Raphaël G. Git Repositories - youtubedl/blob - youtube_dl/downloader/f4m.py

   1 from __future__ import 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 ..utils import (
  13     struct_pack,
  14     struct_unpack,
  15     compat_urlparse,
  16     format_bytes,
  17     encodeFilename,
  18     sanitize_open,
  19     xpath_text,
  20 )
  21
  22
  23 class FlvReader(io.BytesIO):
  24     """
  25     Reader for Flv files
  26     The file format is documented in https://www.adobe.com/devnet/f4v.html
  27     """
  28
  29     # Utility functions for reading numbers and strings
  30     def read_unsigned_long_long(self):
  31         return struct_unpack('!Q', self.read(8))[0]
  32
  33     def read_unsigned_int(self):
  34         return struct_unpack('!I', self.read(4))[0]
  35
  36     def read_unsigned_char(self):
  37         return struct_unpack('!B', self.read(1))[0]
  38
  39     def read_string(self):
  40         res = b''
  41         while True:
  42             char = self.read(1)
  43             if char == b'\x00':
  44                 break
  45             res += char
  46         return res
  47
  48     def read_box_info(self):
  49         """
  50         Read a box and return the info as a tuple: (box_size, box_type, box_data)
  51         """
  52         real_size = size = self.read_unsigned_int()
  53         box_type = self.read(4)
  54         header_end = 8
  55         if size == 1:
  56             real_size = self.read_unsigned_long_long()
  57             header_end = 16
  58         return real_size, box_type, self.read(real_size-header_end)
  59
  60     def read_asrt(self):
  61         # version
  62         self.read_unsigned_char()
  63         # flags
  64         self.read(3)
  65         quality_entry_count = self.read_unsigned_char()
  66         # QualityEntryCount
  67         for i in range(quality_entry_count):
  68             self.read_string()
  69
  70         segment_run_count = self.read_unsigned_int()
  71         segments = []
  72         for i in range(segment_run_count):
  73             first_segment = self.read_unsigned_int()
  74             fragments_per_segment = self.read_unsigned_int()
  75             segments.append((first_segment, fragments_per_segment))
  76
  77         return {
  78             'segment_run': segments,
  79         }
  80
  81     def read_afrt(self):
  82         # version
  83         self.read_unsigned_char()
  84         # flags
  85         self.read(3)
  86         # time scale
  87         self.read_unsigned_int()
  88
  89         quality_entry_count = self.read_unsigned_char()
  90         # QualitySegmentUrlModifiers
  91         for i in range(quality_entry_count):
  92             self.read_string()
  93
  94         fragments_count = self.read_unsigned_int()
  95         fragments = []
  96         for i in range(fragments_count):
  97             first = self.read_unsigned_int()
  98             first_ts = self.read_unsigned_long_long()
  99             duration = self.read_unsigned_int()
 100             if duration == 0:
 101                 discontinuity_indicator = self.read_unsigned_char()
 102             else:
 103                 discontinuity_indicator = None
 104             fragments.append({
 105                 'first': first,
 106                 'ts': first_ts,
 107                 'duration': duration,
 108                 'discontinuity_indicator': discontinuity_indicator,
 109             })
 110
 111         return {
 112             'fragments': fragments,
 113         }
 114
 115     def read_abst(self):
 116         # version
 117         self.read_unsigned_char()
 118         # flags
 119         self.read(3)
 120
 121         self.read_unsigned_int()  # BootstrapinfoVersion
 122         # Profile,Live,Update,Reserved
 123         self.read(1)
 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         }
 163
 164     def read_bootstrap_info(self):
 165         total_size, box_type, box_data = self.read_box_info()
 166         assert box_type == b'abst'
 167         return FlvReader(box_data).read_abst()
 168
 169
 170 def read_bootstrap_info(bootstrap_bytes):
 171     return FlvReader(bootstrap_bytes).read_bootstrap_info()
 172
 173
 174 def build_fragments_list(boot_info):
 175     """ Return a list of (segment, fragment) for each fragment in the video """
 176     res = []
 177     segment_run_table = boot_info['segments'][0]
 178     # I've only found videos with one segment
 179     segment_run_entry = segment_run_table['segment_run'][0]
 180     n_frags = segment_run_entry[1]
 181     fragment_run_entry_table = boot_info['fragments'][0]['fragments']
 182     first_frag_number = fragment_run_entry_table[0]['first']
 183     for (i, frag_number) in zip(range(1, n_frags+1), itertools.count(first_frag_number)):
 184         res.append((1, frag_number))
 185     return res
 186
 187
 188 def write_flv_header(stream, metadata):
 189     """Writes the FLV header and the metadata to stream"""
 190     # FLV header
 191     stream.write(b'FLV\x01')
 192     stream.write(b'\x05')
 193     stream.write(b'\x00\x00\x00\x09')
 194     # FLV File body
 195     stream.write(b'\x00\x00\x00\x00')
 196     # FLVTAG
 197     # Script data
 198     stream.write(b'\x12')
 199     # Size of the metadata with 3 bytes
 200     stream.write(struct_pack('!L', len(metadata))[1:])
 201     stream.write(b'\x00\x00\x00\x00\x00\x00\x00')
 202     stream.write(metadata)
 203     # Magic numbers extracted from the output files produced by AdobeHDS.php
 204     #(https://github.com/K-S-V/Scripts)
 205     stream.write(b'\x00\x00\x01\x73')
 206
 207
 208 def _add_ns(prop):
 209     return '{http://ns.adobe.com/f4m/1.0}%s' % prop
 210
 211
 212 class HttpQuietDownloader(HttpFD):
 213     def to_screen(self, *args, **kargs):
 214         pass
 215
 216
 217 class F4mFD(FileDownloader):
 218     """
 219     A downloader for f4m manifests or AdobeHDS.
 220     """
 221
 222     def real_download(self, filename, info_dict):
 223         man_url = info_dict['url']
 224         requested_bitrate = info_dict.get('tbr')
 225         self.to_screen('[download] Downloading f4m manifest')
 226         manifest = self.ydl.urlopen(man_url).read()
 227         self.report_destination(filename)
 228         http_dl = HttpQuietDownloader(self.ydl,
 229             {
 230                 'continuedl': True,
 231                 'quiet': True,
 232                 'noprogress': True,
 233                 'test': self.params.get('test', False),
 234             })
 235
 236         doc = etree.fromstring(manifest)
 237         formats = [(int(f.attrib.get('bitrate', -1)), f) for f in doc.findall(_add_ns('media'))]
 238         if requested_bitrate is None:
 239             # get the best format
 240             formats = sorted(formats, key=lambda f: f[0])
 241             rate, media = formats[-1]
 242         else:
 243             rate, media = list(filter(
 244                 lambda f: int(f[0]) == requested_bitrate, formats))[0]
 245
 246         base_url = compat_urlparse.urljoin(man_url, media.attrib['url'])
 247         bootstrap_node = doc.find(_add_ns('bootstrapInfo'))
 248         if bootstrap_node.text is None:
 249             bootstrap_url = compat_urlparse.urljoin(
 250                 base_url, bootstrap_node.attrib['url'])
 251             bootstrap = self.ydl.urlopen(bootstrap_url).read()
 252         else:
 253             bootstrap = base64.b64decode(bootstrap_node.text)
 254         metadata = base64.b64decode(media.find(_add_ns('metadata')).text)
 255         boot_info = read_bootstrap_info(bootstrap)
 256
 257         fragments_list = build_fragments_list(boot_info)
 258         if self.params.get('test', False):
 259             # We only download the first fragment
 260             fragments_list = fragments_list[:1]
 261         total_frags = len(fragments_list)
 262         # For some akamai manifests we'll need to add a query to the fragment url
 263         akamai_pv = xpath_text(doc, _add_ns('pv-2.0'))
 264
 265         tmpfilename = self.temp_name(filename)
 266         (dest_stream, tmpfilename) = sanitize_open(tmpfilename, 'wb')
 267         write_flv_header(dest_stream, metadata)
 268
 269         # This dict stores the download progress, it's updated by the progress
 270         # hook
 271         state = {
 272             'downloaded_bytes': 0,
 273             'frag_counter': 0,
 274         }
 275         start = time.time()
 276
 277         def frag_progress_hook(status):
 278             frag_total_bytes = status.get('total_bytes', 0)
 279             estimated_size = (state['downloaded_bytes'] +
 280                 (total_frags - state['frag_counter']) * frag_total_bytes)
 281             if status['status'] == 'finished':
 282                 state['downloaded_bytes'] += frag_total_bytes
 283                 state['frag_counter'] += 1
 284                 progress = self.calc_percent(state['frag_counter'], total_frags)
 285                 byte_counter = state['downloaded_bytes']
 286             else:
 287                 frag_downloaded_bytes = status['downloaded_bytes']
 288                 byte_counter = state['downloaded_bytes'] + frag_downloaded_bytes
 289                 frag_progress = self.calc_percent(frag_downloaded_bytes,
 290                     frag_total_bytes)
 291                 progress = self.calc_percent(state['frag_counter'], total_frags)
 292                 progress += frag_progress / float(total_frags)
 293
 294             eta = self.calc_eta(start, time.time(), estimated_size, byte_counter)
 295             self.report_progress(progress, format_bytes(estimated_size),
 296                 status.get('speed'), eta)
 297         http_dl.add_progress_hook(frag_progress_hook)
 298
 299         frags_filenames = []
 300         for (seg_i, frag_i) in fragments_list:
 301             name = 'Seg%d-Frag%d' % (seg_i, frag_i)
 302             url = base_url + name
 303             if akamai_pv:
 304                 url += '?' + akamai_pv.strip(';')
 305             frag_filename = '%s-%s' % (tmpfilename, name)
 306             success = http_dl.download(frag_filename, {'url': url})
 307             if not success:
 308                 return False
 309             with open(frag_filename, 'rb') as down:
 310                 down_data = down.read()
 311                 reader = FlvReader(down_data)
 312                 while True:
 313                     _, box_type, box_data = reader.read_box_info()
 314                     if box_type == b'mdat':
 315                         dest_stream.write(box_data)
 316                         break
 317             frags_filenames.append(frag_filename)
 318
 319         dest_stream.close()
 320         self.report_finish(format_bytes(state['downloaded_bytes']), time.time() - start)
 321
 322         self.try_rename(tmpfilename, filename)
 323         for frag_file in frags_filenames:
 324             os.remove(frag_file)
 325
 326         fsize = os.path.getsize(encodeFilename(filename))
 327         self._hook_progress({
 328             'downloaded_bytes': fsize,
 329             'total_bytes': fsize,
 330             'filename': filename,
 331             'status': 'finished',
 332         })
 333
 334         return True