]> Raphaƫl G. Git Repositories - youtubedl/blob - youtube_dl/downloader/http.py
Start new release.
[youtubedl] / youtube_dl / downloader / http.py
1 from __future__ import unicode_literals
2
3 import errno
4 import os
5 import socket
6 import time
7 import random
8 import re
9
10 from .common import FileDownloader
11 from ..compat import (
12 compat_str,
13 compat_urllib_error,
14 )
15 from ..utils import (
16 ContentTooShortError,
17 encodeFilename,
18 int_or_none,
19 sanitize_open,
20 sanitized_Request,
21 write_xattr,
22 XAttrMetadataError,
23 XAttrUnavailableError,
24 )
25
26
27 class HttpFD(FileDownloader):
28 def real_download(self, filename, info_dict):
29 url = info_dict['url']
30
31 class DownloadContext(dict):
32 __getattr__ = dict.get
33 __setattr__ = dict.__setitem__
34 __delattr__ = dict.__delitem__
35
36 ctx = DownloadContext()
37 ctx.filename = filename
38 ctx.tmpfilename = self.temp_name(filename)
39 ctx.stream = None
40
41 # Do not include the Accept-Encoding header
42 headers = {'Youtubedl-no-compression': 'True'}
43 add_headers = info_dict.get('http_headers')
44 if add_headers:
45 headers.update(add_headers)
46
47 is_test = self.params.get('test', False)
48 chunk_size = self._TEST_FILE_SIZE if is_test else (
49 info_dict.get('downloader_options', {}).get('http_chunk_size')
50 or self.params.get('http_chunk_size') or 0)
51
52 ctx.open_mode = 'wb'
53 ctx.resume_len = 0
54 ctx.data_len = None
55 ctx.block_size = self.params.get('buffersize', 1024)
56 ctx.start_time = time.time()
57 ctx.chunk_size = None
58
59 if self.params.get('continuedl', True):
60 # Establish possible resume length
61 if os.path.isfile(encodeFilename(ctx.tmpfilename)):
62 ctx.resume_len = os.path.getsize(
63 encodeFilename(ctx.tmpfilename))
64
65 ctx.is_resume = ctx.resume_len > 0
66
67 count = 0
68 retries = self.params.get('retries', 0)
69
70 class SucceedDownload(Exception):
71 pass
72
73 class RetryDownload(Exception):
74 def __init__(self, source_error):
75 self.source_error = source_error
76
77 class NextFragment(Exception):
78 pass
79
80 def set_range(req, start, end):
81 range_header = 'bytes=%d-' % start
82 if end:
83 range_header += compat_str(end)
84 req.add_header('Range', range_header)
85
86 def establish_connection():
87 ctx.chunk_size = (random.randint(int(chunk_size * 0.95), chunk_size)
88 if not is_test and chunk_size else chunk_size)
89 if ctx.resume_len > 0:
90 range_start = ctx.resume_len
91 if ctx.is_resume:
92 self.report_resuming_byte(ctx.resume_len)
93 ctx.open_mode = 'ab'
94 elif ctx.chunk_size > 0:
95 range_start = 0
96 else:
97 range_start = None
98 ctx.is_resume = False
99 range_end = range_start + ctx.chunk_size - 1 if ctx.chunk_size else None
100 if range_end and ctx.data_len is not None and range_end >= ctx.data_len:
101 range_end = ctx.data_len - 1
102 has_range = range_start is not None
103 ctx.has_range = has_range
104 request = sanitized_Request(url, None, headers)
105 if has_range:
106 set_range(request, range_start, range_end)
107 # Establish connection
108 try:
109 ctx.data = self.ydl.urlopen(request)
110 # When trying to resume, Content-Range HTTP header of response has to be checked
111 # to match the value of requested Range HTTP header. This is due to a webservers
112 # that don't support resuming and serve a whole file with no Content-Range
113 # set in response despite of requested Range (see
114 # https://github.com/ytdl-org/youtube-dl/issues/6057#issuecomment-126129799)
115 if has_range:
116 content_range = ctx.data.headers.get('Content-Range')
117 if content_range:
118 content_range_m = re.search(r'bytes (\d+)-(\d+)?(?:/(\d+))?', content_range)
119 # Content-Range is present and matches requested Range, resume is possible
120 if content_range_m:
121 if range_start == int(content_range_m.group(1)):
122 content_range_end = int_or_none(content_range_m.group(2))
123 content_len = int_or_none(content_range_m.group(3))
124 accept_content_len = (
125 # Non-chunked download
126 not ctx.chunk_size
127 # Chunked download and requested piece or
128 # its part is promised to be served
129 or content_range_end == range_end
130 or content_len < range_end)
131 if accept_content_len:
132 ctx.data_len = content_len
133 return
134 # Content-Range is either not present or invalid. Assuming remote webserver is
135 # trying to send the whole file, resume is not possible, so wiping the local file
136 # and performing entire redownload
137 self.report_unable_to_resume()
138 ctx.resume_len = 0
139 ctx.open_mode = 'wb'
140 ctx.data_len = int_or_none(ctx.data.info().get('Content-length', None))
141 return
142 except (compat_urllib_error.HTTPError, ) as err:
143 if err.code == 416:
144 # Unable to resume (requested range not satisfiable)
145 try:
146 # Open the connection again without the range header
147 ctx.data = self.ydl.urlopen(
148 sanitized_Request(url, None, headers))
149 content_length = ctx.data.info()['Content-Length']
150 except (compat_urllib_error.HTTPError, ) as err:
151 if err.code < 500 or err.code >= 600:
152 raise
153 else:
154 # Examine the reported length
155 if (content_length is not None
156 and (ctx.resume_len - 100 < int(content_length) < ctx.resume_len + 100)):
157 # The file had already been fully downloaded.
158 # Explanation to the above condition: in issue #175 it was revealed that
159 # YouTube sometimes adds or removes a few bytes from the end of the file,
160 # changing the file size slightly and causing problems for some users. So
161 # I decided to implement a suggested change and consider the file
162 # completely downloaded if the file size differs less than 100 bytes from
163 # the one in the hard drive.
164 self.report_file_already_downloaded(ctx.filename)
165 self.try_rename(ctx.tmpfilename, ctx.filename)
166 self._hook_progress({
167 'filename': ctx.filename,
168 'status': 'finished',
169 'downloaded_bytes': ctx.resume_len,
170 'total_bytes': ctx.resume_len,
171 })
172 raise SucceedDownload()
173 else:
174 # The length does not match, we start the download over
175 self.report_unable_to_resume()
176 ctx.resume_len = 0
177 ctx.open_mode = 'wb'
178 return
179 elif err.code < 500 or err.code >= 600:
180 # Unexpected HTTP error
181 raise
182 raise RetryDownload(err)
183 except socket.error as err:
184 if err.errno != errno.ECONNRESET:
185 # Connection reset is no problem, just retry
186 raise
187 raise RetryDownload(err)
188
189 def download():
190 data_len = ctx.data.info().get('Content-length', None)
191
192 # Range HTTP header may be ignored/unsupported by a webserver
193 # (e.g. extractor/scivee.py, extractor/bambuser.py).
194 # However, for a test we still would like to download just a piece of a file.
195 # To achieve this we limit data_len to _TEST_FILE_SIZE and manually control
196 # block size when downloading a file.
197 if is_test and (data_len is None or int(data_len) > self._TEST_FILE_SIZE):
198 data_len = self._TEST_FILE_SIZE
199
200 if data_len is not None:
201 data_len = int(data_len) + ctx.resume_len
202 min_data_len = self.params.get('min_filesize')
203 max_data_len = self.params.get('max_filesize')
204 if min_data_len is not None and data_len < min_data_len:
205 self.to_screen('\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len, min_data_len))
206 return False
207 if max_data_len is not None and data_len > max_data_len:
208 self.to_screen('\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len, max_data_len))
209 return False
210
211 byte_counter = 0 + ctx.resume_len
212 block_size = ctx.block_size
213 start = time.time()
214
215 # measure time over whole while-loop, so slow_down() and best_block_size() work together properly
216 now = None # needed for slow_down() in the first loop run
217 before = start # start measuring
218
219 def retry(e):
220 to_stdout = ctx.tmpfilename == '-'
221 if not to_stdout:
222 ctx.stream.close()
223 ctx.stream = None
224 ctx.resume_len = byte_counter if to_stdout else os.path.getsize(encodeFilename(ctx.tmpfilename))
225 raise RetryDownload(e)
226
227 while True:
228 try:
229 # Download and write
230 data_block = ctx.data.read(block_size if not is_test else min(block_size, data_len - byte_counter))
231 # socket.timeout is a subclass of socket.error but may not have
232 # errno set
233 except socket.timeout as e:
234 retry(e)
235 except socket.error as e:
236 if e.errno not in (errno.ECONNRESET, errno.ETIMEDOUT):
237 raise
238 retry(e)
239
240 byte_counter += len(data_block)
241
242 # exit loop when download is finished
243 if len(data_block) == 0:
244 break
245
246 # Open destination file just in time
247 if ctx.stream is None:
248 try:
249 ctx.stream, ctx.tmpfilename = sanitize_open(
250 ctx.tmpfilename, ctx.open_mode)
251 assert ctx.stream is not None
252 ctx.filename = self.undo_temp_name(ctx.tmpfilename)
253 self.report_destination(ctx.filename)
254 except (OSError, IOError) as err:
255 self.report_error('unable to open for writing: %s' % str(err))
256 return False
257
258 if self.params.get('xattr_set_filesize', False) and data_len is not None:
259 try:
260 write_xattr(ctx.tmpfilename, 'user.ytdl.filesize', str(data_len).encode('utf-8'))
261 except (XAttrUnavailableError, XAttrMetadataError) as err:
262 self.report_error('unable to set filesize xattr: %s' % str(err))
263
264 try:
265 ctx.stream.write(data_block)
266 except (IOError, OSError) as err:
267 self.to_stderr('\n')
268 self.report_error('unable to write data: %s' % str(err))
269 return False
270
271 # Apply rate limit
272 self.slow_down(start, now, byte_counter - ctx.resume_len)
273
274 # end measuring of one loop run
275 now = time.time()
276 after = now
277
278 # Adjust block size
279 if not self.params.get('noresizebuffer', False):
280 block_size = self.best_block_size(after - before, len(data_block))
281
282 before = after
283
284 # Progress message
285 speed = self.calc_speed(start, now, byte_counter - ctx.resume_len)
286 if ctx.data_len is None:
287 eta = None
288 else:
289 eta = self.calc_eta(start, time.time(), ctx.data_len - ctx.resume_len, byte_counter - ctx.resume_len)
290
291 self._hook_progress({
292 'status': 'downloading',
293 'downloaded_bytes': byte_counter,
294 'total_bytes': ctx.data_len,
295 'tmpfilename': ctx.tmpfilename,
296 'filename': ctx.filename,
297 'eta': eta,
298 'speed': speed,
299 'elapsed': now - ctx.start_time,
300 })
301
302 if is_test and byte_counter == data_len:
303 break
304
305 if not is_test and ctx.chunk_size and ctx.data_len is not None and byte_counter < ctx.data_len:
306 ctx.resume_len = byte_counter
307 # ctx.block_size = block_size
308 raise NextFragment()
309
310 if ctx.stream is None:
311 self.to_stderr('\n')
312 self.report_error('Did not get any data blocks')
313 return False
314 if ctx.tmpfilename != '-':
315 ctx.stream.close()
316
317 if data_len is not None and byte_counter != data_len:
318 err = ContentTooShortError(byte_counter, int(data_len))
319 if count <= retries:
320 retry(err)
321 raise err
322
323 self.try_rename(ctx.tmpfilename, ctx.filename)
324
325 # Update file modification time
326 if self.params.get('updatetime', True):
327 info_dict['filetime'] = self.try_utime(ctx.filename, ctx.data.info().get('last-modified', None))
328
329 self._hook_progress({
330 'downloaded_bytes': byte_counter,
331 'total_bytes': byte_counter,
332 'filename': ctx.filename,
333 'status': 'finished',
334 'elapsed': time.time() - ctx.start_time,
335 })
336
337 return True
338
339 while count <= retries:
340 try:
341 establish_connection()
342 return download()
343 except RetryDownload as e:
344 count += 1
345 if count <= retries:
346 self.report_retry(e.source_error, count, retries)
347 continue
348 except NextFragment:
349 continue
350 except SucceedDownload:
351 return True
352
353 self.report_error('giving up after %s retries' % retries)
354 return False