ph-color 0.1.1

Fixed-point no_std color math for embedded targets: conversion, transfer functions, matrices, gain, and interpolation
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58

# ph-color

Fixed-point `no_std` color math for embedded targets.

[![crates.io](https://img.shields.io/crates/v/ph-color.svg)](https://crates.io/crates/ph-color)
[![docs.rs](https://docs.rs/ph-color/badge.svg)](https://docs.rs/ph-color)

> [!NOTE]
> **Lifecycle:** Active
> **Distribution:** Version `0.1.1` on
> [crates.io](https://crates.io/crates/ph-color), with API documentation on
> [docs.rs](https://docs.rs/ph-color).
> **Model conformance:** N/A
> **Physical evidence:** N/A

This package applies baked coefficients and tables. It does not derive,
invert, adapt, or solve. Host derivation lives in `ph-color-bake`.

```toml
[dependencies]
ph-color = "0.1"
```

```rust
use ph_color::{Color, Linear, Q0_16, Srgb};

let black = Color::<Srgb, Linear>::new([Q0_16::ZERO; 3]);
let white = Color::<Srgb, Linear>::new([Q0_16::ONE; 3]);
let midpoint = black.lerp(white, Q0_16::from_raw(32_768));
assert!(midpoint.ch[0].to_raw() >= 32_768);
```

The crate currently exposes `quantize` / `quantize_round` / `expand`, `Color`
type-state, numeric `Matrix3` / `Gain` apply, `InterpLut`, `lerp`, `Gradient`,
shipped sRGB EOTF/OETF `InterpLut` constants, feature `oklab` (linear-sRGB ↔
Oklab), and optional `feature = "f32"` (`ColorF32` and `*_f32`). Oklab `f32` is
out of scope. Host `ph-color-bake` derives tables. Frozen golden vectors in
`tests/golden/` are replayed by host tests. Named-consumer host sketches are
in [`examples/`](examples/). Numeric and type-state guarantees are in the
[repository numeric policy](https://github.com/photon-circus/ph-color/blob/main/docs/NUMERICS.md).

## Compatibility and limits

- MSRV: Rust 1.94.0, edition 2024.
- Default build: `no_std`, no allocation, no unsafe code, no `f64`, and zero
  dependencies.
- Optional features: `f32` and `oklab`; Oklab has no `f32` entry points.
- Fixed-point matrix and gain output uses one widening accumulation followed by
  round-half-away-from-zero and saturation. It never wraps.
- Matrices and gains accept linear colors; transfer LUTs cross linear/encoded
  state; encoded interpolation is limited to perceptual color spaces.
- Color processing stops at quantization. Zero-parameter round-to-nearest is
  provided; firmware owns custom thresholds, masks, positional or frame state,
  dithering, residual accumulation, output packing, and hardware I/O.

## License

MIT