def test_check_wavs_bad(wav_file, wav_file_8k, tmpdir, capsys):
def fun(utts):
c = pipeline._init_config(
pipeline.get_default_config('mfcc', with_cmvn=False))
u = pipeline._init_utterances(utts)
pipeline._Manager(c, u)
return u
# build a stereo file and make sure it is not supported by the
# pipeline
audio = Audio.load(wav_file)
stereo = Audio(np.asarray((audio.data, audio.data)).T,
sample_rate=audio.sample_rate)
assert stereo.nchannels == 2
wav_file_2 = str(tmpdir.join('stereo.wav'))
stereo.save(wav_file_2)
with pytest.raises(ValueError) as err:
fun([(wav_file_2, )])
assert 'all wav files are not mono' in str(err)
# ensure we catch differences in sample rates
capsys.readouterr() # clear buffer
w = [(wav_file, ), (wav_file_8k, )]
out = fun(w)
err = capsys.readouterr().err
assert 'several sample rates found in wav files' in err
assert sorted(out.keys()) == ['utt_1', 'utt_2']
# make sure timestamps are ordered
with pytest.raises(ValueError) as err:
fun([('1', wav_file, 1, 0)])
assert 'timestamps are not in increasing order for' in str(err)
def test_save(tmpdir, audio):
p = str(tmpdir.join('test.wav'))
audio.save(p)
# cannot overwrite an existing file
with pytest.raises(ValueError) as err:
audio.save(p)
assert 'file already exist' in str(err)
audio2 = Audio.load(p)
assert audio == audio2
# test with float32 wav
signal = np.zeros((1000,), dtype=np.float32)
signal[10] = 1.0
signal[20] = -1.0
p = str(tmpdir.join('test2.wav'))
audio = Audio(signal, 1000)
audio.save(p)
meta = Audio.scan(p)
assert meta.nsamples == 1000
assert meta.nchannels == 1
audio2 = Audio.load(p)
assert audio2 == audio
assert audio2.data.min() == -1.0
assert audio2.data.max() == 1.0
