Compile-time checks to ensure structs conform to WGSL's memory layout rules

All validation is performed at compile time using declarative macros (macro_rules!) — no proc macros, no build dependencies, no compile-time overhead beyond the checks themselves.

The core of this crate is the [shader_layout], [shader_layout_compat] macros and the [ShaderLayout], [ShaderLayoutCompat] traits.

ShaderLayout corresponds to https://www.w3.org/TR/WGSL/#alignment-and-size.

ShaderLayoutCompat is a stricter subset that enforces the uniform address space layout constraints (without the uniform_buffer_standard_layout extension). Every type implementing ShaderLayoutCompat also implements ShaderLayout.

Both macros validate every field's alignment and the overall struct size at compile time. If a constraint is violated, compilation fails with a clear error message:

shader_layout! {
    pub struct OffsetUnaligned {
        a1: f32,        // align 4
        a4: glam::Vec3, // align 16 — offset 4 is not a multiple of 16!
    }
}
// error[E0080]: evaluation panicked: Failed to implement `ShaderLayout`: field `OffsetUnaligned::a4` (`glam::Vec3`) is not properly aligned. The offset is 4 but required align is 16

use const_shader_layout::{shader_layout, shader_layout_compat, ShaderLayout, ShaderLayoutCompat};
use glam::{Vec2, Vec3, Vec4};

shader_layout! {
    pub struct MyStorage {
        a1: f32,
        a2: [f32; 2],
        a3: [f32; 1],
        a4: Vec3,
        a5: f32,
        a6: Vec3,
        p1: f32, // Padding needed otherwise struct size (44) won't match shader size (48).
    }
}

shader_layout_compat! {
    pub struct Nested {
        a1: [Vec4; 2],
        a2: Vec3,
        a3: f32
    }
}

shader_layout_compat! {
    pub struct MyUniform {
        a1: Nested,
        a2: Vec3,
        // Padding is implicit, because struct size is 64 aligned to 16 in `repr(C)` which matches shader size (64).
    }
}

See https://github.com/beicause/const_shader_layout/tree/master/tests for what this supports and checks.

Features

This crate focuses on layout validation only. For safe cast/transmute utilities, pair it with other crates like bytemuck.