]> Raphaƫl G. Git Repositories - youtubedl/blob - youtube_dl/downloader/f4m.py
Fix extraction from youtube.
[youtubedl] / youtube_dl / downloader / f4m.py
1 from __future__ import division, unicode_literals
2
3 import io
4 import itertools
5 import time
6
7 from .fragment import FragmentFD
8 from ..compat import (
9 compat_b64decode,
10 compat_etree_fromstring,
11 compat_urlparse,
12 compat_urllib_error,
13 compat_urllib_parse_urlparse,
14 compat_struct_pack,
15 compat_struct_unpack,
16 )
17 from ..utils import (
18 fix_xml_ampersands,
19 xpath_text,
20 )
21
22
23 class DataTruncatedError(Exception):
24 pass
25
26
27 class FlvReader(io.BytesIO):
28 """
29 Reader for Flv files
30 The file format is documented in https://www.adobe.com/devnet/f4v.html
31 """
32
33 def read_bytes(self, n):
34 data = self.read(n)
35 if len(data) < n:
36 raise DataTruncatedError(
37 'FlvReader error: need %d bytes while only %d bytes got' % (
38 n, len(data)))
39 return data
40
41 # Utility functions for reading numbers and strings
42 def read_unsigned_long_long(self):
43 return compat_struct_unpack('!Q', self.read_bytes(8))[0]
44
45 def read_unsigned_int(self):
46 return compat_struct_unpack('!I', self.read_bytes(4))[0]
47
48 def read_unsigned_char(self):
49 return compat_struct_unpack('!B', self.read_bytes(1))[0]
50
51 def read_string(self):
52 res = b''
53 while True:
54 char = self.read_bytes(1)
55 if char == b'\x00':
56 break
57 res += char
58 return res
59
60 def read_box_info(self):
61 """
62 Read a box and return the info as a tuple: (box_size, box_type, box_data)
63 """
64 real_size = size = self.read_unsigned_int()
65 box_type = self.read_bytes(4)
66 header_end = 8
67 if size == 1:
68 real_size = self.read_unsigned_long_long()
69 header_end = 16
70 return real_size, box_type, self.read_bytes(real_size - header_end)
71
72 def read_asrt(self):
73 # version
74 self.read_unsigned_char()
75 # flags
76 self.read_bytes(3)
77 quality_entry_count = self.read_unsigned_char()
78 # QualityEntryCount
79 for i in range(quality_entry_count):
80 self.read_string()
81
82 segment_run_count = self.read_unsigned_int()
83 segments = []
84 for i in range(segment_run_count):
85 first_segment = self.read_unsigned_int()
86 fragments_per_segment = self.read_unsigned_int()
87 segments.append((first_segment, fragments_per_segment))
88
89 return {
90 'segment_run': segments,
91 }
92
93 def read_afrt(self):
94 # version
95 self.read_unsigned_char()
96 # flags
97 self.read_bytes(3)
98 # time scale
99 self.read_unsigned_int()
100
101 quality_entry_count = self.read_unsigned_char()
102 # QualitySegmentUrlModifiers
103 for i in range(quality_entry_count):
104 self.read_string()
105
106 fragments_count = self.read_unsigned_int()
107 fragments = []
108 for i in range(fragments_count):
109 first = self.read_unsigned_int()
110 first_ts = self.read_unsigned_long_long()
111 duration = self.read_unsigned_int()
112 if duration == 0:
113 discontinuity_indicator = self.read_unsigned_char()
114 else:
115 discontinuity_indicator = None
116 fragments.append({
117 'first': first,
118 'ts': first_ts,
119 'duration': duration,
120 'discontinuity_indicator': discontinuity_indicator,
121 })
122
123 return {
124 'fragments': fragments,
125 }
126
127 def read_abst(self):
128 # version
129 self.read_unsigned_char()
130 # flags
131 self.read_bytes(3)
132
133 self.read_unsigned_int() # BootstrapinfoVersion
134 # Profile,Live,Update,Reserved
135 flags = self.read_unsigned_char()
136 live = flags & 0x20 != 0
137 # time scale
138 self.read_unsigned_int()
139 # CurrentMediaTime
140 self.read_unsigned_long_long()
141 # SmpteTimeCodeOffset
142 self.read_unsigned_long_long()
143
144 self.read_string() # MovieIdentifier
145 server_count = self.read_unsigned_char()
146 # ServerEntryTable
147 for i in range(server_count):
148 self.read_string()
149 quality_count = self.read_unsigned_char()
150 # QualityEntryTable
151 for i in range(quality_count):
152 self.read_string()
153 # DrmData
154 self.read_string()
155 # MetaData
156 self.read_string()
157
158 segments_count = self.read_unsigned_char()
159 segments = []
160 for i in range(segments_count):
161 box_size, box_type, box_data = self.read_box_info()
162 assert box_type == b'asrt'
163 segment = FlvReader(box_data).read_asrt()
164 segments.append(segment)
165 fragments_run_count = self.read_unsigned_char()
166 fragments = []
167 for i in range(fragments_run_count):
168 box_size, box_type, box_data = self.read_box_info()
169 assert box_type == b'afrt'
170 fragments.append(FlvReader(box_data).read_afrt())
171
172 return {
173 'segments': segments,
174 'fragments': fragments,
175 'live': live,
176 }
177
178 def read_bootstrap_info(self):
179 total_size, box_type, box_data = self.read_box_info()
180 assert box_type == b'abst'
181 return FlvReader(box_data).read_abst()
182
183
184 def read_bootstrap_info(bootstrap_bytes):
185 return FlvReader(bootstrap_bytes).read_bootstrap_info()
186
187
188 def build_fragments_list(boot_info):
189 """ Return a list of (segment, fragment) for each fragment in the video """
190 res = []
191 segment_run_table = boot_info['segments'][0]
192 fragment_run_entry_table = boot_info['fragments'][0]['fragments']
193 first_frag_number = fragment_run_entry_table[0]['first']
194 fragments_counter = itertools.count(first_frag_number)
195 for segment, fragments_count in segment_run_table['segment_run']:
196 # In some live HDS streams (for example Rai), `fragments_count` is
197 # abnormal and causing out-of-memory errors. It's OK to change the
198 # number of fragments for live streams as they are updated periodically
199 if fragments_count == 4294967295 and boot_info['live']:
200 fragments_count = 2
201 for _ in range(fragments_count):
202 res.append((segment, next(fragments_counter)))
203
204 if boot_info['live']:
205 res = res[-2:]
206
207 return res
208
209
210 def write_unsigned_int(stream, val):
211 stream.write(compat_struct_pack('!I', val))
212
213
214 def write_unsigned_int_24(stream, val):
215 stream.write(compat_struct_pack('!I', val)[1:])
216
217
218 def write_flv_header(stream):
219 """Writes the FLV header to stream"""
220 # FLV header
221 stream.write(b'FLV\x01')
222 stream.write(b'\x05')
223 stream.write(b'\x00\x00\x00\x09')
224 stream.write(b'\x00\x00\x00\x00')
225
226
227 def write_metadata_tag(stream, metadata):
228 """Writes optional metadata tag to stream"""
229 SCRIPT_TAG = b'\x12'
230 FLV_TAG_HEADER_LEN = 11
231
232 if metadata:
233 stream.write(SCRIPT_TAG)
234 write_unsigned_int_24(stream, len(metadata))
235 stream.write(b'\x00\x00\x00\x00\x00\x00\x00')
236 stream.write(metadata)
237 write_unsigned_int(stream, FLV_TAG_HEADER_LEN + len(metadata))
238
239
240 def remove_encrypted_media(media):
241 return list(filter(lambda e: 'drmAdditionalHeaderId' not in e.attrib and
242 'drmAdditionalHeaderSetId' not in e.attrib,
243 media))
244
245
246 def _add_ns(prop, ver=1):
247 return '{http://ns.adobe.com/f4m/%d.0}%s' % (ver, prop)
248
249
250 def get_base_url(manifest):
251 base_url = xpath_text(
252 manifest, [_add_ns('baseURL'), _add_ns('baseURL', 2)],
253 'base URL', default=None)
254 if base_url:
255 base_url = base_url.strip()
256 return base_url
257
258
259 class F4mFD(FragmentFD):
260 """
261 A downloader for f4m manifests or AdobeHDS.
262 """
263
264 FD_NAME = 'f4m'
265
266 def _get_unencrypted_media(self, doc):
267 media = doc.findall(_add_ns('media'))
268 if not media:
269 self.report_error('No media found')
270 for e in (doc.findall(_add_ns('drmAdditionalHeader')) +
271 doc.findall(_add_ns('drmAdditionalHeaderSet'))):
272 # If id attribute is missing it's valid for all media nodes
273 # without drmAdditionalHeaderId or drmAdditionalHeaderSetId attribute
274 if 'id' not in e.attrib:
275 self.report_error('Missing ID in f4m DRM')
276 media = remove_encrypted_media(media)
277 if not media:
278 self.report_error('Unsupported DRM')
279 return media
280
281 def _get_bootstrap_from_url(self, bootstrap_url):
282 bootstrap = self.ydl.urlopen(bootstrap_url).read()
283 return read_bootstrap_info(bootstrap)
284
285 def _update_live_fragments(self, bootstrap_url, latest_fragment):
286 fragments_list = []
287 retries = 30
288 while (not fragments_list) and (retries > 0):
289 boot_info = self._get_bootstrap_from_url(bootstrap_url)
290 fragments_list = build_fragments_list(boot_info)
291 fragments_list = [f for f in fragments_list if f[1] > latest_fragment]
292 if not fragments_list:
293 # Retry after a while
294 time.sleep(5.0)
295 retries -= 1
296
297 if not fragments_list:
298 self.report_error('Failed to update fragments')
299
300 return fragments_list
301
302 def _parse_bootstrap_node(self, node, base_url):
303 # Sometimes non empty inline bootstrap info can be specified along
304 # with bootstrap url attribute (e.g. dummy inline bootstrap info
305 # contains whitespace characters in [1]). We will prefer bootstrap
306 # url over inline bootstrap info when present.
307 # 1. http://live-1-1.rutube.ru/stream/1024/HDS/SD/C2NKsS85HQNckgn5HdEmOQ/1454167650/S-s604419906/move/four/dirs/upper/1024-576p.f4m
308 bootstrap_url = node.get('url')
309 if bootstrap_url:
310 bootstrap_url = compat_urlparse.urljoin(
311 base_url, bootstrap_url)
312 boot_info = self._get_bootstrap_from_url(bootstrap_url)
313 else:
314 bootstrap_url = None
315 bootstrap = compat_b64decode(node.text)
316 boot_info = read_bootstrap_info(bootstrap)
317 return boot_info, bootstrap_url
318
319 def real_download(self, filename, info_dict):
320 man_url = info_dict['url']
321 requested_bitrate = info_dict.get('tbr')
322 self.to_screen('[%s] Downloading f4m manifest' % self.FD_NAME)
323
324 urlh = self.ydl.urlopen(self._prepare_url(info_dict, man_url))
325 man_url = urlh.geturl()
326 # Some manifests may be malformed, e.g. prosiebensat1 generated manifests
327 # (see https://github.com/rg3/youtube-dl/issues/6215#issuecomment-121704244
328 # and https://github.com/rg3/youtube-dl/issues/7823)
329 manifest = fix_xml_ampersands(urlh.read().decode('utf-8', 'ignore')).strip()
330
331 doc = compat_etree_fromstring(manifest)
332 formats = [(int(f.attrib.get('bitrate', -1)), f)
333 for f in self._get_unencrypted_media(doc)]
334 if requested_bitrate is None or len(formats) == 1:
335 # get the best format
336 formats = sorted(formats, key=lambda f: f[0])
337 rate, media = formats[-1]
338 else:
339 rate, media = list(filter(
340 lambda f: int(f[0]) == requested_bitrate, formats))[0]
341
342 # Prefer baseURL for relative URLs as per 11.2 of F4M 3.0 spec.
343 man_base_url = get_base_url(doc) or man_url
344
345 base_url = compat_urlparse.urljoin(man_base_url, media.attrib['url'])
346 bootstrap_node = doc.find(_add_ns('bootstrapInfo'))
347 boot_info, bootstrap_url = self._parse_bootstrap_node(
348 bootstrap_node, man_base_url)
349 live = boot_info['live']
350 metadata_node = media.find(_add_ns('metadata'))
351 if metadata_node is not None:
352 metadata = compat_b64decode(metadata_node.text)
353 else:
354 metadata = None
355
356 fragments_list = build_fragments_list(boot_info)
357 test = self.params.get('test', False)
358 if test:
359 # We only download the first fragment
360 fragments_list = fragments_list[:1]
361 total_frags = len(fragments_list)
362 # For some akamai manifests we'll need to add a query to the fragment url
363 akamai_pv = xpath_text(doc, _add_ns('pv-2.0'))
364
365 ctx = {
366 'filename': filename,
367 'total_frags': total_frags,
368 'live': live,
369 }
370
371 self._prepare_frag_download(ctx)
372
373 dest_stream = ctx['dest_stream']
374
375 if ctx['complete_frags_downloaded_bytes'] == 0:
376 write_flv_header(dest_stream)
377 if not live:
378 write_metadata_tag(dest_stream, metadata)
379
380 base_url_parsed = compat_urllib_parse_urlparse(base_url)
381
382 self._start_frag_download(ctx)
383
384 frag_index = 0
385 while fragments_list:
386 seg_i, frag_i = fragments_list.pop(0)
387 frag_index += 1
388 if frag_index <= ctx['fragment_index']:
389 continue
390 name = 'Seg%d-Frag%d' % (seg_i, frag_i)
391 query = []
392 if base_url_parsed.query:
393 query.append(base_url_parsed.query)
394 if akamai_pv:
395 query.append(akamai_pv.strip(';'))
396 if info_dict.get('extra_param_to_segment_url'):
397 query.append(info_dict['extra_param_to_segment_url'])
398 url_parsed = base_url_parsed._replace(path=base_url_parsed.path + name, query='&'.join(query))
399 try:
400 success, down_data = self._download_fragment(ctx, url_parsed.geturl(), info_dict)
401 if not success:
402 return False
403 reader = FlvReader(down_data)
404 while True:
405 try:
406 _, box_type, box_data = reader.read_box_info()
407 except DataTruncatedError:
408 if test:
409 # In tests, segments may be truncated, and thus
410 # FlvReader may not be able to parse the whole
411 # chunk. If so, write the segment as is
412 # See https://github.com/rg3/youtube-dl/issues/9214
413 dest_stream.write(down_data)
414 break
415 raise
416 if box_type == b'mdat':
417 self._append_fragment(ctx, box_data)
418 break
419 except (compat_urllib_error.HTTPError, ) as err:
420 if live and (err.code == 404 or err.code == 410):
421 # We didn't keep up with the live window. Continue
422 # with the next available fragment.
423 msg = 'Fragment %d unavailable' % frag_i
424 self.report_warning(msg)
425 fragments_list = []
426 else:
427 raise
428
429 if not fragments_list and not test and live and bootstrap_url:
430 fragments_list = self._update_live_fragments(bootstrap_url, frag_i)
431 total_frags += len(fragments_list)
432 if fragments_list and (fragments_list[0][1] > frag_i + 1):
433 msg = 'Missed %d fragments' % (fragments_list[0][1] - (frag_i + 1))
434 self.report_warning(msg)
435
436 self._finish_frag_download(ctx)
437
438 return True