AudioOutput.h

Go to the documentation of this file.

#pragma once
#include "AudioTools/CoreAudio/AudioTypes.h"
#include "AudioTools/CoreAudio/BaseConverter.h"
#include "AudioTools/CoreAudio/Buffers.h"
#include "AudioToolsConfig.h"
#if defined(USE_ESP32_DSP)
#include "esp_dsp.h"
#endif
 
namespace audio_tools {
 
#if USE_PRINT_FLUSH
#define PRINT_FLUSH_OVERRIDE override
#else
#define PRINT_FLUSH_OVERRIDE
#endif

class AudioOutput : public Print,
                    public AudioInfoSupport,
                    public AudioInfoSource {
 public:
  virtual ~AudioOutput() = default;
 
  virtual size_t write(const uint8_t *data, size_t len) override = 0;
 
  virtual size_t write(uint8_t ch) override {
    if (tmp.isFull()) {
      flush();
    }
    return tmp.write(ch);
  }
 
  virtual int availableForWrite() override { return DEFAULT_BUFFER_SIZE; }
 
  // removed override because some old implementation did not define this method
  // as virtual
  virtual void flush() PRINT_FLUSH_OVERRIDE {
    if (tmp.available() > 0) {
      write((const uint8_t *)tmp.address(), tmp.available());
    }
  }
 
  // overwrite to do something useful
  virtual void setAudioInfo(AudioInfo newInfo) override {
    TRACED();
    if (cfg != newInfo) {
      cfg = newInfo;
      cfg.logInfo();
    }
    AudioInfo out = audioInfoOut();
    if (out) notifyAudioChange(out);
  }

  virtual bool isDeletable() { return false; }
 
  virtual AudioInfo audioInfo() override { return cfg; }

  virtual void writeSilence(size_t len) {
    int16_t zero = 0;
    for (int j = 0; j < len / 2; j++) {
      write((uint8_t *)&zero, 2);
    }
  }
 
  virtual bool begin(AudioInfo info) {
    setAudioInfo(info);
    return begin();
  }
 
  virtual bool begin() {
    is_active = true;
    return true;
  }
  virtual void end() { is_active = false; }
 
  virtual operator bool() { return is_active; }
 
 protected:
  int tmpPos = 0;
  AudioInfo cfg;
  SingleBuffer<uint8_t> tmp{MAX_SINGLE_CHARS};
  bool is_active = false;
};

 
class ModifyingOutput : public AudioOutput {
 public:
  virtual void setOutput(Print &out) = 0;
};

template <typename T = int16_t>
class CsvOutput : public AudioOutput {
 public:
  CsvOutput(int buffer_size = DEFAULT_BUFFER_SIZE, bool active = true) {
    this->is_active = active;
  }

  CsvOutput(Print &out, int channels = 2, int buffer_size = DEFAULT_BUFFER_SIZE,
            bool active = true) {
    this->out_ptr = &out;
    this->is_active = active;
    cfg.channels = channels;
  }

  void setDelimiter(const char *del) { delimiter_str = del; }

  const char *delimiter() { return delimiter_str; }
 
  AudioInfo defaultConfig(RxTxMode mode) { return defaultConfig(); }

  AudioInfo defaultConfig() {
    AudioInfo info;
    info.channels = 2;
    info.sample_rate = 44100;
    info.bits_per_sample = sizeof(T) * 8;
    return info;
  }

  bool begin(AudioInfo info) override { return begin(info.channels); }

  bool begin(int channels) {
    TRACED();
    cfg.channels = channels;
    return begin();
  }

  bool begin() override {
    this->is_active = true;
    // if (out_ptr == &Serial){
    //   Serial.setTimeout(60000);
    // }
    return true;
  }

  virtual void setAudioInfo(AudioInfo info) override {
    TRACEI();
    this->is_active = true;
    info.logInfo();
    cfg = info;
  };

  virtual size_t write(const uint8_t *data, size_t len) override {
    LOGD("CsvOutput::write: %d", (int)len);
    if (!is_active) {
      LOGE("is not active");
      return 0;
    }
 
    if (len == 0) {
      return 0;
    }
 
    if (cfg.channels == 0) {
      LOGW("Channels not defined: using 2");
      cfg.channels = 2;
    }
    size_t lenChannels = len / (sizeof(T) * cfg.channels);
    if (lenChannels > 0) {
      writeFrames((T *)data, lenChannels);
    } else if (len == sizeof(T)) {
      // if the write contains less then a frame we buffer the data
      T *data_value = (T *)data;
      out_ptr->print(data_value[0]);
      channel++;
      if (channel == cfg.channels) {
        out_ptr->println();
        channel = 0;
      } else {
        out_ptr->print(delimiter_str);
      }
    } else {
      LOGE("Unsupported size: %d for channels %d and bits: %d", (int)len,
           cfg.channels, cfg.bits_per_sample);
    }
#if USE_PRINT_FLUSH
    out_ptr->flush();
#endif
    return len;
  }
 
  int availableForWrite() override { return 1024; }
 
 protected:
  T *data_ptr;
  Print *out_ptr = &Serial;
  int channel = 0;
  const char *delimiter_str = ",";
 
  void writeFrames(T *data_ptr, int frameCount) {
    for (size_t j = 0; j < frameCount; j++) {
      for (int ch = 0; ch < cfg.channels; ch++) {
        if (out_ptr != nullptr && data_ptr != nullptr) {
          T value = *data_ptr;
          out_ptr->print(value);
        }
        data_ptr++;
        if (ch < cfg.channels - 1) this->out_ptr->print(delimiter_str);
      }
      this->out_ptr->println();
    }
  }
};

class HexDumpOutput : public AudioOutput {
 public:
  HexDumpOutput(int buffer_size = DEFAULT_BUFFER_SIZE, bool active = true) {
    this->is_active = active;
  }

  HexDumpOutput(Print &out, int buffer_size = DEFAULT_BUFFER_SIZE,
                bool active = true) {
    this->out_ptr = &out;
    this->is_active = active;
  }
 
  bool begin() override {
    TRACED();
    this->is_active = true;
    pos = 0;
    return is_active;
  }
 
  void flush() override {
    out_ptr->println();
    pos = 0;
  }
 
  virtual size_t write(const uint8_t *data, size_t len) override {
    if (!is_active) return 0;
    TRACED();
    for (size_t j = 0; j < len; j++) {
      out_ptr->print(data[j], HEX);
      out_ptr->print(" ");
      pos++;
      if (pos == 8) {
        out_ptr->print(" - ");
      }
      if (pos == 16) {
        out_ptr->println();
        pos = 0;
      }
    }
    return len;
  }
 
  //
  AudioInfo defaultConfig(RxTxMode mode = TX_MODE) {
    AudioInfo info;
    return info;
  }
 
 protected:
  Print *out_ptr = &Serial;
  int pos = 0;
};


template <typename T>
struct MixGain {
  float value = 0;
  MixGain() = default;
  MixGain(float f) : value(f) {}
  T scale(T sample) const { return static_cast<T>(value * sample); }
};
 
#if PREFER_FIXEDPOINT
template <>
struct MixGain<int16_t> {
  q1_14_t value;
  MixGain() = default;
  MixGain(float f) : value(f) {}
  int16_t scale(int16_t sample) const { return value.scale(sample); }
};
template <>
struct MixGain<int24_t> {
  q1_14_t value;
  MixGain() = default;
  MixGain(float f) : value(f) {}
  int24_t scale(int24_t sample) const { return value.scale(sample); }
};
template <>
struct MixGain<int32_t> {
  q1_14_t value;
  MixGain() = default;
  MixGain(float f) : value(f) {}
  int32_t scale(int32_t sample) const { return value.scale(sample); }
};
#endif
 
template <typename T = int16_t>
class OutputMixer : public Print {
 public:
  OutputMixer(Allocator &allocator = DefaultAllocatorRAM)
      : allocator(allocator) {}

  OutputMixer(Print &finalOutput, int outputStreamCount,
              Allocator &allocator = DefaultAllocatorRAM)
      : OutputMixer(allocator) {
    setOutput(finalOutput);
    setOutputCount(outputStreamCount);
  }
 
  ~OutputMixer() { free_buffers(); }

  void setOutput(Print &finalOutput) { p_final_output = &finalOutput; }

  void setOutputCount(int count) {
    // release any buffers allocated for a previous configuration
    for (int i = 0; i < buffers.size(); i++) {
      if (buffers[i] != nullptr) {
        delete buffers[i];
        buffers[i] = nullptr;
      }
    }
    output_count = count;
    buffers.resize(count);
    for (int i = 0; i < count; i++) {
      buffers[i] = nullptr;
    }
    weights.resize(count);
    for (int i = 0; i < count; i++) {
      weights[i] = 1.0;
    }
 
    update_total_weights();
  }

  void setWeight(int channel, float weight) {
    if (channel >= 0 && channel < size()) {
      weights[channel] = weight;
    } else {
      LOGE("Invalid channel %d - max is %d", channel, size() - 1);
    }
    update_total_weights();
  }

  bool begin(int copy_buffer_size_bytes = DEFAULT_BUFFER_SIZE) {
    is_active = true;
    size_bytes = copy_buffer_size_bytes;
    stream_idx = 0;
    update_total_weights();
    allocate_buffers(size_bytes);
    return true;
  }

  void end() {
    total_weights = 0.0;
    is_active = false;
    // release memory for buffers
    free_buffers();
  }

  int size() { return output_count; }

  size_t write(uint8_t) override { return 0; }

  size_t write(const uint8_t *data, size_t len) override {
    size_t result = write(stream_idx, data, len);
    // after writing the last stream we flush
    if (is_auto_index) {
      stream_idx++;
      if (stream_idx >= output_count) {
        flushMixer();
      }
    }
    return result;
  }

  size_t write(int idx, const uint8_t *buffer_c, size_t bytes) {
    LOGD("write idx %d: %d", idx, (int)bytes);
    size_t result = 0;
    BaseBuffer<T> *p_buffer =
        (idx >= 0 && idx < output_count) ? buffers[idx] : nullptr;
    if (p_buffer == nullptr) {
      LOGE("Invalid stream idx %d - max is %d", idx, output_count - 1);
      return 0;
    }
    size_t samples = bytes / sizeof(T);
    if (p_buffer->availableForWrite() >= samples) {
      result = p_buffer->writeArray((T *)buffer_c, samples) * sizeof(T);
    } else {
      LOGW(
          "Available Buffer %d too small %d: requested: %d -> increase the "
          "buffer size",
          (int)idx, static_cast<int>(p_buffer->availableForWrite() * sizeof(T)),
          (int)bytes);
    }
    return result;
  }

  int availableForWrite() override {
    return is_active ? availableForWrite(stream_idx) : 0;
  }

  int availableForWrite(int idx) {
    if (idx < 0 || idx >= output_count) return 0;
    BaseBuffer<T> *p_buffer = buffers[idx];
    if (p_buffer == nullptr) return 0;
    return p_buffer->availableForWrite() * sizeof(T);
  }

  int available(int idx) {
    if (idx < 0 || idx >= output_count) return 0;
    BaseBuffer<T> *p_buffer = buffers[idx];
    if (p_buffer == nullptr) return 0;
    return p_buffer->available() * sizeof(T);
  }

  int availablePercent(int idx) { return 100.0 * available(idx) / size_bytes; }

  bool flushMixer() {
    LOGD("flush");
    if (p_final_output == nullptr) {
      return false;
    }
    bool result = false;
 
    // determine ringbuffer with mininum available data
    size_t samples = availableSamples();
    // sum up samples
    if (samples > 0) {
      result = true;
      mixSamples(samples);
      // write output
      LOGD("write to final out: %d", static_cast<int>(samples * sizeof(T)));
      p_final_output->write((uint8_t *)output.data(), samples * sizeof(T));
    }
    stream_idx = 0;
    return result;
  }
 
 

  int availableSamples() {
    if (output_count == 0) return 0;
    size_t samples = size_bytes / sizeof(T);
    for (int j = 0; j < output_count; j++) {
      size_t available_samples = buffers[j]->available();
      samples = min(samples, available_samples);
    }
    return samples;
  }

  bool resize(size_t sizeBytes) {
    bool result = true;
    if (sizeBytes != size_bytes) {
      result = allocate_buffers(sizeBytes);
      if (result) {
        size_bytes = sizeBytes;
      }
    }
    return result;
  }

  size_t writeSilence(size_t bytes) {
    if (bytes == 0) return 0;
    uint8_t silence[bytes];
    memset(silence, 0, bytes);
    return write(stream_idx, silence, bytes);
  }

  size_t writeSilence(int idx, size_t bytes) {
    if (bytes == 0) return 0;
    uint8_t silence[bytes];
    memset(silence, 0, bytes);
    return write(idx, silence, bytes);
  }

  void setAutoIndex(bool flag) { is_auto_index = flag; }

  void setIndex(int idx) { stream_idx = idx; }

  void next() { stream_idx++; }

  void setCreateBufferCallback(BaseBuffer<T> *(*cb)(int size)) {
    create_buffer_cb = cb;
  }

  BaseBuffer<T> *getBuffer(int idx) {
    return (idx >= 0 && idx < output_count) ? buffers[idx] : nullptr;
  }
 
 protected:
  Vector<float> weights{0, DefaultAllocatorRAM};
  Vector<BaseBuffer<T> *> buffers{0, DefaultAllocatorRAM};
  Allocator &allocator;
  Vector<T> output{0, allocator};
  Print *p_final_output = nullptr;
  float total_weights = 0.0;
  bool is_active = false;
  int stream_idx = 0;
  int size_bytes = 0;
  int output_count = 0;
  void *p_memory = nullptr;
  bool is_auto_index = true;
  BaseBuffer<T> *(*create_buffer_cb)(int size,
                                     Allocator &allocator) = create_buffer;

  static BaseBuffer<T> *create_buffer(int sizeBytes, Allocator &allocator) {
    return new RingBuffer<T>(sizeBytes / sizeof(T), allocator);
  }

  inline void mixSamples(size_t samples) {
#if defined(USE_ESP32_DSP)
    // Temporary float buffers for mixing
    float mix_out[samples] = {0.0f};
    float temp[samples] = {0.0f};
 
    for (uint8_t j = 0; j < output_count; j++) {
        const float factor = total_weights != 0.0f ? weights[j] / total_weights : 0.0f;
        // Read int16_t samples and convert to float
        for (uint16_t i = 0; i < samples; i++) {
            T s = 0;
            buffers[j]->read(s);
            temp[i] = static_cast<float>(s) * factor;
        }
        // Add to output
        dsps_add_f32(mix_out, temp, mix_out, samples, 1, 1, 1);
    }
    // Convert back to int16_t with clamping
    output.resize(samples);
    for (size_t i = 0; i < samples; i++) {
        float v = mix_out[i];
        output[i] = static_cast<T>(v);
    }
#else
    // Fallback: original scalar code
    output.resize(samples);
    memset(output.data(), 0, samples * sizeof(T));
    for (int j = 0; j < output_count; j++) {
      MixGain<T> factor(total_weights != 0.0f ? weights[j] / total_weights : 0.0f);
      for (int i = 0; i < samples; i++) {
        T sample = 0;
        buffers[j]->read(sample);
        output[i] += factor.scale(sample);
      }
    }
#endif
  }

  void update_total_weights() {
    total_weights = 0.0;
    for (int j = 0; j < weights.size(); j++) {
      total_weights += weights[j];
    }
  }

  bool allocate_buffers(int sizeBytes) {
    // allocate ringbuffers for each output
    for (int j = 0; j < output_count; j++) {
      if (buffers[j] != nullptr) {
        delete buffers[j];
      }
      buffers[j] = create_buffer(sizeBytes, allocator);
      if (buffers[j] == nullptr) {
        return false;
      }
    }
    return true;
  }

  void free_buffers() {
    // allocate ringbuffers for each output
    for (int j = 0; j < output_count; j++) {
      if (buffers[j] != nullptr) {
        delete buffers[j];
        buffers[j] = nullptr;
      }
    }
  }
};

class MemoryOutput : public AudioOutput {
 public:
  MemoryOutput(uint8_t *start, int len) {
    p_start = start;
    p_next = start;
    max_size = len;
    is_active = true;
    if (p_next == nullptr) {
      LOGE("start must not be null");
    }
  }
 
  bool begin() override {
    is_active = true;
    p_next = p_start;
    pos = 0;
    return true;
  }
 
  size_t write(const uint8_t *data, size_t len) override {
    if (p_next == nullptr) return 0;
    if (pos + len <= max_size) {
      memcpy(p_next, data, len);
      pos += len;
      p_next += len;
      return len;
    } else {
      LOGE("Buffer too small: pos:%d, size: %d ", pos, (int)max_size);
      return 0;
    }
  }
 
  int availableForWrite() override { return max_size - pos; }
 
  int size() { return max_size; }
 
 protected:
  int pos = 0;
  uint8_t *p_start = nullptr;
  uint8_t *p_next = nullptr;
  size_t max_size;
};

class ChannelSplitOutput : public AudioOutput {
 public:
  ChannelSplitOutput() = default;
 
  ChannelSplitOutput(Print &out, int channel) { addOutput(out, channel); }

  void addOutput(Print &out, int channel) {
    ChannelSelectionOutputDef def;
    def.channel = channel;
    def.p_out = &out;
    out_channels.push_back(def);
  }
 
  size_t write(const uint8_t *data, size_t len) override {
    switch (cfg.bits_per_sample) {
      case 16:
        return writeT<int16_t>(data, len);
      case 24:
        return writeT<int24_t>(data, len);
      case 32:
        return writeT<int32_t>(data, len);
      default:
        return 0;
    }
  }
 
 protected:
  struct ChannelSelectionOutputDef {
    Print *p_out = nullptr;
    int channel;
  };
  Vector<ChannelSelectionOutputDef> out_channels;
 
  template <typename T = int16_t>
  size_t writeT(const uint8_t *buffer, size_t size) {
    int sample_count = size / sizeof(T);
    int result_size = sample_count / cfg.channels;
    T *data = (T *)buffer;
    T result[result_size];
 
    for (int ch = 0; ch < out_channels.size(); ch++) {
      ChannelSelectionOutputDef &def = out_channels[ch];
      // extract mono result
      int i = 0;
      for (int j = def.channel; j < sample_count; j += cfg.channels) {
        result[i++] = data[j];
      }
      // write mono result
      size_t written =
          def.p_out->write((uint8_t *)result, result_size * sizeof(T));
      if (written != result_size * sizeof(T)) {
        LOGW("Could not write all samples");
      }
    }
    return size;
  }
};
 
}  // namespace audio_tools