def compute(file_obj):
"""Compute a metadata-invariant fingerprint from an MP3 file.
Args:
file_obj: A file-like object
Returns:
A string containing the fingerprint, which takes the form of
a 40-character sequence of hex digits. If no valid MPEG frames
are found, None is returned.
"""
sha1_calc = hashlib.sha1()
saw_a_valid_frame = False
for hdr, data_buffer in mp3_frame.split(file_obj):
if hdr is not None:
sha1_calc.update(data_buffer)
saw_a_valid_frame = True
if saw_a_valid_frame:
return sha1_calc.hexdigest()
else:
return None
def analyze(file_obj, au_file, compute_fingerprint=True, get_payload=True):
"""Populate an AudioFile object with information extracted from a file.
Args:
file_obj: A file-like object.
au_file: An AudioFile object to store the results of the analysis in.
compute_fingerprint: If False, do not compute a fingerprint.
Returns:
The same AudioFile object that was passed in as au_file, which
should now have several fields set.
Raises:
InvalidFileError: if the file appears to be corrupted.
"""
au_file.frame_count = 0
au_file.frame_size = 0
au_file.duration_ms = 0
sha1_calc = hashlib.sha1() # unused if compute_fingerprint is False.
payload = cStringIO.StringIO() # unused if get_payload is False.
bit_rate_kbps_sum = 0
expected_hdr = None
first_bit_rate_kbps = None
is_vbr = False
for hdr, data_buffer in mp3_frame.split(file_obj):
if hdr is None:
continue
au_file.frame_count += 1
au_file.frame_size += len(data_buffer)
au_file.duration_ms += hdr.duration_ms
if compute_fingerprint:
sha1_calc.update(data_buffer)
if get_payload:
payload.write(data_buffer)
# If we've seen a valid header previously, make sure that all of the
# fields that should match do actually match.
if expected_hdr:
if not hdr.match(expected_hdr):
raise InvalidFileError(
"Bad header: found %s, expected %s (path=%s)" %
(hdr, expected_hdr, au_file.path))
# Keep track of if this is a variable bit-rate file.
if hdr.bit_rate_kbps != first_bit_rate_kbps:
is_vbr = True
# Add this frame's bit rate to our sum; we will use this to compute
# the average bit rate.
bit_rate_kbps_sum += hdr.bit_rate_kbps
# If this is the first header we've seen, make a copy and then blank
# out the fields that can vary. All future headers are expected to
# match this template.
if expected_hdr is None:
expected_hdr = hdr
first_bit_rate_kbps = expected_hdr.bit_rate_kbps
expected_hdr.bit_rate_kbps = None # Might be a VBR file.
expected_hdr.padding = None # Not all frames are padded.
expected_hdr.frame_size = None
# You'd think that this would be constant, but MP3s
# encountered in the wild prove otherwise.
expected_hdr.protected = None
if au_file.frame_count < _MINIMUM_FRAMES:
raise InvalidFileError("Found only %d MPEG frames" %
au_file.frame_count)
# Add the bit rate back into the template header, then return it.
# If this is a VBR file, use the average bit rate instead.
if is_vbr:
expected_hdr.bit_rate_kbps = (float(bit_rate_kbps_sum) /
au_file.frame_count)
else:
expected_hdr.bit_rate_kbps = first_bit_rate_kbps
# Finishing populating and then return the AudioFile object.
au_file.mp3_header = expected_hdr
# Round the duration down to an integral number of microseconds.
au_file.duration_ms = int(au_file.duration_ms)
if compute_fingerprint:
au_file.fingerprint = sha1_calc.hexdigest()
if get_payload:
au_file.payload = payload.getvalue()
return au_file
def test_split(self):
raw_hdr, hdr = mp3_header_test.VALID_MP3_HEADERS.items()[0]
frame_data = raw_hdr.ljust(hdr.frame_size, "a")
# Set up a fragment of a header
partial_header = raw_hdr[:3]
short_frame = frame_data[:25]
assert len(short_frame) < len(frame_data)
id3_data = id3_header.create_test_header(77).ljust(77, "b")
# An ID3 tag with a valid frame tag stashed inside.
evil_id3_data = id3_header.create_test_header(50) + raw_hdr
evil_id3_data = evil_id3_data.ljust(50, "c")
for seq in (
[ frame_data ],
[ frame_data, frame_data ],
[ 'junk', frame_data ],
[ 'junk', frame_data, frame_data ],
[ 'junk', frame_data, frame_data, 'junk' ],
[ 'junk', frame_data, frame_data, 'junk', frame_data ],
# Check handling of truncated headers and frames.
[ partial_header ],
[ 'junk', partial_header ],
[ 'junk', short_frame ],
[ frame_data, partial_header ],
[ frame_data, short_frame ],
[ frame_data, 'junk', short_frame ],
[ frame_data, 'junk', partial_header],
# ID3 headers mixed in
[ id3_data, frame_data ],
[ frame_data, id3_data ],
[ id3_data, frame_data ],
[ id3_data, frame_data, id3_data ],
[ evil_id3_data, frame_data, "junk" ],
[ "junk", frame_data, evil_id3_data, frame_data ],
[ "junk", frame_data, evil_id3_data, frame_data, "junk" ],
[ "junk" + evil_id3_data, id3_data, frame_data, evil_id3_data ],
# Some longer sequences
500 * [ frame_data ],
500 * [ "junk", frame_data, id3_data, frame_data ]
):
data = ''.join(seq)
stream = cStringIO.StringIO(data)
split_stream = list(mp3_frame.split(stream))
split_stream_from_blocks = list(mp3_frame.split_blocks(iter(seq)))
split_stream_from_one_block = mp3_frame.split_one_block(data)
# Make sure that the sequences of header/frame data pairs
# returned by mp3_frame.split(), mp3_frame.split_blocks()
# and mp3_frame.split_one_block() matche what we would
# expect.
self.assertEqual(len(seq), len(split_stream))
for expected_data, (actual_hdr, data) in zip(seq, split_stream):
self.assertEqual(expected_data, data)
if expected_data == frame_data:
self.assertTrue(actual_hdr is not None)
self.assertTrue(actual_hdr.match(hdr))
self.assertEqual(hdr.frame_size, len(frame_data))
else:
self.assertTrue(actual_hdr is None)
self.assertEqual(len(seq), len(split_stream_from_blocks))
for (hdr1, data1), (hdr2, data2) in zip(split_stream,
split_stream_from_blocks):
self.assertEqual(str(hdr1), str(hdr2))
self.assertEqual(data1, data2)
self.assertEqual(len(seq), len(split_stream_from_one_block))
for (hdr1, data1), (hdr2, data2) in zip(
split_stream, split_stream_from_one_block):
self.assertEqual(str(hdr1), str(hdr2))
self.assertEqual(data1, data2)
def analyze(file_obj, au_file, compute_fingerprint=True, get_payload=True):
"""Populate an AudioFile object with information extracted from a file.
Args:
file_obj: A file-like object.
au_file: An AudioFile object to store the results of the analysis in.
compute_fingerprint: If False, do not compute a fingerprint.
Returns:
The same AudioFile object that was passed in as au_file, which
should now have several fields set.
Raises:
InvalidFileError: if the file appears to be corrupted.
"""
au_file.frame_count = 0
au_file.frame_size = 0
au_file.duration_ms = 0
sha1_calc = hashlib.sha1() # unused if compute_fingerprint is False.
payload = cStringIO.StringIO() # unused if get_payload is False.
bit_rate_kbps_sum = 0
expected_hdr = None
first_bit_rate_kbps = None
is_vbr = False
for hdr, data_buffer in mp3_frame.split(file_obj):
if hdr is None:
continue
au_file.frame_count += 1
au_file.frame_size += len(data_buffer)
au_file.duration_ms += hdr.duration_ms
if compute_fingerprint:
sha1_calc.update(data_buffer)
if get_payload:
payload.write(data_buffer)
# If we've seen a valid header previously, make sure that all of the
# fields that should match do actually match.
if expected_hdr:
if not hdr.match(expected_hdr):
raise InvalidFileError(
"Bad header: found %s, expected %s (path=%s)" % (
hdr, expected_hdr, au_file.path))
# Keep track of if this is a variable bit-rate file.
if hdr.bit_rate_kbps != first_bit_rate_kbps:
is_vbr = True
# Add this frame's bit rate to our sum; we will use this to compute
# the average bit rate.
bit_rate_kbps_sum += hdr.bit_rate_kbps
# If this is the first header we've seen, make a copy and then blank
# out the fields that can vary. All future headers are expected to
# match this template.
if expected_hdr is None:
expected_hdr = hdr
first_bit_rate_kbps = expected_hdr.bit_rate_kbps
expected_hdr.bit_rate_kbps = None # Might be a VBR file.
expected_hdr.padding = None # Not all frames are padded.
expected_hdr.frame_size = None
# You'd think that this would be constant, but MP3s
# encountered in the wild prove otherwise.
expected_hdr.protected = None
if au_file.frame_count < _MINIMUM_FRAMES:
raise InvalidFileError("Found only %d MPEG frames"
% au_file.frame_count)
# Add the bit rate back into the template header, then return it.
# If this is a VBR file, use the average bit rate instead.
if is_vbr:
expected_hdr.bit_rate_kbps = (
float(bit_rate_kbps_sum) / au_file.frame_count)
else:
expected_hdr.bit_rate_kbps = first_bit_rate_kbps
# Finishing populating and then return the AudioFile object.
au_file.mp3_header = expected_hdr
# Round the duration down to an integral number of microseconds.
au_file.duration_ms = int(au_file.duration_ms)
if compute_fingerprint:
au_file.fingerprint = sha1_calc.hexdigest()
if get_payload:
au_file.payload = payload.getvalue()
return au_file
