def check_format_info(info):
"""
Check that format info is correct.
Args:
info: Format info from fmt chunk.
Raises:
wavy.WaveFileIsCorrupted: If format info is incorrect.
"""
# check that block align matches bits per sample
# BlockAlign == NumChannels * BitsPerSample/8
block_align = int(info.wBitsPerSample * info.nChannels / 8)
if block_align != info.nBlockAlign:
raise wavy.WaveFileIsCorrupted(
f"Block align is incorrect for {info.wBitsPerSample} bits. "
f"Expected: {block_align}, "
f"Actual: {info.nBlockAlign}.")
# check that avg bytes per sample matches expected value
n_avg_bytes_per_sec = info.nSamplesPerSec * info.nBlockAlign
if n_avg_bytes_per_sec != info.nAvgBytesPerSec:
raise wavy.WaveFileIsCorrupted(f"Avg. bytes per sec. is incorrect. "
f"Expected: {n_avg_bytes_per_sec}, "
f"Actual: {info.nAvgBytesPerSec}.")
def get_data_from_chunk(chunk, format, handler):
"""
Read data from data chunk.
Args:
chunk: Data chunk.
format: File format information.
Returns:
numpy.array: Data read from chunk.
"""
# get size of data
size = chunk.getsize()
# this gives us the number of frames
if size % format.nBlockAlign != 0:
# something is wrong here, size should be a multiple
raise wavy.WaveFileIsCorrupted("Data size does not match frame size of"
f" {format.wBitsPerSample} bits")
# number of bytes for data type to be parsed
n_bytes = format.wBitsPerSample // 8
# read data from raw
data = handler.read_data(chunk, size, n_bytes,
format.wFormatTag == WAVE_FORMAT_IEEE_FLOAT)
# check if there is more than one channel, if so reshape
return data if format.nChannels == 1 \
else data.reshape(-1, format.nChannels)
def get_chunk(stream, bigendian):
"""
Get chunk for wave file (always little endian)
"""
try:
return chunk.Chunk(stream, bigendian=bigendian)
except EOFError:
raise wavy.WaveFileIsCorrupted('Reached end of file prematurely.')
def get_fmt_chunk(stream, handler):
"""
Reads the format chunk from the stream, checks that it matches the
specifications for Wave files and returns format information.
Args:
stream: The stream to read.
Returns:
FormatInfo: Format info from fmt chunk.
Raises:
wavy.WaveFileIsCorrupted: If file is corrupted.
wavy.WaveFileNotSupported: If file type is not PCM.
"""
# iterate through chunks until we find format
while True:
chunk = get_chunk(stream, False if handler.little_endian else True)
name = chunk.getname()
# found format
if name == FMT:
break
# we got to the data chunks, something is wrong
if name == DATA:
raise wavy.WaveFileIsCorrupted(
'Found data chunk before fmt chunk.')
chunk.skip()
# chunk size is not supported
if chunk.getsize() not in FMT_CHUNK_SIZES:
raise wavy.WaveFileIsCorrupted(
'Format chunk is of unexpected size: {}.'.format(chunk.getsize()))
# extract common chunk info into tuple
info = FormatInfo(*handler.read('HHLLHH', chunk.read(16)))
format_tag = info.wFormatTag
# if format extensible is used, the format tag
# will be specified in the sub format
if format_tag == WAVE_FORMAT_EXTENSIBLE and \
chunk.getsize() == 40:
# get format tag
format_tag = get_sub_format(chunk, handler)
# replace tag in FormatInfo
info = FormatInfo(format_tag, *list(info)[1:])
# we only support PCM and FLOAT
if format_tag in SUPPORTED_WAVE_FORMATS:
# check that the sample width is supported for type
if info.wBitsPerSample not in \
SUPPORTED_SAMPLE_WIDTH_FOR_FORMAT[format_tag]:
raise wavy.WaveFileNotSupported(
"Sample width '{}' is not supported for "
"given type.".format(info.wBitsPerSample))
chunk.skip()
return info
raise wavy.WaveFileNotSupported(
'The wave format is not of supported type.')