Custom types

Subclass c.AdvancedType<T> when you need a format that is not a primitive, nested struct, or built-in helper. Implement:

• byteSize — fixed width of this slot in bytes
• read(bytes, offset, endian, label) — decode bytes to T
• write(bytes, offset, value, endian, label) — encode T into the buffer

The codec calls these hooks during c.read / c.write; c.sizeof includes byteSize in the struct total.

import { c } from "@craftycodie/cstruct";

/** Three-byte RGB, no padding. */
class Rgb24 extends c.AdvancedType<[number, number, number]> {
  readonly byteSize = 3;

  read(
    bytes: Uint8Array,
    offset: number,
    _endian: c.Endian,
    _label: string
  ): [number, number, number] {
    return [bytes[offset]!, bytes[offset + 1]!, bytes[offset + 2]!];
  }

  write(
    bytes: Uint8Array,
    offset: number,
    value: [number, number, number],
    _endian: c.Endian,
    label: string
  ): void {
    if (value.some((channel) => channel < 0 || channel > 255)) {
      throw new c.CStructError(`${label}: channel out of range`);
    }
    bytes[offset] = value[0];
    bytes[offset + 1] = value[1];
    bytes[offset + 2] = value[2];
  }
}

@c.struct()
class Pixel {
  @c.field(new Rgb24())
  color!: [number, number, number];
}

const pixel = c.read(Pixel, c.write(Pixel, { color: [255, 128, 0] } as Pixel));

C equivalent — three packed bytes, no terminator:

typedef struct {
    uint8_t color[3]; /* or: uint8_t r, g, b; */
} Pixel;

Built-in helpers (CBool, CString, CWString, CTime64) are implemented the same way and are available on c if you need instanceof checks. See the overview for built-in field types.