f4m.py 15 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428
  1. import base64
  2. import io
  3. import itertools
  4. import struct
  5. import time
  6. import urllib.parse
  7. from .fragment import FragmentFD
  8. from ..compat import compat_etree_fromstring
  9. from ..networking.exceptions import HTTPError
  10. from ..utils import fix_xml_ampersands, xpath_text
  11. class DataTruncatedError(Exception):
  12. pass
  13. class FlvReader(io.BytesIO):
  14. """
  15. Reader for Flv files
  16. The file format is documented in https://www.adobe.com/devnet/f4v.html
  17. """
  18. def read_bytes(self, n):
  19. data = self.read(n)
  20. if len(data) < n:
  21. raise DataTruncatedError(
  22. 'FlvReader error: need %d bytes while only %d bytes got' % (
  23. n, len(data)))
  24. return data
  25. # Utility functions for reading numbers and strings
  26. def read_unsigned_long_long(self):
  27. return struct.unpack('!Q', self.read_bytes(8))[0]
  28. def read_unsigned_int(self):
  29. return struct.unpack('!I', self.read_bytes(4))[0]
  30. def read_unsigned_char(self):
  31. return struct.unpack('!B', self.read_bytes(1))[0]
  32. def read_string(self):
  33. res = b''
  34. while True:
  35. char = self.read_bytes(1)
  36. if char == b'\x00':
  37. break
  38. res += char
  39. return res
  40. def read_box_info(self):
  41. """
  42. Read a box and return the info as a tuple: (box_size, box_type, box_data)
  43. """
  44. real_size = size = self.read_unsigned_int()
  45. box_type = self.read_bytes(4)
  46. header_end = 8
  47. if size == 1:
  48. real_size = self.read_unsigned_long_long()
  49. header_end = 16
  50. return real_size, box_type, self.read_bytes(real_size - header_end)
  51. def read_asrt(self):
  52. # version
  53. self.read_unsigned_char()
  54. # flags
  55. self.read_bytes(3)
  56. quality_entry_count = self.read_unsigned_char()
  57. # QualityEntryCount
  58. for i in range(quality_entry_count):
  59. self.read_string()
  60. segment_run_count = self.read_unsigned_int()
  61. segments = []
  62. for i in range(segment_run_count):
  63. first_segment = self.read_unsigned_int()
  64. fragments_per_segment = self.read_unsigned_int()
  65. segments.append((first_segment, fragments_per_segment))
  66. return {
  67. 'segment_run': segments,
  68. }
  69. def read_afrt(self):
  70. # version
  71. self.read_unsigned_char()
  72. # flags
  73. self.read_bytes(3)
  74. # time scale
  75. self.read_unsigned_int()
  76. quality_entry_count = self.read_unsigned_char()
  77. # QualitySegmentUrlModifiers
  78. for i in range(quality_entry_count):
  79. self.read_string()
  80. fragments_count = self.read_unsigned_int()
  81. fragments = []
  82. for i in range(fragments_count):
  83. first = self.read_unsigned_int()
  84. first_ts = self.read_unsigned_long_long()
  85. duration = self.read_unsigned_int()
  86. if duration == 0:
  87. discontinuity_indicator = self.read_unsigned_char()
  88. else:
  89. discontinuity_indicator = None
  90. fragments.append({
  91. 'first': first,
  92. 'ts': first_ts,
  93. 'duration': duration,
  94. 'discontinuity_indicator': discontinuity_indicator,
  95. })
  96. return {
  97. 'fragments': fragments,
  98. }
  99. def read_abst(self):
  100. # version
  101. self.read_unsigned_char()
  102. # flags
  103. self.read_bytes(3)
  104. self.read_unsigned_int() # BootstrapinfoVersion
  105. # Profile,Live,Update,Reserved
  106. flags = self.read_unsigned_char()
  107. live = flags & 0x20 != 0
  108. # time scale
  109. self.read_unsigned_int()
  110. # CurrentMediaTime
  111. self.read_unsigned_long_long()
  112. # SmpteTimeCodeOffset
  113. self.read_unsigned_long_long()
  114. self.read_string() # MovieIdentifier
  115. server_count = self.read_unsigned_char()
  116. # ServerEntryTable
  117. for i in range(server_count):
  118. self.read_string()
  119. quality_count = self.read_unsigned_char()
  120. # QualityEntryTable
  121. for i in range(quality_count):
  122. self.read_string()
  123. # DrmData
  124. self.read_string()
  125. # MetaData
  126. self.read_string()
  127. segments_count = self.read_unsigned_char()
  128. segments = []
  129. for i in range(segments_count):
  130. box_size, box_type, box_data = self.read_box_info()
  131. assert box_type == b'asrt'
  132. segment = FlvReader(box_data).read_asrt()
  133. segments.append(segment)
  134. fragments_run_count = self.read_unsigned_char()
  135. fragments = []
  136. for i in range(fragments_run_count):
  137. box_size, box_type, box_data = self.read_box_info()
  138. assert box_type == b'afrt'
  139. fragments.append(FlvReader(box_data).read_afrt())
  140. return {
  141. 'segments': segments,
  142. 'fragments': fragments,
  143. 'live': live,
  144. }
  145. def read_bootstrap_info(self):
  146. total_size, box_type, box_data = self.read_box_info()
  147. assert box_type == b'abst'
  148. return FlvReader(box_data).read_abst()
  149. def read_bootstrap_info(bootstrap_bytes):
  150. return FlvReader(bootstrap_bytes).read_bootstrap_info()
  151. def build_fragments_list(boot_info):
  152. """ Return a list of (segment, fragment) for each fragment in the video """
  153. res = []
  154. segment_run_table = boot_info['segments'][0]
  155. fragment_run_entry_table = boot_info['fragments'][0]['fragments']
  156. first_frag_number = fragment_run_entry_table[0]['first']
  157. fragments_counter = itertools.count(first_frag_number)
  158. for segment, fragments_count in segment_run_table['segment_run']:
  159. # In some live HDS streams (e.g. Rai), `fragments_count` is
  160. # abnormal and causing out-of-memory errors. It's OK to change the
  161. # number of fragments for live streams as they are updated periodically
  162. if fragments_count == 4294967295 and boot_info['live']:
  163. fragments_count = 2
  164. for _ in range(fragments_count):
  165. res.append((segment, next(fragments_counter)))
  166. if boot_info['live']:
  167. res = res[-2:]
  168. return res
  169. def write_unsigned_int(stream, val):
  170. stream.write(struct.pack('!I', val))
  171. def write_unsigned_int_24(stream, val):
  172. stream.write(struct.pack('!I', val)[1:])
  173. def write_flv_header(stream):
  174. """Writes the FLV header to stream"""
  175. # FLV header
  176. stream.write(b'FLV\x01')
  177. stream.write(b'\x05')
  178. stream.write(b'\x00\x00\x00\x09')
  179. stream.write(b'\x00\x00\x00\x00')
  180. def write_metadata_tag(stream, metadata):
  181. """Writes optional metadata tag to stream"""
  182. SCRIPT_TAG = b'\x12'
  183. FLV_TAG_HEADER_LEN = 11
  184. if metadata:
  185. stream.write(SCRIPT_TAG)
  186. write_unsigned_int_24(stream, len(metadata))
  187. stream.write(b'\x00\x00\x00\x00\x00\x00\x00')
  188. stream.write(metadata)
  189. write_unsigned_int(stream, FLV_TAG_HEADER_LEN + len(metadata))
  190. def remove_encrypted_media(media):
  191. return list(filter(lambda e: 'drmAdditionalHeaderId' not in e.attrib
  192. and 'drmAdditionalHeaderSetId' not in e.attrib,
  193. media))
  194. def _add_ns(prop, ver=1):
  195. return '{http://ns.adobe.com/f4m/%d.0}%s' % (ver, prop)
  196. def get_base_url(manifest):
  197. base_url = xpath_text(
  198. manifest, [_add_ns('baseURL'), _add_ns('baseURL', 2)],
  199. 'base URL', default=None)
  200. if base_url:
  201. base_url = base_url.strip()
  202. return base_url
  203. class F4mFD(FragmentFD):
  204. """
  205. A downloader for f4m manifests or AdobeHDS.
  206. """
  207. def _get_unencrypted_media(self, doc):
  208. media = doc.findall(_add_ns('media'))
  209. if not media:
  210. self.report_error('No media found')
  211. if not self.params.get('allow_unplayable_formats'):
  212. for e in (doc.findall(_add_ns('drmAdditionalHeader'))
  213. + doc.findall(_add_ns('drmAdditionalHeaderSet'))):
  214. # If id attribute is missing it's valid for all media nodes
  215. # without drmAdditionalHeaderId or drmAdditionalHeaderSetId attribute
  216. if 'id' not in e.attrib:
  217. self.report_error('Missing ID in f4m DRM')
  218. media = remove_encrypted_media(media)
  219. if not media:
  220. self.report_error('Unsupported DRM')
  221. return media
  222. def _get_bootstrap_from_url(self, bootstrap_url):
  223. bootstrap = self.ydl.urlopen(bootstrap_url).read()
  224. return read_bootstrap_info(bootstrap)
  225. def _update_live_fragments(self, bootstrap_url, latest_fragment):
  226. fragments_list = []
  227. retries = 30
  228. while (not fragments_list) and (retries > 0):
  229. boot_info = self._get_bootstrap_from_url(bootstrap_url)
  230. fragments_list = build_fragments_list(boot_info)
  231. fragments_list = [f for f in fragments_list if f[1] > latest_fragment]
  232. if not fragments_list:
  233. # Retry after a while
  234. time.sleep(5.0)
  235. retries -= 1
  236. if not fragments_list:
  237. self.report_error('Failed to update fragments')
  238. return fragments_list
  239. def _parse_bootstrap_node(self, node, base_url):
  240. # Sometimes non empty inline bootstrap info can be specified along
  241. # with bootstrap url attribute (e.g. dummy inline bootstrap info
  242. # contains whitespace characters in [1]). We will prefer bootstrap
  243. # url over inline bootstrap info when present.
  244. # 1. http://live-1-1.rutube.ru/stream/1024/HDS/SD/C2NKsS85HQNckgn5HdEmOQ/1454167650/S-s604419906/move/four/dirs/upper/1024-576p.f4m
  245. bootstrap_url = node.get('url')
  246. if bootstrap_url:
  247. bootstrap_url = urllib.parse.urljoin(
  248. base_url, bootstrap_url)
  249. boot_info = self._get_bootstrap_from_url(bootstrap_url)
  250. else:
  251. bootstrap_url = None
  252. bootstrap = base64.b64decode(node.text)
  253. boot_info = read_bootstrap_info(bootstrap)
  254. return boot_info, bootstrap_url
  255. def real_download(self, filename, info_dict):
  256. man_url = info_dict['url']
  257. requested_bitrate = info_dict.get('tbr')
  258. self.to_screen('[%s] Downloading f4m manifest' % self.FD_NAME)
  259. urlh = self.ydl.urlopen(self._prepare_url(info_dict, man_url))
  260. man_url = urlh.url
  261. # Some manifests may be malformed, e.g. prosiebensat1 generated manifests
  262. # (see https://github.com/ytdl-org/youtube-dl/issues/6215#issuecomment-121704244
  263. # and https://github.com/ytdl-org/youtube-dl/issues/7823)
  264. manifest = fix_xml_ampersands(urlh.read().decode('utf-8', 'ignore')).strip()
  265. doc = compat_etree_fromstring(manifest)
  266. formats = [(int(f.attrib.get('bitrate', -1)), f)
  267. for f in self._get_unencrypted_media(doc)]
  268. if requested_bitrate is None or len(formats) == 1:
  269. # get the best format
  270. formats = sorted(formats, key=lambda f: f[0])
  271. rate, media = formats[-1]
  272. else:
  273. rate, media = list(filter(
  274. lambda f: int(f[0]) == requested_bitrate, formats))[0]
  275. # Prefer baseURL for relative URLs as per 11.2 of F4M 3.0 spec.
  276. man_base_url = get_base_url(doc) or man_url
  277. base_url = urllib.parse.urljoin(man_base_url, media.attrib['url'])
  278. bootstrap_node = doc.find(_add_ns('bootstrapInfo'))
  279. boot_info, bootstrap_url = self._parse_bootstrap_node(
  280. bootstrap_node, man_base_url)
  281. live = boot_info['live']
  282. metadata_node = media.find(_add_ns('metadata'))
  283. if metadata_node is not None:
  284. metadata = base64.b64decode(metadata_node.text)
  285. else:
  286. metadata = None
  287. fragments_list = build_fragments_list(boot_info)
  288. test = self.params.get('test', False)
  289. if test:
  290. # We only download the first fragment
  291. fragments_list = fragments_list[:1]
  292. total_frags = len(fragments_list)
  293. # For some akamai manifests we'll need to add a query to the fragment url
  294. akamai_pv = xpath_text(doc, _add_ns('pv-2.0'))
  295. ctx = {
  296. 'filename': filename,
  297. 'total_frags': total_frags,
  298. 'live': bool(live),
  299. }
  300. self._prepare_frag_download(ctx)
  301. dest_stream = ctx['dest_stream']
  302. if ctx['complete_frags_downloaded_bytes'] == 0:
  303. write_flv_header(dest_stream)
  304. if not live:
  305. write_metadata_tag(dest_stream, metadata)
  306. base_url_parsed = urllib.parse.urlparse(base_url)
  307. self._start_frag_download(ctx, info_dict)
  308. frag_index = 0
  309. while fragments_list:
  310. seg_i, frag_i = fragments_list.pop(0)
  311. frag_index += 1
  312. if frag_index <= ctx['fragment_index']:
  313. continue
  314. name = 'Seg%d-Frag%d' % (seg_i, frag_i)
  315. query = []
  316. if base_url_parsed.query:
  317. query.append(base_url_parsed.query)
  318. if akamai_pv:
  319. query.append(akamai_pv.strip(';'))
  320. if info_dict.get('extra_param_to_segment_url'):
  321. query.append(info_dict['extra_param_to_segment_url'])
  322. url_parsed = base_url_parsed._replace(path=base_url_parsed.path + name, query='&'.join(query))
  323. try:
  324. success = self._download_fragment(ctx, url_parsed.geturl(), info_dict)
  325. if not success:
  326. return False
  327. down_data = self._read_fragment(ctx)
  328. reader = FlvReader(down_data)
  329. while True:
  330. try:
  331. _, box_type, box_data = reader.read_box_info()
  332. except DataTruncatedError:
  333. if test:
  334. # In tests, segments may be truncated, and thus
  335. # FlvReader may not be able to parse the whole
  336. # chunk. If so, write the segment as is
  337. # See https://github.com/ytdl-org/youtube-dl/issues/9214
  338. dest_stream.write(down_data)
  339. break
  340. raise
  341. if box_type == b'mdat':
  342. self._append_fragment(ctx, box_data)
  343. break
  344. except HTTPError as err:
  345. if live and (err.status == 404 or err.status == 410):
  346. # We didn't keep up with the live window. Continue
  347. # with the next available fragment.
  348. msg = 'Fragment %d unavailable' % frag_i
  349. self.report_warning(msg)
  350. fragments_list = []
  351. else:
  352. raise
  353. if not fragments_list and not test and live and bootstrap_url:
  354. fragments_list = self._update_live_fragments(bootstrap_url, frag_i)
  355. total_frags += len(fragments_list)
  356. if fragments_list and (fragments_list[0][1] > frag_i + 1):
  357. msg = 'Missed %d fragments' % (fragments_list[0][1] - (frag_i + 1))
  358. self.report_warning(msg)
  359. return self._finish_frag_download(ctx, info_dict)