// 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.

#ifndef LIB_JXL_ENC_ANS_PARAMS_H_

#define LIB_JXL_ENC_ANS_PARAMS_H_

// Encoder-only parameter needed for ANS entropy encoding methods.

#include <cstdint>

#include <cstdlib>

#include <utility>

#include <vector>

#include "lib/jxl/ans_common.h"

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

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

#include "lib/jxl/common.h"

#include "lib/jxl/dec_ans.h"

namespace jxl {

// Forward declaration to break include cycle.

struct CompressParams;

// RebalanceHistogram requires a signed type.

using ANSHistBin = int32_t;

struct HistogramParams {

  enum class ClusteringType {

    kFastest,  // Only 4 clusters.

    kFast,

    kBest,

  };

  enum class HybridUintMethod {

    kNone,        // just use kHybridUint420Config.

    k000,         // force the fastest option.

    kFast,        // just try a couple of options.

    kContextMap,  // fast choice for ctx map.

    kBest,

  };

  enum class LZ77Method {

    kNone,          // do not try lz77.

    kRLE,           // only try doing RLE.

    kLZ77b1w3f,     // lz77 fast without runtime cost comparison

    kLZ77b3w3f,     // lz77

    kLZ77b7w3f,     // lz77

    kLZ77b15w3f,    // lz77

    kLZ77b31w3f,    // lz77 slow

    kLZ77b1w3t,     // lz77 fast with runtime cost comparison (almost always worse)

    kLZ77b3w3t,     // lz77

    kLZ77b7w3t,     // lz77

    kLZ77b15w3t,    // lz77

    kLZ77b31w3t,    // lz77 slow

    kOptc1,         // optimal-matching LZ77 fast.

    kOptc3,         // optimal-matching LZ77

    kOptc8,         // optimal-matching LZ77

    kOptc256,       // optimal-matching LZ77 parsing big chain length.

  };

  enum class ANSHistogramStrategy {

    kFast,         // Only try some methods, early exit.

    kApproximate,  // Only try some methods.

    kPrecise,      // Try all methods.

  };

  HistogramParams() = default;

  HistogramParams(SpeedTier tier, size_t num_ctx) {

    if (tier > SpeedTier::kFalcon) {

      clustering = ClusteringType::kFastest;

      lz77_method = LZ77Method::kNone;

    } else if (tier > SpeedTier::kTortoise) {

      clustering = ClusteringType::kFast;

    } else {

      clustering = ClusteringType::kBest;

    }

    if (tier > SpeedTier::kTortoise) {

      uint_method = HybridUintMethod::kNone;

    }

    if (tier >= SpeedTier::kSquirrel) {

      ans_histogram_strategy = ANSHistogramStrategy::kApproximate;

    }

  }

  static HistogramParams ForModular(

      const CompressParams& cparams,

      const std::vector<uint8_t>& extra_dc_precision, bool streaming_mode);

  HybridUintConfig UintConfig() const {

    if (uint_method == HistogramParams::HybridUintMethod::kContextMap) {

      return HybridUintConfig(2, 0, 1);

    }

    if (uint_method == HistogramParams::HybridUintMethod::k000) {

      return HybridUintConfig(0, 0, 0);

    }

    // Default config for clustering.

    return HybridUintConfig();

  }

  ClusteringType clustering = ClusteringType::kBest;

  HybridUintMethod uint_method = HybridUintMethod::kBest;

  LZ77Method lz77_method = LZ77Method::kRLE;

  ANSHistogramStrategy ans_histogram_strategy = ANSHistogramStrategy::kPrecise;

  std::vector<size_t> image_widths;

  size_t max_histograms = ~0;

  bool force_huffman = false;

  bool initialize_global_state = true;

  bool streaming_mode = false;

  bool add_missing_symbols = false;

  bool add_fixed_histograms = false;

};

struct Histogram {

  Histogram() = default;

  explicit Histogram(size_t length) { EnsureCapacity(length); }

  // Create flat histogram

  static Histogram Flat(int length, int total_count) {

    Histogram flat;

    flat.counts = CreateFlatHistogram(length, total_count);

    flat.total_count = static_cast<size_t>(total_count);

    return flat;

  }

  void Clear() {

    counts.clear();

    total_count = 0;

    entropy = 0.0;

  }

  void Add(size_t symbol) {

    if (counts.size() <= symbol) {

      counts.resize(DivCeil(symbol + 1, kRounding) * kRounding);

    }

    ++counts[symbol];

    ++total_count;

  }

  // Use this before FastAdd sequence.

  void EnsureCapacity(size_t length) {

    counts.resize(DivCeil(length, kRounding) * kRounding);

  }

  // Just increment symbol counter; caller must stretch Histogram beforehead.

  void FastAdd(size_t symbol) { (*(counts.data() + symbol))++; }

  // Should be called after sequence of FastAdd to actualize total_count.

  void Condition();

  void AddHistogram(const Histogram& other) {

    if (other.counts.size() > counts.size()) {

      counts.resize(other.counts.size());

    }

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

      counts[i] += other.counts[i];

    }

    total_count += other.total_count;

  }

  size_t alphabet_size() const {

    for (int i = counts.size() - 1; i >= 0; --i) {

      if (counts[i] > 0) {

        return i + 1;

      }

    }

    return 0;

  }

  size_t MaxSymbol() const {

    if (total_count == 0) return 0;

    for (int i = counts.size() - 1; i > 0; --i) {

      if (counts[i]) return i;

    }

    return 0;

  }

  // Returns an estimate of the number of bits required to encode the given

  // histogram (header bits plus data bits).

  StatusOr<float> ANSPopulationCost() const;

  float ShannonEntropy() const;

  void swap(Histogram& other) {

    counts.swap(other.counts);

    std::swap(total_count, other.total_count);

    std::swap(entropy, other.entropy);

  }

  std::vector<ANSHistBin> counts;

  size_t total_count = 0;

  mutable float entropy = 0;  // WARNING: not kept up-to-date.

  static constexpr size_t kRounding = 8;

};

}  // namespace jxl

#endif  // LIB_JXL_ENC_ANS_PARAMS_H_