alac.c

/*
 * ALAC (Apple Lossless Audio Codec) decoder
 * Copyright (c) 2005 David Hammerton
 * All rights reserved.
 *
 * This is the actual decoder.
 *
 * http://crazney.net/programs/itunes/alac.html
 *
 * Permission is hereby granted, free of charge, to any person
 * obtaining a copy of this software and associated documentation
 * files (the "Software"), to deal in the Software without
 * restriction, including without limitation the rights to use,
 * copy, modify, merge, publish, distribute, sublicense, and/or
 * sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be
 * included in all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
 * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
 * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
 * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
 * OTHER DEALINGS IN THE SOFTWARE.
 *
 */

static const int host_bigendian = 0;

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#ifdef _WIN32
#include "stdint_win.h"
#else
#include <stdint.h>
#endif

#include "alac.h"

#define _Swap32(v)                                                                                 \
  do {                                                                                             \
    v = (((v)&0x000000FF) << 0x18) | (((v)&0x0000FF00) << 0x08) | (((v)&0x00FF0000) >> 0x08) |     \
        (((v)&0xFF000000) >> 0x18);                                                                \
  } while (0)

#define _Swap16(v)                                                                                 \
  do {                                                                                             \
    v = (((v)&0x00FF) << 0x08) | (((v)&0xFF00) >> 0x08);                                           \
  } while (0)

struct {
  signed int x : 24;
} se_struct_24;
#define SignExtend24(val) (se_struct_24.x = val)

void alac_free(alac_file *alac) {
  if (alac->predicterror_buffer_a)
    free(alac->predicterror_buffer_a);
  if (alac->predicterror_buffer_b)
    free(alac->predicterror_buffer_b);

  if (alac->outputsamples_buffer_a)
    free(alac->outputsamples_buffer_a);
  if (alac->outputsamples_buffer_b)
    free(alac->outputsamples_buffer_b);

  if (alac->uncompressed_bytes_buffer_a)
    free(alac->uncompressed_bytes_buffer_a);
  if (alac->uncompressed_bytes_buffer_b)
    free(alac->uncompressed_bytes_buffer_b);

  free(alac);
}

void alac_allocate_buffers(alac_file *alac) {
  alac->predicterror_buffer_a = malloc(alac->setinfo_max_samples_per_frame * 4);
  alac->predicterror_buffer_b = malloc(alac->setinfo_max_samples_per_frame * 4);

  alac->outputsamples_buffer_a = malloc(alac->setinfo_max_samples_per_frame * 4);
  alac->outputsamples_buffer_b = malloc(alac->setinfo_max_samples_per_frame * 4);

  alac->uncompressed_bytes_buffer_a = malloc(alac->setinfo_max_samples_per_frame * 4);
  alac->uncompressed_bytes_buffer_b = malloc(alac->setinfo_max_samples_per_frame * 4);
}

void alac_set_info(alac_file *alac, char *inputbuffer) {
  char *ptr = inputbuffer;
  ptr += 4; /* size */
  ptr += 4; /* frma */
  ptr += 4; /* alac */
  ptr += 4; /* size */
  ptr += 4; /* alac */

  ptr += 4; /* 0 ? */

  alac->setinfo_max_samples_per_frame = *(uint32_t *)ptr; /* buffer size / 2 ? */
  if (!host_bigendian)
    _Swap32(alac->setinfo_max_samples_per_frame);
  ptr += 4;
  alac->setinfo_7a = *(uint8_t *)ptr;
  ptr += 1;
  alac->setinfo_sample_size = *(uint8_t *)ptr;
  ptr += 1;
  alac->setinfo_rice_historymult = *(uint8_t *)ptr;
  ptr += 1;
  alac->setinfo_rice_initialhistory = *(uint8_t *)ptr;
  ptr += 1;
  alac->setinfo_rice_kmodifier = *(uint8_t *)ptr;
  ptr += 1;
  alac->setinfo_7f = *(uint8_t *)ptr;
  ptr += 1;
  alac->setinfo_80 = *(uint16_t *)ptr;
  if (!host_bigendian)
    _Swap16(alac->setinfo_80);
  ptr += 2;
  alac->setinfo_82 = *(uint32_t *)ptr;
  if (!host_bigendian)
    _Swap32(alac->setinfo_82);
  ptr += 4;
  alac->setinfo_86 = *(uint32_t *)ptr;
  if (!host_bigendian)
    _Swap32(alac->setinfo_86);
  ptr += 4;
  alac->setinfo_8a_rate = *(uint32_t *)ptr;
  if (!host_bigendian)
    _Swap32(alac->setinfo_8a_rate);

  alac_allocate_buffers(alac);
}

/* stream reading */

/* supports reading 1 to 16 bits, in big endian format */
static uint32_t readbits_16(alac_file *alac, int bits) {
  uint32_t result;
  int new_accumulator;

  result = (alac->input_buffer[0] << 16) | (alac->input_buffer[1] << 8) | (alac->input_buffer[2]);

  /* shift left by the number of bits we've already read,
   * so that the top 'n' bits of the 24 bits we read will
   * be the return bits */
  result = result << alac->input_buffer_bitaccumulator;

  result = result & 0x00ffffff;

  /* and then only want the top 'n' bits from that, where
   * n is 'bits' */
  result = result >> (24 - bits);

  new_accumulator = (alac->input_buffer_bitaccumulator + bits);

  /* increase the buffer pointer if we've read over n bytes. */
  alac->input_buffer += (new_accumulator >> 3);

  /* and the remainder goes back into the bit accumulator */
  alac->input_buffer_bitaccumulator = (new_accumulator & 7);

  return result;
}

/* supports reading 1 to 32 bits, in big endian format */
static uint32_t readbits(alac_file *alac, int bits) {
  int32_t result = 0;

  if (bits > 16) {
    bits -= 16;
    result = readbits_16(alac, 16) << bits;
  }

  result |= readbits_16(alac, bits);

  return result;
}

/* reads a single bit */
static int readbit(alac_file *alac) {
  int result;
  int new_accumulator;

  result = alac->input_buffer[0];

  result = result << alac->input_buffer_bitaccumulator;

  result = result >> 7 & 1;

  new_accumulator = (alac->input_buffer_bitaccumulator + 1);

  alac->input_buffer += (new_accumulator / 8);

  alac->input_buffer_bitaccumulator = (new_accumulator % 8);

  return result;
}

static void unreadbits(alac_file *alac, int bits) {
  int new_accumulator = (alac->input_buffer_bitaccumulator - bits);

  alac->input_buffer += (new_accumulator >> 3);

  alac->input_buffer_bitaccumulator = (new_accumulator & 7);
  if (alac->input_buffer_bitaccumulator < 0)
    alac->input_buffer_bitaccumulator *= -1;
}

/* various implementations of count_leading_zero:
 * the first one is the original one, the simplest and most
 * obvious for what it's doing. never use this.
 * then there are the asm ones. fill in as necessary
 * and finally an unrolled and optimised c version
 * to fall back to
 */
#if 0
/* hideously inefficient. could use a bitmask search,
 * alternatively bsr on x86,
 */
static int count_leading_zeros(int32_t input)
{
    int i = 0;
    while (!(0x80000000 & input) && i < 32)
    {
        i++;
        input = input << 1;
    }
    return i;
}
#elif defined(__GNUC__)
/* for some reason the unrolled version (below) is
 * actually faster than this. yay intel!
 */
static int count_leading_zeros(int input) { return __builtin_clz(input); }
#elif defined(_MSC_VER) && defined(_M_IX86)
static int count_leading_zeros(int input) {
  int output = 0;
  if (!input)
    return 32;
  __asm
  {
        mov eax, input;
        mov edx, 0x1f;
        bsr ecx, eax;
        sub edx, ecx;
        mov output, edx;
  }
  return output;
}
#else
#warning using generic count leading zeroes. You may wish to write one for your CPU / compiler
static int count_leading_zeros(int input) {
  int output = 0;
  int curbyte = 0;

  curbyte = input >> 24;
  if (curbyte)
    goto found;
  output += 8;

  curbyte = input >> 16;
  if (curbyte & 0xff)
    goto found;
  output += 8;

  curbyte = input >> 8;
  if (curbyte & 0xff)
    goto found;
  output += 8;

  curbyte = input;
  if (curbyte & 0xff)
    goto found;
  output += 8;

  return output;

found:
  if (!(curbyte & 0xf0)) {
    output += 4;
  } else
    curbyte >>= 4;

  if (curbyte & 0x8)
    return output;
  if (curbyte & 0x4)
    return output + 1;
  if (curbyte & 0x2)
    return output + 2;
  if (curbyte & 0x1)
    return output + 3;

  /* shouldn't get here: */
  return output + 4;
}
#endif

#define RICE_THRESHOLD 8 // maximum number of bits for a rice prefix.

static int32_t entropy_decode_value(alac_file *alac, int readSampleSize, int k,
                                    int rice_kmodifier_mask) {
  int32_t x = 0; // decoded value

  // read x, number of 1s before 0 represent the rice value.
  while (x <= RICE_THRESHOLD && readbit(alac)) {
    x++;
  }

  if (x > RICE_THRESHOLD) {
    // read the number from the bit stream (raw value)
    int32_t value;

    value = readbits(alac, readSampleSize);

    // mask value
    value &= (((uint32_t)0xffffffff) >> (32 - readSampleSize));

    x = value;
  } else {
    if (k != 1) {
      int extraBits = readbits(alac, k);

      // x = x * (2^k - 1)
      x *= (((1 << k) - 1) & rice_kmodifier_mask);

      if (extraBits > 1)
        x += extraBits - 1;
      else
        unreadbits(alac, 1);
    }
  }

  return x;
}

static void entropy_rice_decode(alac_file *alac, int32_t *outputBuffer, int outputSize,
                                int readSampleSize, int rice_initialhistory, int rice_kmodifier,
                                int rice_historymult, int rice_kmodifier_mask) {
  int outputCount;
  int history = rice_initialhistory;
  int signModifier = 0;

  for (outputCount = 0; outputCount < outputSize; outputCount++) {
    int32_t decodedValue;
    int32_t finalValue;
    int32_t k;

    k = 31 - rice_kmodifier - count_leading_zeros((history >> 9) + 3);

    if (k < 0)
      k += rice_kmodifier;
    else
      k = rice_kmodifier;

    // note: don't use rice_kmodifier_mask here (set mask to 0xFFFFFFFF)
    decodedValue = entropy_decode_value(alac, readSampleSize, k, 0xFFFFFFFF);

    decodedValue += signModifier;
    finalValue = (decodedValue + 1) / 2; // inc by 1 and shift out sign bit
    if (decodedValue & 1)                // the sign is stored in the low bit
      finalValue *= -1;

    outputBuffer[outputCount] = finalValue;

    signModifier = 0;

    // update history
    history += (decodedValue * rice_historymult) - ((history * rice_historymult) >> 9);

    if (decodedValue > 0xFFFF)
      history = 0xFFFF;

    // special case, for compressed blocks of 0
    if ((history < 128) && (outputCount + 1 < outputSize)) {
      int32_t blockSize;

      signModifier = 1;

      k = count_leading_zeros(history) + ((history + 16) / 64) - 24;

      // note: blockSize is always 16bit
      blockSize = entropy_decode_value(alac, 16, k, rice_kmodifier_mask);

      // got blockSize 0s
      if (blockSize > 0) {
        memset(&outputBuffer[outputCount + 1], 0, blockSize * sizeof(*outputBuffer));
        outputCount += blockSize;
      }

      if (blockSize > 0xFFFF)
        signModifier = 0;

      history = 0;
    }
  }
}

#define SIGN_EXTENDED32(val, bits) ((val << (32 - bits)) >> (32 - bits))

#define SIGN_ONLY(v) ((v < 0) ? (-1) : ((v > 0) ? (1) : (0)))

static void predictor_decompress_fir_adapt(int32_t *error_buffer, int32_t *buffer_out,
                                           int output_size, int readsamplesize,
                                           int16_t *predictor_coef_table, int predictor_coef_num,
                                           int predictor_quantitization) {
  int i;

  /* first sample always copies */
  *buffer_out = *error_buffer;

  if (!predictor_coef_num) {
    if (output_size <= 1)
      return;
    memcpy(buffer_out + 1, error_buffer + 1, (output_size - 1) * 4);
    return;
  }

  if (predictor_coef_num == 0x1f) /* 11111 - max value of predictor_coef_num */
  {                               /* second-best case scenario for fir decompression,
                                   * error describes a small difference from the previous sample only
                                   */
    if (output_size <= 1)
      return;
    for (i = 0; i < output_size - 1; i++) {
      int32_t prev_value;
      int32_t error_value;

      prev_value = buffer_out[i];
      error_value = error_buffer[i + 1];
      buffer_out[i + 1] = SIGN_EXTENDED32((prev_value + error_value), readsamplesize);
    }
    return;
  }

  /* read warm-up samples */
  if (predictor_coef_num > 0) {
    int i;
    for (i = 0; i < predictor_coef_num; i++) {
      int32_t val;

      val = buffer_out[i] + error_buffer[i + 1];

      val = SIGN_EXTENDED32(val, readsamplesize);

      buffer_out[i + 1] = val;
    }
  }

#if 0
    /* 4 and 8 are very common cases (the only ones i've seen). these
     * should be unrolled and optimised
     */
    if (predictor_coef_num == 4)
    {
        /* FIXME: optimised general case */
        return;
    }

    if (predictor_coef_table == 8)
    {
        /* FIXME: optimised general case */
        return;
    }
#endif

  /* general case */
  if (predictor_coef_num > 0) {
    for (i = predictor_coef_num + 1; i < output_size; i++) {
      int j;
      int sum = 0;
      int outval;
      int error_val = error_buffer[i];

      for (j = 0; j < predictor_coef_num; j++) {
        sum += (buffer_out[predictor_coef_num - j] - buffer_out[0]) * predictor_coef_table[j];
      }

      outval = (1 << (predictor_quantitization - 1)) + sum;
      outval = outval >> predictor_quantitization;
      outval = outval + buffer_out[0] + error_val;
      outval = SIGN_EXTENDED32(outval, readsamplesize);

      buffer_out[predictor_coef_num + 1] = outval;

      if (error_val > 0) {
        int predictor_num = predictor_coef_num - 1;

        while (predictor_num >= 0 && error_val > 0) {
          int val = buffer_out[0] - buffer_out[predictor_coef_num - predictor_num];
          int sign = SIGN_ONLY(val);

          predictor_coef_table[predictor_num] -= sign;

          val *= sign; /* absolute value */

          error_val -= ((val >> predictor_quantitization) * (predictor_coef_num - predictor_num));

          predictor_num--;
        }
      } else if (error_val < 0) {
        int predictor_num = predictor_coef_num - 1;

        while (predictor_num >= 0 && error_val < 0) {
          int val = buffer_out[0] - buffer_out[predictor_coef_num - predictor_num];
          int sign = -SIGN_ONLY(val);

          predictor_coef_table[predictor_num] -= sign;

          val *= sign; /* neg value */

          error_val -= ((val >> predictor_quantitization) * (predictor_coef_num - predictor_num));

          predictor_num--;
        }
      }

      buffer_out++;
    }
  }
}

static void deinterlace_16(int32_t *buffer_a, int32_t *buffer_b, int16_t *buffer_out,
                           int numchannels, int numsamples, uint8_t interlacing_shift,
                           uint8_t interlacing_leftweight) {
  int i;
  if (numsamples <= 0)
    return;

  /* weighted interlacing */
  if (interlacing_leftweight) {
    for (i = 0; i < numsamples; i++) {
      int32_t difference, midright;
      int16_t left;
      int16_t right;

      midright = buffer_a[i];
      difference = buffer_b[i];

      right = midright - ((difference * interlacing_leftweight) >> interlacing_shift);
      left = right + difference;

      /* output is always little endian */
      if (host_bigendian) {
        _Swap16(left);
        _Swap16(right);
      }

      buffer_out[i * numchannels] = left;
      buffer_out[i * numchannels + 1] = right;
    }

    return;
  }

  /* otherwise basic interlacing took place */
  for (i = 0; i < numsamples; i++) {
    int16_t left, right;

    left = buffer_a[i];
    right = buffer_b[i];

    /* output is always little endian */
    if (host_bigendian) {
      _Swap16(left);
      _Swap16(right);
    }

    buffer_out[i * numchannels] = left;
    buffer_out[i * numchannels + 1] = right;
  }
}

static void deinterlace_24(int32_t *buffer_a, int32_t *buffer_b, int uncompressed_bytes,
                           int32_t *uncompressed_bytes_buffer_a,
                           int32_t *uncompressed_bytes_buffer_b, void *buffer_out, int numchannels,
                           int numsamples, uint8_t interlacing_shift,
                           uint8_t interlacing_leftweight) {
  int i;
  if (numsamples <= 0)
    return;

  /* weighted interlacing */
  if (interlacing_leftweight) {
    for (i = 0; i < numsamples; i++) {
      int32_t difference, midright;
      int32_t left;
      int32_t right;

      midright = buffer_a[i];
      difference = buffer_b[i];

      right = midright - ((difference * interlacing_leftweight) >> interlacing_shift);
      left = right + difference;

      if (uncompressed_bytes) {
        uint32_t mask = ~(0xFFFFFFFF << (uncompressed_bytes * 8));
        left <<= (uncompressed_bytes * 8);
        right <<= (uncompressed_bytes * 8);

        left |= uncompressed_bytes_buffer_a[i] & mask;
        right |= uncompressed_bytes_buffer_b[i] & mask;
      }

      ((uint8_t *)buffer_out)[i * numchannels * 3] = (left)&0xFF;
      ((uint8_t *)buffer_out)[i * numchannels * 3 + 1] = (left >> 8) & 0xFF;
      ((uint8_t *)buffer_out)[i * numchannels * 3 + 2] = (left >> 16) & 0xFF;

      ((uint8_t *)buffer_out)[i * numchannels * 3 + 3] = (right)&0xFF;
      ((uint8_t *)buffer_out)[i * numchannels * 3 + 4] = (right >> 8) & 0xFF;
      ((uint8_t *)buffer_out)[i * numchannels * 3 + 5] = (right >> 16) & 0xFF;
    }

    return;
  }

  /* otherwise basic interlacing took place */
  for (i = 0; i < numsamples; i++) {
    int32_t left, right;

    left = buffer_a[i];
    right = buffer_b[i];

    if (uncompressed_bytes) {
      uint32_t mask = ~(0xFFFFFFFF << (uncompressed_bytes * 8));
      left <<= (uncompressed_bytes * 8);
      right <<= (uncompressed_bytes * 8);

      left |= uncompressed_bytes_buffer_a[i] & mask;
      right |= uncompressed_bytes_buffer_b[i] & mask;
    }

    ((uint8_t *)buffer_out)[i * numchannels * 3] = (left)&0xFF;
    ((uint8_t *)buffer_out)[i * numchannels * 3 + 1] = (left >> 8) & 0xFF;
    ((uint8_t *)buffer_out)[i * numchannels * 3 + 2] = (left >> 16) & 0xFF;

    ((uint8_t *)buffer_out)[i * numchannels * 3 + 3] = (right)&0xFF;
    ((uint8_t *)buffer_out)[i * numchannels * 3 + 4] = (right >> 8) & 0xFF;
    ((uint8_t *)buffer_out)[i * numchannels * 3 + 5] = (right >> 16) & 0xFF;
  }
}

void alac_decode_frame(alac_file *alac, unsigned char *inbuffer, void *outbuffer, int *outputsize) {
  int outbuffer_allocation_size = *outputsize; // initial value
  int channels;
  int32_t outputsamples = alac->setinfo_max_samples_per_frame;

  /* setup the stream */
  alac->input_buffer = inbuffer;
  alac->input_buffer_bitaccumulator = 0;

  channels = readbits(alac, 3);

  *outputsize = outputsamples * alac->bytespersample;
  if (*outputsize > outbuffer_allocation_size) {
    fprintf(stderr, "FIXME: Not enough space if the output buffer for audio frame - E1.\n");
    *outputsize = 0;
    return;
  }

  switch (channels) {
  case 0: /* 1 channel */
  {
    int hassize;
    int isnotcompressed;
    int readsamplesize;

    int uncompressed_bytes;
    int ricemodifier;

    /* 2^result = something to do with output waiting.
     * perhaps matters if we read > 1 frame in a pass?
     */
    readbits(alac, 4);

    readbits(alac, 12); /* unknown, skip 12 bits */

    hassize = readbits(alac, 1); /* the output sample size is stored soon */

    uncompressed_bytes =
        readbits(alac, 2); /* number of bytes in the (compressed) stream that are not compressed */

    isnotcompressed = readbits(alac, 1); /* whether the frame is compressed */

    if (hassize) {
      /* now read the number of samples,
       * as a 32bit integer */
      outputsamples = readbits(alac, 32);
      *outputsize = outputsamples * alac->bytespersample;
      if (*outputsize > outbuffer_allocation_size) {
        fprintf(stderr, "FIXME: Not enough space if the output buffer for audio frame - E2.\n");
        *outputsize = 0;
        return;
      }
    }

    readsamplesize = alac->setinfo_sample_size - (uncompressed_bytes * 8);

    if (!isnotcompressed) { /* so it is compressed */
      int16_t predictor_coef_table[32];
      int predictor_coef_num;
      int prediction_type;
      int prediction_quantitization;
      int i;

      /* skip 16 bits, not sure what they are. seem to be used in
       * two channel case */
      readbits(alac, 8);
      readbits(alac, 8);

      prediction_type = readbits(alac, 4);
      prediction_quantitization = readbits(alac, 4);

      ricemodifier = readbits(alac, 3);
      predictor_coef_num = readbits(alac, 5);

      /* read the predictor table */
      for (i = 0; i < predictor_coef_num; i++) {
        predictor_coef_table[i] = (int16_t)readbits(alac, 16);
      }

      if (uncompressed_bytes) {
        int i;
        for (i = 0; i < outputsamples; i++) {
          alac->uncompressed_bytes_buffer_a[i] = readbits(alac, uncompressed_bytes * 8);
        }
      }

      entropy_rice_decode(alac, alac->predicterror_buffer_a, outputsamples, readsamplesize,
                          alac->setinfo_rice_initialhistory, alac->setinfo_rice_kmodifier,
                          ricemodifier * alac->setinfo_rice_historymult / 4,
                          (1 << alac->setinfo_rice_kmodifier) - 1);

      if (prediction_type == 0) { /* adaptive fir */
        predictor_decompress_fir_adapt(alac->predicterror_buffer_a, alac->outputsamples_buffer_a,
                                       outputsamples, readsamplesize, predictor_coef_table,
                                       predictor_coef_num, prediction_quantitization);
      } else {
        fprintf(stderr, "FIXME: unhandled prediction type for compressed case: %i\n",
                prediction_type);
        /* i think the only other prediction type (or perhaps this is just a
         * boolean?) runs adaptive fir twice.. like:
         * predictor_decompress_fir_adapt(predictor_error, tempout, ...)
         * predictor_decompress_fir_adapt(predictor_error, outputsamples ...)
         * little strange..
         */
      }

    } else { /* not compressed, easy case */
      if (alac->setinfo_sample_size <= 16) {
        int i;
        for (i = 0; i < outputsamples; i++) {
          int32_t audiobits = readbits(alac, alac->setinfo_sample_size);

          audiobits = SIGN_EXTENDED32(audiobits, alac->setinfo_sample_size);

          alac->outputsamples_buffer_a[i] = audiobits;
        }
      } else {
        int i;
        for (i = 0; i < outputsamples; i++) {
          int32_t audiobits;

          audiobits = readbits(alac, 16);
          /* special case of sign extension..
           * as we'll be ORing the low 16bits into this */
          audiobits = audiobits << (alac->setinfo_sample_size - 16);
          audiobits |= readbits(alac, alac->setinfo_sample_size - 16);
          audiobits = SignExtend24(audiobits);

          alac->outputsamples_buffer_a[i] = audiobits;
        }
      }
      uncompressed_bytes = 0; // always 0 for uncompressed
    }

    switch (alac->setinfo_sample_size) {
    case 16: {
      int i;
      for (i = 0; i < outputsamples; i++) {
        int16_t sample = alac->outputsamples_buffer_a[i];
        if (host_bigendian)
          _Swap16(sample);
        ((int16_t *)outbuffer)[i * alac->numchannels] = sample;
      }
      break;
    }
    case 24: {
      int i;
      for (i = 0; i < outputsamples; i++) {
        int32_t sample = alac->outputsamples_buffer_a[i];

        if (uncompressed_bytes) {
          uint32_t mask;
          sample = sample << (uncompressed_bytes * 8);
          mask = ~(0xFFFFFFFF << (uncompressed_bytes * 8));
          sample |= alac->uncompressed_bytes_buffer_a[i] & mask;
        }

        ((uint8_t *)outbuffer)[i * alac->numchannels * 3] = (sample)&0xFF;
        ((uint8_t *)outbuffer)[i * alac->numchannels * 3 + 1] = (sample >> 8) & 0xFF;
        ((uint8_t *)outbuffer)[i * alac->numchannels * 3 + 2] = (sample >> 16) & 0xFF;
      }
      break;
    }
    case 20:
    case 32:
      fprintf(stderr, "FIXME: unimplemented sample size %i\n", alac->setinfo_sample_size);
      break;
    default:
      break;
    }
    break;
  }
  case 1: /* 2 channels */
  {
    int hassize;
    int isnotcompressed;
    int readsamplesize;

    int uncompressed_bytes;

    uint8_t interlacing_shift;
    uint8_t interlacing_leftweight;

    /* 2^result = something to do with output waiting.
     * perhaps matters if we read > 1 frame in a pass?
     */
    readbits(alac, 4);

    readbits(alac, 12); /* unknown, skip 12 bits */

    hassize = readbits(alac, 1); /* the output sample size is stored soon */

    uncompressed_bytes = readbits(
        alac, 2); /* the number of bytes in the (compressed) stream that are not compressed */

    isnotcompressed = readbits(alac, 1); /* whether the frame is compressed */

    if (hassize) {
      /* now read the number of samples,
       * as a 32bit integer */
      outputsamples = readbits(alac, 32);
      *outputsize = outputsamples * alac->bytespersample;
      if (*outputsize > outbuffer_allocation_size) {
        fprintf(stderr, "FIXME: Not enough space if the output buffer for audio frame - E3.\n");
        *outputsize = 0;
        return;
      }
    }

    readsamplesize = alac->setinfo_sample_size - (uncompressed_bytes * 8) + 1;

    if (!isnotcompressed) { /* compressed */
      int16_t predictor_coef_table_a[32];
      int predictor_coef_num_a;
      int prediction_type_a;
      int prediction_quantitization_a;
      int ricemodifier_a;

      int16_t predictor_coef_table_b[32];
      int predictor_coef_num_b;
      int prediction_type_b;
      int prediction_quantitization_b;
      int ricemodifier_b;

      int i;

      interlacing_shift = readbits(alac, 8);
      interlacing_leftweight = readbits(alac, 8);

      /******** channel 1 ***********/
      prediction_type_a = readbits(alac, 4);
      prediction_quantitization_a = readbits(alac, 4);

      ricemodifier_a = readbits(alac, 3);
      predictor_coef_num_a = readbits(alac, 5);

      /* read the predictor table */
      for (i = 0; i < predictor_coef_num_a; i++) {
        predictor_coef_table_a[i] = (int16_t)readbits(alac, 16);
      }

      /******** channel 2 *********/
      prediction_type_b = readbits(alac, 4);
      prediction_quantitization_b = readbits(alac, 4);

      ricemodifier_b = readbits(alac, 3);
      predictor_coef_num_b = readbits(alac, 5);

      /* read the predictor table */
      for (i = 0; i < predictor_coef_num_b; i++) {
        predictor_coef_table_b[i] = (int16_t)readbits(alac, 16);
      }

      /*********************/
      if (uncompressed_bytes) { /* see mono case */
        int i;
        for (i = 0; i < outputsamples; i++) {
          alac->uncompressed_bytes_buffer_a[i] = readbits(alac, uncompressed_bytes * 8);
          alac->uncompressed_bytes_buffer_b[i] = readbits(alac, uncompressed_bytes * 8);
        }
      }

      /* channel 1 */
      entropy_rice_decode(alac, alac->predicterror_buffer_a, outputsamples, readsamplesize,
                          alac->setinfo_rice_initialhistory, alac->setinfo_rice_kmodifier,
                          ricemodifier_a * alac->setinfo_rice_historymult / 4,
                          (1 << alac->setinfo_rice_kmodifier) - 1);

      if (prediction_type_a == 0) { /* adaptive fir */
        predictor_decompress_fir_adapt(alac->predicterror_buffer_a, alac->outputsamples_buffer_a,
                                       outputsamples, readsamplesize, predictor_coef_table_a,
                                       predictor_coef_num_a, prediction_quantitization_a);
      } else { /* see mono case */
        fprintf(stderr, "FIXME: unhandled prediction type on channel 1: %i\n", prediction_type_a);
      }

      /* channel 2 */
      entropy_rice_decode(alac, alac->predicterror_buffer_b, outputsamples, readsamplesize,
                          alac->setinfo_rice_initialhistory, alac->setinfo_rice_kmodifier,
                          ricemodifier_b * alac->setinfo_rice_historymult / 4,
                          (1 << alac->setinfo_rice_kmodifier) - 1);

      if (prediction_type_b == 0) { /* adaptive fir */
        predictor_decompress_fir_adapt(alac->predicterror_buffer_b, alac->outputsamples_buffer_b,
                                       outputsamples, readsamplesize, predictor_coef_table_b,
                                       predictor_coef_num_b, prediction_quantitization_b);
      } else {
        fprintf(stderr, "FIXME: unhandled prediction type on channel 2: %i\n", prediction_type_b);
      }
    } else { /* not compressed, easy case */
      if (alac->setinfo_sample_size <= 16) {
        int i;
        for (i = 0; i < outputsamples; i++) {
          int32_t audiobits_a, audiobits_b;

          audiobits_a = readbits(alac, alac->setinfo_sample_size);
          audiobits_b = readbits(alac, alac->setinfo_sample_size);

          audiobits_a = SIGN_EXTENDED32(audiobits_a, alac->setinfo_sample_size);
          audiobits_b = SIGN_EXTENDED32(audiobits_b, alac->setinfo_sample_size);

          alac->outputsamples_buffer_a[i] = audiobits_a;
          alac->outputsamples_buffer_b[i] = audiobits_b;
        }
      } else {
        int i;
        for (i = 0; i < outputsamples; i++) {
          int32_t audiobits_a, audiobits_b;

          audiobits_a = readbits(alac, 16);
          audiobits_a = audiobits_a << (alac->setinfo_sample_size - 16);
          audiobits_a |= readbits(alac, alac->setinfo_sample_size - 16);
          audiobits_a = SignExtend24(audiobits_a);

          audiobits_b = readbits(alac, 16);
          audiobits_b = audiobits_b << (alac->setinfo_sample_size - 16);
          audiobits_b |= readbits(alac, alac->setinfo_sample_size - 16);
          audiobits_b = SignExtend24(audiobits_b);

          alac->outputsamples_buffer_a[i] = audiobits_a;
          alac->outputsamples_buffer_b[i] = audiobits_b;
        }
      }
      uncompressed_bytes = 0; // always 0 for uncompressed
      interlacing_shift = 0;
      interlacing_leftweight = 0;
    }

    switch (alac->setinfo_sample_size) {
    case 16: {
      deinterlace_16(alac->outputsamples_buffer_a, alac->outputsamples_buffer_b,
                     (int16_t *)outbuffer, alac->numchannels, outputsamples, interlacing_shift,
                     interlacing_leftweight);
      break;
    }
    case 24: {
      deinterlace_24(alac->outputsamples_buffer_a, alac->outputsamples_buffer_b, uncompressed_bytes,
                     alac->uncompressed_bytes_buffer_a, alac->uncompressed_bytes_buffer_b,
                     (int16_t *)outbuffer, alac->numchannels, outputsamples, interlacing_shift,
                     interlacing_leftweight);
      break;
    }
    case 20:
    case 32:
      fprintf(stderr, "FIXME: unimplemented sample size %i\n", alac->setinfo_sample_size);
      break;
    default:
      break;
    }

    break;
  }
  }
}

alac_file *alac_create(int samplesize, int numchannels) {
  alac_file *newfile = malloc(sizeof(alac_file));

  memset(newfile, 0, sizeof(alac_file));

  newfile->samplesize = samplesize;
  newfile->numchannels = numchannels;
  newfile->bytespersample = (samplesize / 8) * numchannels;

  return newfile;
}