//! Partial / dirty-region compositing for the HCIE engine.
//!
//! ## Purpose
//! Contains the functions that turn layer pixel data into a flat composite
//! image: `render_composite_region()` for GUI partial updates,
//! `get_composite_pixels()` for full-canvas exports, and the incremental
//! tile synchroniser `sync_dirty_tiles()`. These are the most performance-
//! sensitive composite paths.
//!
//! ## Logic & Workflow
//! 1. Render dirty vector layers into their pixel buffers.
//! 2. Apply layer effects/styles when `effects_dirty` is set, caching the
//!    result in `layer.effects_cache`.
//! 3. Sync the sparse tile cache for any dirty layers (`sync_dirty_tiles`).
//! 4. Composite either the dirty sub-region or the full canvas using the
//!    tiled compositor, optionally reusing the `below_cache` snapshot of
//!    layers below the active layer.
//!
//! ## Side Effects / Dependencies
//! Mutates layer pixels, `tile_layers`, `composite_scratch`, `raw_pixel_backup`,
//! and dirty flags. Depends on `hcie_tile::TiledLayer`, `hcie_fx`, and the
//! dynamic `tiled::composite_tiled_into` / `composite_layers` compositors.

use crate::dynamic_loader::vector::render_vector_shapes;
use crate::dynamic_loader::{composite_layers, tiled};
use crate::Engine;
use hcie_tile::TiledLayer;
use rayon::prelude::*;

impl Engine {
    /// **Purpose:**
    /// Computes the flat composite RGBA pixel buffer of all layers in the document.
    ///
    /// **Logic & Workflow:**
    /// 1. Renders vector shapes for dirty vector layers.
    /// 2. Applies layer effects/styles if they are dirty, caching the result.
    /// 3. Synchronizes `self.tile_layers` for any dirty layers.
    /// 4. Clears dirty flags.
    /// 5. Blends all layers into a flat caller-allocated buffer.
    pub fn get_composite_pixels(&mut self) -> Vec<u8> {
        log::trace!("[get_composite_pixels] ===== START =====");
        for (i, l) in self.document.layers.iter().enumerate() {
            log::trace!(
                "[get_composite_pixels] layer[{}] id={} name='{}' visible={} dirty={} opacity={} blend={:?}",
                i, l.id, l.name, l.visible, l.dirty, l.opacity, l.blend_mode
            );
        }
        self.apply_effects_and_sync_tiles();
        self.document.clear_dirty();
        let output = composite_layers(
            &self.document.layers,
            self.document.canvas_width,
            self.document.canvas_height,
        );
        let non_zero_alpha = output.iter().skip(3).step_by(4).filter(|&&a| a > 0).count();
        let total_pixels = output.len() / 4;
        log::trace!(
            "[get_composite_pixels] completed full composite: size={} bytes, non_zero_alpha={}/{} ({}%), first_pixel={:?}",
            output.len(),
            non_zero_alpha,
            total_pixels,
            if total_pixels > 0 { non_zero_alpha * 100 / total_pixels } else { 0 },
            if output.len() >= 4 { [output[0], output[1], output[2], output[3]] } else { [0; 4] }
        );
        for (i, l) in self.document.layers.iter().enumerate() {
            let opaque_count = l
                .pixels
                .iter()
                .skip(3)
                .step_by(4)
                .filter(|&&a| a > 0)
                .count();
            let total = l.pixels.len() / 4;
            log::trace!(
                "[get_composite_pixels] LAYER CONTENT: layer[{}] id={} name='{}' visible={} pixels_total={} opaque_pixels={}/{} ({}%)",
                i, l.id, l.name, l.visible, l.pixels.len(),
                opaque_count, total,
                if total > 0 { opaque_count * 100 / total } else { 0 }
            );
        }
        output
    }

    pub fn render_composite(&mut self) -> Vec<u8> {
        self.get_composite_pixels()
    }

    /// Composite only the dirty region into a pooled internal scratch buffer.
    /// Returns the dirty bounds `[x0, y0, x1, y1]` that were updated,
    /// or `None` if nothing was dirty (full composite needed).
    pub fn render_composite_region(&mut self) -> (Option<[u32; 4]>, *const u8, usize) {
        let has_dirty = self.document.composite_dirty;
        let dirty = self.document.dirty_bounds;
        let w = self.document.canvas_width;
        let h = self.document.canvas_height;

        if !has_dirty && dirty.is_none() {
            log::trace!("[render_composite_region] nothing dirty — returning None");
            return (None, std::ptr::null(), 0);
        }

        if has_dirty && dirty.is_none() {
            // composite_dirty was set (e.g., after loading/importing) but no layer-level
            // dirty_bounds exist. Force a full-canvas composite so the caller sees content
            // instead of an empty/null region.
            self.document.dirty_bounds = Some([0, 0, w, h]);
        }
        log::trace!(
            "[render_composite_region] ===== START composite_dirty={}, dirty_bounds={:?} =====",
            has_dirty,
            dirty
        );
        let w = self.document.canvas_width;
        let h = self.document.canvas_height;
        let buf_size = (w * h * 4) as usize;

        self.apply_effects_and_sync_tiles();

        let (x0, y0, x1, y1) = match dirty {
            Some([x0, y0, x1, y1]) => (x0.min(w), y0.min(h), x1.min(w), y1.min(h)),
            None => (0, 0, w, h),
        };

        // Now safe to borrow composite_scratch — no more mutable calls to self
        // that could touch this buffer until we return.
        if self
            .composite_scratch
            .as_ref()
            .map_or(true, |b| b.len() != buf_size)
        {
            self.composite_scratch = Some(vec![0u8; buf_size]);
        }
        let buf = self.composite_scratch.as_mut().unwrap();

        if x1 > x0 && y1 > y0 {
            let wu = w as usize;
            let x0u = x0 as usize;

            let active_idx = self.document.active_layer;
            let cache_valid = !self.below_cache_dirty
                && self.below_cache.is_some()
                && self.below_cache_active_idx == Some(active_idx)
                && active_idx > 0;

            log::trace!(
                "[render_composite_region] BEFORE composite: active_idx={}, cache_valid={}, below_cache={}, below_cache_active_idx={:?}, tile_layers_len={}, dirty_rect=[{},{},{},{}]",
                active_idx, cache_valid, self.below_cache.is_some(), self.below_cache_active_idx, self.tile_layers.len(), x0, y0, x1, y1
            );
            if log::log_enabled!(log::Level::Trace) {
                for (i, l) in self.document.layers.iter().enumerate() {
                    let tc = self
                        .tile_layers
                        .get(i)
                        .and_then(|t| t.as_ref().map(|tl| tl.tile_count()))
                        .unwrap_or(0);
                    log::trace!(
                        "[render_composite_region] layer[{}] id={} visible={} dirty={} opacity={} blend={:?} tile_count={}",
                        i, l.id, l.visible, l.dirty, l.opacity, l.blend_mode, tc
                    );
                }
            }

            if cache_valid {
                let cache = self.below_cache.as_ref().unwrap();
                if log::log_enabled!(log::Level::Trace) {
                    let below_non_zero = cache.iter().filter(|&&b| b != 0).count();
                    log::trace!(
                        "[render_composite_region] CACHE HIT: using below_cache ({} non-zero bytes), compositing layers[{}..{}] on top",
                        below_non_zero, active_idx, self.document.layers.len()
                    );
                }
                for y in y0..y1 {
                    let start = (y as usize * wu + x0u) * 4;
                    let end = start + ((x1 - x0) as usize) * 4;
                    if end <= cache.len() && end <= buf.len() {
                        buf[start..end].copy_from_slice(&cache[start..end]);
                    }
                }
                let tile_start = active_idx.min(self.tile_layers.len());
                tiled::composite_tiled_into(
                    &self.document.layers[active_idx..],
                    &self.tile_layers[tile_start..],
                    w,
                    h,
                    x0,
                    y0,
                    x1,
                    y1,
                    buf,
                );
                log::trace!(
                    "[render_composite_region] cache hit DONE: {} above layers composited, dirty_rect=[{},{},{},{}]",
                    self.document.layers.len() - active_idx, x0, y0, x1, y1
                );
            } else {
                log::trace!(
                    "[render_composite_region] CACHE MISS: full composite of all {} layers, dirty_rect=[{},{},{},{}]",
                    self.document.layers.len(), x0, y0, x1, y1
                );
                let is_full_canvas = x0 == 0 && y0 == 0 && x1 == w && y1 == h;
                if is_full_canvas {
                    buf.par_chunks_mut(65536).for_each(|chunk| chunk.fill(0));
                } else {
                    for y in y0..y1 {
                        let start = (y as usize * wu + x0u) * 4;
                        let end = start + ((x1 - x0) as usize) * 4;
                        buf[start..end].fill(0);
                    }
                }
                tiled::composite_tiled_into(
                    &self.document.layers,
                    &self.tile_layers,
                    w,
                    h,
                    x0,
                    y0,
                    x1,
                    y1,
                    buf,
                );
                log::trace!(
                    "[render_composite_region] full composite DONE, dirty_rect=[{},{},{},{}]",
                    x0,
                    y0,
                    x1,
                    y1
                );
            }
        }

        self.document.clear_dirty();
        let ptr = buf.as_ptr();
        (Some([x0, y0, x1, y1]), ptr, buf_size)
    }

    /// Shared preprocessing for full and partial compositing:
    /// - render dirty vector shapes,
    /// - apply dirty layer effects/styles,
    /// - sync the sparse tile cache.
    fn apply_effects_and_sync_tiles(&mut self) {
        // Render vector shapes before compositing — only dirty vector layers
        for layer in &mut self.document.layers {
            if let hcie_protocol::LayerData::Vector { shapes: _ } = &layer.data {
                if layer.dirty {
                    layer.pixels.fill(0);
                    render_vector_shapes(layer);
                }
            }
        }

        // ── Effects pipeline (two-pass) ─────────────────────────────────────────────
        // Pass 1: Snapshot raw pixels for any effect-bearing layer that does NOT yet
        // have a backup.
        {
            let ids_needing_backup: Vec<(u64, Vec<u8>)> = self
                .document
                .layers
                .iter()
                .filter(|l| {
                    (!l.effects.is_empty() || !l.styles.is_empty())
                        && !l.pixels.is_empty()
                        && !self.raw_pixel_backup.contains_key(&l.id)
                })
                .map(|l| (l.id, l.pixels.clone()))
                .collect();
            for (id, pixels) in ids_needing_backup {
                self.raw_pixel_backup.insert(id, pixels);
            }
        }

        // Pass 2: Apply layer effects. Restores from backup first so that each
        // slider edit always applies on top of the original untouched pixels.
        // CRITICAL: Do NOT modify layer.pixels with effects output. layer.pixels
        // must always contain the raw drawing data so that new strokes are drawn
        // on top of clean pixels (not on effects-applied pixels). The composite
        // uses effects_cache for layers with effects, not layer.pixels.
        for layer in &mut self.document.layers {
            if layer.effects.is_empty() && layer.styles.is_empty() {
                continue;
            }

            // Check if there are any active/enabled effects or styles
            let has_enabled = layer.effects.iter().any(|e| match e {
                hcie_protocol::effects::LayerEffect::DropShadow { enabled, .. }
                | hcie_protocol::effects::LayerEffect::InnerShadow { enabled, .. }
                | hcie_protocol::effects::LayerEffect::OuterGlow { enabled, .. }
                | hcie_protocol::effects::LayerEffect::InnerGlow { enabled, .. }
                | hcie_protocol::effects::LayerEffect::BevelEmboss { enabled, .. }
                | hcie_protocol::effects::LayerEffect::Satin { enabled, .. }
                | hcie_protocol::effects::LayerEffect::ColorOverlay { enabled, .. }
                | hcie_protocol::effects::LayerEffect::GradientOverlay { enabled, .. }
                | hcie_protocol::effects::LayerEffect::PatternOverlay { enabled, .. }
                | hcie_protocol::effects::LayerEffect::Stroke { enabled, .. } => *enabled,
            }) || layer.styles.iter().any(|s| match s {
                hcie_protocol::LayerStyle::DropShadow { enabled, .. }
                | hcie_protocol::LayerStyle::InnerShadow { enabled, .. }
                | hcie_protocol::LayerStyle::OuterGlow { enabled, .. }
                | hcie_protocol::LayerStyle::InnerGlow { enabled, .. }
                | hcie_protocol::LayerStyle::BevelEmboss { enabled, .. }
                | hcie_protocol::LayerStyle::Satin { enabled, .. }
                | hcie_protocol::LayerStyle::ColorOverlay { enabled, .. }
                | hcie_protocol::LayerStyle::GradientOverlay { enabled, .. }
                | hcie_protocol::LayerStyle::PatternOverlay { enabled, .. }
                | hcie_protocol::LayerStyle::Stroke { enabled, .. } => *enabled,
            });

            if !has_enabled {
                // No active effects: clean up cached rendering and backup to restore raw performance.
                *layer.effects_cache.lock().unwrap() = None;
                self.raw_pixel_backup.remove(&layer.id);
                layer
                    .effects_dirty
                    .store(false, std::sync::atomic::Ordering::Release);
                layer.dirty = true;
                continue;
            }

            if layer
                .effects_dirty
                .load(std::sync::atomic::Ordering::Acquire)
            {
                log::trace!(
                    "Applying active effects/styles for layer ID {} because effects are dirty",
                    layer.id
                );
                // DON'T restore raw pixels from backup! Restoring raw pixels from backup overwrites
                // active stroke drawing and wipes out new strokes. Since we never overwrite layer.pixels
                // with the effects output, layer.pixels already contains the clean, raw pixels.
                let mut effects: Vec<hcie_fx::LayerEffect> = layer.effects.iter()
                    .filter(|e| !matches!(e, hcie_protocol::effects::LayerEffect::DropShadow { noise, .. } if *noise == -999.0))
                    .map(|e| hcie_fx::protocol_to_hcie_fx_effect(e))
                    .collect();
                effects.extend(
                    layer
                        .styles
                        .iter()
                        .filter_map(|s| hcie_fx::layer_style_to_effect(s)),
                );
                if effects.is_empty() {
                    continue;
                }
                let processed = hcie_fx::apply_layer_effects(
                    &layer.pixels,
                    layer.width,
                    layer.height,
                    &effects,
                    layer.fill_opacity,
                );
                *layer.effects_cache.lock().unwrap() = Some(hcie_protocol::LayerEffects {
                    rendered: processed.clone(),
                    width: layer.width,
                    height: layer.height,
                });
                layer
                    .effects_dirty
                    .store(false, std::sync::atomic::Ordering::Release);
                layer.dirty = true;
            }
        }

        self.sync_dirty_tiles();
    }

    /// Incremental tile cache update.
    /// For dirty layers, only re-tile the tiles overlapping the global
    /// dirty_bounds region instead of scanning the entire 33MB dense buffer.
    /// Falls back to full from_dense() when dirty_bounds is None.
    fn sync_dirty_tiles(&mut self) {
        let count = self.document.layers.len();
        if self.tile_layers.len() < count {
            self.tile_layers.resize_with(count, || None);
        }
        let db = self.document.dirty_bounds;
        let mut synced_count = 0usize;
        for (i, layer) in self.document.layers.iter().enumerate() {
            if !layer.dirty || layer.pixels.is_empty() {
                continue;
            }
            if let Some([dx0, dy0, dx1, dy1]) = db {
                if let Some(ref mut tl) = self.tile_layers[i] {
                    if tl.width() == layer.width && tl.height() == layer.height {
                        tl.update_tiles_in_region(&layer.pixels, layer.width, dx0, dy0, dx1, dy1);
                        log::trace!(
                            "[sync_dirty_tiles] incremental update layer[{}] id={} visible={} dirty_bounds=[{},{},{},{}]",
                            i, layer.id, layer.visible, dx0, dy0, dx1, dy1
                        );
                    } else {
                        *tl = TiledLayer::from_dense(&layer.pixels, layer.width, layer.height);
                        log::trace!(
                            "[sync_dirty_tiles] size mismatch - full rebuild TiledLayer for layer[{}] id={} visible={} bounds=[{},{}]",
                            i, layer.id, layer.visible, layer.width, layer.height
                        );
                    }
                } else {
                    let mut tl = TiledLayer::new(layer.width, layer.height);
                    tl.update_tiles_in_region(&layer.pixels, layer.width, dx0, dy0, dx1, dy1);
                    self.tile_layers[i] = Some(tl);
                    log::trace!(
                        "[sync_dirty_tiles] created new TiledLayer for layer[{}] id={} visible={} dirty_bounds=[{},{},{},{}]",
                        i, layer.id, layer.visible, dx0, dy0, dx1, dy1
                    );
                }
            } else {
                self.tile_layers[i] = Some(TiledLayer::from_dense(
                    &layer.pixels,
                    layer.width,
                    layer.height,
                ));
                log::trace!(
                    "[sync_dirty_tiles] full rebuild TiledLayer for layer[{}] id={} visible={} (no dirty_bounds)",
                    i, layer.id, layer.visible
                );
            }
            synced_count += 1;
        }
        self.tile_layers.truncate(count);
        if synced_count > 0 {
            log::trace!(
                "[sync_dirty_tiles] synced {} layers, db={:?}, tile_layers_len={}",
                synced_count,
                db,
                self.tile_layers.len()
            );
        }
    }
}

/// Snapshot of layer data needed for background compositing.
///
/// # Purpose
/// Captures all state required to composite the canvas without borrowing
/// `&mut Engine`. The background thread uses this snapshot to compute the
/// composite independently of the UI thread.
///
/// # Logic & Workflow
/// 1. `Engine::create_composite_snapshot()` clones layer pixels, properties,
///    and tile data into this struct.
/// 2. `composite_from_snapshot()` reads the snapshot and produces the flat
///    RGBA composite buffer on a background thread.
pub struct CompositeSnapshot {
    pub layers: Vec<hcie_protocol::Layer>,
    pub tile_layers: Vec<Option<TiledLayer>>,
    pub canvas_width: u32,
    pub canvas_height: u32,
    pub active_layer: usize,
}

impl Engine {
    /// Create a snapshot of the current layer state for background compositing.
    ///
    /// # Purpose
    /// Clones all layer pixels, properties, and tile data so a background
    /// thread can compute the composite without borrowing the engine.
    ///
    /// # Side Effects
    /// Calls `apply_effects_and_sync_tiles()` to ensure dirty layers are
    /// up-to-date before cloning. This mutates engine state but is safe
    /// because it's called on the UI thread before spawning the background.
    pub fn create_composite_snapshot(&mut self) -> CompositeSnapshot {
        // Ensure all dirty layers are up-to-date before cloning
        self.apply_effects_and_sync_tiles();

        let layers = self.document.layers.clone();
        let tile_layers = self.tile_layers.clone();
        let canvas_width = self.document.canvas_width;
        let canvas_height = self.document.canvas_height;
        let active_layer = self.document.active_layer;

        CompositeSnapshot {
            layers,
            tile_layers,
            canvas_width,
            canvas_height,
            active_layer,
        }
    }
}

/// Compute the flat composite RGBA buffer from a snapshot.
///
/// # Purpose
/// Performs the full canvas composite on a background thread using the
/// cloned layer data from `CompositeSnapshot`. This function is standalone
/// (not a method on `Engine`) so it can be called from a background thread
/// without borrowing the engine.
///
/// # Arguments
/// * `snapshot` — Cloned layer data from `create_composite_snapshot()`.
///
/// # Returns
/// A `Vec<u8>` of size `width * height * 4` containing the flat RGBA composite.
pub fn composite_from_snapshot(snapshot: CompositeSnapshot) -> Vec<u8> {
    let w = snapshot.canvas_width;
    let h = snapshot.canvas_height;
    let buf_size = (w * h * 4) as usize;

    // Build tile slices for the compositor
    let tile_slice_len = snapshot.active_layer.min(snapshot.tile_layers.len());
    let _visible_below: Vec<(usize, bool)> = snapshot.layers[..snapshot.active_layer]
        .iter()
        .enumerate()
        .map(|(i, l)| (i, l.visible))
        .collect();

    let mut buf = vec![0u8; buf_size];

    // Composite below layers first (if any)
    if snapshot.active_layer > 0 {
        tiled::composite_tiled_into(
            &snapshot.layers[..snapshot.active_layer],
            &snapshot.tile_layers[..tile_slice_len],
            w,
            h,
            0,
            0,
            w,
            h,
            &mut buf,
        );
    }

    // Composite active layer and above
    let tile_start = snapshot.active_layer.min(snapshot.tile_layers.len());
    tiled::composite_tiled_into(
        &snapshot.layers[snapshot.active_layer..],
        &snapshot.tile_layers[tile_start..],
        w,
        h,
        0,
        0,
        w,
        h,
        &mut buf,
    );

    buf
}