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