def test_start_seconds(start_seconds):
parser_zero = GenericSubtitleParser(start_seconds=0)
parser_zero.fit(BytesIO(fake_srt))
parser = GenericSubtitleParser(start_seconds=start_seconds)
parser.fit(BytesIO(fake_srt))
expected = [sub for sub in parser_zero.subs_ if sub.start >= timedelta(seconds=start_seconds)]
assert all(esub == psub for esub, psub in zip(expected, parser.subs_))
def timestamps_roughly_match(f1, f2):
parser = GenericSubtitleParser()
extractor = SubtitleSpeechTransformer(sample_rate=ffsubsync.DEFAULT_FRAME_RATE)
pipe = make_pipeline(parser, extractor)
f1_bitstring = pipe.fit_transform(f1).astype(bool)
f2_bitstring = pipe.fit_transform(f2).astype(bool)
return np.alltrue(f1_bitstring == f2_bitstring)
def test_same_encoding(encoding):
parser = GenericSubtitleParser(encoding=encoding)
offseter = SubtitleShifter(1)
pipe = make_pipeline(parser, offseter)
pipe.fit(BytesIO(fake_srt))
assert parser.subs_._encoding == encoding
assert offseter.subs_._encoding == parser.subs_._encoding
assert offseter.subs_.set_encoding('same')._encoding == encoding
assert offseter.subs_.set_encoding('utf-8')._encoding == 'utf-8'
def test_offset(offset):
parser = GenericSubtitleParser()
offseter = SubtitleShifter(offset)
pipe = make_pipeline(parser, offseter)
pipe.fit(BytesIO(fake_srt))
for sub_orig, sub_offset in zip(parser.subs_, offseter.subs_):
assert abs(sub_offset.start.total_seconds() -
sub_orig.start.total_seconds() - offset) < 1e-6
assert abs(sub_offset.end.total_seconds() -
sub_orig.end.total_seconds() - offset) < 1e-6
def test_speech_extraction(sample_rate, start_seconds):
parser = GenericSubtitleParser(start_seconds=start_seconds)
extractor = SubtitleSpeechTransformer(sample_rate=sample_rate, start_seconds=start_seconds)
pipe = make_pipeline(parser, extractor)
bitstring = pipe.fit_transform(BytesIO(fake_srt)).astype(bool)
bitstring_shifted_left = np.append(bitstring[1:], [False])
bitstring_shifted_right = np.append([False], bitstring[:-1])
bitstring_cumsum = np.cumsum(bitstring)
consec_ones_end_pos = np.nonzero(bitstring_cumsum *
(bitstring ^ bitstring_shifted_left) *
(bitstring_cumsum != np.cumsum(bitstring_shifted_right)))[0]
prev = 0
for pos, sub in zip(consec_ones_end_pos, parser.subs_):
start = int(round(sub.start.total_seconds() * sample_rate))
duration = sub.end.total_seconds() - sub.start.total_seconds()
stop = start + int(round(duration * sample_rate))
assert bitstring_cumsum[pos] - prev == stop - start
prev = bitstring_cumsum[pos]
def test_max_time_found():
parser = GenericSubtitleParser()
extractor = SubtitleSpeechTransformer(sample_rate=100)
pipe = make_pipeline(parser, extractor)
pipe.fit(BytesIO(fake_srt))
assert extractor.max_time_ == 6.062
def test_max_seconds(max_seconds):
parser = GenericSubtitleParser(max_subtitle_seconds=max_seconds)
parser.fit(BytesIO(fake_srt))
assert max(sub.end - sub.start for sub in parser.subs_) <= timedelta(seconds=max_seconds)
def detected_encoding(fname):
parser = GenericSubtitleParser()
parser.fit(fname)
return parser.detected_encoding_
def detected_encoding(fname):
parser = GenericSubtitleParser(skip_ssa_info=True)
parser.fit(fname)
return parser.detected_encoding_