yt_dlp/downloader/f4m.py

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