wallswitch 0.61.1

//! Mandelbrot Set fractal generator overlay.
//!
//! Renders Mandelbrot Set overlays using the distance estimator colouring method
//! shared with the Julia generator via [`color_distance_estimator`]. The
//! `random` constructor runs a parallelised entropy sweep to find the zoom level
//! that maximises structural detail, then applies `dynamic_autofocus` to refine
//! the centre point and derive a level-of-detail iteration count.
//!
//! Generator function: `z(n+1) = z(n)^2 + c`, starting from `z = 0`,
//! where `c` varies over the viewport grid coordinates.

use crate::{
    ColorRGB, Complex, FractalConfig, FractalDescriptor, FractalPreset, MAX_ITERATIONS,
    MIN_ITERATIONS, Monitor, NEON_PALETTES, ProceduralEffect, ROTATION_STEPS, RandomExt, Viewport,
    ViewportSpecs, WallSwitchResult, color_distance_estimator, get_random_integer,
    mandelbrot_escape,
};
use rayon::prelude::*;
use std::cmp::Ordering;

// ============================================================================
// PRESET TABLE
// ============================================================================

/// All available Mandelbrot coordinate presets.
const MANDELBROT_PRESETS: &[FractalPreset] = &[
    FractalPreset {
        center: Complex {
            re: -0.8115,
            im: 0.2014,
        },
        fractal_name: "Tendril Valley Filaments",
        effect_name: ProceduralEffect::Mandelbrot,
    },
    FractalPreset {
        center: Complex {
            re: -0.156,
            im: 1.033,
        },
        fractal_name: "Dreadlock Valley Basin",
        effect_name: ProceduralEffect::Mandelbrot,
    },
    FractalPreset {
        center: Complex {
            re: -0.38,
            im: 0.66,
        },
        fractal_name: "Starburst Star Valley",
        effect_name: ProceduralEffect::Mandelbrot,
    },
    FractalPreset {
        center: Complex {
            re: -0.56226,
            im: 0.64273,
        },
        fractal_name: "Feathered Filament Cascades",
        effect_name: ProceduralEffect::Mandelbrot,
    },
    FractalPreset {
        center: Complex {
            re: -0.77568377,
            im: 0.13646737,
        },
        fractal_name: "Deep Seahorse Tail Spiral",
        effect_name: ProceduralEffect::Mandelbrot,
    },
    FractalPreset {
        center: Complex { re: -1.45, im: 0.0 },
        fractal_name: "West Needle Crown Filaments",
        effect_name: ProceduralEffect::Mandelbrot,
    },
    FractalPreset {
        center: Complex {
            re: -0.55,
            im: 0.62,
        },
        fractal_name: "Pentagonal Star Valley",
        effect_name: ProceduralEffect::Mandelbrot,
    },
    FractalPreset {
        center: Complex {
            re: -1.625,
            im: 0.0,
        },
        fractal_name: "Bi-Directional Filament",
        effect_name: ProceduralEffect::Mandelbrot,
    },
    FractalPreset {
        center: Complex {
            re: -0.70176,
            im: 0.35422,
        },
        fractal_name: "Peacock Tail Valley Plumes",
        effect_name: ProceduralEffect::Mandelbrot,
    },
    FractalPreset {
        center: Complex {
            re: -0.8113,
            im: 0.2015,
        },
        fractal_name: "Medusa Tendril Clusters",
        effect_name: ProceduralEffect::Mandelbrot,
    },
];

// ============================================================================
// GENERATOR
// ============================================================================

/// A procedural generator for rendering Mandelbrot Set fractals onto desktop backgrounds.
pub struct MandelbrotGenerator {
    /// The selected coordinate preset defining the viewport focus point.
    pub preset: FractalPreset,
    /// Shared viewport and rendering configuration.
    pub config: FractalConfig,
}

impl FractalDescriptor for MandelbrotGenerator {
    #[inline(always)]
    fn config(&self) -> &FractalConfig {
        &self.config
    }

    #[inline(always)]
    fn center(&self) -> Complex {
        self.preset.center
    }

    // Mandelbrot: `is_julia` defaults to `false` — the viewport maps `c`, not `z`.

    #[inline(always)]
    fn render_pixel(&self, c: Complex, scale: f64, _max_radius: f64) -> (ColorRGB, f64, f64) {
        let (i, z, dz) = mandelbrot_escape(c, self.config.scan_iterations);
        color_distance_estimator(
            i,
            self.config.scan_iterations,
            z,
            dz,
            scale,
            self.config.color_palette,
        )
    }

    fn info_text(&self) -> String {
        format!(
            "fractal [{}]\n\
             f(z) = z^2 + c, where c = {:8.5} {} {:7.5}i (iter = {:4}, zoom = {:.5}), color: {}",
            self.preset.fractal_name,
            self.preset.center.re,
            if self.preset.center.im >= 0.0 {
                "+"
            } else {
                "-"
            },
            self.preset.center.im.abs(),
            self.config.scan_iterations,
            self.config.zoom,
            self.config.color_palette
        )
    }
}

impl MandelbrotGenerator {
    /// Constructs a randomized, non-fitted Mandelbrot Generator.
    ///
    /// This acts as a base constructor, returning an error if
    /// the preset or color palette tables are empty.
    pub fn new() -> WallSwitchResult<Self> {
        let preset = MANDELBROT_PRESETS.get_random_sample()?;
        let color_palette = NEON_PALETTES.get_random_sample()?;

        Ok(Self {
            preset,
            config: FractalConfig {
                scan_iterations: MIN_ITERATIONS,
                color_palette,
                zoom: 0.0025,
                rotation: Complex::one(),
            },
        })
    }

    /// Constructs a randomised, monitor-fitted Mandelbrot generator.
    ///
    /// Reuses [`new`](Self::new) to construct the base randomized generator.
    /// Runs a parallelised entropy sweep over a geometric zoom sequence and
    /// `ROTATION_STEPS` rotation angles, selecting the combination that
    /// maximises information content. `dynamic_autofocus` then refines the
    /// centre and derives the level-of-detail iteration count.
    pub fn random(monitor: &Monitor) -> WallSwitchResult<Self> {
        let (width, height) = (
            monitor.resolution.width as u32,
            monitor.resolution.height as u32,
        );

        // 1. Initialize base randomized generator configuration
        let mut mandelbrot = Self::new()?;

        // 2. Perform parallelized entropy sweep to optimize zoom and rotation
        let rotation_phasors: Vec<Complex> = Complex::rotation_phasors(ROTATION_STEPS).collect();
        let zooms_count = get_random_integer(30, 50);
        let candidates = generate_zoom_candidates(zooms_count, ROTATION_STEPS);
        let preset_center = mandelbrot.preset.center;

        let (best_base_zoom, best_rotation, _) = candidates
            .par_iter()
            .map(|&(base_zoom, r_idx)| {
                let aspect_ratio = width as f64 / height as f64;
                let adjusted_zoom = if aspect_ratio > 1.0 {
                    base_zoom * aspect_ratio.sqrt()
                } else {
                    base_zoom
                };

                let rotation = rotation_phasors[r_idx];
                let entropy = calculate_entropy(
                    preset_center,
                    adjusted_zoom,
                    rotation,
                    MIN_ITERATIONS,
                    width,
                    height,
                );
                (base_zoom, rotation, entropy)
            })
            .max_by(|a, b| a.2.partial_cmp(&b.2).unwrap_or(Ordering::Equal))
            .unwrap_or((0.0002, Complex::one(), 0.0));

        mandelbrot.config.zoom = best_base_zoom;
        mandelbrot.config.rotation = best_rotation;

        // 3. Apply aspect ratio scaling and focus refinement
        mandelbrot.optimize_fit(width, height);
        mandelbrot.dynamic_autofocus(width, height);
        Ok(mandelbrot)
    }

    /// Scales the zoom for wide-aspect-ratio monitors.
    pub fn optimize_fit(&mut self, width: u32, height: u32) {
        let aspect_ratio = width as f64 / height as f64;
        if aspect_ratio > 1.0 {
            self.config.zoom *= aspect_ratio.sqrt();
        }
    }

    /// Refines the viewport centre and derives a level-of-detail iteration count.
    ///
    /// Uses a boundary-direction heuristic to align the centre with the nearest
    /// interior segment, then performs a local entropy-hill-climb across
    /// `ROTATION_STEPS` offsets to maximise structural detail. The iteration
    /// count is derived from the zoom level using a logarithmic LOD formula.
    pub fn dynamic_autofocus(&mut self, width: u32, height: u32) {
        let search_radius = self.config.zoom * 0.25;
        let branch_phasor = find_branch_phasor(
            self.preset.center,
            search_radius,
            self.config.scan_iterations,
        );

        let aligned_center = locked_interior_grid_alignment(
            self.preset.center,
            branch_phasor,
            search_radius,
            self.config.scan_iterations,
        );
        self.preset.center = aligned_center;

        let best_entropy = calculate_entropy(
            self.preset.center,
            self.config.zoom,
            self.config.rotation,
            self.config.scan_iterations,
            width,
            height,
        );

        let climb_radius = self.config.zoom * 0.05;
        let search_directions: Vec<Complex> = std::iter::once(Complex::zero())
            .chain(Complex::rotation_phasors(ROTATION_STEPS).map(|p| p * climb_radius))
            .collect();

        let (best_center, _) = search_directions
            .par_iter()
            .map(|&offset| {
                let candidate = self.preset.center + offset;
                let entropy = calculate_entropy(
                    candidate,
                    self.config.zoom,
                    self.config.rotation,
                    self.config.scan_iterations,
                    width,
                    height,
                );
                (candidate, entropy)
            })
            .max_by(|a, b| a.1.partial_cmp(&b.1).unwrap_or(Ordering::Equal))
            .unwrap_or((self.preset.center, best_entropy));

        self.preset.center = best_center;

        // LOD: scale iterations logarithmically with zoom level.
        let scale = self.config.zoom / (width.min(height) as f64);
        let lod = (150.0 + 45.0 * (1.0 / scale).ln()) as u32;
        self.config.scan_iterations = lod.clamp(MIN_ITERATIONS, MAX_ITERATIONS);
    }
}

// ============================================================================
// PRIVATE PURE HELPERS
// ============================================================================

/// Identifies the phasor direction pointing toward the steepest boundary gradient.
fn find_branch_phasor(center: Complex, search_radius: f64, scan_iterations: u32) -> Complex {
    let mut best_phasor = Complex::one();
    let mut max_variation = -1.0_f64;

    for phasor in Complex::rotation_phasors(ROTATION_STEPS) {
        let mut total_variation = 0.0;
        let mut prev_i = 0;

        for k in 1..=4 {
            let sample = center + phasor * (search_radius * k as f64 * 0.25);
            let (i, _, _) = mandelbrot_escape(sample, scan_iterations);
            if k > 1 {
                total_variation += (i as f64 - prev_i as f64).abs();
            }
            prev_i = i;
        }

        if total_variation > max_variation {
            max_variation = total_variation;
            best_phasor = phasor;
        }
    }
    best_phasor
}

/// Locates the mid-point of the target interior segment along the branch direction.
fn locked_interior_grid_alignment(
    center: Complex,
    phasor: Complex,
    search_radius: f64,
    scan_iterations: u32,
) -> Complex {
    const STEPS: usize = 64;
    let mut interior_segments: Vec<(usize, usize)> = Vec::new();
    let mut in_interior = false;
    let mut segment_start = 0;

    for step in 0..STEPS {
        let t = -search_radius + (step as f64 / (STEPS - 1) as f64) * (2.0 * search_radius);
        let (i, _, _) = mandelbrot_escape(center + phasor * t, scan_iterations);
        let is_interior = i >= scan_iterations;

        match (is_interior, in_interior) {
            (true, false) => {
                in_interior = true;
                segment_start = step;
            }
            (false, true) => {
                in_interior = false;
                interior_segments.push((segment_start, step - 1));
            }
            _ => {}
        }
    }
    if in_interior {
        interior_segments.push((segment_start, STEPS - 1));
    }

    let target = if interior_segments.len() >= 4 {
        Some(interior_segments[3])
    } else {
        interior_segments.last().copied()
    };

    if let Some((start_idx, end_idx)) = target {
        let mid_step = (start_idx + end_idx) as f64 / 2.0;
        let t_mid = -search_radius + (mid_step / (STEPS - 1) as f64) * (2.0 * search_radius);
        center + phasor * t_mid
    } else {
        center
    }
}

/// Computes the Shannon entropy of the iteration histogram at a given viewport configuration.
///
/// Higher entropy indicates more varied iteration counts and therefore more structural
/// detail visible in the rendered image.
fn calculate_entropy(
    center: Complex,
    zoom: f64,
    rotation: Complex,
    scan_iterations: u32,
    width: u32,
    height: u32,
) -> f64 {
    const GRID: usize = 32;
    let mut histogram = vec![0u32; (scan_iterations + 1) as usize];

    let specs = ViewportSpecs {
        center,
        zoom,
        rotation,
        is_julia: false,
    };
    let viewport = Viewport::new(width as f64, height as f64, &specs);
    let (step_x, step_y) = (width as f64 / GRID as f64, height as f64 / GRID as f64);

    for gy in 0..GRID {
        let y_f = gy as f64 * step_y;
        for gx in 0..GRID {
            let (i, _, _) =
                mandelbrot_escape(viewport.map(gx as f64 * step_x, y_f), scan_iterations);
            histogram[i as usize] += 1;
        }
    }

    let total = (GRID * GRID) as f64;
    histogram.iter().filter(|&&c| c > 0).fold(0.0, |acc, &c| {
        let p = c as f64 / total;
        acc - p * p.ln()
    })
}

/// Generates a geometric (log-spaced) sequence of zoom candidates.
///
/// Samples `zooms_count * rotations_count` `(zoom, rotation_index)` pairs
/// uniformly distributed in log-space between `2e-6` (microscopic close-up)
/// and `9.0` (panoramic macro view).
///
/// For a pool of `N` zooms and index `i`:
///
/// ```text
/// zoom_i = min_zoom * (max_zoom / min_zoom)^(i / (N - 1))
/// ```
///
/// This guarantees exact boundary alignment and smooth spatial transitions
/// between candidates regardless of pool size.
pub fn generate_zoom_candidates(zooms_count: usize, rotations_count: usize) -> Vec<(f64, usize)> {
    if zooms_count == 0 || rotations_count == 0 {
        return Vec::new();
    }

    const MIN_ZOOM: f64 = 2e-6;
    const MAX_ZOOM: f64 = 9.0;
    let log_ratio = MAX_ZOOM / MIN_ZOOM;

    let mut candidates = Vec::with_capacity(zooms_count * rotations_count);

    for z_idx in 0..zooms_count {
        let t = if zooms_count > 1 {
            z_idx as f64 / (zooms_count - 1) as f64
        } else {
            0.0
        };
        let zoom = MIN_ZOOM * log_ratio.powf(t);
        for r_idx in 0..rotations_count {
            candidates.push((zoom, r_idx));
        }
    }
    candidates
}

// ============================================================================
// TESTS
// ============================================================================

#[cfg(test)]
mod tests_mandelbrot {
    use super::*;

    #[test]
    fn test_mandelbrot_new_sanity() -> WallSwitchResult<()> {
        let m = MandelbrotGenerator::new()?;
        assert!(m.config.zoom > 0.0, "zoom must be positive");
        assert_eq!(m.preset.effect_name, ProceduralEffect::Mandelbrot);
        Ok(())
    }

    #[test]
    fn test_random_generator_sanity() -> WallSwitchResult<()> {
        let monitor = Monitor::default();
        let m = MandelbrotGenerator::random(&monitor)?;
        assert!(m.config.zoom > 0.0, "zoom must be positive");
        assert_eq!(m.preset.effect_name, ProceduralEffect::Mandelbrot);
        assert!(m.config.scan_iterations >= MIN_ITERATIONS);
        assert!(m.config.scan_iterations <= MAX_ITERATIONS);
        Ok(())
    }

    #[test]
    fn test_all_presets_correct_effect_name() {
        for p in MANDELBROT_PRESETS {
            assert_eq!(
                p.effect_name,
                ProceduralEffect::Mandelbrot,
                "wrong effect_name for '{}'",
                p.fractal_name
            );
        }
    }

    #[test]
    fn test_zoom_candidates_boundaries() {
        let zooms_count = 50;
        let rotations_count = 16;
        let candidates = generate_zoom_candidates(zooms_count, rotations_count);

        assert_eq!(candidates.len(), zooms_count * rotations_count);

        // First candidate must equal MIN_ZOOM exactly (t = 0).
        assert!((candidates[0].0 - 2e-6).abs() < 1e-12, "min bound mismatch");

        // Last candidate must equal MAX_ZOOM exactly (t = 1).
        let last = candidates.last().unwrap().0;
        assert!((last - 9.0).abs() < 1e-12, "max bound mismatch: {last}");
    }

    #[test]
    fn test_zoom_candidates_monotonically_increasing() {
        let candidates = generate_zoom_candidates(20, 1);
        let zooms: Vec<f64> = candidates.iter().map(|&(z, _)| z).collect();
        for w in zooms.windows(2) {
            assert!(
                w[0] <= w[1],
                "zoom sequence not monotone: {} > {}",
                w[0],
                w[1]
            );
        }
    }

    #[test]
    fn test_zoom_candidates_empty_on_zero_count() {
        assert!(generate_zoom_candidates(0, 16).is_empty());
        assert!(generate_zoom_candidates(16, 0).is_empty());
    }
}