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 )
  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         self.read(1)
 123         # time scale
 124         self.read_unsigned_int()
 125         # CurrentMediaTime
 126         self.read_unsigned_long_long()
 127         # SmpteTimeCodeOffset
 128         self.read_unsigned_long_long()
 129
 130         self.read_string()  # MovieIdentifier
 131         server_count = self.read_unsigned_char()
 132         # ServerEntryTable
 133         for i in range(server_count):
 134             self.read_string()
 135         quality_count = self.read_unsigned_char()
 136         # QualityEntryTable
 137         for i in range(quality_count):
 138             self.read_string()
 139         # DrmData
 140         self.read_string()
 141         # MetaData
 142         self.read_string()
 143
 144         segments_count = self.read_unsigned_char()
 145         segments = []
 146         for i in range(segments_count):
 147             box_size, box_type, box_data = self.read_box_info()
 148             assert box_type == b'asrt'
 149             segment = FlvReader(box_data).read_asrt()
 150             segments.append(segment)
 151         fragments_run_count = self.read_unsigned_char()
 152         fragments = []
 153         for i in range(fragments_run_count):
 154             box_size, box_type, box_data = self.read_box_info()
 155             assert box_type == b'afrt'
 156             fragments.append(FlvReader(box_data).read_afrt())
 157
 158         return {
 159             'segments': segments,
 160             'fragments': fragments,
 161         }
 162
 163     def read_bootstrap_info(self):
 164         total_size, box_type, box_data = self.read_box_info()
 165         assert box_type == b'abst'
 166         return FlvReader(box_data).read_abst()
 167
 168
 169 def read_bootstrap_info(bootstrap_bytes):
 170     return FlvReader(bootstrap_bytes).read_bootstrap_info()
 171
 172
 173 def build_fragments_list(boot_info):
 174     """ Return a list of (segment, fragment) for each fragment in the video """
 175     res = []
 176     segment_run_table = boot_info['segments'][0]
 177     # I've only found videos with one segment
 178     segment_run_entry = segment_run_table['segment_run'][0]
 179     n_frags = segment_run_entry[1]
 180     fragment_run_entry_table = boot_info['fragments'][0]['fragments']
 181     first_frag_number = fragment_run_entry_table[0]['first']
 182     for (i, frag_number) in zip(range(1, n_frags+1), itertools.count(first_frag_number)):
 183         res.append((1, frag_number))
 184     return res
 185
 186
 187 def write_flv_header(stream, metadata):
 188     """Writes the FLV header and the metadata to stream"""
 189     # FLV header
 190     stream.write(b'FLV\x01')
 191     stream.write(b'\x05')
 192     stream.write(b'\x00\x00\x00\x09')
 193     # FLV File body
 194     stream.write(b'\x00\x00\x00\x00')
 195     # FLVTAG
 196     # Script data
 197     stream.write(b'\x12')
 198     # Size of the metadata with 3 bytes
 199     stream.write(struct_pack('!L', len(metadata))[1:])
 200     stream.write(b'\x00\x00\x00\x00\x00\x00\x00')
 201     stream.write(metadata)
 202     # Magic numbers extracted from the output files produced by AdobeHDS.php
 203     #(https://github.com/K-S-V/Scripts)
 204     stream.write(b'\x00\x00\x01\x73')
 205
 206
 207 def _add_ns(prop):
 208     return '{http://ns.adobe.com/f4m/1.0}%s' % prop
 209
 210
 211 class HttpQuietDownloader(HttpFD):
 212     def to_screen(self, *args, **kargs):
 213         pass
 214
 215
 216 class F4mFD(FileDownloader):
 217     """
 218     A downloader for f4m manifests or AdobeHDS.
 219     """
 220
 221     def real_download(self, filename, info_dict):
 222         man_url = info_dict['url']
 223         requested_bitrate = info_dict.get('tbr')
 224         self.to_screen('[download] Downloading f4m manifest')
 225         manifest = self.ydl.urlopen(man_url).read()
 226         self.report_destination(filename)
 227         http_dl = HttpQuietDownloader(self.ydl,
 228             {
 229                 'continuedl': True,
 230                 'quiet': True,
 231                 'noprogress': True,
 232                 'test': self.params.get('test', False),
 233             })
 234
 235         doc = etree.fromstring(manifest)
 236         formats = [(int(f.attrib.get('bitrate', -1)), f) for f in doc.findall(_add_ns('media'))]
 237         if requested_bitrate is None:
 238             # get the best format
 239             formats = sorted(formats, key=lambda f: f[0])
 240             rate, media = formats[-1]
 241         else:
 242             rate, media = list(filter(
 243                 lambda f: int(f[0]) == requested_bitrate, formats))[0]
 244
 245         base_url = compat_urlparse.urljoin(man_url, media.attrib['url'])
 246         bootstrap = base64.b64decode(doc.find(_add_ns('bootstrapInfo')).text)
 247         metadata = base64.b64decode(media.find(_add_ns('metadata')).text)
 248         boot_info = read_bootstrap_info(bootstrap)
 249         fragments_list = build_fragments_list(boot_info)
 250         if self.params.get('test', False):
 251             # We only download the first fragment
 252             fragments_list = fragments_list[:1]
 253         total_frags = len(fragments_list)
 254
 255         tmpfilename = self.temp_name(filename)
 256         (dest_stream, tmpfilename) = sanitize_open(tmpfilename, 'wb')
 257         write_flv_header(dest_stream, metadata)
 258
 259         # This dict stores the download progress, it's updated by the progress
 260         # hook
 261         state = {
 262             'downloaded_bytes': 0,
 263             'frag_counter': 0,
 264         }
 265         start = time.time()
 266
 267         def frag_progress_hook(status):
 268             frag_total_bytes = status.get('total_bytes', 0)
 269             estimated_size = (state['downloaded_bytes'] +
 270                 (total_frags - state['frag_counter']) * frag_total_bytes)
 271             if status['status'] == 'finished':
 272                 state['downloaded_bytes'] += frag_total_bytes
 273                 state['frag_counter'] += 1
 274                 progress = self.calc_percent(state['frag_counter'], total_frags)
 275                 byte_counter = state['downloaded_bytes']
 276             else:
 277                 frag_downloaded_bytes = status['downloaded_bytes']
 278                 byte_counter = state['downloaded_bytes'] + frag_downloaded_bytes
 279                 frag_progress = self.calc_percent(frag_downloaded_bytes,
 280                     frag_total_bytes)
 281                 progress = self.calc_percent(state['frag_counter'], total_frags)
 282                 progress += frag_progress / float(total_frags)
 283
 284             eta = self.calc_eta(start, time.time(), estimated_size, byte_counter)
 285             self.report_progress(progress, format_bytes(estimated_size),
 286                 status.get('speed'), eta)
 287         http_dl.add_progress_hook(frag_progress_hook)
 288
 289         frags_filenames = []
 290         for (seg_i, frag_i) in fragments_list:
 291             name = 'Seg%d-Frag%d' % (seg_i, frag_i)
 292             url = base_url + name
 293             frag_filename = '%s-%s' % (tmpfilename, name)
 294             success = http_dl.download(frag_filename, {'url': url})
 295             if not success:
 296                 return False
 297             with open(frag_filename, 'rb') as down:
 298                 down_data = down.read()
 299                 reader = FlvReader(down_data)
 300                 while True:
 301                     _, box_type, box_data = reader.read_box_info()
 302                     if box_type == b'mdat':
 303                         dest_stream.write(box_data)
 304                         break
 305             frags_filenames.append(frag_filename)
 306
 307         dest_stream.close()
 308         self.report_finish(format_bytes(state['downloaded_bytes']), time.time() - start)
 309
 310         self.try_rename(tmpfilename, filename)
 311         for frag_file in frags_filenames:
 312             os.remove(frag_file)
 313
 314         fsize = os.path.getsize(encodeFilename(filename))
 315         self._hook_progress({
 316             'downloaded_bytes': fsize,
 317             'total_bytes': fsize,
 318             'filename': filename,
 319             'status': 'finished',
 320         })
 321
 322         return True