youtube_dl/downloader/f4m.py

   1 from __future__ import division, unicode_literals
   2
   3 import base64
   4 import io
   5 import itertools
   6 import os
   7 import time
   8
   9 from .fragment import FragmentFD
  10 from ..compat import (
  11     compat_etree_fromstring,
  12     compat_urlparse,
  13     compat_urllib_error,
  14     compat_urllib_parse_urlparse,
  15     compat_struct_pack,
  16     compat_struct_unpack,
  17 )
  18 from ..utils import (
  19     encodeFilename,
  20     fix_xml_ampersands,
  21     sanitize_open,
  22     xpath_text,
  23 )
  24
  25
  26 class DataTruncatedError(Exception):
  27     pass
  28
  29
  30 class FlvReader(io.BytesIO):
  31     """
  32     Reader for Flv files
  33     The file format is documented in https://www.adobe.com/devnet/f4v.html
  34     """
  35
  36     def read_bytes(self, n):
  37         data = self.read(n)
  38         if len(data) < n:
  39             raise DataTruncatedError(
  40                 'FlvReader error: need %d bytes while only %d bytes got' % (
  41                     n, len(data)))
  42         return data
  43
  44     # Utility functions for reading numbers and strings
  45     def read_unsigned_long_long(self):
  46         return compat_struct_unpack('!Q', self.read_bytes(8))[0]
  47
  48     def read_unsigned_int(self):
  49         return compat_struct_unpack('!I', self.read_bytes(4))[0]
  50
  51     def read_unsigned_char(self):
  52         return compat_struct_unpack('!B', self.read_bytes(1))[0]
  53
  54     def read_string(self):
  55         res = b''
  56         while True:
  57             char = self.read_bytes(1)
  58             if char == b'\x00':
  59                 break
  60             res += char
  61         return res
  62
  63     def read_box_info(self):
  64         """
  65         Read a box and return the info as a tuple: (box_size, box_type, box_data)
  66         """
  67         real_size = size = self.read_unsigned_int()
  68         box_type = self.read_bytes(4)
  69         header_end = 8
  70         if size == 1:
  71             real_size = self.read_unsigned_long_long()
  72             header_end = 16
  73         return real_size, box_type, self.read_bytes(real_size - header_end)
  74
  75     def read_asrt(self):
  76         # version
  77         self.read_unsigned_char()
  78         # flags
  79         self.read_bytes(3)
  80         quality_entry_count = self.read_unsigned_char()
  81         # QualityEntryCount
  82         for i in range(quality_entry_count):
  83             self.read_string()
  84
  85         segment_run_count = self.read_unsigned_int()
  86         segments = []
  87         for i in range(segment_run_count):
  88             first_segment = self.read_unsigned_int()
  89             fragments_per_segment = self.read_unsigned_int()
  90             segments.append((first_segment, fragments_per_segment))
  91
  92         return {
  93             'segment_run': segments,
  94         }
  95
  96     def read_afrt(self):
  97         # version
  98         self.read_unsigned_char()
  99         # flags
 100         self.read_bytes(3)
 101         # time scale
 102         self.read_unsigned_int()
 103
 104         quality_entry_count = self.read_unsigned_char()
 105         # QualitySegmentUrlModifiers
 106         for i in range(quality_entry_count):
 107             self.read_string()
 108
 109         fragments_count = self.read_unsigned_int()
 110         fragments = []
 111         for i in range(fragments_count):
 112             first = self.read_unsigned_int()
 113             first_ts = self.read_unsigned_long_long()
 114             duration = self.read_unsigned_int()
 115             if duration == 0:
 116                 discontinuity_indicator = self.read_unsigned_char()
 117             else:
 118                 discontinuity_indicator = None
 119             fragments.append({
 120                 'first': first,
 121                 'ts': first_ts,
 122                 'duration': duration,
 123                 'discontinuity_indicator': discontinuity_indicator,
 124             })
 125
 126         return {
 127             'fragments': fragments,
 128         }
 129
 130     def read_abst(self):
 131         # version
 132         self.read_unsigned_char()
 133         # flags
 134         self.read_bytes(3)
 135
 136         self.read_unsigned_int()  # BootstrapinfoVersion
 137         # Profile,Live,Update,Reserved
 138         flags = self.read_unsigned_char()
 139         live = flags & 0x20 != 0
 140         # time scale
 141         self.read_unsigned_int()
 142         # CurrentMediaTime
 143         self.read_unsigned_long_long()
 144         # SmpteTimeCodeOffset
 145         self.read_unsigned_long_long()
 146
 147         self.read_string()  # MovieIdentifier
 148         server_count = self.read_unsigned_char()
 149         # ServerEntryTable
 150         for i in range(server_count):
 151             self.read_string()
 152         quality_count = self.read_unsigned_char()
 153         # QualityEntryTable
 154         for i in range(quality_count):
 155             self.read_string()
 156         # DrmData
 157         self.read_string()
 158         # MetaData
 159         self.read_string()
 160
 161         segments_count = self.read_unsigned_char()
 162         segments = []
 163         for i in range(segments_count):
 164             box_size, box_type, box_data = self.read_box_info()
 165             assert box_type == b'asrt'
 166             segment = FlvReader(box_data).read_asrt()
 167             segments.append(segment)
 168         fragments_run_count = self.read_unsigned_char()
 169         fragments = []
 170         for i in range(fragments_run_count):
 171             box_size, box_type, box_data = self.read_box_info()
 172             assert box_type == b'afrt'
 173             fragments.append(FlvReader(box_data).read_afrt())
 174
 175         return {
 176             'segments': segments,
 177             'fragments': fragments,
 178             'live': live,
 179         }
 180
 181     def read_bootstrap_info(self):
 182         total_size, box_type, box_data = self.read_box_info()
 183         assert box_type == b'abst'
 184         return FlvReader(box_data).read_abst()
 185
 186
 187 def read_bootstrap_info(bootstrap_bytes):
 188     return FlvReader(bootstrap_bytes).read_bootstrap_info()
 189
 190
 191 def build_fragments_list(boot_info):
 192     """ Return a list of (segment, fragment) for each fragment in the video """
 193     res = []
 194     segment_run_table = boot_info['segments'][0]
 195     fragment_run_entry_table = boot_info['fragments'][0]['fragments']
 196     first_frag_number = fragment_run_entry_table[0]['first']
 197     fragments_counter = itertools.count(first_frag_number)
 198     for segment, fragments_count in segment_run_table['segment_run']:
 199         # In some live HDS streams (for example Rai), `fragments_count` is
 200         # abnormal and causing out-of-memory errors. It's OK to change the
 201         # number of fragments for live streams as they are updated periodically
 202         if fragments_count == 4294967295 and boot_info['live']:
 203             fragments_count = 2
 204         for _ in range(fragments_count):
 205             res.append((segment, next(fragments_counter)))
 206
 207     if boot_info['live']:
 208         res = res[-2:]
 209
 210     return res
 211
 212
 213 def write_unsigned_int(stream, val):
 214     stream.write(compat_struct_pack('!I', val))
 215
 216
 217 def write_unsigned_int_24(stream, val):
 218     stream.write(compat_struct_pack('!I', val)[1:])
 219
 220
 221 def write_flv_header(stream):
 222     """Writes the FLV header to stream"""
 223     # FLV header
 224     stream.write(b'FLV\x01')
 225     stream.write(b'\x05')
 226     stream.write(b'\x00\x00\x00\x09')
 227     stream.write(b'\x00\x00\x00\x00')
 228
 229
 230 def write_metadata_tag(stream, metadata):
 231     """Writes optional metadata tag to stream"""
 232     SCRIPT_TAG = b'\x12'
 233     FLV_TAG_HEADER_LEN = 11
 234
 235     if metadata:
 236         stream.write(SCRIPT_TAG)
 237         write_unsigned_int_24(stream, len(metadata))
 238         stream.write(b'\x00\x00\x00\x00\x00\x00\x00')
 239         stream.write(metadata)
 240         write_unsigned_int(stream, FLV_TAG_HEADER_LEN + len(metadata))
 241
 242
 243 def remove_encrypted_media(media):
 244     return list(filter(lambda e: 'drmAdditionalHeaderId' not in e.attrib and
 245                                  'drmAdditionalHeaderSetId' not in e.attrib,
 246                        media))
 247
 248
 249 def _add_ns(prop):
 250     return '{http://ns.adobe.com/f4m/1.0}%s' % prop
 251
 252
 253 class F4mFD(FragmentFD):
 254     """
 255     A downloader for f4m manifests or AdobeHDS.
 256     """
 257
 258     FD_NAME = 'f4m'
 259
 260     def _get_unencrypted_media(self, doc):
 261         media = doc.findall(_add_ns('media'))
 262         if not media:
 263             self.report_error('No media found')
 264         for e in (doc.findall(_add_ns('drmAdditionalHeader')) +
 265                   doc.findall(_add_ns('drmAdditionalHeaderSet'))):
 266             # If id attribute is missing it's valid for all media nodes
 267             # without drmAdditionalHeaderId or drmAdditionalHeaderSetId attribute
 268             if 'id' not in e.attrib:
 269                 self.report_error('Missing ID in f4m DRM')
 270         media = remove_encrypted_media(media)
 271         if not media:
 272             self.report_error('Unsupported DRM')
 273         return media
 274
 275     def _get_bootstrap_from_url(self, bootstrap_url):
 276         bootstrap = self.ydl.urlopen(bootstrap_url).read()
 277         return read_bootstrap_info(bootstrap)
 278
 279     def _update_live_fragments(self, bootstrap_url, latest_fragment):
 280         fragments_list = []
 281         retries = 30
 282         while (not fragments_list) and (retries > 0):
 283             boot_info = self._get_bootstrap_from_url(bootstrap_url)
 284             fragments_list = build_fragments_list(boot_info)
 285             fragments_list = [f for f in fragments_list if f[1] > latest_fragment]
 286             if not fragments_list:
 287                 # Retry after a while
 288                 time.sleep(5.0)
 289                 retries -= 1
 290
 291         if not fragments_list:
 292             self.report_error('Failed to update fragments')
 293
 294         return fragments_list
 295
 296     def _parse_bootstrap_node(self, node, base_url):
 297         # Sometimes non empty inline bootstrap info can be specified along
 298         # with bootstrap url attribute (e.g. dummy inline bootstrap info
 299         # contains whitespace characters in [1]). We will prefer bootstrap
 300         # url over inline bootstrap info when present.
 301         # 1. http://live-1-1.rutube.ru/stream/1024/HDS/SD/C2NKsS85HQNckgn5HdEmOQ/1454167650/S-s604419906/move/four/dirs/upper/1024-576p.f4m
 302         bootstrap_url = node.get('url')
 303         if bootstrap_url:
 304             bootstrap_url = compat_urlparse.urljoin(
 305                 base_url, bootstrap_url)
 306             boot_info = self._get_bootstrap_from_url(bootstrap_url)
 307         else:
 308             bootstrap_url = None
 309             bootstrap = base64.b64decode(node.text.encode('ascii'))
 310             boot_info = read_bootstrap_info(bootstrap)
 311         return boot_info, bootstrap_url
 312
 313     def real_download(self, filename, info_dict):
 314         man_url = info_dict['url']
 315         requested_bitrate = info_dict.get('tbr')
 316         self.to_screen('[%s] Downloading f4m manifest' % self.FD_NAME)
 317
 318         urlh = self.ydl.urlopen(self._prepare_url(info_dict, man_url))
 319         man_url = urlh.geturl()
 320         # Some manifests may be malformed, e.g. prosiebensat1 generated manifests
 321         # (see https://github.com/rg3/youtube-dl/issues/6215#issuecomment-121704244
 322         # and https://github.com/rg3/youtube-dl/issues/7823)
 323         manifest = fix_xml_ampersands(urlh.read().decode('utf-8', 'ignore')).strip()
 324
 325         doc = compat_etree_fromstring(manifest)
 326         formats = [(int(f.attrib.get('bitrate', -1)), f)
 327                    for f in self._get_unencrypted_media(doc)]
 328         if requested_bitrate is None or len(formats) == 1:
 329             # get the best format
 330             formats = sorted(formats, key=lambda f: f[0])
 331             rate, media = formats[-1]
 332         else:
 333             rate, media = list(filter(
 334                 lambda f: int(f[0]) == requested_bitrate, formats))[0]
 335
 336         base_url = compat_urlparse.urljoin(man_url, media.attrib['url'])
 337         bootstrap_node = doc.find(_add_ns('bootstrapInfo'))
 338         # From Adobe F4M 3.0 spec:
 339         # The <baseURL> element SHALL be the base URL for all relative
 340         # (HTTP-based) URLs in the manifest. If <baseURL> is not present, said
 341         # URLs should be relative to the location of the containing document.
 342         boot_info, bootstrap_url = self._parse_bootstrap_node(bootstrap_node, man_url)
 343         live = boot_info['live']
 344         metadata_node = media.find(_add_ns('metadata'))
 345         if metadata_node is not None:
 346             metadata = base64.b64decode(metadata_node.text.encode('ascii'))
 347         else:
 348             metadata = None
 349
 350         fragments_list = build_fragments_list(boot_info)
 351         test = self.params.get('test', False)
 352         if test:
 353             # We only download the first fragment
 354             fragments_list = fragments_list[:1]
 355         total_frags = len(fragments_list)
 356         # For some akamai manifests we'll need to add a query to the fragment url
 357         akamai_pv = xpath_text(doc, _add_ns('pv-2.0'))
 358
 359         ctx = {
 360             'filename': filename,
 361             'total_frags': total_frags,
 362             'live': live,
 363         }
 364
 365         self._prepare_frag_download(ctx)
 366
 367         dest_stream = ctx['dest_stream']
 368
 369         write_flv_header(dest_stream)
 370         if not live:
 371             write_metadata_tag(dest_stream, metadata)
 372
 373         base_url_parsed = compat_urllib_parse_urlparse(base_url)
 374
 375         self._start_frag_download(ctx)
 376
 377         frags_filenames = []
 378         while fragments_list:
 379             seg_i, frag_i = fragments_list.pop(0)
 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             frag_filename = '%s-%s' % (ctx['tmpfilename'], name)
 390             try:
 391                 success = ctx['dl'].download(frag_filename, {
 392                     'url': url_parsed.geturl(),
 393                     'http_headers': info_dict.get('http_headers'),
 394                 })
 395                 if not success:
 396                     return False
 397                 (down, frag_sanitized) = sanitize_open(frag_filename, 'rb')
 398                 down_data = down.read()
 399                 down.close()
 400                 reader = FlvReader(down_data)
 401                 while True:
 402                     try:
 403                         _, box_type, box_data = reader.read_box_info()
 404                     except DataTruncatedError:
 405                         if test:
 406                             # In tests, segments may be truncated, and thus
 407                             # FlvReader may not be able to parse the whole
 408                             # chunk. If so, write the segment as is
 409                             # See https://github.com/rg3/youtube-dl/issues/9214
 410                             dest_stream.write(down_data)
 411                             break
 412                         raise
 413                     if box_type == b'mdat':
 414                         dest_stream.write(box_data)
 415                         break
 416                 if live:
 417                     os.remove(encodeFilename(frag_sanitized))
 418                 else:
 419                     frags_filenames.append(frag_sanitized)
 420             except (compat_urllib_error.HTTPError, ) as err:
 421                 if live and (err.code == 404 or err.code == 410):
 422                     # We didn't keep up with the live window. Continue
 423                     # with the next available fragment.
 424                     msg = 'Fragment %d unavailable' % frag_i
 425                     self.report_warning(msg)
 426                     fragments_list = []
 427                 else:
 428                     raise
 429
 430             if not fragments_list and not test and live and bootstrap_url:
 431                 fragments_list = self._update_live_fragments(bootstrap_url, frag_i)
 432                 total_frags += len(fragments_list)
 433                 if fragments_list and (fragments_list[0][1] > frag_i + 1):
 434                     msg = 'Missed %d fragments' % (fragments_list[0][1] - (frag_i + 1))
 435                     self.report_warning(msg)
 436
 437         self._finish_frag_download(ctx)
 438
 439         for frag_file in frags_filenames:
 440             os.remove(encodeFilename(frag_file))
 441
 442         return True