// NTSC post-process effect: fisheye distortion, scanlines, and color bleeding.
// Produces a retro CRT/NTSC look using YIQ color space and C64-style dithering.
#include "sequence_uniforms"
#include "render/fullscreen_uv_vs"
#include "math/noise"
#include "math/color"
#include "math/color_c64"
#include "debug/debug_print"

const PI = 3.14159265;
const TAU = 6.28318530718;

const XSIZE = 54.0 * 8.;
const YSIZE = 33.0 * 8.;
const SCAN_FLICKER = 2.33;
const X_INTERFERENCE = 1.1;
const Y_INTERFERENCE = 0.101;
const LUMA_BRIGHTNESS = 1.1;
const CHROMA_SATURATION = 1.6;
const BLUR_SIZE = 0.2;
const LUMA_BLUR = 1.7;
const CHROMA_BLUR = 0.7;
const CHROMA_SIZE = 6.0;
const SUB_CARRIER = 2.1;
const CROSS_TALK = 0.1;
const CROSS_Y_INTERFERENCE = 30.;
const CHROMA_MOD_FREQ = (0.4 * PI);


@group(0) @binding(0) var input_sampler: sampler;
@group(0) @binding(1) var input_texture: texture_2d<f32>;
@group(0) @binding(2) var<uniform> uniforms: UniformsSequenceParams;

// Barrel (fisheye) distortion: strength > 0 = barrel, < 0 = pincushion
fn fisheye(uv: vec2f, scale: f32) -> vec2f {
  const strength = vec2f(0.1, 0.24);
  return uv * (1.0 + scale * strength * (uv * uv).yx) * 0.60 + .50;
}

fn vignette(uv: vec2f) -> f32 {
  const vignetteRounding = 160.0;
  const vignetteSmoothness = 0.7;
  let uv2 = 2.0 * uv - 1.0;
  let amount = 1.0 - sqrt(pow(abs(uv2.x), vignetteRounding) + pow(abs(uv2.y), vignetteRounding));
  return smoothstep(0., vignetteSmoothness, amount);
}

// returns Luma, chroma subcarrier level, phase, transparency
fn get_luma_chroma_phase_a(uv: vec2f) -> vec4f {
  let rgba = textureSample(input_texture, input_sampler, uv);
  return rgba_to_luma_chroma_phase(rgba, uv.y, YSIZE);
}

fn get_value(uv: vec2f, off: f32, yscale: f32) -> vec4f {
  return get_luma_chroma_phase_a(uv + off * vec2f(1., yscale));
}

fn peak(y: f32, ypos: f32, scale: f32) -> f32 {
  return clamp((y - 1.) * scale * log(abs(y - ypos)), 0.0, 1.0);
}

// 6-taps Luma horizontal filtering
//   fs = 3.84 MHz (Nyquist 1.92 MHz)
//   Passband: 0–2.8 MHz
//   Stopband: 3.4–3.84 MHz (>20 dB atten.)
//   => firpm(12, [0 2.8/3.842 3.4/3.842 1], [1 1 0 0])
const luma_filter = array<f32, 2 * 6 + 1>(
   0.0102, 0.0214, 0.0387, -0.0018, -0.0785, -0.1572,
  -0.1698,
   0.1275, 0.4924, 0.5381,  0.4924,  0.1275, -0.1698
);
// Chroma:
//   fs = 3.84 MHz
//   Passband: 3.3–3.7 MHz (around fsc)
//   Stopbands: 0–2.5 MHz (>40 dB) and 3.9+ MHz
//   => firpm(12, [0 2.5/1.92 3.3/1.92 3.7/1.92 1], [0 0 1 1 0])
const chroma_filter = array<f32, 2 * 6 + 1>(
  -0.0123, -0.0456, -0.0892,  0.0234,  0.1678,  0.2984,
   0.3456,
   0.0000,  0.3456,  0.2984,  0.1678,  0.0234, -0.0892
);

fn get_signal(uv: vec2f, d: f32) -> vec4f {
  var signal = vec4f(0.0);
  for (var i = 0; i <= 12; i += 1) {
    let offset = f32(i) - 6.0;
    let suml = luma_filter[i] * get_value(uv, offset * d, 0.67);
    let sumc = chroma_filter[i] * get_value(uv, offset * d * CHROMA_SIZE, 0.67);
    signal += vec4f(suml.x, sumc.y, sumc.z, suml.a);
  }
  let base = get_luma_chroma_phase_a(uv);
  return mix(signal, base, vec4f(LUMA_BLUR, CHROMA_BLUR, CHROMA_BLUR, 1.));
}

fn randomized_f32(p: vec2f, t: f32) -> f32 {
  return hash_2f_alt(vec2f(p * 0.152 + t * 1500. + 50.0));
}

@fragment fn fs_main(in: VertexOutput) -> @location(0) vec4f {
    let t = uniforms.time;
    let bt = uniforms.beat_phase;

    // Fisheye/barrel distortion
    var uv = fisheye(in.st, 0.18);
    let mframe = sin(bt * 32.) * 1.2;
    uv.y += mframe * SCAN_FLICKER / uniforms.resolution.y;  // flicker at target resolution

    // interference
    let cur_line = round(uv.y * YSIZE);
    var r = randomized_f32(vec2f(0.0, cur_line), uniforms.time);
    if (r > 0.995) { r *= 3.0; }

    let x_interf = X_INTERFERENCE * r / XSIZE;
    let y_interf = Y_INTERFERENCE * r * peak(uv.y, 0.2, 0.03);
    uv.x += x_interf - y_interf;

    // luma fringing
    let d = (BLUR_SIZE + y_interf * 100.0) / XSIZE;
    var signal = get_signal(uv, d);

    // luma / chroma saturation
    let lchroma = signal.y * CHROMA_SATURATION;
    let phase = signal.z * TAU;
    signal.x *= LUMA_BRIGHTNESS;
    signal.y = lchroma * sin(phase);
    signal.z = lchroma * cos(phase);

    // color subcarrier signal, crosstalk
    let chroma_phase = t * 60.0 * PI * 0.6667;
    let mod_phase = chroma_phase + dot(uv, vec2f(1.0, 0.1)) * CHROMA_MOD_FREQ * XSIZE * 2.0;
    let scarrier = SUB_CARRIER * lchroma;
    let i_mod = cos(mod_phase);
    let q_mod = sin(mod_phase);

    signal.x *= 1.0 + CROSS_TALK * scarrier * q_mod - y_interf * CROSS_Y_INTERFERENCE;
    signal.y *= 1.0 + scarrier * i_mod;
    signal.z *= 1.0 + scarrier * q_mod;

    // convert back to rgb
    var col = yiqa_to_rgba(signal);
    // Slight NTSC warm tint (boost red/green, attenuate blue)
    col *= vec4f(1.04, 1.01, .94, 1.);
//    col = Dither(col, uv, XSIZE, YSIZE);

    let border_col = get_border_c64(uv, uniforms.beat_time, YSIZE);

    let v_strength = vignette(uv);
    let scanl = 0.82 + 0.5 * sin(PI * uv.y * uniforms.resolution.y / 2.);
    col = scanl * mix(border_col, col, v_strength);
    col = clamp(col, vec4f(0.), vec4f(1.0));

    // Black outside screen edges
    if (uv.x <= 0.0 || uv.x >= 1.0 || uv.y <= 0.0 || uv.y >= 1.0) {
	// discard;
    }

    col = debug_f32(col, in.position.xy / 2., vec2f(100., 75.), uniforms.beat_time);
    col = debug_str(col, in.position.xy / 2., vec2f(100., 150.), vec4u(0x48656C6Cu, 0x6F000000u, 0u, 0u), 5u);
    return col;
}