_ Git - youtube-dl/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 import xml.etree.ElementTree as etree
   9
  10 from .common import FileDownloader
  11 from .http import HttpFD
  12 from ..compat import (
  13     compat_urlparse,
  14 )
  15 from ..utils import (
  16     struct_pack,
  17     struct_unpack,
  18     encodeFilename,
  19     sanitize_open,
  20     xpath_text,
  21 )
  22
  23
  24 class FlvReader(io.BytesIO):
  25     """
  26     Reader for Flv files
  27     The file format is documented in https://www.adobe.com/devnet/f4v.html
  28     """
  29
  30     # Utility functions for reading numbers and strings
  31     def read_unsigned_long_long(self):
  32         return struct_unpack('!Q', self.read(8))[0]
  33
  34     def read_unsigned_int(self):
  35         return struct_unpack('!I', self.read(4))[0]
  36
  37     def read_unsigned_char(self):
  38         return struct_unpack('!B', self.read(1))[0]
  39
  40     def read_string(self):
  41         res = b''
  42         while True:
  43             char = self.read(1)
  44             if char == b'\x00':
  45                 break
  46             res += char
  47         return res
  48
  49     def read_box_info(self):
  50         """
  51         Read a box and return the info as a tuple: (box_size, box_type, box_data)
  52         """
  53         real_size = size = self.read_unsigned_int()
  54         box_type = self.read(4)
  55         header_end = 8
  56         if size == 1:
  57             real_size = self.read_unsigned_long_long()
  58             header_end = 16
  59         return real_size, box_type, self.read(real_size - header_end)
  60
  61     def read_asrt(self):
  62         # version
  63         self.read_unsigned_char()
  64         # flags
  65         self.read(3)
  66         quality_entry_count = self.read_unsigned_char()
  67         # QualityEntryCount
  68         for i in range(quality_entry_count):
  69             self.read_string()
  70
  71         segment_run_count = self.read_unsigned_int()
  72         segments = []
  73         for i in range(segment_run_count):
  74             first_segment = self.read_unsigned_int()
  75             fragments_per_segment = self.read_unsigned_int()
  76             segments.append((first_segment, fragments_per_segment))
  77
  78         return {
  79             'segment_run': segments,
  80         }
  81
  82     def read_afrt(self):
  83         # version
  84         self.read_unsigned_char()
  85         # flags
  86         self.read(3)
  87         # time scale
  88         self.read_unsigned_int()
  89
  90         quality_entry_count = self.read_unsigned_char()
  91         # QualitySegmentUrlModifiers
  92         for i in range(quality_entry_count):
  93             self.read_string()
  94
  95         fragments_count = self.read_unsigned_int()
  96         fragments = []
  97         for i in range(fragments_count):
  98             first = self.read_unsigned_int()
  99             first_ts = self.read_unsigned_long_long()
 100             duration = self.read_unsigned_int()
 101             if duration == 0:
 102                 discontinuity_indicator = self.read_unsigned_char()
 103             else:
 104                 discontinuity_indicator = None
 105             fragments.append({
 106                 'first': first,
 107                 'ts': first_ts,
 108                 'duration': duration,
 109                 'discontinuity_indicator': discontinuity_indicator,
 110             })
 111
 112         return {
 113             'fragments': fragments,
 114         }
 115
 116     def read_abst(self):
 117         # version
 118         self.read_unsigned_char()
 119         # flags
 120         self.read(3)
 121
 122         self.read_unsigned_int()  # BootstrapinfoVersion
 123         # Profile,Live,Update,Reserved
 124         flags = self.read_unsigned_char()
 125         live = flags & 0x20 != 0
 126         # time scale
 127         self.read_unsigned_int()
 128         # CurrentMediaTime
 129         self.read_unsigned_long_long()
 130         # SmpteTimeCodeOffset
 131         self.read_unsigned_long_long()
 132
 133         self.read_string()  # MovieIdentifier
 134         server_count = self.read_unsigned_char()
 135         # ServerEntryTable
 136         for i in range(server_count):
 137             self.read_string()
 138         quality_count = self.read_unsigned_char()
 139         # QualityEntryTable
 140         for i in range(quality_count):
 141             self.read_string()
 142         # DrmData
 143         self.read_string()
 144         # MetaData
 145         self.read_string()
 146
 147         segments_count = self.read_unsigned_char()
 148         segments = []
 149         for i in range(segments_count):
 150             box_size, box_type, box_data = self.read_box_info()
 151             assert box_type == b'asrt'
 152             segment = FlvReader(box_data).read_asrt()
 153             segments.append(segment)
 154         fragments_run_count = self.read_unsigned_char()
 155         fragments = []
 156         for i in range(fragments_run_count):
 157             box_size, box_type, box_data = self.read_box_info()
 158             assert box_type == b'afrt'
 159             fragments.append(FlvReader(box_data).read_afrt())
 160
 161         return {
 162             'segments': segments,
 163             'fragments': fragments,
 164             'live': live,
 165         }
 166
 167     def read_bootstrap_info(self):
 168         total_size, box_type, box_data = self.read_box_info()
 169         assert box_type == b'abst'
 170         return FlvReader(box_data).read_abst()
 171
 172
 173 def read_bootstrap_info(bootstrap_bytes):
 174     return FlvReader(bootstrap_bytes).read_bootstrap_info()
 175
 176
 177 def build_fragments_list(boot_info):
 178     """ Return a list of (segment, fragment) for each fragment in the video """
 179     res = []
 180     segment_run_table = boot_info['segments'][0]
 181     fragment_run_entry_table = boot_info['fragments'][0]['fragments']
 182     first_frag_number = fragment_run_entry_table[0]['first']
 183     fragments_counter = itertools.count(first_frag_number)
 184     for segment, fragments_count in segment_run_table['segment_run']:
 185         for _ in range(fragments_count):
 186             res.append((segment, next(fragments_counter)))
 187
 188     if boot_info['live']:
 189         res = res[-2:]
 190
 191     return res
 192
 193
 194 def write_unsigned_int(stream, val):
 195     stream.write(struct_pack('!I', val))
 196
 197
 198 def write_unsigned_int_24(stream, val):
 199     stream.write(struct_pack('!I', val)[1:])
 200
 201
 202 def write_flv_header(stream):
 203     """Writes the FLV header to stream"""
 204     # FLV header
 205     stream.write(b'FLV\x01')
 206     stream.write(b'\x05')
 207     stream.write(b'\x00\x00\x00\x09')
 208     stream.write(b'\x00\x00\x00\x00')
 209
 210
 211 def write_metadata_tag(stream, metadata):
 212     """Writes optional metadata tag to stream"""
 213     SCRIPT_TAG = b'\x12'
 214     FLV_TAG_HEADER_LEN = 11
 215
 216     if metadata:
 217         stream.write(SCRIPT_TAG)
 218         write_unsigned_int_24(stream, len(metadata))
 219         stream.write(b'\x00\x00\x00\x00\x00\x00\x00')
 220         stream.write(metadata)
 221         write_unsigned_int(stream, FLV_TAG_HEADER_LEN + len(metadata))
 222
 223
 224 def _add_ns(prop):
 225     return '{http://ns.adobe.com/f4m/1.0}%s' % prop
 226
 227
 228 class HttpQuietDownloader(HttpFD):
 229     def to_screen(self, *args, **kargs):
 230         pass
 231
 232
 233 class F4mFD(FileDownloader):
 234     """
 235     A downloader for f4m manifests or AdobeHDS.
 236     """
 237
 238     def _get_unencrypted_media(self, doc):
 239         media = doc.findall(_add_ns('media'))
 240         if not media:
 241             self.report_error('No media found')
 242         for e in (doc.findall(_add_ns('drmAdditionalHeader')) +
 243                   doc.findall(_add_ns('drmAdditionalHeaderSet'))):
 244             # If id attribute is missing it's valid for all media nodes
 245             # without drmAdditionalHeaderId or drmAdditionalHeaderSetId attribute
 246             if 'id' not in e.attrib:
 247                 self.report_error('Missing ID in f4m DRM')
 248         media = list(filter(lambda e: 'drmAdditionalHeaderId' not in e.attrib and
 249                                       'drmAdditionalHeaderSetId' not in e.attrib,
 250                             media))
 251         if not media:
 252             self.report_error('Unsupported DRM')
 253         return media
 254
 255     def _get_bootstrap_from_url(self, bootstrap_url):
 256         bootstrap = self.ydl.urlopen(bootstrap_url).read()
 257         return read_bootstrap_info(bootstrap)
 258
 259     def _update_live_fragments(self, bootstrap_url, latest_fragment):
 260         fragments_list = []
 261         retries = 30
 262         while (not fragments_list) and (retries > 0):
 263             boot_info = self._get_bootstrap_from_url(bootstrap_url)
 264             fragments_list = build_fragments_list(boot_info)
 265             fragments_list = [f for f in fragments_list if f[1] > latest_fragment]
 266             if not fragments_list:
 267                 # Retry after a while
 268                 time.sleep(5.0)
 269                 retries -= 1
 270
 271         if not fragments_list:
 272             self.report_error('Failed to update fragments')
 273
 274         return fragments_list
 275
 276     def _parse_bootstrap_node(self, node, base_url):
 277         if node.text is None:
 278             bootstrap_url = compat_urlparse.urljoin(
 279                 base_url, node.attrib['url'])
 280             boot_info = self._get_bootstrap_from_url(bootstrap_url)
 281         else:
 282             bootstrap_url = None
 283             bootstrap = base64.b64decode(node.text)
 284             boot_info = read_bootstrap_info(bootstrap)
 285         return (boot_info, bootstrap_url)
 286
 287     def real_download(self, filename, info_dict):
 288         man_url = info_dict['url']
 289         requested_bitrate = info_dict.get('tbr')
 290         self.to_screen('[download] Downloading f4m manifest')
 291         manifest = self.ydl.urlopen(man_url).read()
 292
 293         doc = etree.fromstring(manifest)
 294         formats = [(int(f.attrib.get('bitrate', -1)), f)
 295                    for f in self._get_unencrypted_media(doc)]
 296         if requested_bitrate is None:
 297             # get the best format
 298             formats = sorted(formats, key=lambda f: f[0])
 299             rate, media = formats[-1]
 300         else:
 301             rate, media = list(filter(
 302                 lambda f: int(f[0]) == requested_bitrate, formats))[0]
 303
 304         base_url = compat_urlparse.urljoin(man_url, media.attrib['url'])
 305         bootstrap_node = doc.find(_add_ns('bootstrapInfo'))
 306         boot_info, bootstrap_url = self._parse_bootstrap_node(bootstrap_node, base_url)
 307         live = boot_info['live']
 308         metadata_node = media.find(_add_ns('metadata'))
 309         if metadata_node is not None:
 310             metadata = base64.b64decode(metadata_node.text)
 311         else:
 312             metadata = None
 313
 314         fragments_list = build_fragments_list(boot_info)
 315         if self.params.get('test', False):
 316             # We only download the first fragment
 317             fragments_list = fragments_list[:1]
 318         total_frags = len(fragments_list)
 319         # For some akamai manifests we'll need to add a query to the fragment url
 320         akamai_pv = xpath_text(doc, _add_ns('pv-2.0'))
 321
 322         self.report_destination(filename)
 323         http_dl = HttpQuietDownloader(
 324             self.ydl,
 325             {
 326                 'continuedl': True,
 327                 'quiet': True,
 328                 'noprogress': True,
 329                 'ratelimit': self.params.get('ratelimit', None),
 330                 'test': self.params.get('test', False),
 331             }
 332         )
 333         tmpfilename = self.temp_name(filename)
 334         (dest_stream, tmpfilename) = sanitize_open(tmpfilename, 'wb')
 335
 336         write_flv_header(dest_stream)
 337         if not live:
 338             write_metadata_tag(dest_stream, metadata)
 339
 340         # This dict stores the download progress, it's updated by the progress
 341         # hook
 342         state = {
 343             'status': 'downloading',
 344             'downloaded_bytes': 0,
 345             'frag_index': 0,
 346             'frag_count': total_frags,
 347             'filename': filename,
 348             'tmpfilename': tmpfilename,
 349         }
 350         start = time.time()
 351
 352         def frag_progress_hook(s):
 353             if s['status'] not in ('downloading', 'finished'):
 354                 return
 355
 356             frag_total_bytes = s.get('total_bytes', 0)
 357             if s['status'] == 'finished':
 358                 state['downloaded_bytes'] += frag_total_bytes
 359                 state['frag_index'] += 1
 360
 361             estimated_size = (
 362                 (state['downloaded_bytes'] + frag_total_bytes) /
 363                 (state['frag_index'] + 1) * total_frags)
 364             time_now = time.time()
 365             state['total_bytes_estimate'] = estimated_size
 366             state['elapsed'] = time_now - start
 367
 368             if s['status'] == 'finished':
 369                 progress = self.calc_percent(state['frag_index'], total_frags)
 370             else:
 371                 frag_downloaded_bytes = s['downloaded_bytes']
 372                 frag_progress = self.calc_percent(frag_downloaded_bytes,
 373                                                   frag_total_bytes)
 374                 progress = self.calc_percent(state['frag_index'], total_frags)
 375                 progress += frag_progress / float(total_frags)
 376
 377                 state['eta'] = self.calc_eta(
 378                     start, time_now, estimated_size, state['downloaded_bytes'] + frag_downloaded_bytes)
 379                 state['speed'] = s.get('speed')
 380             self._hook_progress(state)
 381
 382         http_dl.add_progress_hook(frag_progress_hook)
 383
 384         frags_filenames = []
 385         while fragments_list:
 386             seg_i, frag_i = fragments_list.pop(0)
 387             name = 'Seg%d-Frag%d' % (seg_i, frag_i)
 388             url = base_url + name
 389             if akamai_pv:
 390                 url += '?' + akamai_pv.strip(';')
 391             frag_filename = '%s-%s' % (tmpfilename, name)
 392             success = http_dl.download(frag_filename, {'url': url})
 393             if not success:
 394                 return False
 395             with open(frag_filename, 'rb') as down:
 396                 down_data = down.read()
 397                 reader = FlvReader(down_data)
 398                 while True:
 399                     _, box_type, box_data = reader.read_box_info()
 400                     if box_type == b'mdat':
 401                         dest_stream.write(box_data)
 402                         break
 403             frags_filenames.append(frag_filename)
 404
 405             if not fragments_list and live and bootstrap_url:
 406                 fragments_list = self._update_live_fragments(bootstrap_url, frag_i)
 407                 self.to_screen('Updated available fragments: %d' % len(fragments_list))
 408
 409         dest_stream.close()
 410
 411         elapsed = time.time() - start
 412         self.try_rename(tmpfilename, filename)
 413         for frag_file in frags_filenames:
 414             os.remove(frag_file)
 415
 416         fsize = os.path.getsize(encodeFilename(filename))
 417         self._hook_progress({
 418             'downloaded_bytes': fsize,
 419             'total_bytes': fsize,
 420             'filename': filename,
 421             'status': 'finished',
 422             'elapsed': elapsed,
 423         })
 424
 425         return True