1 from __future__ import unicode_literals
8 import xml.etree.ElementTree as etree
10 from .common import FileDownloader
11 from .http import HttpFD
12 from ..compat import (
25 class FlvReader(io.BytesIO):
28 The file format is documented in https://www.adobe.com/devnet/f4v.html
31 # Utility functions for reading numbers and strings
32 def read_unsigned_long_long(self):
33 return struct_unpack('!Q', self.read(8))[0]
35 def read_unsigned_int(self):
36 return struct_unpack('!I', self.read(4))[0]
38 def read_unsigned_char(self):
39 return struct_unpack('!B', self.read(1))[0]
41 def read_string(self):
50 def read_box_info(self):
52 Read a box and return the info as a tuple: (box_size, box_type, box_data)
54 real_size = size = self.read_unsigned_int()
55 box_type = self.read(4)
58 real_size = self.read_unsigned_long_long()
60 return real_size, box_type, self.read(real_size - header_end)
64 self.read_unsigned_char()
67 quality_entry_count = self.read_unsigned_char()
69 for i in range(quality_entry_count):
72 segment_run_count = self.read_unsigned_int()
74 for i in range(segment_run_count):
75 first_segment = self.read_unsigned_int()
76 fragments_per_segment = self.read_unsigned_int()
77 segments.append((first_segment, fragments_per_segment))
80 'segment_run': segments,
85 self.read_unsigned_char()
89 self.read_unsigned_int()
91 quality_entry_count = self.read_unsigned_char()
92 # QualitySegmentUrlModifiers
93 for i in range(quality_entry_count):
96 fragments_count = self.read_unsigned_int()
98 for i in range(fragments_count):
99 first = self.read_unsigned_int()
100 first_ts = self.read_unsigned_long_long()
101 duration = self.read_unsigned_int()
103 discontinuity_indicator = self.read_unsigned_char()
105 discontinuity_indicator = None
109 'duration': duration,
110 'discontinuity_indicator': discontinuity_indicator,
114 'fragments': fragments,
119 self.read_unsigned_char()
123 self.read_unsigned_int() # BootstrapinfoVersion
124 # Profile,Live,Update,Reserved
127 self.read_unsigned_int()
129 self.read_unsigned_long_long()
130 # SmpteTimeCodeOffset
131 self.read_unsigned_long_long()
133 self.read_string() # MovieIdentifier
134 server_count = self.read_unsigned_char()
136 for i in range(server_count):
138 quality_count = self.read_unsigned_char()
140 for i in range(quality_count):
147 segments_count = self.read_unsigned_char()
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()
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())
162 'segments': segments,
163 'fragments': fragments,
166 def read_bootstrap_info(self):
167 total_size, box_type, box_data = self.read_box_info()
168 assert box_type == b'abst'
169 return FlvReader(box_data).read_abst()
172 def read_bootstrap_info(bootstrap_bytes):
173 return FlvReader(bootstrap_bytes).read_bootstrap_info()
176 def build_fragments_list(boot_info):
177 """ Return a list of (segment, fragment) for each fragment in the video """
179 segment_run_table = boot_info['segments'][0]
180 # I've only found videos with one segment
181 segment_run_entry = segment_run_table['segment_run'][0]
182 n_frags = segment_run_entry[1]
183 fragment_run_entry_table = boot_info['fragments'][0]['fragments']
184 first_frag_number = fragment_run_entry_table[0]['first']
185 for (i, frag_number) in zip(range(1, n_frags + 1), itertools.count(first_frag_number)):
186 res.append((1, frag_number))
190 def write_flv_header(stream, metadata):
191 """Writes the FLV header and the metadata to stream"""
193 stream.write(b'FLV\x01')
194 stream.write(b'\x05')
195 stream.write(b'\x00\x00\x00\x09')
197 stream.write(b'\x00\x00\x00\x00')
200 stream.write(b'\x12')
201 # Size of the metadata with 3 bytes
202 stream.write(struct_pack('!L', len(metadata))[1:])
203 stream.write(b'\x00\x00\x00\x00\x00\x00\x00')
204 stream.write(metadata)
205 # Magic numbers extracted from the output files produced by AdobeHDS.php
206 # (https://github.com/K-S-V/Scripts)
207 stream.write(b'\x00\x00\x01\x73')
211 return '{http://ns.adobe.com/f4m/1.0}%s' % prop
214 class HttpQuietDownloader(HttpFD):
215 def to_screen(self, *args, **kargs):
219 class F4mFD(FileDownloader):
221 A downloader for f4m manifests or AdobeHDS.
224 def real_download(self, filename, info_dict):
225 man_url = info_dict['url']
226 requested_bitrate = info_dict.get('tbr')
227 self.to_screen('[download] Downloading f4m manifest')
228 manifest = self.ydl.urlopen(man_url).read()
229 self.report_destination(filename)
230 http_dl = HttpQuietDownloader(
236 'ratelimit': self.params.get('ratelimit', None),
237 'test': self.params.get('test', False),
241 doc = etree.fromstring(manifest)
242 formats = [(int(f.attrib.get('bitrate', -1)), f) for f in doc.findall(_add_ns('media'))]
243 if requested_bitrate is None:
244 # get the best format
245 formats = sorted(formats, key=lambda f: f[0])
246 rate, media = formats[-1]
248 rate, media = list(filter(
249 lambda f: int(f[0]) == requested_bitrate, formats))[0]
251 base_url = compat_urlparse.urljoin(man_url, media.attrib['url'])
252 bootstrap_node = doc.find(_add_ns('bootstrapInfo'))
253 if bootstrap_node.text is None:
254 bootstrap_url = compat_urlparse.urljoin(
255 base_url, bootstrap_node.attrib['url'])
256 bootstrap = self.ydl.urlopen(bootstrap_url).read()
258 bootstrap = base64.b64decode(bootstrap_node.text)
259 metadata = base64.b64decode(media.find(_add_ns('metadata')).text)
260 boot_info = read_bootstrap_info(bootstrap)
262 fragments_list = build_fragments_list(boot_info)
263 if self.params.get('test', False):
264 # We only download the first fragment
265 fragments_list = fragments_list[:1]
266 total_frags = len(fragments_list)
267 # For some akamai manifests we'll need to add a query to the fragment url
268 akamai_pv = xpath_text(doc, _add_ns('pv-2.0'))
270 tmpfilename = self.temp_name(filename)
271 (dest_stream, tmpfilename) = sanitize_open(tmpfilename, 'wb')
272 write_flv_header(dest_stream, metadata)
274 # This dict stores the download progress, it's updated by the progress
277 'downloaded_bytes': 0,
282 def frag_progress_hook(status):
283 frag_total_bytes = status.get('total_bytes', 0)
284 estimated_size = (state['downloaded_bytes'] +
285 (total_frags - state['frag_counter']) * frag_total_bytes)
286 if status['status'] == 'finished':
287 state['downloaded_bytes'] += frag_total_bytes
288 state['frag_counter'] += 1
289 progress = self.calc_percent(state['frag_counter'], total_frags)
290 byte_counter = state['downloaded_bytes']
292 frag_downloaded_bytes = status['downloaded_bytes']
293 byte_counter = state['downloaded_bytes'] + frag_downloaded_bytes
294 frag_progress = self.calc_percent(frag_downloaded_bytes,
296 progress = self.calc_percent(state['frag_counter'], total_frags)
297 progress += frag_progress / float(total_frags)
299 eta = self.calc_eta(start, time.time(), estimated_size, byte_counter)
300 self.report_progress(progress, format_bytes(estimated_size),
301 status.get('speed'), eta)
302 http_dl.add_progress_hook(frag_progress_hook)
305 for (seg_i, frag_i) in fragments_list:
306 name = 'Seg%d-Frag%d' % (seg_i, frag_i)
307 url = base_url + name
309 url += '?' + akamai_pv.strip(';')
310 frag_filename = '%s-%s' % (tmpfilename, name)
311 success = http_dl.download(frag_filename, {'url': url})
314 with open(frag_filename, 'rb') as down:
315 down_data = down.read()
316 reader = FlvReader(down_data)
318 _, box_type, box_data = reader.read_box_info()
319 if box_type == b'mdat':
320 dest_stream.write(box_data)
322 frags_filenames.append(frag_filename)
325 self.report_finish(format_bytes(state['downloaded_bytes']), time.time() - start)
327 self.try_rename(tmpfilename, filename)
328 for frag_file in frags_filenames:
331 fsize = os.path.getsize(encodeFilename(filename))
332 self._hook_progress({
333 'downloaded_bytes': fsize,
334 'total_bytes': fsize,
335 'filename': filename,
336 'status': 'finished',