def process(self, event):
if not isinstance(event, dict):
raise ValueError("Expected event of type dict, got one of type %s" % (type(event),))
# We want the bindings in the event to override those in the static
# dict, so we make of copy of the static dict and update the copy with
# the event
bindings = dict(self._static_variable_dict)
bindings.update(event)
command_args = self._make_command_sequence(bindings)
# This will throw with a useful error message if the command can't be
# run or if it runs and fails.
if self.notify_stream:
note = '[' + timestamp.get_timestamp() + '] Starting: ' + ' '.join(command_args) + '\n'
self.notify_stream.write(note)
(self._last_stdout_output,
self._last_stderr_output,
self._last_command_line) = process.subprocess(command_args)
# If we got here, the subprocess finished successfully.
if self.notify_stream:
note = '[' + timestamp.get_timestamp() + '] Finished: ' + ' '.join(command_args) + '\n'
self.notify_stream.write(note)
self.send(event)
def process(self, event):
if not isinstance(event, dict):
raise ValueError(
"Expected event of type dict, got one of type %s" %
(type(event), ))
# We want the bindings in the event to override those in the static
# dict, so we make of copy of the static dict and update the copy with
# the event
bindings = dict(self._static_variable_dict)
bindings.update(event)
command_args = self._make_command_sequence(bindings)
# This will throw with a useful error message if the command can't be
# run or if it runs and fails.
if self.notify_stream:
note = '[' + timestamp.get_timestamp() + '] Starting: ' + ' '.join(
command_args) + '\n'
self.notify_stream.write(note)
(self._last_stdout_output, self._last_stderr_output,
self._last_command_line) = process.subprocess(command_args)
# If we got here, the subprocess finished successfully.
if self.notify_stream:
note = '[' + timestamp.get_timestamp() + '] Finished: ' + ' '.join(
command_args) + '\n'
self.notify_stream.write(note)
self.send(event)
def _get_sphere_fd(fileno, file_num_bytes, filename, wave_format,
filename_abs):
"""
Low-level reading of NIST Sphere audio data.
>>> module_dir, module_name = os.path.split(__file__)
>>> zero_sph = os.path.join(module_dir, 'zero.sph')
>>> shorten_sph = os.path.join(module_dir, 'shorten.sph')
>>> with open(zero_sph, 'rb') as audio_file:
... file_info, audio_info, wave = _get_sphere_fd(audio_file.fileno(), os.path.getsize(zero_sph), zero_sph, 'int16', os.path.abspath(zero_sph))
>>> for key in sorted(file_info): print "%-24s %r" % (key, file_info[key])
file_item_bytes 2
file_item_coding 'int16'
file_sndfile_extension 'sph'
file_sndfile_format ''
file_sndfile_type 'pcm SPH (NIST Sphere)'
>>> for key in sorted(audio_info): print "%-24s %r" % (key, audio_info[key])
audio_num_channels 2
audio_num_samples 128064
audio_sample_rate 44100
>>> print str(wave)
[[ 0 -1 -3 ..., 203 211 199]
[ 0 0 -3 ..., 225 225 221]]
>>> with open(shorten_sph, 'rb') as audio_file:
... file_info2, audio_info2, wave2 = _get_sphere_fd(audio_file.fileno(), os.path.getsize(shorten_sph), shorten_sph, 'int16', os.path.abspath(shorten_sph))
>>> for key in sorted(file_info2): print "%-24s %r" % (key, file_info2[key])
file_item_bytes 2
file_item_coding 'int16'
file_sndfile_extension 'sph'
file_sndfile_format ''
file_sndfile_type 'pcm,embedded-shorten-v2.00 SPH (NIST Sphere)'
>>> for key in sorted(audio_info2): print "%-24s %r" % (key, audio_info2[key])
audio_num_channels 1
audio_num_samples 37120
audio_sample_rate 20000
>>> print str(wave2)
[[-1 1 1 ..., -4 -8 -5]]
"""
assert wave_format is None or wave_format in wave_c_formats
# As of 2009-04-20 see the following for NIST's underspecified
# format description:
# http://ftp.cwi.nl/audio/NIST-SPHERE
# http://www.ldc.upenn.edu/Catalog/docs/LDC93S5/WAV_SPEC.TXT
nist_1a = 'NIST_1A'
header1 = read_fd_strict(fileno, 128, filename_abs)
if not header1.startswith(nist_1a):
raise AudioTypeError(
"did not find %r in header of purported NIST Sphere file %r" %
(nist_1a, filename_abs))
nist, header_size, rest = header1.split(None, 2)
assert nist == nist_1a
header_size = int(header_size)
rest += read_fd_strict(fileno, header_size - len(header1), filename_abs)
# For now, we require the following fields:
# sample_count -i 128064
# sample_n_bytes -i 2
# channel_count -i 2
# sample_byte_format -s2 01
# sample_rate -i 44100
# sample_coding -s3 pcm
info = builtin.attrdict()
for line in cStringIO.StringIO(rest):
parts = line.split()
if not parts or parts[0][0] == ';': continue
if parts[0] == 'end_head': break
if len(parts) < 3:
raise AudioTypeError(
"expected at least three white-space-separated fields in NIST header line %r in file %r"
% (line.strip(), filename_abs))
field_name, field_type, field_value = line.split(None, 3)
#print field_name, field_type, field_value
if field_type in ('-i', '-r'):
field_value, _, _ = field_value.partition(';')
info[field_name] = (int
if field_type == '-i' else float)(field_value)
elif field_type.startswith('-s'):
# here we do a stricter interpretation of the spec
prefix_len = len(field_name + ' ' + field_type + ' ')
str_len = int(field_type[2:])
info[field_name] = line[prefix_len:prefix_len + str_len]
else:
raise (AudioTypeError,
"unhandled field_type %r for field_name %r" %
(field_type, field_name))
missing = set(
('sample_count', 'sample_n_bytes', 'channel_count',
'sample_byte_format', 'sample_rate', 'sample_coding')) - set(info)
if missing:
raise AudioTypeError(
"missing required header fields (%s) in NIST Sphere file %r" %
(', '.join(sorted(missing)), filename_abs))
# this is a blunt check against bogus data
if info.sample_n_bytes > 2:
raise AudioTypeError(
"unhandled sample_n_bytes, %d, in NIST Sphere file %r" %
(info.sample_n_bytes, filename_abs))
audio_info = builtin.attrdict((('audio_num_channels', info.channel_count),
('audio_num_samples', info.sample_count),
('audio_sample_rate', info.sample_rate)))
check_positive_ints(audio_info, filename_abs)
file_info = builtin.attrdict((
('file_item_bytes', info.sample_n_bytes),
('file_item_coding', ('int' + str(8 * info.sample_n_bytes))
if info.sample_coding.lower().startswith('pcm') else
info.sample_coding[:4].lower()),
('file_sndfile_extension', 'sph'),
('file_sndfile_format', ''),
('file_sndfile_type', info.sample_coding + ' SPH (NIST Sphere)'),
))
check_positive_ints(file_info, filename_abs)
if wave_format is not None:
args = 'sph2pipe', '-p', '-f', 'raw', filename_abs
stdout, stderr, cmd = process.subprocess(args)
if stderr:
raise onyx.ExternalError(
"unexpected stderr from command '%s': '%s'" %
(cmd, stderr.strip()))
num_audio_bytes = audio_info.audio_num_channels * audio_info.audio_num_samples * numpy.int16(
).itemsize
if len(stdout) != num_audio_bytes:
raise onyx.DataFormatError(
"expected %d bytes of audio data, got %d" %
(num_audio_bytes, len(stdout)))
wave = numpy.fromstring(stdout, dtype=numpy.int16)
assert wave.shape == (audio_info.audio_num_channels *
audio_info.audio_num_samples, )
# reshape it, etc; then construct a new ndarray
wave = wave.reshape((-1, audio_info.audio_num_channels)).transpose()
wave = numpy.array(wave, dtype=wave_numpy_formats[wave_format])
# do scaling the same way libsndfile does
if wave_format == 'int32':
wave <<= 16
elif wave_format in ('float32', 'float64'):
wave *= (1 / (1 << 15))
assert wave.shape == (audio_info.audio_num_channels,
audio_info.audio_num_samples)
else:
wave = None
return file_info, audio_info, wave
def _get_sphere_fd(fileno, file_num_bytes, filename, wave_format, filename_abs):
"""
Low-level reading of NIST Sphere audio data.
>>> module_dir, module_name = os.path.split(__file__)
>>> zero_sph = os.path.join(module_dir, 'zero.sph')
>>> shorten_sph = os.path.join(module_dir, 'shorten.sph')
>>> with open(zero_sph, 'rb') as audio_file:
... file_info, audio_info, wave = _get_sphere_fd(audio_file.fileno(), os.path.getsize(zero_sph), zero_sph, 'int16', os.path.abspath(zero_sph))
>>> for key in sorted(file_info): print "%-24s %r" % (key, file_info[key])
file_item_bytes 2
file_item_coding 'int16'
file_sndfile_extension 'sph'
file_sndfile_format ''
file_sndfile_type 'pcm SPH (NIST Sphere)'
>>> for key in sorted(audio_info): print "%-24s %r" % (key, audio_info[key])
audio_num_channels 2
audio_num_samples 128064
audio_sample_rate 44100
>>> print str(wave)
[[ 0 -1 -3 ..., 203 211 199]
[ 0 0 -3 ..., 225 225 221]]
>>> with open(shorten_sph, 'rb') as audio_file:
... file_info2, audio_info2, wave2 = _get_sphere_fd(audio_file.fileno(), os.path.getsize(shorten_sph), shorten_sph, 'int16', os.path.abspath(shorten_sph))
>>> for key in sorted(file_info2): print "%-24s %r" % (key, file_info2[key])
file_item_bytes 2
file_item_coding 'int16'
file_sndfile_extension 'sph'
file_sndfile_format ''
file_sndfile_type 'pcm,embedded-shorten-v2.00 SPH (NIST Sphere)'
>>> for key in sorted(audio_info2): print "%-24s %r" % (key, audio_info2[key])
audio_num_channels 1
audio_num_samples 37120
audio_sample_rate 20000
>>> print str(wave2)
[[-1 1 1 ..., -4 -8 -5]]
"""
assert wave_format is None or wave_format in wave_c_formats
# As of 2009-04-20 see the following for NIST's underspecified
# format description:
# http://ftp.cwi.nl/audio/NIST-SPHERE
# http://www.ldc.upenn.edu/Catalog/docs/LDC93S5/WAV_SPEC.TXT
nist_1a = 'NIST_1A'
header1 = read_fd_strict(fileno, 128, filename_abs)
if not header1.startswith(nist_1a):
raise AudioTypeError("did not find %r in header of purported NIST Sphere file %r" % (nist_1a, filename_abs))
nist, header_size, rest = header1.split(None, 2)
assert nist == nist_1a
header_size = int(header_size)
rest += read_fd_strict(fileno, header_size - len(header1), filename_abs)
# For now, we require the following fields:
# sample_count -i 128064
# sample_n_bytes -i 2
# channel_count -i 2
# sample_byte_format -s2 01
# sample_rate -i 44100
# sample_coding -s3 pcm
info = builtin.attrdict()
for line in cStringIO.StringIO(rest):
parts = line.split()
if not parts or parts[0][0] == ';': continue
if parts[0] == 'end_head': break
if len(parts) < 3:
raise AudioTypeError("expected at least three white-space-separated fields in NIST header line %r in file %r" % (line.strip(), filename_abs))
field_name, field_type, field_value = line.split(None, 3)
#print field_name, field_type, field_value
if field_type in ('-i', '-r'):
field_value, _, _ = field_value.partition(';')
info[field_name] = (int if field_type == '-i' else float)(field_value)
elif field_type.startswith('-s'):
# here we do a stricter interpretation of the spec
prefix_len = len(field_name + ' ' + field_type + ' ')
str_len = int(field_type[2:])
info[field_name] = line[prefix_len:prefix_len+str_len]
else:
raise (AudioTypeError, "unhandled field_type %r for field_name %r" % (field_type, field_name))
missing = set(('sample_count', 'sample_n_bytes', 'channel_count', 'sample_byte_format', 'sample_rate', 'sample_coding')) - set(info)
if missing:
raise AudioTypeError("missing required header fields (%s) in NIST Sphere file %r" % (', '.join(sorted(missing)), filename_abs))
# this is a blunt check against bogus data
if info.sample_n_bytes > 2:
raise AudioTypeError("unhandled sample_n_bytes, %d, in NIST Sphere file %r" % (info.sample_n_bytes, filename_abs))
audio_info = builtin.attrdict((('audio_num_channels', info.channel_count),
('audio_num_samples', info.sample_count),
('audio_sample_rate', info.sample_rate)))
check_positive_ints(audio_info, filename_abs)
file_info = builtin.attrdict((('file_item_bytes', info.sample_n_bytes),
('file_item_coding', ('int' + str(8*info.sample_n_bytes)) if info.sample_coding.lower().startswith('pcm') else info.sample_coding[:4].lower()),
('file_sndfile_extension', 'sph'),
('file_sndfile_format', ''),
('file_sndfile_type', info.sample_coding + ' SPH (NIST Sphere)'),
))
check_positive_ints(file_info, filename_abs)
if wave_format is not None:
args = 'sph2pipe', '-p', '-f', 'raw', filename_abs
stdout, stderr, cmd = process.subprocess(args)
if stderr:
raise onyx.ExternalError("unexpected stderr from command '%s': '%s'" % (cmd, stderr.strip()))
num_audio_bytes = audio_info.audio_num_channels * audio_info.audio_num_samples * numpy.int16().itemsize
if len(stdout) != num_audio_bytes:
raise onyx.DataFormatError("expected %d bytes of audio data, got %d" % (num_audio_bytes, len(stdout)))
wave = numpy.fromstring(stdout, dtype=numpy.int16)
assert wave.shape == (audio_info.audio_num_channels * audio_info.audio_num_samples,)
# reshape it, etc; then construct a new ndarray
wave = wave.reshape((-1, audio_info.audio_num_channels)).transpose()
wave = numpy.array(wave, dtype=wave_numpy_formats[wave_format])
# do scaling the same way libsndfile does
if wave_format == 'int32':
wave <<= 16
elif wave_format in ('float32', 'float64'):
wave *= (1 / (1 << 15))
assert wave.shape == (audio_info.audio_num_channels, audio_info.audio_num_samples)
else:
wave = None
return file_info, audio_info, wave
