// Copyright (c) the JPEG XL Project Authors. All rights reserved.

//

// Use of this source code is governed by a BSD-style

// license that can be found in the LICENSE file.

#include "lib/jxl/enc_patch_dictionary.h"

#include <jxl/cms_interface.h>

#include <jxl/memory_manager.h>

#include <jxl/types.h>

#include <algorithm>

#include <atomic>

#include <cmath>

#include <cstdint>

#include <cstdlib>

#include <utility>

#include <vector>

#include "lib/jxl/base/common.h"

#include "lib/jxl/base/compiler_specific.h"

#include "lib/jxl/base/data_parallel.h"

#include "lib/jxl/base/override.h"

#include "lib/jxl/base/printf_macros.h"

#include "lib/jxl/base/random.h"

#include "lib/jxl/base/rect.h"

#include "lib/jxl/base/span.h"

#include "lib/jxl/base/status.h"

#include "lib/jxl/common.h"

#include "lib/jxl/dec_cache.h"

#include "lib/jxl/dec_frame.h"

#include "lib/jxl/dec_patch_dictionary.h"

#include "lib/jxl/enc_ans.h"

#include "lib/jxl/enc_ans_params.h"

#include "lib/jxl/enc_aux_out.h"

#include "lib/jxl/enc_bit_writer.h"

#include "lib/jxl/enc_cache.h"

#include "lib/jxl/enc_debug_image.h"

#include "lib/jxl/enc_dot_dictionary.h"

#include "lib/jxl/enc_frame.h"

#include "lib/jxl/enc_params.h"

#include "lib/jxl/frame_header.h"

#include "lib/jxl/image.h"

#include "lib/jxl/image_bundle.h"

#include "lib/jxl/image_ops.h"

#include "lib/jxl/modular/options.h"

#include "lib/jxl/pack_signed.h"

#include "lib/jxl/patch_dictionary_internal.h"

namespace jxl {

static constexpr size_t kPatchFrameReferenceId = 3;

// static

Status PatchDictionaryEncoder::Encode(const PatchDictionary& pdic,

                                      BitWriter* writer, LayerType layer,

                                      AuxOut* aux_out) {

  JXL_ENSURE(pdic.HasAny());

  JxlMemoryManager* memory_manager = writer->memory_manager();

  std::vector<std::vector<Token>> tokens(1);

  auto add_num = [&](int context, size_t num) {

    tokens[0].emplace_back(context, static_cast<uint32_t>(num));

  };

  size_t num_ref_patch = 0;

  for (size_t i = 0; i < pdic.positions_.size();) {

    size_t ref_pos_idx = pdic.positions_[i].ref_pos_idx;

    while (i < pdic.positions_.size() &&

           pdic.positions_[i].ref_pos_idx == ref_pos_idx) {

      i++;

    }

    num_ref_patch++;

  }

  add_num(kNumRefPatchContext, num_ref_patch);

  size_t blend_pos = 0;

  size_t blending_stride = pdic.blendings_stride_;

  // blending_stride == num_ec + 1; num_ec > 1 =>

  bool choose_alpha = (blending_stride > 1 + 1);

  for (size_t i = 0; i < pdic.positions_.size();) {

    size_t i_start = i;

    size_t ref_pos_idx = pdic.positions_[i].ref_pos_idx;

    const auto& ref_pos = pdic.ref_positions_[ref_pos_idx];

    while (i < pdic.positions_.size() &&

           pdic.positions_[i].ref_pos_idx == ref_pos_idx) {

      i++;

    }

    size_t num = i - i_start;

    JXL_ENSURE(num > 0);

    add_num(kReferenceFrameContext, ref_pos.ref);

    add_num(kPatchReferencePositionContext, ref_pos.x0);

    add_num(kPatchReferencePositionContext, ref_pos.y0);

    add_num(kPatchSizeContext, ref_pos.xsize - 1);

    add_num(kPatchSizeContext, ref_pos.ysize - 1);

    add_num(kPatchCountContext, num - 1);

    for (size_t j = i_start; j < i; j++) {

      const PatchPosition& pos = pdic.positions_[j];

      if (j == i_start) {

        add_num(kPatchPositionContext, pos.x);

        add_num(kPatchPositionContext, pos.y);

      } else {

        add_num(kPatchOffsetContext,

                PackSigned(pos.x - pdic.positions_[j - 1].x));

        add_num(kPatchOffsetContext,

                PackSigned(pos.y - pdic.positions_[j - 1].y));

      }

      for (size_t k = 0; k < blending_stride; ++k, ++blend_pos) {

        const PatchBlending& info = pdic.blendings_[blend_pos];

        add_num(kPatchBlendModeContext, static_cast<uint32_t>(info.mode));

        if (UsesAlpha(info.mode) && choose_alpha) {

          add_num(kPatchAlphaChannelContext, info.alpha_channel);

        }

        if (UsesClamp(info.mode)) {

          add_num(kPatchClampContext, TO_JXL_BOOL(info.clamp));

        }

      }

    }

  }

  EntropyEncodingData codes;

  JXL_ASSIGN_OR_RETURN(

      size_t cost, BuildAndEncodeHistograms(memory_manager, HistogramParams(),

                                            kNumPatchDictionaryContexts, tokens,

                                            &codes, writer, layer, aux_out));

  (void)cost;

  JXL_RETURN_IF_ERROR(WriteTokens(tokens[0], codes, 0, writer, layer, aux_out));

  return true;

}

// static

Status PatchDictionaryEncoder::SubtractFrom(const PatchDictionary& pdic,

                                            Image3F* opsin) {

  // TODO(veluca): this can likely be optimized knowing it runs on full images.

  for (size_t y = 0; y < opsin->ysize(); y++) {

    float* JXL_RESTRICT rows[3] = {

        opsin->PlaneRow(0, y),

        opsin->PlaneRow(1, y),

        opsin->PlaneRow(2, y),

    };

    size_t blending_stride = pdic.blendings_stride_;

    for (size_t pos_idx : pdic.GetPatchesForRow(y)) {

      const size_t blending_idx = pos_idx * blending_stride;

      const PatchPosition& pos = pdic.positions_[pos_idx];

      const PatchReferencePosition& ref_pos =

          pdic.ref_positions_[pos.ref_pos_idx];

      const PatchBlendMode mode = pdic.blendings_[blending_idx].mode;

      size_t by = pos.y;

      size_t bx = pos.x;

      size_t xsize = ref_pos.xsize;

      JXL_ENSURE(y >= by);

      JXL_ENSURE(y < by + ref_pos.ysize);

      size_t iy = y - by;

      size_t ref = ref_pos.ref;

      const float* JXL_RESTRICT ref_rows[3] = {

          pdic.reference_frames_->at(ref).frame->color()->ConstPlaneRow(

              0, ref_pos.y0 + iy) +

              ref_pos.x0,

          pdic.reference_frames_->at(ref).frame->color()->ConstPlaneRow(

              1, ref_pos.y0 + iy) +

              ref_pos.x0,

          pdic.reference_frames_->at(ref).frame->color()->ConstPlaneRow(

              2, ref_pos.y0 + iy) +

              ref_pos.x0,

      };

      for (size_t ix = 0; ix < xsize; ix++) {

        for (size_t c = 0; c < 3; c++) {

          if (mode == PatchBlendMode::kAdd) {

            rows[c][bx + ix] -= ref_rows[c][ix];

          } else if (mode == PatchBlendMode::kReplace) {

            rows[c][bx + ix] = 0;

          } else if (mode == PatchBlendMode::kNone) {

            // Nothing to do.

          } else {

            return JXL_UNREACHABLE("blending mode %u not yet implemented",

                                   static_cast<uint32_t>(mode));

          }

        }

      }

    }

  }

  return true;

}

namespace {

struct PatchColorspaceInfo {

  float kChannelDequant[3];

  float kChannelWeights[3];

  explicit PatchColorspaceInfo(bool is_xyb) {

    if (is_xyb) {

      kChannelDequant[0] = 0.01615;

      kChannelDequant[1] = 0.08875;

      kChannelDequant[2] = 0.1922;

      kChannelWeights[0] = 30.0;

      kChannelWeights[1] = 3.0;

      kChannelWeights[2] = 1.0;

    } else {

      kChannelDequant[0] = 20.0f / 255;

      kChannelDequant[1] = 22.0f / 255;

      kChannelDequant[2] = 20.0f / 255;

      kChannelWeights[0] = 0.017 * 255;

      kChannelWeights[1] = 0.02 * 255;

      kChannelWeights[2] = 0.017 * 255;

    }

  }

  float ScaleForQuantization(float val, size_t c) {

    return val / kChannelDequant[c];

  }

  int Quantize(float val, size_t c) {

    float scaled = ScaleForQuantization(val, c);

    // Clamping allows values outside of target range (int8_t); caller should

    // deal with out-of-range values.

    scaled = jxl::Clamp1(scaled, -32768.0f, 32767.0f);

    return std::trunc(scaled);

  }

  bool is_similar_v(const Color& v1, const Color& v2, float threshold) {

    float distance = 0;

    for (size_t c = 0; c < 3; c++) {

      distance += std::abs(v1[c] - v2[c]) * kChannelWeights[c];

    }

    return distance <= threshold;

  }

};

using XY = std::pair<int32_t, int32_t>;

constexpr const size_t kPatchSide = 4;

StatusOr<std::vector<PatchInfo>> FindTextLikePatches(

    const CompressParams& cparams, const Image3F& opsin,

    const PassesEncoderState* JXL_RESTRICT state, ThreadPool* pool,

    AuxOut* aux_out, bool is_xyb) {

  std::vector<PatchInfo> info;

  if (state->cparams.patches == Override::kOff) return info;

  const auto& frame_dim = state->shared.frame_dim;

  JxlMemoryManager* memory_manager = opsin.memory_manager();

  PatchColorspaceInfo pci(is_xyb);

  float kSimilarThreshold = 0.8f;

  auto is_similar_impl = [&pci](const XY& p1, const XY& p2,

                                const float* JXL_RESTRICT rows[3],

                                size_t stride, float threshold) {

    size_t offset1 = p1.second * stride + p1.first;

    Color v1{rows[0][offset1], rows[1][offset1], rows[2][offset1]};

    size_t offset2 = p2.second * stride + p2.first;

    Color v2{rows[0][offset2], rows[1][offset2], rows[2][offset2]};

    return pci.is_similar_v(v1, v2, threshold);

  };

  std::atomic<uint32_t> screenshot_area_seeds{0};

  const size_t opsin_stride = opsin.PixelsPerRow();

  const float* JXL_RESTRICT opsin_rows[3] = {opsin.ConstPlaneRow(0, 0),

                                             opsin.ConstPlaneRow(1, 0),

                                             opsin.ConstPlaneRow(2, 0)};

  const auto pick = [&opsin_rows, opsin_stride](const XY& p) -> Color {

    size_t offset = p.second * opsin_stride + p.first;

    return {opsin_rows[0][offset], opsin_rows[1][offset],

            opsin_rows[2][offset]};

  };

  const auto is_same_color = [&opsin_rows, opsin_stride](

                                 const XY& p, const Color& c) -> size_t {

    const size_t offset = p.second * opsin_stride + p.first;

    for (size_t i = 0; i < c.size(); ++i) {

      if (std::fabs(c[i] - opsin_rows[i][offset]) > 1e-4) {

        return 0;

      }

    }

    return 1;

  };

  auto is_similar = [&](const XY& p1, const XY& p2) {

    return is_similar_impl(p1, p2, opsin_rows, opsin_stride, kSimilarThreshold);

  };

  // Look for kPatchSide size squares, naturally aligned, that all have the same

  // pixel values.

  JXL_ASSIGN_OR_RETURN(

      ImageB is_screenshot_like,

      ImageB::Create(memory_manager, DivCeil(frame_dim.xsize, kPatchSide),

                     DivCeil(frame_dim.ysize, kPatchSide)));

  ZeroFillImage(&is_screenshot_like);

  const size_t pw = frame_dim.xsize / kPatchSide;

  const size_t ph = frame_dim.ysize / kPatchSide;

  const auto flat_patch = [&](const XY& o, const Color& base) -> bool {

    for (size_t iy = 0; iy < kPatchSide; iy++) {

      for (size_t ix = 0; ix < kPatchSide; ix++) {

        XY p = {static_cast<int32_t>(o.first + ix),

                static_cast<int32_t>(o.second + iy)};

        if (!is_same_color(p, base)) {

          return false;

        }

      }

    }

    return true;

  };

  // TODO(eustas): should do this in 2 phases:

  //   1) if patches are not enabled do sampling run for has_screenshot_areas

  //   2) if patches forced or not disables + has_screenshot_areas do

  //      SIMDified full scan for is_screenshot_like

  const auto process_row = [&](const uint32_t py,

                               size_t /* thread */) -> Status {

    uint32_t found = 0;

    for (size_t px = 1; px <= pw - 2; px++) {

      XY o = {static_cast<uint32_t>(px * kPatchSide),

              static_cast<uint32_t>(py * kPatchSide)};

      Color base = pick(o);

      if (!flat_patch(o, base)) continue;

      size_t num_same = 0;

      for (size_t y = (py - 1) * kPatchSide; y <= (py + 1) * kPatchSide;

           y += kPatchSide) {

        for (size_t x = (px - 1) * kPatchSide; x <= (px + 1) * kPatchSide;

             x += kPatchSide) {

          XY p = {static_cast<uint32_t>(x), static_cast<uint32_t>(y)};

          num_same += is_same_color(p, base);

        }

      }

      // Too few equal pixels nearby.

      if (num_same < 8) continue;

      is_screenshot_like.Row(py)[px] = 1;

      found++;

    }

    screenshot_area_seeds.fetch_add(found);

    return true;

  };

  bool can_have_seeds = ((pw >= 3) && (ph >= 3));

  if (can_have_seeds) {

    JXL_RETURN_IF_ERROR(RunOnPool(pool, 1, ph - 2, ThreadPool::NoInit,

                                  process_row, "IsScreenshotLike"));

  }

  // TODO(veluca): also parallelize the rest of this function.

  if (WantDebugOutput(cparams)) {

    JXL_RETURN_IF_ERROR(

        DumpPlaneNormalized(cparams, "screenshot_like", is_screenshot_like));

  }

  constexpr int kSearchRadius = 1;

  size_t num_seeds = screenshot_area_seeds.load();

  if (!ApplyOverride(state->cparams.patches, (num_seeds > 0))) {

    return info;

  }

  // Search for "similar enough" pixels near the screenshot-like areas.

  JXL_ASSIGN_OR_RETURN(

      ImageB is_background,

      ImageB::Create(memory_manager, frame_dim.xsize, frame_dim.ysize));

  ZeroFillImage(&is_background);

  JXL_ASSIGN_OR_RETURN(

      Image3F background,

      Image3F::Create(memory_manager, frame_dim.xsize, frame_dim.ysize));

  ZeroFillImage(&background);

  constexpr size_t kDistanceLimit = 50;

  float* JXL_RESTRICT background_rows[3] = {

      background.PlaneRow(0, 0),

      background.PlaneRow(1, 0),

      background.PlaneRow(2, 0),

  };

  const size_t background_stride = background.PixelsPerRow();

  uint8_t* JXL_RESTRICT is_background_row = is_background.Row(0);

  const size_t is_background_stride = is_background.PixelsPerRow();

  const auto is_bg = [&](const XY& p) -> uint8_t& {

    return is_background_row[p.second * is_background_stride + p.first];

  };

  std::vector<std::pair<XY, XY>> queue;

  queue.reserve(2 * num_seeds * kPatchSide * kPatchSide);

  size_t queue_front = 0;

  // TODO(eustas): coalesce neighbours, leave only border.

  if (can_have_seeds) {

    for (size_t py = 1; py < ph - 1; py++) {

      uint8_t* JXL_RESTRICT screenshot_row = is_screenshot_like.Row(py);

      for (size_t px = 1; px < pw - 1; px++) {

        if (!screenshot_row[px]) continue;

        for (size_t y = py * kPatchSide; y < (py + 1) * kPatchSide; ++y) {

          for (size_t x = px * kPatchSide; x < (px + 1) * kPatchSide; ++x) {

            XY p = {static_cast<uint32_t>(x), static_cast<uint32_t>(y)};

            queue.emplace_back(p, p);

            is_bg(p) = 1;

          }

        }

      }

    }

  }

  while (queue_front < queue.size()) {

    XY cur = queue[queue_front].first;

    XY src = queue[queue_front].second;

    queue_front++;

    Color src_color;

    for (size_t c = 0; c < 3; c++) {

      float clr = opsin_rows[c][src.second * opsin_stride + src.first];

      src_color[c] = clr;

      background_rows[c][cur.second * background_stride + cur.first] = clr;

    }

    for (int dx = -kSearchRadius; dx <= kSearchRadius; dx++) {

      for (int dy = -kSearchRadius; dy <= kSearchRadius; dy++) {

        XY next{cur.first + dx, cur.second + dy};

        if (next.first < 0 || next.second < 0 ||

            static_cast<uint32_t>(next.first) >= frame_dim.xsize ||

            static_cast<uint32_t>(next.second) >= frame_dim.ysize) {

          continue;

        }

        uint8_t& bg = is_bg(next);

        if (bg) continue;

        if (static_cast<uint32_t>(

                std::abs(next.first - static_cast<int>(src.first)) +

                std::abs(next.second - static_cast<int>(src.second))) >

            kDistanceLimit) {

          continue;

        }

        if (is_similar(src, next)) {

          queue.emplace_back(next, src);

          bg = 1;

        }

      }

    }

  }

  queue.clear();

  ImageF ccs;

  Rng rng(0);

  bool paint_ccs = false;

  if (WantDebugOutput(cparams)) {

    JXL_RETURN_IF_ERROR(

        DumpPlaneNormalized(cparams, "is_background", is_background));

    if (is_xyb) {

      JXL_RETURN_IF_ERROR(DumpXybImage(cparams, "background", background));

    } else {

      JXL_RETURN_IF_ERROR(DumpImage(cparams, "background", background));

    }

    JXL_ASSIGN_OR_RETURN(

        ccs, ImageF::Create(memory_manager, frame_dim.xsize, frame_dim.ysize));

    ZeroFillImage(&ccs);

    paint_ccs = true;

  }

  constexpr float kVerySimilarThreshold = 0.03f;

  constexpr float kHasSimilarThreshold = 0.03f;

  const float* JXL_RESTRICT const_background_rows[3] = {

      background_rows[0], background_rows[1], background_rows[2]};

  auto is_similar_b = [&](std::pair<int, int> p1, std::pair<int, int> p2) {

    return is_similar_impl(p1, p2, const_background_rows, background_stride,

                           kVerySimilarThreshold);

  };

  constexpr int kMinPeak = 2;

  constexpr int kHasSimilarRadius = 2;

  // Find small CC outside the "similar enough" areas, compute bounding boxes,

  // and run heuristics to exclude some patches.

  JXL_ASSIGN_OR_RETURN(

      ImageB visited,

      ImageB::Create(memory_manager, frame_dim.xsize, frame_dim.ysize));

  ZeroFillImage(&visited);

  uint8_t* JXL_RESTRICT visited_row = visited.Row(0);

  const size_t visited_stride = visited.PixelsPerRow();

  std::vector<std::pair<uint32_t, uint32_t>> cc;

  std::vector<std::pair<uint32_t, uint32_t>> stack;

  for (size_t y = 0; y < frame_dim.ysize; y++) {

    for (size_t x = 0; x < frame_dim.xsize; x++) {

      if (is_background_row[y * is_background_stride + x]) continue;

      cc.clear();

      stack.clear();

      stack.emplace_back(static_cast<uint32_t>(x), static_cast<uint32_t>(y));

      size_t min_x = x;

      size_t max_x = x;

      size_t min_y = y;

      size_t max_y = y;

      std::pair<uint32_t, uint32_t> reference;

      bool found_border = false;

      bool all_similar = true;

      while (!stack.empty()) {

        std::pair<uint32_t, uint32_t> cur = stack.back();

        stack.pop_back();

        if (visited_row[cur.second * visited_stride + cur.first]) continue;

        visited_row[cur.second * visited_stride + cur.first] = 1;

        if (cur.first < min_x) min_x = cur.first;

        if (cur.first > max_x) max_x = cur.first;

        if (cur.second < min_y) min_y = cur.second;

        if (cur.second > max_y) max_y = cur.second;

        if (paint_ccs) {

          cc.push_back(cur);

        }

        for (int dx = -kSearchRadius; dx <= kSearchRadius; dx++) {

          for (int dy = -kSearchRadius; dy <= kSearchRadius; dy++) {

            if (dx == 0 && dy == 0) continue;

            int next_first = static_cast<int32_t>(cur.first) + dx;

            int next_second = static_cast<int32_t>(cur.second) + dy;

            if (next_first < 0 || next_second < 0 ||

                static_cast<uint32_t>(next_first) >= frame_dim.xsize ||

                static_cast<uint32_t>(next_second) >= frame_dim.ysize) {

              continue;

            }

            std::pair<uint32_t, uint32_t> next{next_first, next_second};

            if (!is_background_row[next.second * is_background_stride +

                                   next.first]) {

              stack.push_back(next);

            } else {

              if (!found_border) {

                reference = next;

                found_border = true;

              } else {

                if (!is_similar_b(next, reference)) all_similar = false;

              }

            }

          }

        }

      }

      if (!found_border || !all_similar || max_x - min_x >= kMaxPatchSize ||

          max_y - min_y >= kMaxPatchSize) {

        continue;

      }

      size_t bpos = background_stride * reference.second + reference.first;

      Color ref = {background_rows[0][bpos], background_rows[1][bpos],

                   background_rows[2][bpos]};

      bool has_similar = false;

      for (size_t iy = std::max<int>(

               static_cast<int32_t>(min_y) - kHasSimilarRadius, 0);

           iy < std::min(max_y + kHasSimilarRadius + 1, frame_dim.ysize);

           iy++) {

        for (size_t ix = std::max<int>(

                 static_cast<int32_t>(min_x) - kHasSimilarRadius, 0);

             ix < std::min(max_x + kHasSimilarRadius + 1, frame_dim.xsize);

             ix++) {

          size_t opos = opsin_stride * iy + ix;

          Color px = {opsin_rows[0][opos], opsin_rows[1][opos],

                      opsin_rows[2][opos]};

          if (pci.is_similar_v(ref, px, kHasSimilarThreshold)) {

            has_similar = true;

          }

        }

      }

      if (!has_similar) continue;

      info.emplace_back();

      info.back().second.emplace_back(static_cast<uint32_t>(min_x),

                                      static_cast<uint32_t>(min_y));

      QuantizedPatch& patch = info.back().first;

      patch.xsize = max_x - min_x + 1;

      patch.ysize = max_y - min_y + 1;

      bool too_big = false;

      bool too_small = true;

      for (size_t c : {1, 0, 2}) {

        for (size_t iy = min_y; iy <= max_y; iy++) {

          for (size_t ix = min_x; ix <= max_x; ix++) {

            size_t offset = (iy - min_y) * patch.xsize + ix - min_x;

            float fval = opsin_rows[c][iy * opsin_stride + ix] - ref[c];

            patch.fpixels[c][offset] = fval;

            int val = pci.Quantize(patch.fpixels[c][offset], c);

            int8_t qval = static_cast<int8_t>(val);

            patch.pixels[c][offset] = qval;

            too_big |= (val != static_cast<int>(qval));

            too_small &= (val < kMinPeak) && (val > -kMinPeak);

          }

        }

      }

      if (too_small || too_big) {

        info.pop_back();

        continue;

      }

      if (paint_ccs) {

        float cc_color = rng.UniformF(0.5, 1.0);

        for (std::pair<uint32_t, uint32_t> p : cc) {

          ccs.Row(p.second)[p.first] = cc_color;

        }

      }

    }

  }

  if (paint_ccs) {

    JXL_ENSURE(WantDebugOutput(cparams));

    JXL_RETURN_IF_ERROR(DumpPlaneNormalized(cparams, "ccs", ccs));

  }

  if (info.empty()) {

    return info;

  }

  // Remove duplicates.

  constexpr size_t kMinPatchOccurrences = 2;

  std::sort(info.begin(), info.end());

  size_t unique = 0;

  for (size_t i = 1; i < info.size(); i++) {

    if (info[i].first == info[unique].first) {

      info[unique].second.insert(info[unique].second.end(),

                                 info[i].second.begin(), info[i].second.end());

    } else {

      if (info[unique].second.size() >= kMinPatchOccurrences) {

        unique++;

      }

      info[unique] = info[i];

    }

  }

  if (info[unique].second.size() >= kMinPatchOccurrences) {

    unique++;

  }

  info.resize(unique);

  size_t max_patch_size = 0;

  for (const auto& patch : info) {

    size_t pixels = patch.first.xsize * patch.first.ysize;

    if (pixels > max_patch_size) max_patch_size = pixels;

  }

  // don't use patches if all patches are smaller than this

  constexpr size_t kMinMaxPatchSize = 20;

  if (max_patch_size < kMinMaxPatchSize) {

    info.clear();

  }

  return info;

}

}  // namespace

Status FindBestPatchDictionary(const Image3F& opsin,

                               PassesEncoderState* JXL_RESTRICT state,

                               const JxlCmsInterface& cms, ThreadPool* pool,

                               AuxOut* aux_out, bool is_xyb) {

  JXL_ASSIGN_OR_RETURN(

      std::vector<PatchInfo> info,

      FindTextLikePatches(state->cparams, opsin, state, pool, aux_out, is_xyb));

  JxlMemoryManager* memory_manager = opsin.memory_manager();

  // TODO(veluca): this doesn't work if both dots and patches are enabled.

  // For now, since dots and patches are not likely to occur in the same kind of

  // images, disable dots if some patches were found.

  if (info.empty() &&

      ApplyOverride(

          state->cparams.dots,

          state->cparams.speed_tier <= SpeedTier::kSquirrel &&

              state->cparams.butteraugli_distance >= kMinButteraugliForDots &&

              !state->cparams.disable_perceptual_optimizations)) {

    Rect rect(0, 0, state->shared.frame_dim.xsize,

              state->shared.frame_dim.ysize);

    JXL_ASSIGN_OR_RETURN(info,

                         FindDotDictionary(state->cparams, opsin, rect,

                                           state->shared.cmap.base(), pool));

  }

  if (info.empty()) return true;

  std::sort(

      info.begin(), info.end(), [&](const PatchInfo& a, const PatchInfo& b) {

        return a.first.xsize * a.first.ysize > b.first.xsize * b.first.ysize;

      });

  size_t max_x_size = 0;

  size_t max_y_size = 0;

  size_t total_pixels = 0;

  for (const auto& patch : info) {

    size_t pixels = patch.first.xsize * patch.first.ysize;

    if (max_x_size < patch.first.xsize) max_x_size = patch.first.xsize;

    if (max_y_size < patch.first.ysize) max_y_size = patch.first.ysize;

    total_pixels += pixels;

  }

  // Bin-packing & conversion of patches.

  constexpr float kBinPackingSlackness = 1.05f;

  size_t ref_xsize = std::max<float>(max_x_size, std::sqrt(total_pixels));

  size_t ref_ysize = std::max<float>(max_y_size, std::sqrt(total_pixels));

  std::vector<std::pair<size_t, size_t>> ref_positions(info.size());

  // TODO(veluca): allow partial overlaps of patches that have the same pixels.

  size_t max_y = 0;

  do {

    max_y = 0;

    // Increase packed image size.

    ref_xsize = ref_xsize * kBinPackingSlackness + 1;

    ref_ysize = ref_ysize * kBinPackingSlackness + 1;

    JXL_ASSIGN_OR_RETURN(ImageB occupied,

                         ImageB::Create(memory_manager, ref_xsize, ref_ysize));

    ZeroFillImage(&occupied);

    uint8_t* JXL_RESTRICT occupied_rows = occupied.Row(0);

    size_t occupied_stride = occupied.PixelsPerRow();

    bool success = true;

    // For every patch...

    for (size_t patch = 0; patch < info.size(); patch++) {

      size_t x0 = 0;

      size_t y0 = 0;

      size_t xsize = info[patch].first.xsize;

      size_t ysize = info[patch].first.ysize;

      bool found = false;

      // For every possible start position ...

      for (; y0 + ysize <= ref_ysize; y0++) {

        x0 = 0;

        for (; x0 + xsize <= ref_xsize; x0++) {

          bool has_occupied_pixel = false;

          size_t x = x0;

          // Check if it is possible to place the patch in this position in the

          // reference frame.

          for (size_t y = y0; y < y0 + ysize; y++) {

            x = x0;

            for (; x < x0 + xsize; x++) {

              if (occupied_rows[y * occupied_stride + x]) {

                has_occupied_pixel = true;

                break;

              }

            }

          }  // end of positioning check

          if (!has_occupied_pixel) {

            found = true;

            break;

          }

          x0 = x;  // Jump to next pixel after the occupied one.

        }

        if (found) break;

      }  // end of start position checking

      // We didn't find a possible position: repeat from the beginning with a

      // larger reference frame size.

      if (!found) {

        success = false;

        break;

      }

      // We found a position: mark the corresponding positions in the reference

      // image as used.

      ref_positions[patch] = {x0, y0};

      for (size_t y = y0; y < y0 + ysize; y++) {

        for (size_t x = x0; x < x0 + xsize; x++) {

          occupied_rows[y * occupied_stride + x] = JXL_TRUE;

        }

      }

      max_y = std::max(max_y, y0 + ysize);

    }

    if (success) break;

  } while (true);

  JXL_ENSURE(ref_ysize >= max_y);

  ref_ysize = max_y;

  JXL_ASSIGN_OR_RETURN(Image3F reference_frame,

                       Image3F::Create(memory_manager, ref_xsize, ref_ysize));

  // TODO(veluca): figure out a better way to fill the image.

  ZeroFillImage(&reference_frame);

  std::vector<PatchPosition> positions;

  std::vector<PatchReferencePosition> pref_positions;

  std::vector<PatchBlending> blendings;

  float* JXL_RESTRICT ref_rows[3] = {

      reference_frame.PlaneRow(0, 0),

      reference_frame.PlaneRow(1, 0),

      reference_frame.PlaneRow(2, 0),

  };

  size_t ref_stride = reference_frame.PixelsPerRow();

  size_t num_ec = state->shared.metadata->m.num_extra_channels;

  for (size_t i = 0; i < info.size(); i++) {

    PatchReferencePosition ref_pos;

    ref_pos.xsize = info[i].first.xsize;

    ref_pos.ysize = info[i].first.ysize;

    ref_pos.x0 = ref_positions[i].first;

    ref_pos.y0 = ref_positions[i].second;

    ref_pos.ref = kPatchFrameReferenceId;

    for (size_t y = 0; y < ref_pos.ysize; y++) {

      for (size_t x = 0; x < ref_pos.xsize; x++) {

        for (size_t c = 0; c < 3; c++) {

          ref_rows[c][(y + ref_pos.y0) * ref_stride + x + ref_pos.x0] =

              info[i].first.fpixels[c][y * ref_pos.xsize + x];

        }

      }

    }

    for (const auto& pos : info[i].second) {

      JXL_DEBUG_V(4, "Patch %" PRIuS "x%" PRIuS " at position %u,%u",

                  ref_pos.xsize, ref_pos.ysize, pos.first, pos.second);

      positions.emplace_back(

          PatchPosition{pos.first, pos.second, pref_positions.size()});

      // Add blending for color channels, ignore other channels.

      blendings.push_back({PatchBlendMode::kAdd, 0, false});

      for (size_t j = 0; j < num_ec; ++j) {

        blendings.push_back({PatchBlendMode::kNone, 0, false});

      }

    }

    pref_positions.emplace_back(ref_pos);

  }

  CompressParams cparams = state->cparams;

  // Recursive application of patches could create very weird issues.

  cparams.patches = Override::kOff;

  if (WantDebugOutput(cparams)) {

    if (is_xyb) {

      JXL_RETURN_IF_ERROR(

          DumpXybImage(cparams, "patch_reference", reference_frame));

    } else {

      JXL_RETURN_IF_ERROR(

          DumpImage(cparams, "patch_reference", reference_frame));

    }

  }

  JXL_RETURN_IF_ERROR(RoundtripPatchFrame(&reference_frame, state,

                                          kPatchFrameReferenceId, cparams, cms,

                                          pool, aux_out, /*subtract=*/true));

  // TODO(veluca): this assumes that applying patches is commutative, which is

  // not true for all blending modes. This code only produces kAdd patches, so

  // this works out.

  PatchDictionaryEncoder::SetPositions(

      &state->shared.image_features.patches, std::move(positions),

      std::move(pref_positions), std::move(blendings), num_ec + 1);

  return true;

}

Status RoundtripPatchFrame(Image3F* reference_frame,

                           PassesEncoderState* JXL_RESTRICT state, int idx,

                           CompressParams& cparams, const JxlCmsInterface& cms,

                           ThreadPool* pool, AuxOut* aux_out, bool subtract) {

  JxlMemoryManager* memory_manager = state->memory_manager();

  FrameInfo patch_frame_info;

  cparams.resampling = 1;

  cparams.ec_resampling = 1;

  cparams.dots = Override::kOff;

  cparams.noise = Override::kOff;

  cparams.modular_mode = true;

  cparams.responsive = 0;

  cparams.progressive_dc = 0;

  cparams.progressive_mode = Override::kOff;

  cparams.qprogressive_mode = Override::kOff;

  // Use gradient predictor and not Predictor::Best.

  cparams.options.predictor = Predictor::Gradient;

  patch_frame_info.save_as_reference = idx;  // always saved.

  patch_frame_info.frame_type = FrameType::kReferenceOnly;

  patch_frame_info.save_before_color_transform = true;

  ImageBundle ib(memory_manager, &state->shared.metadata->m);

  // TODO(veluca): metadata.color_encoding is a lie: ib is in XYB, but there is

  // no simple way to express that yet.

  patch_frame_info.ib_needs_color_transform = false;

  JXL_RETURN_IF_ERROR(ib.SetFromImage(

      std::move(*reference_frame), state->shared.metadata->m.color_encoding));

  if (!ib.metadata()->extra_channel_info.empty()) {

    // Add placeholder extra channels to the patch image: patch encoding does

    // not yet support extra channels, but the codec expects that the amount of

    // extra channels in frames matches that in the metadata of the codestream.

    std::vector<ImageF> extra_channels;

    extra_channels.reserve(ib.metadata()->extra_channel_info.size());

    for (size_t i = 0; i < ib.metadata()->extra_channel_info.size(); i++) {

      JXL_ASSIGN_OR_RETURN(

          ImageF ch, ImageF::Create(memory_manager, ib.xsize(), ib.ysize()));

      extra_channels.emplace_back(std::move(ch));

      // Must initialize the image with data to not affect blending with

      // uninitialized memory.

      // TODO(lode): patches must copy and use the real extra channels instead.

      ZeroFillImage(&extra_channels.back());

    }

    JXL_RETURN_IF_ERROR(ib.SetExtraChannels(std::move(extra_channels)));

  }

  auto special_frame = jxl::make_unique<BitWriter>(memory_manager);

  AuxOut patch_aux_out;

  JXL_RETURN_IF_ERROR(EncodeFrame(

      memory_manager, cparams, patch_frame_info, state->shared.metadata, ib,

      cms, pool, special_frame.get(), aux_out ? &patch_aux_out : nullptr));

  if (aux_out) {

    for (const auto& l : patch_aux_out.layers) {

      aux_out->layer(LayerType::Dictionary).Assimilate(l);

    }

  }

  const Span<const uint8_t> encoded = special_frame->GetSpan();

  state->special_frames.emplace_back(std::move(special_frame));

  if (subtract) {

    ImageBundle decoded(memory_manager, &state->shared.metadata->m);

    auto dec_state = jxl::make_unique<PassesDecoderState>(memory_manager);

    JXL_RETURN_IF_ERROR(dec_state->output_encoding_info.SetFromMetadata(

        *state->shared.metadata));

    const uint8_t* frame_start = encoded.data();

    size_t encoded_size = encoded.size();

    JXL_RETURN_IF_ERROR(DecodeFrame(

        dec_state.get(), pool, frame_start, encoded_size,

        /*frame_header=*/nullptr, &decoded, *state->shared.metadata));

    frame_start += decoded.decoded_bytes();

    encoded_size -= decoded.decoded_bytes();

    size_t ref_xsize =

        dec_state->shared_storage.reference_frames[idx].frame->color()->xsize();

    // if the frame itself uses patches, we need to decode another frame

    if (!ref_xsize) {

      JXL_RETURN_IF_ERROR(DecodeFrame(

          dec_state.get(), pool, frame_start, encoded_size,

          /*frame_header=*/nullptr, &decoded, *state->shared.metadata));

    }

    JXL_ENSURE(encoded_size == 0);

    state->shared.reference_frames[idx] =

        std::move(dec_state->shared_storage.reference_frames[idx]);

  } else {

    *state->shared.reference_frames[idx].frame = std::move(ib);

  }

  return true;

}

}  // namespace jxl