
+impl Vector2i {
+    /// Constructs a new `Vector2i` from the given `x` and `y`.
+    pub const fn new(x: i32, y: i32) -> Self {
+        Self { x, y }
+    }
+
+    /// Constructs a new `Vector2i` with all components set to `v`.
+    pub const fn splat(v: i32) -> Self {
+        Self { x: v, y: v }
+    }
+
+    /// Constructs a new `Vector2i` from a [`Vector2`]. The floating point coordinates will be
+    /// truncated.
+    pub const fn from_vector2(v: Vector2) -> Self {
+        Self { x: v.x as i32, y: v.y as i32 }
+    }
+
+    /// Zero vector, a vector with all components set to `0`.
+    pub const ZERO: Self = Self::splat(0);
+
+    /// One vector, a vector with all components set to `1`.
+    pub const ONE: Self = Self::splat(1);
+
+    /// Left unit vector. Represents the direction of left.
+    pub const LEFT: Self = Self::new(-1, 0);
+
+    /// Right unit vector. Represents the direction of right.
+    pub const RIGHT: Self = Self::new(1, 0);
+
+    /// Up unit vector. Y is down in 2D, so this vector points -Y.
+    pub const UP: Self = Self::new(0, -1);
+
+    /// Down unit vector. Y is down in 2D, so this vector points +Y.
+    pub const DOWN: Self = Self::new(0, 1);
+
+    /// Converts the corresponding `glam` type to `Self`.
+    fn from_glam(v: glam::IVec2) -> Self {
+        Self::new(v.x, v.y)
+    }
+
+    /// Converts `self` to the corresponding `glam` type.
+    fn to_glam(self) -> glam::IVec2 {
+        glam::IVec2::new(self.x, self.y)
+    }
+}
+
+/// Formats this vector in the same way the Godot engine would.
+impl fmt::Display for Vector2i {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "({}, {})", self.x, self.y)
+    }
+}
+
+impl GodotFfi for Vector2i {
+    ffi_methods! { type sys::GDExtensionTypePtr = *mut Self; .. }
+}
+
+/// Enumerates the axes in a [`Vector2i`].
+#[derive(Copy, Clone, Debug, Eq, Ord, PartialEq, PartialOrd)]
+#[repr(i32)]
+pub enum Vector2iAxis {
+    /// The X axis.
+    X,
+    /// The Y axis.
+    Y,
+}
+
+impl GodotFfi for Vector2iAxis {
+    ffi_methods! { type sys::GDExtensionTypePtr = *mut Self; .. }
+}