Raphaël G. Git Repositories - youtubedl/blob - 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 )
  16 from ..utils import (
  17     encodeFilename,
  18     fix_xml_ampersands,
  19     sanitize_open,
  20     struct_pack,
  21     struct_unpack,
  22     xpath_text,
  23 )
  24
  25
  26 class FlvReader(io.BytesIO):
  27     """
  28     Reader for Flv files
  29     The file format is documented in https://www.adobe.com/devnet/f4v.html
  30     """
  31
  32     # Utility functions for reading numbers and strings
  33     def read_unsigned_long_long(self):
  34         return struct_unpack('!Q', self.read(8))[0]
  35
  36     def read_unsigned_int(self):
  37         return struct_unpack('!I', self.read(4))[0]
  38
  39     def read_unsigned_char(self):
  40         return struct_unpack('!B', self.read(1))[0]
  41
  42     def read_string(self):
  43         res = b''
  44         while True:
  45             char = self.read(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(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(real_size - header_end)
  62
  63     def read_asrt(self):
  64         # version
  65         self.read_unsigned_char()
  66         # flags
  67         self.read(3)
  68         quality_entry_count = self.read_unsigned_char()
  69         # QualityEntryCount
  70         for i in range(quality_entry_count):
  71             self.read_string()
  72
  73         segment_run_count = self.read_unsigned_int()
  74         segments = []
  75         for i 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(3)
  89         # time scale
  90         self.read_unsigned_int()
  91
  92         quality_entry_count = self.read_unsigned_char()
  93         # QualitySegmentUrlModifiers
  94         for i in range(quality_entry_count):
  95             self.read_string()
  96
  97         fragments_count = self.read_unsigned_int()
  98         fragments = []
  99         for i 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(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 i in range(server_count):
 139             self.read_string()
 140         quality_count = self.read_unsigned_char()
 141         # QualityEntryTable
 142         for i 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 i 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 i 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         for _ in range(fragments_count):
 188             res.append((segment, next(fragments_counter)))
 189
 190     if boot_info['live']:
 191         res = res[-2:]
 192
 193     return res
 194
 195
 196 def write_unsigned_int(stream, val):
 197     stream.write(struct_pack('!I', val))
 198
 199
 200 def write_unsigned_int_24(stream, val):
 201     stream.write(struct_pack('!I', val)[1:])
 202
 203
 204 def write_flv_header(stream):
 205     """Writes the FLV header to stream"""
 206     # FLV header
 207     stream.write(b'FLV\x01')
 208     stream.write(b'\x05')
 209     stream.write(b'\x00\x00\x00\x09')
 210     stream.write(b'\x00\x00\x00\x00')
 211
 212
 213 def write_metadata_tag(stream, metadata):
 214     """Writes optional metadata tag to stream"""
 215     SCRIPT_TAG = b'\x12'
 216     FLV_TAG_HEADER_LEN = 11
 217
 218     if metadata:
 219         stream.write(SCRIPT_TAG)
 220         write_unsigned_int_24(stream, len(metadata))
 221         stream.write(b'\x00\x00\x00\x00\x00\x00\x00')
 222         stream.write(metadata)
 223         write_unsigned_int(stream, FLV_TAG_HEADER_LEN + len(metadata))
 224
 225
 226 def _add_ns(prop):
 227     return '{http://ns.adobe.com/f4m/1.0}%s' % prop
 228
 229
 230 class F4mFD(FragmentFD):
 231     """
 232     A downloader for f4m manifests or AdobeHDS.
 233     """
 234
 235     FD_NAME = 'f4m'
 236
 237     def _get_unencrypted_media(self, doc):
 238         media = doc.findall(_add_ns('media'))
 239         if not media:
 240             self.report_error('No media found')
 241         for e in (doc.findall(_add_ns('drmAdditionalHeader')) +
 242                   doc.findall(_add_ns('drmAdditionalHeaderSet'))):
 243             # If id attribute is missing it's valid for all media nodes
 244             # without drmAdditionalHeaderId or drmAdditionalHeaderSetId attribute
 245             if 'id' not in e.attrib:
 246                 self.report_error('Missing ID in f4m DRM')
 247         media = list(filter(lambda e: 'drmAdditionalHeaderId' not in e.attrib and
 248                                       'drmAdditionalHeaderSetId' not in e.attrib,
 249                             media))
 250         if not media:
 251             self.report_error('Unsupported DRM')
 252         return media
 253
 254     def _get_bootstrap_from_url(self, bootstrap_url):
 255         bootstrap = self.ydl.urlopen(bootstrap_url).read()
 256         return read_bootstrap_info(bootstrap)
 257
 258     def _update_live_fragments(self, bootstrap_url, latest_fragment):
 259         fragments_list = []
 260         retries = 30
 261         while (not fragments_list) and (retries > 0):
 262             boot_info = self._get_bootstrap_from_url(bootstrap_url)
 263             fragments_list = build_fragments_list(boot_info)
 264             fragments_list = [f for f in fragments_list if f[1] > latest_fragment]
 265             if not fragments_list:
 266                 # Retry after a while
 267                 time.sleep(5.0)
 268                 retries -= 1
 269
 270         if not fragments_list:
 271             self.report_error('Failed to update fragments')
 272
 273         return fragments_list
 274
 275     def _parse_bootstrap_node(self, node, base_url):
 276         # Sometimes non empty inline bootstrap info can be specified along
 277         # with bootstrap url attribute (e.g. dummy inline bootstrap info
 278         # contains whitespace characters in [1]). We will prefer bootstrap
 279         # url over inline bootstrap info when present.
 280         # 1. http://live-1-1.rutube.ru/stream/1024/HDS/SD/C2NKsS85HQNckgn5HdEmOQ/1454167650/S-s604419906/move/four/dirs/upper/1024-576p.f4m
 281         bootstrap_url = node.get('url')
 282         if bootstrap_url:
 283             bootstrap_url = compat_urlparse.urljoin(
 284                 base_url, bootstrap_url)
 285             boot_info = self._get_bootstrap_from_url(bootstrap_url)
 286         else:
 287             bootstrap_url = None
 288             bootstrap = base64.b64decode(node.text.encode('ascii'))
 289             boot_info = read_bootstrap_info(bootstrap)
 290         return boot_info, bootstrap_url
 291
 292     def real_download(self, filename, info_dict):
 293         man_url = info_dict['url']
 294         requested_bitrate = info_dict.get('tbr')
 295         self.to_screen('[%s] Downloading f4m manifest' % self.FD_NAME)
 296         urlh = self.ydl.urlopen(man_url)
 297         man_url = urlh.geturl()
 298         # Some manifests may be malformed, e.g. prosiebensat1 generated manifests
 299         # (see https://github.com/rg3/youtube-dl/issues/6215#issuecomment-121704244
 300         # and https://github.com/rg3/youtube-dl/issues/7823)
 301         manifest = fix_xml_ampersands(urlh.read().decode('utf-8', 'ignore')).strip()
 302
 303         doc = compat_etree_fromstring(manifest)
 304         formats = [(int(f.attrib.get('bitrate', -1)), f)
 305                    for f in self._get_unencrypted_media(doc)]
 306         if requested_bitrate is None:
 307             # get the best format
 308             formats = sorted(formats, key=lambda f: f[0])
 309             rate, media = formats[-1]
 310         else:
 311             rate, media = list(filter(
 312                 lambda f: int(f[0]) == requested_bitrate, formats))[0]
 313
 314         base_url = compat_urlparse.urljoin(man_url, media.attrib['url'])
 315         bootstrap_node = doc.find(_add_ns('bootstrapInfo'))
 316         boot_info, bootstrap_url = self._parse_bootstrap_node(bootstrap_node, base_url)
 317         live = boot_info['live']
 318         metadata_node = media.find(_add_ns('metadata'))
 319         if metadata_node is not None:
 320             metadata = base64.b64decode(metadata_node.text.encode('ascii'))
 321         else:
 322             metadata = None
 323
 324         fragments_list = build_fragments_list(boot_info)
 325         test = self.params.get('test', False)
 326         if test:
 327             # We only download the first fragment
 328             fragments_list = fragments_list[:1]
 329         total_frags = len(fragments_list)
 330         # For some akamai manifests we'll need to add a query to the fragment url
 331         akamai_pv = xpath_text(doc, _add_ns('pv-2.0'))
 332
 333         ctx = {
 334             'filename': filename,
 335             'total_frags': total_frags,
 336             'live': live,
 337         }
 338
 339         self._prepare_frag_download(ctx)
 340
 341         dest_stream = ctx['dest_stream']
 342
 343         write_flv_header(dest_stream)
 344         if not live:
 345             write_metadata_tag(dest_stream, metadata)
 346
 347         base_url_parsed = compat_urllib_parse_urlparse(base_url)
 348
 349         self._start_frag_download(ctx)
 350
 351         frags_filenames = []
 352         while fragments_list:
 353             seg_i, frag_i = fragments_list.pop(0)
 354             name = 'Seg%d-Frag%d' % (seg_i, frag_i)
 355             query = []
 356             if base_url_parsed.query:
 357                 query.append(base_url_parsed.query)
 358             if akamai_pv:
 359                 query.append(akamai_pv.strip(';'))
 360             if info_dict.get('extra_param_to_segment_url'):
 361                 query.append(info_dict['extra_param_to_segment_url'])
 362             url_parsed = base_url_parsed._replace(path=base_url_parsed.path + name, query='&'.join(query))
 363             frag_filename = '%s-%s' % (ctx['tmpfilename'], name)
 364             try:
 365                 success = ctx['dl'].download(frag_filename, {'url': url_parsed.geturl()})
 366                 if not success:
 367                     return False
 368                 (down, frag_sanitized) = sanitize_open(frag_filename, 'rb')
 369                 down_data = down.read()
 370                 down.close()
 371                 reader = FlvReader(down_data)
 372                 while True:
 373                     _, box_type, box_data = reader.read_box_info()
 374                     if box_type == b'mdat':
 375                         dest_stream.write(box_data)
 376                         break
 377                 if live:
 378                     os.remove(encodeFilename(frag_sanitized))
 379                 else:
 380                     frags_filenames.append(frag_sanitized)
 381             except (compat_urllib_error.HTTPError, ) as err:
 382                 if live and (err.code == 404 or err.code == 410):
 383                     # We didn't keep up with the live window. Continue
 384                     # with the next available fragment.
 385                     msg = 'Fragment %d unavailable' % frag_i
 386                     self.report_warning(msg)
 387                     fragments_list = []
 388                 else:
 389                     raise
 390
 391             if not fragments_list and not test and live and bootstrap_url:
 392                 fragments_list = self._update_live_fragments(bootstrap_url, frag_i)
 393                 total_frags += len(fragments_list)
 394                 if fragments_list and (fragments_list[0][1] > frag_i + 1):
 395                     msg = 'Missed %d fragments' % (fragments_list[0][1] - (frag_i + 1))
 396                     self.report_warning(msg)
 397
 398         self._finish_frag_download(ctx)
 399
 400         for frag_file in frags_filenames:
 401             os.remove(encodeFilename(frag_file))
 402
 403         return True