mdchaney · January 10, 2024 00:13
diff --git a/audio_time.rb b/audio_time.rb
 module AudioTime
  # AIFF stores the samples per second as an IEEE 80-bit float, big-endian.
  # This code will convert it to a 32-bit integer.  This is not generalized
  # code for this operation as it ignores the sign bit, part of the mantissa,
  # etc.
  def AudioTime.ieee80_to_long(raw_number)
    discard, mantissa, hidden, exponent, sign = raw_number.reverse.unpack('b80').first.unpack('a32 a31 a1 a15 a1')
    # Get the real exponent - the exponent is really exponent + 1
    exponent = [exponent].pack('b15').unpack('v').first - 16382
    # Now, use the exponent to pull the proper bits from the mantissa.
    # The implicit "1" is added.
    new_mantissa = mantissa.slice(exponent,999) + '1'
    new_mantissa = [new_mantissa].pack('b*').unpack('v').first
    new_mantissa += 1 if exponent > 0 && mantissa.slice(exponent-1) == '1'
    return new_mantissa
  end

  def AudioTime.parse_aiff_comm_chunk(raw_comm_chunk)
    number_of_channels, number_of_sample_frames, sample_size, raw_sample_rate = raw_comm_chunk.unpack('n N n a10')
    sample_rate = ieee80_to_long(raw_sample_rate)
    bytes_per_sample = sample_size / 8
    return { number_of_channels: number_of_channels, number_of_sample_frames: number_of_sample_frames, sample_size: sample_size, bits_per_sample: sample_size, sample_rate: sample_rate, bytes_per_sample: bytes_per_sample }
  end

  def AudioTime.aiff_info(filename)
    raw_comm_chunk = comm_info = data_length = file_seconds = actual_file_size = nil

    chunk_info = []

    File.open(filename,'rb') do |file|

      begin

        raw_header = file.read(12)

        if !raw_header || raw_header.size < 12
          return { error: "Not an AIFF", valid: false, filename: filename }
        end

        form, filelen, aiff = raw_header.unpack('A4NA4')

        if form == 'FORM' && aiff =~ /\AAIF[FC]\z/
    
          # This odd code two lines below actually allows the difference
          # in actual file size and expected file size to differ by 1.
          # Some files add a padding byte if the last chunk is odd length,
          # but don't add this to to the file size in the header, causing
          # this minor inconsistency.
          actual_file_size = File.size(file)
          if (filelen + 8 - actual_file_size).abs > 1
            return { error: "File size error for AIFF, header says #{filelen+8}, actual is #{actual_file_size}", valid: false, filename: filename }
          end

          # Read chunks, get the comm and ssnd chunks
          until file.eof?
    
            raw_chunk_header = file.read(8)
            # This will force it to be "eof" above
            next unless raw_chunk_header.size == 8
    
            chunk_type, chunk_length = raw_chunk_header.unpack('A4N')
    
            chunk_info.push({chunk_type: chunk_type, position: file.pos, length: chunk_length})

            if chunk_type == 'COMM'
              raw_comm_chunk = file.read(chunk_length)
              comm_info = parse_aiff_comm_chunk(raw_comm_chunk)
    
            else

              data_length = chunk_length if chunk_type == 'SSND'
      
              file.seek(chunk_length + (chunk_length.odd? ? 1 : 0),IO::SEEK_CUR)

            end
    
          end
    
        else
          return { error: "Not an AIFF", valid: false, filename: filename }
        end

      end
    end

    if !data_length.nil? && !comm_info.nil?
      # Make sure chunk lengths add up.  Each chunk has an 8-byte header,
      # plus the file has a 12 byte header.
      total_chunk_length = 12 + chunk_info.size * 8 + chunk_info.map { |ci| ci[:length] }.inject(0) { |s,n| s+n } + chunk_info.select { |ci| ci[:length].odd? }.size
      if total_chunk_length == actual_file_size || (total_chunk_length - 1 == actual_file_size && chunk_info.last[:length].odd?)
        average_bytes_per_second = comm_info[:bytes_per_sample] * comm_info[:number_of_channels] * comm_info[:sample_rate]
        file_seconds = data_length.to_f / average_bytes_per_second.to_f
        return comm_info.merge({format: 'aif', seconds: file_seconds, average_bytes_per_second: average_bytes_per_second, data_length: data_length, valid: true, filename: filename, chunk_info: chunk_info })
      else
        return { format: 'aif', error: "AIFF chunks total length is different than file size: #{total_chunk_length} vs. #{actual_file_size}", valid: false, filename: filename }
      end
    else
      return { error: "Malformed AIFF", valid: false, filename: filename }
    end
  end

  def AudioTime.parse_wav_fmt_chunk(raw_fmt_chunk)
    # Expand it out to 40 characters
    raw_fmt_chunk += (0.chr*(40-raw_fmt_chunk.size))
    format_tag, number_of_channels, sample_rate, average_bytes_per_second, block_align, bits_per_sample, extended_size, valid_bits_per_sample, channel_mask, subformat = raw_fmt_chunk.unpack('v v V V v v v v V a16')
    # In this case, the format tag is the first two bytes of the subformat
    if extended_size == 22 && format_tag == 65534
      format_tag = subformat.unpack('v').first
    end
    bytes_per_sample = bits_per_sample / 8
    return { format_tag: format_tag, number_of_channels: number_of_channels, sample_rate: sample_rate, average_bytes_per_second: average_bytes_per_second, block_align: block_align, bits_per_sample: bits_per_sample, bytes_per_sample: bytes_per_sample, has_extension: (extended_size > 0), valid_bits_per_sample: valid_bits_per_sample, channel_mask: channel_mask, subformat: subformat }
  end

  def AudioTime.wav_info(filename)
    raw_fmt_chunk = fmt_info = data_length = file_seconds = actual_file_size = nil

    chunk_info = []

    File.open(filename,'rb') do |file|

      begin

        raw_header = file.read(12)

        if !raw_header || raw_header.size < 12
          return { error: "Not a WAV file", valid: false, filename: filename }
        end

        riff, filelen, wave = raw_header.unpack('A4VA4')
        if riff == 'RIFF' && wave == 'WAVE'

          # This odd code two lines below actually allows the difference
          # in actual file size and expected file size to differ by 1.
          # Some files add a padding byte if the last chunk is odd length,
          # but don't add this to to the file size in the header, causing
          # this minor inconsistency.
          actual_file_size = File.size(file)
          if (filelen + 8 - actual_file_size).abs > 1
            return { error: "File size error for WAV, header says #{filelen+8}, actual is #{actual_file_size}", valid: false, filename: filename }
          end

          # Read chunks, get the fmt and data chunks
          until file.eof?

            raw_chunk_header = file.read(8)
            # This will force it to be "eof" above
            next unless raw_chunk_header.size == 8

            chunk_type, chunk_length = raw_chunk_header.unpack('A4V')

            chunk_info.push({chunk_type: chunk_type, position: file.pos, length: chunk_length})

            if chunk_type == 'fmt'
              raw_fmt_chunk = file.read(chunk_length)
              fmt_info = parse_wav_fmt_chunk(raw_fmt_chunk)

            else

              data_length = chunk_length if chunk_type == 'data'

              # If the chunk length is odd, there should be a padding byte.
              file.seek(chunk_length + (chunk_length.odd? ? 1 : 0),IO::SEEK_CUR)

            end

          end

        else
          return { error: "Not a WAV", valid: false, filename: filename }
        end

      end
    end

    if !data_length.nil? && !fmt_info.nil?
      # Make sure chunk lengths add up.  Each chunk has an 8-byte header,
      # plus the file has a 12 byte header.  Odd sized chunks should be
      # followed by a padding byte, so that is added in at the end.
      # Note: the last chunk won't have a padding byte.
      total_chunk_length = 12 + chunk_info.size * 8 + chunk_info.map { |ci| ci[:length] }.inject(0) { |s,n| s+n } + chunk_info.select { |ci| ci[:length].odd? }.size
      if total_chunk_length == actual_file_size || (total_chunk_length - 1 == actual_file_size && chunk_info.last[:length].odd?)
        number_of_sample_frames = data_length / (fmt_info[:bytes_per_sample] * fmt_info[:number_of_channels])
        file_seconds = number_of_sample_frames.to_f / fmt_info[:sample_rate].to_f
        return fmt_info.merge({format: 'wav', number_of_sample_frames: number_of_sample_frames, seconds: file_seconds, data_length: data_length, valid: true, filename: filename, chunk_info: chunk_info })
      else
        return { format: 'wav', error: "WAV chunks total length is different than file size: #{total_chunk_length} vs. #{actual_file_size}", valid: false, filename: filename }
      end
    else
      return { error: "Malformed WAV", valid: false, filename: filename }
    end
  end

  def AudioTime.parse_id3v2_header(header)
    id, major_version, minor_version, flags, sizeb = header.unpack('A3 C C b8 b32')
    size = [sizeb[24..30]+sizeb[16..22]+sizeb[8..14]+sizeb[0..6]+'0000'].pack('b32').unpack('l').first
    return [id, major_version, minor_version, flags, size]
  end

  MP3_LOOKUPS = {
    mpeg_audio_version_descriptions: {
      '00' => 'MPEG Version 2.5',
      '01' => 'Reserved',
      '10' => 'MPEG Version 2 (ISO/IEC 13818-3)',
      '11' => 'MPEG Version 1 (ISO/IEC 11172-3)'
    },
    mpeg_audio_version_ids: {
      '00' => 2.5,
      '01' => nil,
      '10' => 2,
      '11' => 1
    },
    layer_descriptions: {
      '00' => 'Reserved',
      '01' => 'Layer III',
      '10' => 'Layer II',
      '11' => 'Layer I'
    },
    layers: {
      '00' => nil,
      '01' => 3,
      '10' => 2,
      '11' => 1
    },
    # Yeah, I know it's just binary, this is as easy as anything
    bitrate_indices: {
      '0000' =>  0, '0001' =>  1, '0010' =>  2, '0011' =>  3,
      '0100' =>  4, '0101' =>  5, '0110' =>  6, '0111' =>  7,
      '1000' =>  8, '1001' =>  9, '1010' => 10, '1011' => 11,
      '1100' => 12, '1101' => 13, '1110' => 14, '1111' => 15
    },
    # Lookup for bitrate is mpeg_audio_version_id, layer, and bitrate_index
    bitrates: {
      1 => {
        1 => [ 'free',  32,  64,  96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448, nil ],
        2 => [ 'free',  32,  48,  56,  64,  80,  96, 112, 128, 160, 192, 224, 256, 320, 384, nil ],
        3 => [ 'free',  32,  40,  48,  56,  64,  80,  96, 112, 128, 160, 192, 224, 256, 320, nil ]
      },
      2 => {
        1 => [ 'free',  32,  48,  56,  64,  80,  96, 112, 128, 144, 160, 176, 192, 224, 256, nil ],
        2 => [ 'free',   8,  16,  24,  32,  40,  48,  56,  64,  80,  96, 112, 128, 144, 160, nil ],
        3 => [ 'free',   8,  16,  24,  32,  40,  48,  56,  64,  80,  96, 112, 128, 144, 160, nil ],
      },
      2.5 => {
        1 => [ 'free',  32,  48,  56,  64,  80,  96, 112, 128, 144, 160, 176, 192, 224, 256, nil ],
        2 => [ 'free',   8,  16,  24,  32,  40,  48,  56,  64,  80,  96, 112, 128, 144, 160, nil ],
        3 => [ 'free',   8,  16,  24,  32,  40,  48,  56,  64,  80,  96, 112, 128, 144, 160, nil ],
      }
    },
    # Lookup for sampling_rate_frequencies is sampling_rate_frequency_index
    # and mpeg_audio_version_id
    sampling_rate_frequencies: {
      '00' => { 1 => 44100, 2 => 22050, 2.5 => 11025 },
      '01' => { 1 => 48000, 2 => 24000, 2.5 => 12000 },
      '10' => { 1 => 32000, 2 => 16000, 2.5 =>  8000 },
      '11' => { 1 => nil,   2 => nil,   2.5 => nil   }
    },
    channel_modes: {
      '00' => 'stereo',
      '01' => 'joint stereo (stereo)',
      '10' => 'dual channel (stereo)',
      '11' => 'single channel (mono)'
    },
    channel_counts: {
      '00' => 2,
      '01' => 2,
      '10' => 2,
      '11' => 1
    },
    emphases: {
      '00' => 'none',
      '01' => '50/15 ms',
      '10' => 'reserved',
      '11' => 'CCIT J.17'
    },
    # Lookup for samples_per_frame is mpeg_audio_version_id and layer
    samples_per_frame: {
      1   => { 1 =>  384, 2 => 1152, 3 => 1152 },
      2   => { 1 =>  384, 2 => 1152, 3 =>  576 },
      2.5 => { 1 =>  384, 2 => 1152, 3 =>  576 }
    },
    # Lookup for mode extensions - layer I and II only
    mode_extensions: {
      '00' => { low_band:  4, high_band: 31 },
      '01' => { low_band:  8, high_band: 31 },
      '10' => { low_band: 12, high_band: 31 },
      '11' => { low_band: 16, high_band: 31 }
    }
  }

  def AudioTime.parse_mp3_frame_header(frame_header)
    # MP3 frame is 4 bytes.  Info for parsing is here:
    # http://mpgedit.org/mpgedit/mpeg_format/mpeghdr.htm
    # http://www.codeproject.com/Articles/8295/MPEG-Audio-Frame-Header
    #
    # Hopefully those will stay good for a long time :)
    # Fields are of these lengths in bits: 11, 2, 2, 1, 4, 2, 1, 1, 2, 2, 1, 1, 2
    frame_sync, raw_mpeg_audio_version_id, raw_layer_description, protection_bit, raw_bitrate_index, sampling_rate_frequency_index, padding_bit, private_bit, raw_channel_mode, raw_mode_extension, copyright_flag, original_flag, raw_emphasis = frame_header.unpack('B32').first.unpack('a11 a2 a2 a1 a4 a2 a1 a1 a2 a2 a1 a1 a2')

    #STDERR.puts frame_sync, raw_mpeg_audio_version_id, raw_layer_description, protection_bit, raw_bitrate_index, sampling_rate_frequency_index, padding_bit, private_bit, raw_channel_mode, raw_mode_extension, copyright_flag, original_flag, raw_emphasis

    return nil unless frame_sync == '11111111111' || frame_sync == '11111111110'

    has_crc = (protection_bit == '0')
    mpeg_audio_version_id = MP3_LOOKUPS[:mpeg_audio_version_ids][raw_mpeg_audio_version_id]
    mpeg_audio_version_description = MP3_LOOKUPS[:mpeg_audio_version_descriptions][raw_mpeg_audio_version_id]
    layer = MP3_LOOKUPS[:layers][raw_layer_description]
    layer_description = MP3_LOOKUPS[:layer_descriptions][raw_layer_description]
    bitrate_index = MP3_LOOKUPS[:bitrate_indices][raw_bitrate_index]
    bitrate = MP3_LOOKUPS[:bitrates][mpeg_audio_version_id][layer][bitrate_index]
    sampling_rate_frequency = MP3_LOOKUPS[:sampling_rate_frequencies][sampling_rate_frequency_index][mpeg_audio_version_id]
    samples_per_frame = MP3_LOOKUPS[:samples_per_frame][mpeg_audio_version_id][layer]
    channel_mode = MP3_LOOKUPS[:channel_modes][raw_channel_mode]
    channel_count = MP3_LOOKUPS[:channel_counts][raw_channel_mode]
    emphasis = MP3_LOOKUPS[:emphases][raw_emphasis]

    if layer == 3
      mode_extension = nil
      ms_stereo = raw_mode_extension[0]
      intensity_stereo =  raw_mode_extension[1]
    else
      mode_extension = MP3_LOOKUPS[:mode_extensions][raw_mode_extension]
      intensity_stereo = ms_stereo = nil
    end

    frame_slot_count = (samples_per_frame / 8) * (bitrate * 1000) / sampling_rate_frequency + padding_bit.to_i
    frame_length =
      if layer == 1
        frame_slot_count * 4
      else
        frame_slot_count
      end

    return { frame_sync: frame_sync, raw_mpeg_audio_version_id: raw_mpeg_audio_version_id, raw_layer_description: raw_layer_description, protection_bit: protection_bit, raw_bitrate_index: raw_bitrate_index, sampling_rate_frequency_index: sampling_rate_frequency_index, padding_bit: padding_bit, private_bit: private_bit, raw_channel_mode: raw_channel_mode, raw_mode_extension: raw_mode_extension, copyright_flag: copyright_flag, original_flag: original_flag, raw_emphasis: raw_emphasis, mpeg_audio_version_id: mpeg_audio_version_id, mpeg_audio_version_description: mpeg_audio_version_description, layer_description: layer_description, has_crc: has_crc, sampling_rate_frequency: sampling_rate_frequency, channel_mode: channel_mode, channel_count: channel_count, emphasis: emphasis, frame_length: frame_length, bitrate: bitrate, layer: layer, mode_extension: mode_extension, ms_stereo: ms_stereo, intensity_stereo: intensity_stereo }
  end

  def AudioTime.mp3_info(filename)
    # Need to find all possible id3 tag sets - id3v2 and beginning and/or
    # end, and id3v1 at end.  The idea is to separate out the actual MP3
    # data so we can read and handle the frames.

    File.open(filename, 'rb') do |file|

      begin

        mp3_offset = 0
        file_size = mp3_length = file.size
        has_v1 = has_pre_v2 = has_post_v2 = false

        chunk = file.read(10)
        if !chunk || chunk.size < 10
          return { error: "Not an MP3 file", valid: false, filename: filename }
        end

        if chunk[0..2] == 'ID3'
          id, major_version, minor_version, flags, size = parse_id3v2_header(chunk)
          mp3_offset += size + 10
          mp3_length -= size + 10
          has_pre_v2 = true
          #STDERR.puts "Has a v2 tag at start, #{size+10} bytes" if DEBUG
        end

        file.seek(-128, IO::SEEK_END)
        chunk = file.read(3)

        if !chunk || chunk.size < 3
          return { error: "Not an MP3 file", valid: false, filename: filename }
        end

        # This is a v1 tag at the end
        if chunk[0..2] == 'TAG'
          mp3_length -= 128
          has_v1 = true
          #STDERR.puts "Has a v1 tag at end, 128 bytes" if DEBUG
        end

        # Check for v2 tag at the end if there wasn't a v2 tag at the
        # beginning
        unless has_pre_v2
          if has_v1
            file.seek(-138, IO::SEEK_END)
          else
            file.seek(-10, IO::SEEK_END)
          end

          chunk = file.read(10)
          if !chunk || chunk.size < 10
            return { error: "Not an MP3 file", valid: false, filename: filename }
          end

          if chunk[0..2] == '3DI'
            id, major_version, minor_version, flags, size = parse_id3v2_header(chunk)
            mp3_length -= size + 10
            has_post_v2 = true
            #STDERR.puts "Has a v2 tag at end, #{size+10} bytes" if DEBUG
          end
        end

        # Sanity checks
        if mp3_offset < 0 || mp3_offset > file_size || mp3_length < 0
          return { error: "MP3 information is confused: offset #{mp3_offset} length #{mp3_length}", valid: false, filename: filename }
        end

        mp3_end_offset = mp3_offset + mp3_length
        total_seconds = 0.0

        #STDERR.puts "mp3_offset is #{mp3_offset}, mp3_length is #{mp3_length}" if DEBUG

        # At this point, we can parse the MP3 file starting at mp3_offset and
        # continuing for mp3_length bytes.  Of course, we need to make sure
        # frame is good.

        low_bitrate = high_bitrate = nil

        first_frame_info = nil
        file_size_seconds = nil

        mpeg_audio_version_id = nil
        mpeg_audio_version_description = nil
        layer = nil
        layer_description = nil
        sampling_rate_frequency = nil
        ms_stereo = nil
        intensity_stereo = nil
        channel_mode = nil

        frame_infos = []

        file.seek(mp3_offset, IO::SEEK_SET)

        until file.pos >= mp3_end_offset || file.eof?

          frame_header = file.read(4)

          if !frame_header || frame_header.size < 4
            return { error: "Invalid MP3 file", valid: false, filename: filename }
          end

          frame_info = parse_mp3_frame_header(frame_header)

          return { error: 'Invalid frame sync in MP3 file', valid: false, filename: filename } unless frame_info

          frame_info[:offset] = file.pos

          # TODO: Check CRC if present

          low_bitrate = frame_info[:bitrate] if !low_bitrate || low_bitrate > frame_info[:bitrate]
          high_bitrate = frame_info[:bitrate] if !high_bitrate || high_bitrate < frame_info[:bitrate]

          mpeg_audio_version_id = frame_info[:mpeg_audio_version_id]
          mpeg_audio_version_description = frame_info[:mpeg_audio_version_description]
          layer = frame_info[:layer]
          layer_description = frame_info[:layer_description]
          sampling_rate_frequency = frame_info[:sampling_rate_frequency]
          ms_stereo = frame_info[:ms_stereo]
          intensity_stereo = frame_info[:intensity_stereo]
          channel_mode = frame_info[:channel_mode]

          total_seconds += frame_info[:frame_length].to_f * 8.0 / frame_info[:bitrate].to_f

          first_frame_info = frame_info unless first_frame_info

          frame_infos.push(frame_info)

          #pp frame_info.inspect

          file.seek(frame_info[:frame_length] - 4, IO::SEEK_CUR)
        end

        cbr = (low_bitrate == high_bitrate)

        file_size_seconds = mp3_length.to_f * 8.0 / (low_bitrate.to_f * 1000.0) if cbr

        mp3_info = first_frame_info

        mp3_info[:format] = 'mp3'

        mp3_info[:total_seconds] = total_seconds / 1000.0
        mp3_info[:file_size_seconds] = file_size_seconds
        mp3_info[:seconds] = mp3_info[:file_size_seconds] || mp3_info[:total_seconds]
        mp3_info[:cbr] = cbr
        mp3_info[:low_bitrate] = low_bitrate
        mp3_info[:high_bitrate] = high_bitrate

        mp3_info[:mpeg_audio_version_id] = mpeg_audio_version_id
        mp3_info[:mpeg_audio_version_description] = mpeg_audio_version_description
        mp3_info[:layer] = layer
        mp3_info[:layer_description] = layer_description
        mp3_info[:sampling_rate_frequency] = sampling_rate_frequency
        mp3_info[:ms_stereo] = ms_stereo
        mp3_info[:intensity_stereo] = intensity_stereo
        mp3_info[:channel_mode] = channel_mode

        mp3_info[:mp3_offset] = mp3_offset
        mp3_info[:mp3_length] = mp3_length
        mp3_info[:has_v1] = has_v1
        mp3_info[:has_pre_v2] = has_pre_v2
        mp3_info[:has_post_v2] = has_post_v2

        mp3_info[:valid] = true
        mp3_info[:filename] = filename

        return mp3_info

      rescue Errno::EINVAL => e
        # Invalid seek
        return { error: "Cannot seek in MP3 file", valid: false, filename: filename }

      end
    end
  end

  def AudioTime.aiff_time(filename)
    info = aiff_info(filename)
    return info[:seconds]
  end

  def AudioTime.wav_time(filename)
    info = wav_info(filename)
    return info[:seconds]
  end

  def AudioTime.mp3_time(filename)
    info = mp3_info(filename)
    return info[:seconds]
  end

  def AudioTime.aiff_lint(filename)
    info = aiff_info(filename)
    return info[:valid]
  end

  def AudioTime.wav_lint(filename)
    info = wav_info(filename)
    return info[:valid]
  end

  def AudioTime.mp3_lint(filename)
    info = mp3_info(filename)
    return info[:valid]
  end

  def AudioTime.valid_audio_file?(filename, type=nil)
    if type
      type = type.gsub(/\A\./,'').downcase
      if type == 'aif' || type == 'aiff'
        aiff_lint(filename)
      elsif type == 'wav'
        wav_lint(filename)
      elsif type == 'mp3'
        mp3_lint(filename)
      else
        nil
      end
    else
      raw_header = nil
      File.open(filename,'rb') do |file|
        raw_header = file.read(12)
      end
      if !raw_header || raw_header.size < 12
        false
      elsif raw_header =~ /\AFORM....AIF[FC]/
        aiff_lint(filename)
      elsif raw_header =~ /\ARIFF....WAVE/
        wav_lint(filename)
      else
        # MP3s are difficult to easily determine
        mp3_lint(filename)
      end
    end
  end
 end
	module AudioTime
	# AIFF stores the samples per second as an IEEE 80-bit float, big-endian.
	# This code will convert it to a 32-bit integer. This is not generalized
	# code for this operation as it ignores the sign bit, part of the mantissa,
	# etc.
	def AudioTime.ieee80_to_long(raw_number)
	discard, mantissa, hidden, exponent, sign = raw_number.reverse.unpack('b80').first.unpack('a32 a31 a1 a15 a1')
	# Get the real exponent - the exponent is really exponent + 1
	exponent = [exponent].pack('b15').unpack('v').first - 16382
	# Now, use the exponent to pull the proper bits from the mantissa.
	# The implicit "1" is added.
	new_mantissa = mantissa.slice(exponent,999) + '1'
	new_mantissa = [new_mantissa].pack('b*').unpack('v').first
	new_mantissa += 1 if exponent > 0 && mantissa.slice(exponent-1) == '1'
	return new_mantissa
	end

	def AudioTime.parse_aiff_comm_chunk(raw_comm_chunk)
	number_of_channels, number_of_sample_frames, sample_size, raw_sample_rate = raw_comm_chunk.unpack('n N n a10')
	sample_rate = ieee80_to_long(raw_sample_rate)
	bytes_per_sample = sample_size / 8
	return { number_of_channels: number_of_channels, number_of_sample_frames: number_of_sample_frames, sample_size: sample_size, bits_per_sample: sample_size, sample_rate: sample_rate, bytes_per_sample: bytes_per_sample }
	end

	def AudioTime.aiff_info(filename)
	raw_comm_chunk = comm_info = data_length = file_seconds = actual_file_size = nil

	chunk_info = []

	File.open(filename,'rb') do \|file\|

	begin

	raw_header = file.read(12)

	if !raw_header \|\| raw_header.size < 12
	return { error: "Not an AIFF", valid: false, filename: filename }
	end

	form, filelen, aiff = raw_header.unpack('A4NA4')

	if form == 'FORM' && aiff =~ /\AAIF[FC]\z/

	# This odd code two lines below actually allows the difference
	# in actual file size and expected file size to differ by 1.
	# Some files add a padding byte if the last chunk is odd length,
	# but don't add this to to the file size in the header, causing
	# this minor inconsistency.
	actual_file_size = File.size(file)
	if (filelen + 8 - actual_file_size).abs > 1
	return { error: "File size error for AIFF, header says #{filelen+8}, actual is #{actual_file_size}", valid: false, filename: filename }
	end

	# Read chunks, get the comm and ssnd chunks
	until file.eof?

	raw_chunk_header = file.read(8)
	# This will force it to be "eof" above
	next unless raw_chunk_header.size == 8

	chunk_type, chunk_length = raw_chunk_header.unpack('A4N')

	chunk_info.push({chunk_type: chunk_type, position: file.pos, length: chunk_length})

	if chunk_type == 'COMM'
	raw_comm_chunk = file.read(chunk_length)
	comm_info = parse_aiff_comm_chunk(raw_comm_chunk)

	else

	data_length = chunk_length if chunk_type == 'SSND'

	file.seek(chunk_length + (chunk_length.odd? ? 1 : 0),IO::SEEK_CUR)

	end

	end

	else
	return { error: "Not an AIFF", valid: false, filename: filename }
	end

	end
	end

	if !data_length.nil? && !comm_info.nil?
	# Make sure chunk lengths add up. Each chunk has an 8-byte header,
	# plus the file has a 12 byte header.
	total_chunk_length = 12 + chunk_info.size * 8 + chunk_info.map { \|ci\| ci[:length] }.inject(0) { \|s,n\| s+n } + chunk_info.select { \|ci\| ci[:length].odd? }.size
	if total_chunk_length == actual_file_size \|\| (total_chunk_length - 1 == actual_file_size && chunk_info.last[:length].odd?)
	average_bytes_per_second = comm_info[:bytes_per_sample] * comm_info[:number_of_channels] * comm_info[:sample_rate]
	file_seconds = data_length.to_f / average_bytes_per_second.to_f
	return comm_info.merge({format: 'aif', seconds: file_seconds, average_bytes_per_second: average_bytes_per_second, data_length: data_length, valid: true, filename: filename, chunk_info: chunk_info })
	else
	return { format: 'aif', error: "AIFF chunks total length is different than file size: #{total_chunk_length} vs. #{actual_file_size}", valid: false, filename: filename }
	end
	else
	return { error: "Malformed AIFF", valid: false, filename: filename }
	end
	end

	def AudioTime.parse_wav_fmt_chunk(raw_fmt_chunk)
	# Expand it out to 40 characters
	raw_fmt_chunk += (0.chr*(40-raw_fmt_chunk.size))
	format_tag, number_of_channels, sample_rate, average_bytes_per_second, block_align, bits_per_sample, extended_size, valid_bits_per_sample, channel_mask, subformat = raw_fmt_chunk.unpack('v v V V v v v v V a16')
	# In this case, the format tag is the first two bytes of the subformat
	if extended_size == 22 && format_tag == 65534
	format_tag = subformat.unpack('v').first
	end
	bytes_per_sample = bits_per_sample / 8
	return { format_tag: format_tag, number_of_channels: number_of_channels, sample_rate: sample_rate, average_bytes_per_second: average_bytes_per_second, block_align: block_align, bits_per_sample: bits_per_sample, bytes_per_sample: bytes_per_sample, has_extension: (extended_size > 0), valid_bits_per_sample: valid_bits_per_sample, channel_mask: channel_mask, subformat: subformat }
	end

	def AudioTime.wav_info(filename)
	raw_fmt_chunk = fmt_info = data_length = file_seconds = actual_file_size = nil

	chunk_info = []

	File.open(filename,'rb') do \|file\|

	begin

	raw_header = file.read(12)

	if !raw_header \|\| raw_header.size < 12
	return { error: "Not a WAV file", valid: false, filename: filename }
	end

	riff, filelen, wave = raw_header.unpack('A4VA4')
	if riff == 'RIFF' && wave == 'WAVE'

	# This odd code two lines below actually allows the difference
	# in actual file size and expected file size to differ by 1.
	# Some files add a padding byte if the last chunk is odd length,
	# but don't add this to to the file size in the header, causing
	# this minor inconsistency.
	actual_file_size = File.size(file)
	if (filelen + 8 - actual_file_size).abs > 1
	return { error: "File size error for WAV, header says #{filelen+8}, actual is #{actual_file_size}", valid: false, filename: filename }
	end

	# Read chunks, get the fmt and data chunks
	until file.eof?

	raw_chunk_header = file.read(8)
	# This will force it to be "eof" above
	next unless raw_chunk_header.size == 8

	chunk_type, chunk_length = raw_chunk_header.unpack('A4V')

	chunk_info.push({chunk_type: chunk_type, position: file.pos, length: chunk_length})

	if chunk_type == 'fmt'
	raw_fmt_chunk = file.read(chunk_length)
	fmt_info = parse_wav_fmt_chunk(raw_fmt_chunk)

	else

	data_length = chunk_length if chunk_type == 'data'

	# If the chunk length is odd, there should be a padding byte.
	file.seek(chunk_length + (chunk_length.odd? ? 1 : 0),IO::SEEK_CUR)

	end

	end

	else
	return { error: "Not a WAV", valid: false, filename: filename }
	end

	end
	end

	if !data_length.nil? && !fmt_info.nil?
	# Make sure chunk lengths add up. Each chunk has an 8-byte header,
	# plus the file has a 12 byte header. Odd sized chunks should be
	# followed by a padding byte, so that is added in at the end.
	# Note: the last chunk won't have a padding byte.
	total_chunk_length = 12 + chunk_info.size * 8 + chunk_info.map { \|ci\| ci[:length] }.inject(0) { \|s,n\| s+n } + chunk_info.select { \|ci\| ci[:length].odd? }.size
	if total_chunk_length == actual_file_size \|\| (total_chunk_length - 1 == actual_file_size && chunk_info.last[:length].odd?)
	number_of_sample_frames = data_length / (fmt_info[:bytes_per_sample] * fmt_info[:number_of_channels])
	file_seconds = number_of_sample_frames.to_f / fmt_info[:sample_rate].to_f
	return fmt_info.merge({format: 'wav', number_of_sample_frames: number_of_sample_frames, seconds: file_seconds, data_length: data_length, valid: true, filename: filename, chunk_info: chunk_info })
	else
	return { format: 'wav', error: "WAV chunks total length is different than file size: #{total_chunk_length} vs. #{actual_file_size}", valid: false, filename: filename }
	end
	else
	return { error: "Malformed WAV", valid: false, filename: filename }
	end
	end

	def AudioTime.parse_id3v2_header(header)
	id, major_version, minor_version, flags, sizeb = header.unpack('A3 C C b8 b32')
	size = [sizeb[24..30]+sizeb[16..22]+sizeb[8..14]+sizeb[0..6]+'0000'].pack('b32').unpack('l').first
	return [id, major_version, minor_version, flags, size]
	end

	MP3_LOOKUPS = {
	mpeg_audio_version_descriptions: {
	'00' => 'MPEG Version 2.5',
	'01' => 'Reserved',
	'10' => 'MPEG Version 2 (ISO/IEC 13818-3)',
	'11' => 'MPEG Version 1 (ISO/IEC 11172-3)'
	},
	mpeg_audio_version_ids: {
	'00' => 2.5,
	'01' => nil,
	'10' => 2,
	'11' => 1
	},
	layer_descriptions: {
	'00' => 'Reserved',
	'01' => 'Layer III',
	'10' => 'Layer II',
	'11' => 'Layer I'
	},
	layers: {
	'00' => nil,
	'01' => 3,
	'10' => 2,
	'11' => 1
	},
	# Yeah, I know it's just binary, this is as easy as anything
	bitrate_indices: {
	'0000' => 0, '0001' => 1, '0010' => 2, '0011' => 3,
	'0100' => 4, '0101' => 5, '0110' => 6, '0111' => 7,
	'1000' => 8, '1001' => 9, '1010' => 10, '1011' => 11,
	'1100' => 12, '1101' => 13, '1110' => 14, '1111' => 15
	},
	# Lookup for bitrate is mpeg_audio_version_id, layer, and bitrate_index
	bitrates: {
	1 => {
	1 => [ 'free', 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448, nil ],
	2 => [ 'free', 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384, nil ],
	3 => [ 'free', 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, nil ]
	},
	2 => {
	1 => [ 'free', 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256, nil ],
	2 => [ 'free', 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160, nil ],
	3 => [ 'free', 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160, nil ],
	},
	2.5 => {
	1 => [ 'free', 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256, nil ],
	2 => [ 'free', 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160, nil ],
	3 => [ 'free', 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160, nil ],
	}
	},
	# Lookup for sampling_rate_frequencies is sampling_rate_frequency_index
	# and mpeg_audio_version_id
	sampling_rate_frequencies: {
	'00' => { 1 => 44100, 2 => 22050, 2.5 => 11025 },
	'01' => { 1 => 48000, 2 => 24000, 2.5 => 12000 },
	'10' => { 1 => 32000, 2 => 16000, 2.5 => 8000 },
	'11' => { 1 => nil, 2 => nil, 2.5 => nil }
	},
	channel_modes: {
	'00' => 'stereo',
	'01' => 'joint stereo (stereo)',
	'10' => 'dual channel (stereo)',
	'11' => 'single channel (mono)'
	},
	channel_counts: {
	'00' => 2,
	'01' => 2,
	'10' => 2,
	'11' => 1
	},
	emphases: {
	'00' => 'none',
	'01' => '50/15 ms',
	'10' => 'reserved',
	'11' => 'CCIT J.17'
	},
	# Lookup for samples_per_frame is mpeg_audio_version_id and layer
	samples_per_frame: {
	1 => { 1 => 384, 2 => 1152, 3 => 1152 },
	2 => { 1 => 384, 2 => 1152, 3 => 576 },
	2.5 => { 1 => 384, 2 => 1152, 3 => 576 }
	},
	# Lookup for mode extensions - layer I and II only
	mode_extensions: {
	'00' => { low_band: 4, high_band: 31 },
	'01' => { low_band: 8, high_band: 31 },
	'10' => { low_band: 12, high_band: 31 },
	'11' => { low_band: 16, high_band: 31 }
	}
	}

	def AudioTime.parse_mp3_frame_header(frame_header)
	# MP3 frame is 4 bytes. Info for parsing is here:
	# http://mpgedit.org/mpgedit/mpeg_format/mpeghdr.htm
	# http://www.codeproject.com/Articles/8295/MPEG-Audio-Frame-Header
	#
	# Hopefully those will stay good for a long time :)
	# Fields are of these lengths in bits: 11, 2, 2, 1, 4, 2, 1, 1, 2, 2, 1, 1, 2
	frame_sync, raw_mpeg_audio_version_id, raw_layer_description, protection_bit, raw_bitrate_index, sampling_rate_frequency_index, padding_bit, private_bit, raw_channel_mode, raw_mode_extension, copyright_flag, original_flag, raw_emphasis = frame_header.unpack('B32').first.unpack('a11 a2 a2 a1 a4 a2 a1 a1 a2 a2 a1 a1 a2')

	#STDERR.puts frame_sync, raw_mpeg_audio_version_id, raw_layer_description, protection_bit, raw_bitrate_index, sampling_rate_frequency_index, padding_bit, private_bit, raw_channel_mode, raw_mode_extension, copyright_flag, original_flag, raw_emphasis

	return nil unless frame_sync == '11111111111' \|\| frame_sync == '11111111110'

	has_crc = (protection_bit == '0')
	mpeg_audio_version_id = MP3_LOOKUPS[:mpeg_audio_version_ids][raw_mpeg_audio_version_id]
	mpeg_audio_version_description = MP3_LOOKUPS[:mpeg_audio_version_descriptions][raw_mpeg_audio_version_id]
	layer = MP3_LOOKUPS[:layers][raw_layer_description]
	layer_description = MP3_LOOKUPS[:layer_descriptions][raw_layer_description]
	bitrate_index = MP3_LOOKUPS[:bitrate_indices][raw_bitrate_index]
	bitrate = MP3_LOOKUPS[:bitrates][mpeg_audio_version_id][layer][bitrate_index]
	sampling_rate_frequency = MP3_LOOKUPS[:sampling_rate_frequencies][sampling_rate_frequency_index][mpeg_audio_version_id]
	samples_per_frame = MP3_LOOKUPS[:samples_per_frame][mpeg_audio_version_id][layer]
	channel_mode = MP3_LOOKUPS[:channel_modes][raw_channel_mode]
	channel_count = MP3_LOOKUPS[:channel_counts][raw_channel_mode]
	emphasis = MP3_LOOKUPS[:emphases][raw_emphasis]

	if layer == 3
	mode_extension = nil
	ms_stereo = raw_mode_extension[0]
	intensity_stereo = raw_mode_extension[1]
	else
	mode_extension = MP3_LOOKUPS[:mode_extensions][raw_mode_extension]
	intensity_stereo = ms_stereo = nil
	end

	frame_slot_count = (samples_per_frame / 8) * (bitrate * 1000) / sampling_rate_frequency + padding_bit.to_i
	frame_length =
	if layer == 1
	frame_slot_count * 4
	else
	frame_slot_count
	end

	return { frame_sync: frame_sync, raw_mpeg_audio_version_id: raw_mpeg_audio_version_id, raw_layer_description: raw_layer_description, protection_bit: protection_bit, raw_bitrate_index: raw_bitrate_index, sampling_rate_frequency_index: sampling_rate_frequency_index, padding_bit: padding_bit, private_bit: private_bit, raw_channel_mode: raw_channel_mode, raw_mode_extension: raw_mode_extension, copyright_flag: copyright_flag, original_flag: original_flag, raw_emphasis: raw_emphasis, mpeg_audio_version_id: mpeg_audio_version_id, mpeg_audio_version_description: mpeg_audio_version_description, layer_description: layer_description, has_crc: has_crc, sampling_rate_frequency: sampling_rate_frequency, channel_mode: channel_mode, channel_count: channel_count, emphasis: emphasis, frame_length: frame_length, bitrate: bitrate, layer: layer, mode_extension: mode_extension, ms_stereo: ms_stereo, intensity_stereo: intensity_stereo }
	end

	def AudioTime.mp3_info(filename)
	# Need to find all possible id3 tag sets - id3v2 and beginning and/or
	# end, and id3v1 at end. The idea is to separate out the actual MP3
	# data so we can read and handle the frames.

	File.open(filename, 'rb') do \|file\|

	begin

	mp3_offset = 0
	file_size = mp3_length = file.size
	has_v1 = has_pre_v2 = has_post_v2 = false

	chunk = file.read(10)
	if !chunk \|\| chunk.size < 10
	return { error: "Not an MP3 file", valid: false, filename: filename }
	end

	if chunk[0..2] == 'ID3'
	id, major_version, minor_version, flags, size = parse_id3v2_header(chunk)
	mp3_offset += size + 10
	mp3_length -= size + 10
	has_pre_v2 = true
	#STDERR.puts "Has a v2 tag at start, #{size+10} bytes" if DEBUG
	end

	file.seek(-128, IO::SEEK_END)
	chunk = file.read(3)

	if !chunk \|\| chunk.size < 3
	return { error: "Not an MP3 file", valid: false, filename: filename }
	end

	# This is a v1 tag at the end
	if chunk[0..2] == 'TAG'
	mp3_length -= 128
	has_v1 = true
	#STDERR.puts "Has a v1 tag at end, 128 bytes" if DEBUG
	end

	# Check for v2 tag at the end if there wasn't a v2 tag at the
	# beginning
	unless has_pre_v2
	if has_v1
	file.seek(-138, IO::SEEK_END)
	else
	file.seek(-10, IO::SEEK_END)
	end

	chunk = file.read(10)
	if !chunk \|\| chunk.size < 10
	return { error: "Not an MP3 file", valid: false, filename: filename }
	end

	if chunk[0..2] == '3DI'
	id, major_version, minor_version, flags, size = parse_id3v2_header(chunk)
	mp3_length -= size + 10
	has_post_v2 = true
	#STDERR.puts "Has a v2 tag at end, #{size+10} bytes" if DEBUG
	end
	end

	# Sanity checks
	if mp3_offset < 0 \|\| mp3_offset > file_size \|\| mp3_length < 0
	return { error: "MP3 information is confused: offset #{mp3_offset} length #{mp3_length}", valid: false, filename: filename }
	end

	mp3_end_offset = mp3_offset + mp3_length
	total_seconds = 0.0

	#STDERR.puts "mp3_offset is #{mp3_offset}, mp3_length is #{mp3_length}" if DEBUG

	# At this point, we can parse the MP3 file starting at mp3_offset and
	# continuing for mp3_length bytes. Of course, we need to make sure
	# frame is good.

	low_bitrate = high_bitrate = nil

	first_frame_info = nil
	file_size_seconds = nil

	mpeg_audio_version_id = nil
	mpeg_audio_version_description = nil
	layer = nil
	layer_description = nil
	sampling_rate_frequency = nil
	ms_stereo = nil
	intensity_stereo = nil
	channel_mode = nil

	frame_infos = []

	file.seek(mp3_offset, IO::SEEK_SET)

	until file.pos >= mp3_end_offset \|\| file.eof?

	frame_header = file.read(4)

	if !frame_header \|\| frame_header.size < 4
	return { error: "Invalid MP3 file", valid: false, filename: filename }
	end

	frame_info = parse_mp3_frame_header(frame_header)

	return { error: 'Invalid frame sync in MP3 file', valid: false, filename: filename } unless frame_info

	frame_info[:offset] = file.pos

	# TODO: Check CRC if present

	low_bitrate = frame_info[:bitrate] if !low_bitrate \|\| low_bitrate > frame_info[:bitrate]
	high_bitrate = frame_info[:bitrate] if !high_bitrate \|\| high_bitrate < frame_info[:bitrate]

	mpeg_audio_version_id = frame_info[:mpeg_audio_version_id]
	mpeg_audio_version_description = frame_info[:mpeg_audio_version_description]
	layer = frame_info[:layer]
	layer_description = frame_info[:layer_description]
	sampling_rate_frequency = frame_info[:sampling_rate_frequency]
	ms_stereo = frame_info[:ms_stereo]
	intensity_stereo = frame_info[:intensity_stereo]
	channel_mode = frame_info[:channel_mode]

	total_seconds += frame_info[:frame_length].to_f * 8.0 / frame_info[:bitrate].to_f

	first_frame_info = frame_info unless first_frame_info

	frame_infos.push(frame_info)

	#pp frame_info.inspect

	file.seek(frame_info[:frame_length] - 4, IO::SEEK_CUR)
	end

	cbr = (low_bitrate == high_bitrate)

	file_size_seconds = mp3_length.to_f * 8.0 / (low_bitrate.to_f * 1000.0) if cbr

	mp3_info = first_frame_info

	mp3_info[:format] = 'mp3'

	mp3_info[:total_seconds] = total_seconds / 1000.0
	mp3_info[:file_size_seconds] = file_size_seconds
	mp3_info[:seconds] = mp3_info[:file_size_seconds] \|\| mp3_info[:total_seconds]
	mp3_info[:cbr] = cbr
	mp3_info[:low_bitrate] = low_bitrate
	mp3_info[:high_bitrate] = high_bitrate

	mp3_info[:mpeg_audio_version_id] = mpeg_audio_version_id
	mp3_info[:mpeg_audio_version_description] = mpeg_audio_version_description
	mp3_info[:layer] = layer
	mp3_info[:layer_description] = layer_description
	mp3_info[:sampling_rate_frequency] = sampling_rate_frequency
	mp3_info[:ms_stereo] = ms_stereo
	mp3_info[:intensity_stereo] = intensity_stereo
	mp3_info[:channel_mode] = channel_mode

	mp3_info[:mp3_offset] = mp3_offset
	mp3_info[:mp3_length] = mp3_length
	mp3_info[:has_v1] = has_v1
	mp3_info[:has_pre_v2] = has_pre_v2
	mp3_info[:has_post_v2] = has_post_v2

	mp3_info[:valid] = true
	mp3_info[:filename] = filename

	return mp3_info

	rescue Errno::EINVAL => e
	# Invalid seek
	return { error: "Cannot seek in MP3 file", valid: false, filename: filename }

	end
	end
	end

	def AudioTime.aiff_time(filename)
	info = aiff_info(filename)
	return info[:seconds]
	end

	def AudioTime.wav_time(filename)
	info = wav_info(filename)
	return info[:seconds]
	end

	def AudioTime.mp3_time(filename)
	info = mp3_info(filename)
	return info[:seconds]
	end

	def AudioTime.aiff_lint(filename)
	info = aiff_info(filename)
	return info[:valid]
	end

	def AudioTime.wav_lint(filename)
	info = wav_info(filename)
	return info[:valid]
	end

	def AudioTime.mp3_lint(filename)
	info = mp3_info(filename)
	return info[:valid]
	end

	def AudioTime.valid_audio_file?(filename, type=nil)
	if type
	type = type.gsub(/\A\./,'').downcase
	if type == 'aif' \|\| type == 'aiff'
	aiff_lint(filename)
	elsif type == 'wav'
	wav_lint(filename)
	elsif type == 'mp3'
	mp3_lint(filename)
	else
	nil
	end
	else
	raw_header = nil
	File.open(filename,'rb') do \|file\|
	raw_header = file.read(12)
	end
	if !raw_header \|\| raw_header.size < 12
	false
	elsif raw_header =~ /\AFORM....AIF[FC]/
	aiff_lint(filename)
	elsif raw_header =~ /\ARIFF....WAVE/
	wav_lint(filename)
	else
	# MP3s are difficult to easily determine
	mp3_lint(filename)
	end
	end
	end
	end