yt_dlp/downloader/f4m.py

   1 import base64
   2 import io
   3 import itertools
   4 import struct
   5 import time
   6 import urllib.parse
   7
   8 from .fragment import FragmentFD
   9 from ..compat import compat_etree_fromstring
  10 from ..networking.exceptions import HTTPError
  11 from ..utils import fix_xml_ampersands, xpath_text
  12
  13
  14 class DataTruncatedError(Exception):
  15     pass
  16
  17
  18 class FlvReader(io.BytesIO):
  19     """
  20     Reader for Flv files
  21     The file format is documented in https://www.adobe.com/devnet/f4v.html
  22     """
  23
  24     def read_bytes(self, n):
  25         data = self.read(n)
  26         if len(data) < n:
  27             raise DataTruncatedError(
  28                 'FlvReader error: need %d bytes while only %d bytes got' % (
  29                     n, len(data)))
  30         return data
  31
  32     # Utility functions for reading numbers and strings
  33     def read_unsigned_long_long(self):
  34         return struct.unpack('!Q', self.read_bytes(8))[0]
  35
  36     def read_unsigned_int(self):
  37         return struct.unpack('!I', self.read_bytes(4))[0]
  38
  39     def read_unsigned_char(self):
  40         return struct.unpack('!B', self.read_bytes(1))[0]
  41
  42     def read_string(self):
  43         res = b''
  44         while True:
  45             char = self.read_bytes(1)
  46             if char == b'\x00':
  47                 break
  48             res += char
  49         return res
  50
  51     def read_box_info(self):
  52         """
  53         Read a box and return the info as a tuple: (box_size, box_type, box_data)
  54         """
  55         real_size = size = self.read_unsigned_int()
  56         box_type = self.read_bytes(4)
  57         header_end = 8
  58         if size == 1:
  59             real_size = self.read_unsigned_long_long()
  60             header_end = 16
  61         return real_size, box_type, self.read_bytes(real_size - header_end)
  62
  63     def read_asrt(self):
  64         # version
  65         self.read_unsigned_char()
  66         # flags
  67         self.read_bytes(3)
  68         quality_entry_count = self.read_unsigned_char()
  69         # QualityEntryCount
  70         for _ in range(quality_entry_count):
  71             self.read_string()
  72
  73         segment_run_count = self.read_unsigned_int()
  74         segments = []
  75         for _ in range(segment_run_count):
  76             first_segment = self.read_unsigned_int()
  77             fragments_per_segment = self.read_unsigned_int()
  78             segments.append((first_segment, fragments_per_segment))
  79
  80         return {
  81             'segment_run': segments,
  82         }
  83
  84     def read_afrt(self):
  85         # version
  86         self.read_unsigned_char()
  87         # flags
  88         self.read_bytes(3)
  89         # time scale
  90         self.read_unsigned_int()
  91
  92         quality_entry_count = self.read_unsigned_char()
  93         # QualitySegmentUrlModifiers
  94         for _ in range(quality_entry_count):
  95             self.read_string()
  96
  97         fragments_count = self.read_unsigned_int()
  98         fragments = []
  99         for _ in range(fragments_count):
 100             first = self.read_unsigned_int()
 101             first_ts = self.read_unsigned_long_long()
 102             duration = self.read_unsigned_int()
 103             if duration == 0:
 104                 discontinuity_indicator = self.read_unsigned_char()
 105             else:
 106                 discontinuity_indicator = None
 107             fragments.append({
 108                 'first': first,
 109                 'ts': first_ts,
 110                 'duration': duration,
 111                 'discontinuity_indicator': discontinuity_indicator,
 112             })
 113
 114         return {
 115             'fragments': fragments,
 116         }
 117
 118     def read_abst(self):
 119         # version
 120         self.read_unsigned_char()
 121         # flags
 122         self.read_bytes(3)
 123
 124         self.read_unsigned_int()  # BootstrapinfoVersion
 125         # Profile,Live,Update,Reserved
 126         flags = self.read_unsigned_char()
 127         live = flags & 0x20 != 0
 128         # time scale
 129         self.read_unsigned_int()
 130         # CurrentMediaTime
 131         self.read_unsigned_long_long()
 132         # SmpteTimeCodeOffset
 133         self.read_unsigned_long_long()
 134
 135         self.read_string()  # MovieIdentifier
 136         server_count = self.read_unsigned_char()
 137         # ServerEntryTable
 138         for _ in range(server_count):
 139             self.read_string()
 140         quality_count = self.read_unsigned_char()
 141         # QualityEntryTable
 142         for _ in range(quality_count):
 143             self.read_string()
 144         # DrmData
 145         self.read_string()
 146         # MetaData
 147         self.read_string()
 148
 149         segments_count = self.read_unsigned_char()
 150         segments = []
 151         for _ in range(segments_count):
 152             box_size, box_type, box_data = self.read_box_info()
 153             assert box_type == b'asrt'
 154             segment = FlvReader(box_data).read_asrt()
 155             segments.append(segment)
 156         fragments_run_count = self.read_unsigned_char()
 157         fragments = []
 158         for _ in range(fragments_run_count):
 159             box_size, box_type, box_data = self.read_box_info()
 160             assert box_type == b'afrt'
 161             fragments.append(FlvReader(box_data).read_afrt())
 162
 163         return {
 164             'segments': segments,
 165             'fragments': fragments,
 166             'live': live,
 167         }
 168
 169     def read_bootstrap_info(self):
 170         total_size, box_type, box_data = self.read_box_info()
 171         assert box_type == b'abst'
 172         return FlvReader(box_data).read_abst()
 173
 174
 175 def read_bootstrap_info(bootstrap_bytes):
 176     return FlvReader(bootstrap_bytes).read_bootstrap_info()
 177
 178
 179 def build_fragments_list(boot_info):
 180     """ Return a list of (segment, fragment) for each fragment in the video """
 181     res = []
 182     segment_run_table = boot_info['segments'][0]
 183     fragment_run_entry_table = boot_info['fragments'][0]['fragments']
 184     first_frag_number = fragment_run_entry_table[0]['first']
 185     fragments_counter = itertools.count(first_frag_number)
 186     for segment, fragments_count in segment_run_table['segment_run']:
 187         # In some live HDS streams (e.g. Rai), `fragments_count` is
 188         # abnormal and causing out-of-memory errors. It's OK to change the
 189         # number of fragments for live streams as they are updated periodically
 190         if fragments_count == 4294967295 and boot_info['live']:
 191             fragments_count = 2
 192         for _ in range(fragments_count):
 193             res.append((segment, next(fragments_counter)))
 194
 195     if boot_info['live']:
 196         res = res[-2:]
 197
 198     return res
 199
 200
 201 def write_unsigned_int(stream, val):
 202     stream.write(struct.pack('!I', val))
 203
 204
 205 def write_unsigned_int_24(stream, val):
 206     stream.write(struct.pack('!I', val)[1:])
 207
 208
 209 def write_flv_header(stream):
 210     """Writes the FLV header to stream"""
 211     # FLV header
 212     stream.write(b'FLV\x01')
 213     stream.write(b'\x05')
 214     stream.write(b'\x00\x00\x00\x09')
 215     stream.write(b'\x00\x00\x00\x00')
 216
 217
 218 def write_metadata_tag(stream, metadata):
 219     """Writes optional metadata tag to stream"""
 220     SCRIPT_TAG = b'\x12'
 221     FLV_TAG_HEADER_LEN = 11
 222
 223     if metadata:
 224         stream.write(SCRIPT_TAG)
 225         write_unsigned_int_24(stream, len(metadata))
 226         stream.write(b'\x00\x00\x00\x00\x00\x00\x00')
 227         stream.write(metadata)
 228         write_unsigned_int(stream, FLV_TAG_HEADER_LEN + len(metadata))
 229
 230
 231 def remove_encrypted_media(media):
 232     return list(filter(lambda e: 'drmAdditionalHeaderId' not in e.attrib
 233                                  and 'drmAdditionalHeaderSetId' not in e.attrib,
 234                        media))
 235
 236
 237 def _add_ns(prop, ver=1):
 238     return '{http://ns.adobe.com/f4m/%d.0}%s' % (ver, prop)
 239
 240
 241 def get_base_url(manifest):
 242     base_url = xpath_text(
 243         manifest, [_add_ns('baseURL'), _add_ns('baseURL', 2)],
 244         'base URL', default=None)
 245     if base_url:
 246         base_url = base_url.strip()
 247     return base_url
 248
 249
 250 class F4mFD(FragmentFD):
 251     """
 252     A downloader for f4m manifests or AdobeHDS.
 253     """
 254
 255     def _get_unencrypted_media(self, doc):
 256         media = doc.findall(_add_ns('media'))
 257         if not media:
 258             self.report_error('No media found')
 259         if not self.params.get('allow_unplayable_formats'):
 260             for e in (doc.findall(_add_ns('drmAdditionalHeader'))
 261                       + doc.findall(_add_ns('drmAdditionalHeaderSet'))):
 262                 # If id attribute is missing it's valid for all media nodes
 263                 # without drmAdditionalHeaderId or drmAdditionalHeaderSetId attribute
 264                 if 'id' not in e.attrib:
 265                     self.report_error('Missing ID in f4m DRM')
 266             media = remove_encrypted_media(media)
 267         if not media:
 268             self.report_error('Unsupported DRM')
 269         return media
 270
 271     def _get_bootstrap_from_url(self, bootstrap_url):
 272         bootstrap = self.ydl.urlopen(bootstrap_url).read()
 273         return read_bootstrap_info(bootstrap)
 274
 275     def _update_live_fragments(self, bootstrap_url, latest_fragment):
 276         fragments_list = []
 277         retries = 30
 278         while (not fragments_list) and (retries > 0):
 279             boot_info = self._get_bootstrap_from_url(bootstrap_url)
 280             fragments_list = build_fragments_list(boot_info)
 281             fragments_list = [f for f in fragments_list if f[1] > latest_fragment]
 282             if not fragments_list:
 283                 # Retry after a while
 284                 time.sleep(5.0)
 285                 retries -= 1
 286
 287         if not fragments_list:
 288             self.report_error('Failed to update fragments')
 289
 290         return fragments_list
 291
 292     def _parse_bootstrap_node(self, node, base_url):
 293         # Sometimes non empty inline bootstrap info can be specified along
 294         # with bootstrap url attribute (e.g. dummy inline bootstrap info
 295         # contains whitespace characters in [1]). We will prefer bootstrap
 296         # url over inline bootstrap info when present.
 297         # 1. http://live-1-1.rutube.ru/stream/1024/HDS/SD/C2NKsS85HQNckgn5HdEmOQ/1454167650/S-s604419906/move/four/dirs/upper/1024-576p.f4m
 298         bootstrap_url = node.get('url')
 299         if bootstrap_url:
 300             bootstrap_url = urllib.parse.urljoin(
 301                 base_url, bootstrap_url)
 302             boot_info = self._get_bootstrap_from_url(bootstrap_url)
 303         else:
 304             bootstrap_url = None
 305             bootstrap = base64.b64decode(node.text)
 306             boot_info = read_bootstrap_info(bootstrap)
 307         return boot_info, bootstrap_url
 308
 309     def real_download(self, filename, info_dict):
 310         man_url = info_dict['url']
 311         requested_bitrate = info_dict.get('tbr')
 312         self.to_screen(f'[{self.FD_NAME}] Downloading f4m manifest')
 313
 314         urlh = self.ydl.urlopen(self._prepare_url(info_dict, man_url))
 315         man_url = urlh.url
 316         # Some manifests may be malformed, e.g. prosiebensat1 generated manifests
 317         # (see https://github.com/ytdl-org/youtube-dl/issues/6215#issuecomment-121704244
 318         # and https://github.com/ytdl-org/youtube-dl/issues/7823)
 319         manifest = fix_xml_ampersands(urlh.read().decode('utf-8', 'ignore')).strip()
 320
 321         doc = compat_etree_fromstring(manifest)
 322         formats = [(int(f.attrib.get('bitrate', -1)), f)
 323                    for f in self._get_unencrypted_media(doc)]
 324         if requested_bitrate is None or len(formats) == 1:
 325             # get the best format
 326             formats = sorted(formats, key=lambda f: f[0])
 327             rate, media = formats[-1]
 328         else:
 329             rate, media = next(filter(
 330                 lambda f: int(f[0]) == requested_bitrate, formats))
 331
 332         # Prefer baseURL for relative URLs as per 11.2 of F4M 3.0 spec.
 333         man_base_url = get_base_url(doc) or man_url
 334
 335         base_url = urllib.parse.urljoin(man_base_url, media.attrib['url'])
 336         bootstrap_node = doc.find(_add_ns('bootstrapInfo'))
 337         boot_info, bootstrap_url = self._parse_bootstrap_node(
 338             bootstrap_node, man_base_url)
 339         live = boot_info['live']
 340         metadata_node = media.find(_add_ns('metadata'))
 341         if metadata_node is not None:
 342             metadata = base64.b64decode(metadata_node.text)
 343         else:
 344             metadata = None
 345
 346         fragments_list = build_fragments_list(boot_info)
 347         test = self.params.get('test', False)
 348         if test:
 349             # We only download the first fragment
 350             fragments_list = fragments_list[:1]
 351         total_frags = len(fragments_list)
 352         # For some akamai manifests we'll need to add a query to the fragment url
 353         akamai_pv = xpath_text(doc, _add_ns('pv-2.0'))
 354
 355         ctx = {
 356             'filename': filename,
 357             'total_frags': total_frags,
 358             'live': bool(live),
 359         }
 360
 361         self._prepare_frag_download(ctx)
 362
 363         dest_stream = ctx['dest_stream']
 364
 365         if ctx['complete_frags_downloaded_bytes'] == 0:
 366             write_flv_header(dest_stream)
 367             if not live:
 368                 write_metadata_tag(dest_stream, metadata)
 369
 370         base_url_parsed = urllib.parse.urlparse(base_url)
 371
 372         self._start_frag_download(ctx, info_dict)
 373
 374         frag_index = 0
 375         while fragments_list:
 376             seg_i, frag_i = fragments_list.pop(0)
 377             frag_index += 1
 378             if frag_index <= ctx['fragment_index']:
 379                 continue
 380             name = 'Seg%d-Frag%d' % (seg_i, frag_i)
 381             query = []
 382             if base_url_parsed.query:
 383                 query.append(base_url_parsed.query)
 384             if akamai_pv:
 385                 query.append(akamai_pv.strip(';'))
 386             if info_dict.get('extra_param_to_segment_url'):
 387                 query.append(info_dict['extra_param_to_segment_url'])
 388             url_parsed = base_url_parsed._replace(path=base_url_parsed.path + name, query='&'.join(query))
 389             try:
 390                 success = self._download_fragment(ctx, url_parsed.geturl(), info_dict)
 391                 if not success:
 392                     return False
 393                 down_data = self._read_fragment(ctx)
 394                 reader = FlvReader(down_data)
 395                 while True:
 396                     try:
 397                         _, box_type, box_data = reader.read_box_info()
 398                     except DataTruncatedError:
 399                         if test:
 400                             # In tests, segments may be truncated, and thus
 401                             # FlvReader may not be able to parse the whole
 402                             # chunk. If so, write the segment as is
 403                             # See https://github.com/ytdl-org/youtube-dl/issues/9214
 404                             dest_stream.write(down_data)
 405                             break
 406                         raise
 407                     if box_type == b'mdat':
 408                         self._append_fragment(ctx, box_data)
 409                         break
 410             except HTTPError as err:
 411                 if live and (err.status == 404 or err.status == 410):
 412                     # We didn't keep up with the live window. Continue
 413                     # with the next available fragment.
 414                     msg = 'Fragment %d unavailable' % frag_i
 415                     self.report_warning(msg)
 416                     fragments_list = []
 417                 else:
 418                     raise
 419
 420             if not fragments_list and not test and live and bootstrap_url:
 421                 fragments_list = self._update_live_fragments(bootstrap_url, frag_i)
 422                 total_frags += len(fragments_list)
 423                 if fragments_list and (fragments_list[0][1] > frag_i + 1):
 424                     msg = 'Missed %d fragments' % (fragments_list[0][1] - (frag_i + 1))
 425                     self.report_warning(msg)
 426
 427         return self._finish_frag_download(ctx, info_dict)