RawBuffer

Use RawBuffer when Volten should not pack your data.

import { RawBuffer } from '@volten/core';

const packed = new ArrayBuffer(64);
const data = new RawBuffer(packed, 'array<vec4f, 4>');

You provide the bytes and the WGSL type string.

That makes RawBuffer useful for shapes Volten does not model directly yet, such as structs with runtime-sized arrays at the end of the struct:

struct Particle {
  position: vec3f,
  mass: f32,
}

struct ParticleBuffer {
  count: u32,
  particles: array<Particle>,
}

const packed = packParticleBufferSomehow(particles);
const data = new RawBuffer(packed, 'ParticleBuffer', 'rw');

Access Mode

RawBuffer uses the same access modes as Buffer:

const input = new RawBuffer(packed, 'array<f32>', 'r');
const inout = new RawBuffer(packed, 'array<f32>', 'rw');

'r' means shader read-only. 'rw' means shader read-write.

If omitted, the access mode defaults to 'rw'.

Use set() to replace the full byte contents:

data.set(new Uint8Array(64));

Use update() to write a byte range. The offset is measured in bytes:

data.update(new Uint8Array([1, 2, 3, 4]), 16);

Reach for RawBuffer when:

you have a custom packed binary layout
you need a WGSL type that Volten’s struct() and array() helpers cannot express cleanly yet
you want to handle packing and unpacking yourself