def generate_word_contexts(self, length):
self.word_contexts = {}
self.len_windows = []
delta = SubRipTime.from_ordinal(int(length) *
1000) # (ordinal is milliseconds)
if not self.all_frames:
self.full_tokens()
for i, f in enumerate(self.all_frames):
# Get data from frame
f_start = f["start"]
f_end = f["end"]
f_tokens = f["tokens"]
start_of_window = f_start - delta
end_of_window = f_end + delta
if not f_tokens: # The frame has no tokens
continue
f_context = f["tokens"].copy() # Initialization of the context
# Add tokens of preceding frames
j = -1
while (i + j
) >= 0 and self.all_frames[i + j]["end"] >= start_of_window:
f_context.extend(self.all_frames[i + j]["tokens"])
j -= 1
# Add tokens of later frames
j = 1
while (i + j) <= (len(self.all_frames) - 1) and self.all_frames[
i + j]["start"] <= end_of_window:
f_context.extend(self.all_frames[i + j]["tokens"])
j += 1
# Add to context dictionary
for t in f["tokens"]:
self.len_windows.append(
len(f_context) -
1) # This is for the length of windows analysis
if t not in self.word_contexts:
self.word_contexts[t] = {}
for c in f_context:
self.word_contexts[t][c] = self.word_contexts[t].get(c,
0) + 1
self.word_contexts[t][t] -= 1
if self.word_contexts[t][t] == 0:
del self.word_contexts[t][t]
if not self.check_correct_start_end(
): # If end and star are not correct, the matrices are not symmmetric
self.correct_symmetry()
return (self.word_contexts)
def __init__(self, index: int, media_path: MediaPath,
pds: PaletteDefinitionSegment, ods: ObjectDefinitionSegment, wds: WindowDefinitionSegment):
self.index = index
self.start = SubRipTime.from_ordinal(ods.presentation_timestamp)
self.end: Optional[SubRipTime] = None
self.pds = pds
self.ods = ods
self.wds = wds
self.media_path = media_path
self.image = PgsImage(ods.img_data, pds.palettes)
self.text: Optional[str] = None
self.place: Optional[Tuple[int, int, int, int]] = None
def merge_subtitle(sub_a, sub_b, delta, encoding='utf-8'):
"""
合并两种不同言语的srt字幕
因为两个字幕文件的时间轴不一样，所以合并后的字幕会在某一字幕文件转换时生成新的一条字幕，
导致双语字幕并不是同时变化，不过这也是没有办法的事，无法避免
参考https://github.com/byroot/pysrt/issues/17
https://github.com/byroot/pysrt/issues/15
:param sub_a: 使用sub_a = SubRipFile.open(sub_a_path, encoding=encoding)
:param sub_b:
:param delta:
:return:
"""
out = SubRipFile()
intervals = [item.start.ordinal for item in sub_a]
intervals.extend([item.end.ordinal for item in sub_a])
intervals.extend([item.start.ordinal for item in sub_b])
intervals.extend([item.end.ordinal for item in sub_b])
intervals.sort()
j = k = 0
for i in xrange(1, len(intervals)):
start = SubRipTime.from_ordinal(intervals[i - 1])
end = SubRipTime.from_ordinal(intervals[i])
if (end - start) > delta:
text_a, j = find_subtitle(sub_a, start, end, j)
text_b, k = find_subtitle(sub_b, start, end, k)
text = join_lines(text_a, text_b)
if len(text) > 0:
item = SubRipItem(0, start, end, text)
out.append(item)
out.clean_indexes()
return out
def merge_subtitle(sub_a, sub_b, delta, encoding='utf-8'):
"""
合并两种不同言语的srt字幕
因为两个字幕文件的时间轴不一样，所以合并后的字幕会在某一字幕文件转换时生成新的一条字幕，
导致双语字幕并不是同时变化，不过这也是没有办法的事，无法避免
参考https://github.com/byroot/pysrt/issues/17
https://github.com/byroot/pysrt/issues/15
:param sub_a: 使用sub_a = SubRipFile.open(sub_a_path, encoding=encoding)
:param sub_b:
:param delta:
:return:
"""
out = SubRipFile()
intervals = [item.start.ordinal for item in sub_a]
intervals.extend([item.end.ordinal for item in sub_a])
intervals.extend([item.start.ordinal for item in sub_b])
intervals.extend([item.end.ordinal for item in sub_b])
intervals.sort()
j = k = 0
for i in xrange(1, len(intervals)):
start = SubRipTime.from_ordinal(intervals[i - 1])
end = SubRipTime.from_ordinal(intervals[i])
if (end - start) > delta:
text_a, j = find_subtitle(sub_a, start, end, j)
text_b, k = find_subtitle(sub_b, start, end, k)
text = join_lines(text_a, text_b)
if len(text) > 0:
item = SubRipItem(0, start, end, text)
out.append(item)
out.clean_indexes()
return out
def merge_subtitle(sub_a, sub_b, delta):
out = SubRipFile()
intervals = [item.start.ordinal for item in sub_a]
intervals.extend([item.end.ordinal for item in sub_a])
intervals.extend([item.start.ordinal for item in sub_b])
intervals.extend([item.end.ordinal for item in sub_b])
intervals.sort()
j = k = 0
for i in xrange(1, len(intervals)):
start = SubRipTime.from_ordinal(intervals[i-1])
end = SubRipTime.from_ordinal(intervals[i])
if (end-start) > delta:
text_a, j = find_subtitle(sub_a, start, end, j)
text_b, k = find_subtitle(sub_b, start, end, k)
text = join_lines(text_a, text_b)
if len(text) > 0:
item = SubRipItem(0, start, end, text)
out.append(item)
out.clean_indexes()
return out
def merge_subtitle(sub_a, sub_b, delta):
out = SubRipFile()
intervals = [item.start.ordinal for item in sub_a]
intervals.extend([item.end.ordinal for item in sub_a])
intervals.extend([item.start.ordinal for item in sub_b])
intervals.extend([item.end.ordinal for item in sub_b])
intervals.sort()
j = k = 0
for i in range(1, len(intervals)):
start = SubRipTime.from_ordinal(intervals[i - 1])
end = SubRipTime.from_ordinal(intervals[i])
if (end - start) > delta:
text_a, j = find_subtitle(sub_a, start, end, j)
text_b, k = find_subtitle(sub_b, start, end, k)
text = join_lines(text_a, text_b)
if len(text) > 0:
item = SubRipItem(0, start, end, text)
out.append(item)
out.clean_indexes()
return out
def decode(cls, data: bytes, media_path: MediaPath):
display_sets = PgsReader.decode(data, media_path)
index = 0
items = []
for display_set in display_sets:
if items and not display_set.has_image and display_set.wds:
items[-1].end = SubRipTime.from_ordinal(display_set.wds[-1].presentation_timestamp)
continue
for (pds, ods, wds) in zip(display_set.pds, display_set.ods, display_set.wds):
item = PgsSubtitleItem(index, media_path, pds, ods, wds)
if items and items[-1].end is None and items[-1].start + 10000 >= item.start:
items[-1].end = max(items[-1].start, item.start - 1)
items.append(item)
index += 1
for item in items:
item.validate()
return items
def test_from_ordinal(self):
self.assertEquals(SubRipTime.from_ordinal(3600000), {'hours': 1})
self.assertEquals(SubRipTime(1), 3600000)
srt = SubRipFile(eol='\n', encoding='utf-8')
i = 1
for line in sublog:
line = line.split(",", 1)
if (line[0] and line[0][0] == '-'):
if (START_TIME == None and line[0][:8] == '- start '):
START_TIME = datetime.strptime(line[0], '- start ' + TIMEFORMAT +
'\n')
continue
no = datetime.strptime(line[0], TIMEFORMAT) - START_TIME
if (abs(no) > timedelta(1)):
print("\nCan't go over a day in a subtitle! Delete non-used lines in" + \
" log.\nLet there only be one '- start' line at the top of" + \
" the log-file.")
sys.exit(1)
time = SubRipTime.from_ordinal(no.seconds*1000 + no.microseconds*0.001)
item = SubRipItem(i, start=time, end=time + 30*1000,
text=unicode(line[1], 'utf-8'))
srt.append(item)
i += 1
srt.clean_indexes()
#srt.save(path=sys.stdout)
for line in srt:
sys.stdout.write(unicode(line).encode('utf-8'))
def test_from_ordinal(self):
self.assertEqual(SubRipTime.from_ordinal(3600000), {'hours': 1})
self.assertEqual(SubRipTime(1), 3600000)
