Merge pull request #3927 from qrtt1/master
[ytdl] / 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 ..compat import (
13     compat_urlparse,
14 )
15 from ..utils import (
16     struct_pack,
17     struct_unpack,
18     format_bytes,
19     encodeFilename,
20     sanitize_open,
21     xpath_text,
22 )
23
24
25 class FlvReader(io.BytesIO):
26     """
27     Reader for Flv files
28     The file format is documented in https://www.adobe.com/devnet/f4v.html
29     """
30
31     # Utility functions for reading numbers and strings
32     def read_unsigned_long_long(self):
33         return struct_unpack('!Q', self.read(8))[0]
34
35     def read_unsigned_int(self):
36         return struct_unpack('!I', self.read(4))[0]
37
38     def read_unsigned_char(self):
39         return struct_unpack('!B', self.read(1))[0]
40
41     def read_string(self):
42         res = b''
43         while True:
44             char = self.read(1)
45             if char == b'\x00':
46                 break
47             res += char
48         return res
49
50     def read_box_info(self):
51         """
52         Read a box and return the info as a tuple: (box_size, box_type, box_data)
53         """
54         real_size = size = self.read_unsigned_int()
55         box_type = self.read(4)
56         header_end = 8
57         if size == 1:
58             real_size = self.read_unsigned_long_long()
59             header_end = 16
60         return real_size, box_type, self.read(real_size - header_end)
61
62     def read_asrt(self):
63         # version
64         self.read_unsigned_char()
65         # flags
66         self.read(3)
67         quality_entry_count = self.read_unsigned_char()
68         # QualityEntryCount
69         for i in range(quality_entry_count):
70             self.read_string()
71
72         segment_run_count = self.read_unsigned_int()
73         segments = []
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))
78
79         return {
80             'segment_run': segments,
81         }
82
83     def read_afrt(self):
84         # version
85         self.read_unsigned_char()
86         # flags
87         self.read(3)
88         # time scale
89         self.read_unsigned_int()
90
91         quality_entry_count = self.read_unsigned_char()
92         # QualitySegmentUrlModifiers
93         for i in range(quality_entry_count):
94             self.read_string()
95
96         fragments_count = self.read_unsigned_int()
97         fragments = []
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()
102             if duration == 0:
103                 discontinuity_indicator = self.read_unsigned_char()
104             else:
105                 discontinuity_indicator = None
106             fragments.append({
107                 'first': first,
108                 'ts': first_ts,
109                 'duration': duration,
110                 'discontinuity_indicator': discontinuity_indicator,
111             })
112
113         return {
114             'fragments': fragments,
115         }
116
117     def read_abst(self):
118         # version
119         self.read_unsigned_char()
120         # flags
121         self.read(3)
122
123         self.read_unsigned_int()  # BootstrapinfoVersion
124         # Profile,Live,Update,Reserved
125         self.read(1)
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         }
165
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()
170
171
172 def read_bootstrap_info(bootstrap_bytes):
173     return FlvReader(bootstrap_bytes).read_bootstrap_info()
174
175
176 def build_fragments_list(boot_info):
177     """ Return a list of (segment, fragment) for each fragment in the video """
178     res = []
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))
187     return res
188
189
190 def write_flv_header(stream, metadata):
191     """Writes the FLV header and the metadata to stream"""
192     # FLV header
193     stream.write(b'FLV\x01')
194     stream.write(b'\x05')
195     stream.write(b'\x00\x00\x00\x09')
196     # FLV File body
197     stream.write(b'\x00\x00\x00\x00')
198     # FLVTAG
199     # Script data
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')
208
209
210 def _add_ns(prop):
211     return '{http://ns.adobe.com/f4m/1.0}%s' % prop
212
213
214 class HttpQuietDownloader(HttpFD):
215     def to_screen(self, *args, **kargs):
216         pass
217
218
219 class F4mFD(FileDownloader):
220     """
221     A downloader for f4m manifests or AdobeHDS.
222     """
223
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(
231             self.ydl,
232             {
233                 'continuedl': True,
234                 'quiet': True,
235                 'noprogress': True,
236                 'ratelimit': self.params.get('ratelimit', None),
237                 'test': self.params.get('test', False),
238             }
239         )
240
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]
247         else:
248             rate, media = list(filter(
249                 lambda f: int(f[0]) == requested_bitrate, formats))[0]
250
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()
257         else:
258             bootstrap = base64.b64decode(bootstrap_node.text)
259         metadata = base64.b64decode(media.find(_add_ns('metadata')).text)
260         boot_info = read_bootstrap_info(bootstrap)
261
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'))
269
270         tmpfilename = self.temp_name(filename)
271         (dest_stream, tmpfilename) = sanitize_open(tmpfilename, 'wb')
272         write_flv_header(dest_stream, metadata)
273
274         # This dict stores the download progress, it's updated by the progress
275         # hook
276         state = {
277             'downloaded_bytes': 0,
278             'frag_counter': 0,
279         }
280         start = time.time()
281
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']
291             else:
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,
295                                                   frag_total_bytes)
296                 progress = self.calc_percent(state['frag_counter'], total_frags)
297                 progress += frag_progress / float(total_frags)
298
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)
303
304         frags_filenames = []
305         for (seg_i, frag_i) in fragments_list:
306             name = 'Seg%d-Frag%d' % (seg_i, frag_i)
307             url = base_url + name
308             if akamai_pv:
309                 url += '?' + akamai_pv.strip(';')
310             frag_filename = '%s-%s' % (tmpfilename, name)
311             success = http_dl.download(frag_filename, {'url': url})
312             if not success:
313                 return False
314             with open(frag_filename, 'rb') as down:
315                 down_data = down.read()
316                 reader = FlvReader(down_data)
317                 while True:
318                     _, box_type, box_data = reader.read_box_info()
319                     if box_type == b'mdat':
320                         dest_stream.write(box_data)
321                         break
322             frags_filenames.append(frag_filename)
323
324         dest_stream.close()
325         self.report_finish(format_bytes(state['downloaded_bytes']), time.time() - start)
326
327         self.try_rename(tmpfilename, filename)
328         for frag_file in frags_filenames:
329             os.remove(frag_file)
330
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',
337         })
338
339         return True