1 from __future__
import unicode_literals
8 from .common
import FileDownloader
23 XAttrUnavailableError
,
26 RESPONSE_READ_EXCEPTIONS
= (TimeoutError
, ConnectionError
, ssl
.SSLError
, compat_http_client
.HTTPException
)
29 class HttpFD(FileDownloader
):
30 def real_download(self
, filename
, info_dict
):
31 url
= info_dict
['url']
32 request_data
= info_dict
.get('request_data', None)
34 class DownloadContext(dict):
35 __getattr__
= dict.get
36 __setattr__
= dict.__setitem
__
37 __delattr__
= dict.__delitem
__
39 ctx
= DownloadContext()
40 ctx
.filename
= filename
41 ctx
.tmpfilename
= self
.temp_name(filename
)
44 # Do not include the Accept-Encoding header
45 headers
= {'Youtubedl-no-compression': 'True'}
46 add_headers
= info_dict
.get('http_headers')
48 headers
.update(add_headers
)
50 is_test
= self
.params
.get('test', False)
51 chunk_size
= self
._TEST
_FILE
_SIZE
if is_test
else (
52 self
.params
.get('http_chunk_size')
53 or info_dict
.get('downloader_options', {}).get('http_chunk_size')
58 ctx
.block_size
= self
.params
.get('buffersize', 1024)
59 ctx
.start_time
= time
.time()
62 req_start
, req_end
, _
= parse_http_range(headers
.get('Range'))
64 if self
.params
.get('continuedl', True):
65 # Establish possible resume length
66 if os
.path
.isfile(encodeFilename(ctx
.tmpfilename
)):
67 ctx
.resume_len
= os
.path
.getsize(
68 encodeFilename(ctx
.tmpfilename
))
70 ctx
.is_resume
= ctx
.resume_len
> 0
73 retries
= self
.params
.get('retries', 0)
75 class SucceedDownload(Exception):
78 class RetryDownload(Exception):
79 def __init__(self
, source_error
):
80 self
.source_error
= source_error
82 class NextFragment(Exception):
85 def establish_connection():
86 ctx
.chunk_size
= (random
.randint(int(chunk_size
* 0.95), chunk_size
)
87 if not is_test
and chunk_size
else chunk_size
)
88 if ctx
.resume_len
> 0:
89 range_start
= ctx
.resume_len
90 if req_start
is not None:
91 # offset the beginning of Range to be within request
92 range_start
+= req_start
94 self
.report_resuming_byte(ctx
.resume_len
)
96 elif req_start
is not None:
97 range_start
= req_start
98 elif ctx
.chunk_size
> 0:
102 ctx
.is_resume
= False
105 chunk_aware_end
= range_start
+ ctx
.chunk_size
- 1
106 # we're not allowed to download outside Range
107 range_end
= chunk_aware_end
if req_end
is None else min(chunk_aware_end
, req_end
)
108 elif req_end
is not None:
109 # there's no need for chunked downloads, so download until the end of Range
114 if try_call(lambda: range_start
> range_end
):
117 raise RetryDownload(Exception(f
'Conflicting range. (start={range_start} > end={range_end})'))
119 if try_call(lambda: range_end
>= ctx
.content_len
):
120 range_end
= ctx
.content_len
- 1
122 request
= sanitized_Request(url
, request_data
, headers
)
123 has_range
= range_start
is not None
125 request
.add_header('Range', f
'bytes={int(range_start)}-{int_or_none(range_end) or ""}')
126 # Establish connection
128 ctx
.data
= self
.ydl
.urlopen(request
)
129 # When trying to resume, Content-Range HTTP header of response has to be checked
130 # to match the value of requested Range HTTP header. This is due to a webservers
131 # that don't support resuming and serve a whole file with no Content-Range
132 # set in response despite of requested Range (see
133 # https://github.com/ytdl-org/youtube-dl/issues/6057#issuecomment-126129799)
135 content_range
= ctx
.data
.headers
.get('Content-Range')
136 content_range_start
, content_range_end
, content_len
= parse_http_range(content_range
)
137 if content_range_start
is not None and range_start
== content_range_start
:
138 # Content-Range is present and matches requested Range, resume is possible
139 accept_content_len
= (
140 # Non-chunked download
142 # Chunked download and requested piece or
143 # its part is promised to be served
144 or content_range_end
== range_end
145 or content_len
< range_end
)
146 if accept_content_len
:
147 ctx
.content_len
= content_len
148 if content_len
or req_end
:
149 ctx
.data_len
= min(content_len
or req_end
, req_end
or content_len
) - (req_start
or 0)
151 # Content-Range is either not present or invalid. Assuming remote webserver is
152 # trying to send the whole file, resume is not possible, so wiping the local file
153 # and performing entire redownload
154 self
.report_unable_to_resume()
157 ctx
.data_len
= ctx
.content_len
= int_or_none(ctx
.data
.info().get('Content-length', None))
158 except (compat_urllib_error
.HTTPError
, ) as err
:
160 # Unable to resume (requested range not satisfiable)
162 # Open the connection again without the range header
163 ctx
.data
= self
.ydl
.urlopen(
164 sanitized_Request(url
, request_data
, headers
))
165 content_length
= ctx
.data
.info()['Content-Length']
166 except (compat_urllib_error
.HTTPError
, ) as err
:
167 if err
.code
< 500 or err
.code
>= 600:
170 # Examine the reported length
171 if (content_length
is not None
172 and (ctx
.resume_len
- 100 < int(content_length
) < ctx
.resume_len
+ 100)):
173 # The file had already been fully downloaded.
174 # Explanation to the above condition: in issue #175 it was revealed that
175 # YouTube sometimes adds or removes a few bytes from the end of the file,
176 # changing the file size slightly and causing problems for some users. So
177 # I decided to implement a suggested change and consider the file
178 # completely downloaded if the file size differs less than 100 bytes from
179 # the one in the hard drive.
180 self
.report_file_already_downloaded(ctx
.filename
)
181 self
.try_rename(ctx
.tmpfilename
, ctx
.filename
)
182 self
._hook
_progress
({
183 'filename': ctx
.filename
,
184 'status': 'finished',
185 'downloaded_bytes': ctx
.resume_len
,
186 'total_bytes': ctx
.resume_len
,
188 raise SucceedDownload()
190 # The length does not match, we start the download over
191 self
.report_unable_to_resume()
195 elif err
.code
< 500 or err
.code
>= 600:
196 # Unexpected HTTP error
198 raise RetryDownload(err
)
199 except compat_urllib_error
.URLError
as err
:
200 if isinstance(err
.reason
, ssl
.CertificateError
):
202 raise RetryDownload(err
)
203 # In urllib.request.AbstractHTTPHandler, the response is partially read on request.
204 # Any errors that occur during this will not be wrapped by URLError
205 except RESPONSE_READ_EXCEPTIONS
as err
:
206 raise RetryDownload(err
)
209 data_len
= ctx
.data
.info().get('Content-length', None)
211 # Range HTTP header may be ignored/unsupported by a webserver
212 # (e.g. extractor/scivee.py, extractor/bambuser.py).
213 # However, for a test we still would like to download just a piece of a file.
214 # To achieve this we limit data_len to _TEST_FILE_SIZE and manually control
215 # block size when downloading a file.
216 if is_test
and (data_len
is None or int(data_len
) > self
._TEST
_FILE
_SIZE
):
217 data_len
= self
._TEST
_FILE
_SIZE
219 if data_len
is not None:
220 data_len
= int(data_len
) + ctx
.resume_len
221 min_data_len
= self
.params
.get('min_filesize')
222 max_data_len
= self
.params
.get('max_filesize')
223 if min_data_len
is not None and data_len
< min_data_len
:
224 self
.to_screen('\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len
, min_data_len
))
226 if max_data_len
is not None and data_len
> max_data_len
:
227 self
.to_screen('\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len
, max_data_len
))
230 byte_counter
= 0 + ctx
.resume_len
231 block_size
= ctx
.block_size
234 # measure time over whole while-loop, so slow_down() and best_block_size() work together properly
235 now
= None # needed for slow_down() in the first loop run
236 before
= start
# start measuring
239 to_stdout
= ctx
.tmpfilename
== '-'
240 if ctx
.stream
is not None:
244 ctx
.resume_len
= byte_counter
if to_stdout
else os
.path
.getsize(encodeFilename(ctx
.tmpfilename
))
245 raise RetryDownload(e
)
250 data_block
= ctx
.data
.read(block_size
if not is_test
else min(block_size
, data_len
- byte_counter
))
251 except RESPONSE_READ_EXCEPTIONS
as err
:
254 byte_counter
+= len(data_block
)
256 # exit loop when download is finished
257 if len(data_block
) == 0:
260 # Open destination file just in time
261 if ctx
.stream
is None:
263 ctx
.stream
, ctx
.tmpfilename
= self
.sanitize_open(
264 ctx
.tmpfilename
, ctx
.open_mode
)
265 assert ctx
.stream
is not None
266 ctx
.filename
= self
.undo_temp_name(ctx
.tmpfilename
)
267 self
.report_destination(ctx
.filename
)
268 except (OSError, IOError) as err
:
269 self
.report_error('unable to open for writing: %s' % str(err
))
272 if self
.params
.get('xattr_set_filesize', False) and data_len
is not None:
274 write_xattr(ctx
.tmpfilename
, 'user.ytdl.filesize', str(data_len
).encode('utf-8'))
275 except (XAttrUnavailableError
, XAttrMetadataError
) as err
:
276 self
.report_error('unable to set filesize xattr: %s' % str(err
))
279 ctx
.stream
.write(data_block
)
280 except (IOError, OSError) as err
:
282 self
.report_error('unable to write data: %s' % str(err
))
286 self
.slow_down(start
, now
, byte_counter
- ctx
.resume_len
)
288 # end measuring of one loop run
293 if not self
.params
.get('noresizebuffer', False):
294 block_size
= self
.best_block_size(after
- before
, len(data_block
))
299 speed
= self
.calc_speed(start
, now
, byte_counter
- ctx
.resume_len
)
300 if ctx
.data_len
is None:
303 eta
= self
.calc_eta(start
, time
.time(), ctx
.data_len
- ctx
.resume_len
, byte_counter
- ctx
.resume_len
)
305 self
._hook
_progress
({
306 'status': 'downloading',
307 'downloaded_bytes': byte_counter
,
308 'total_bytes': ctx
.data_len
,
309 'tmpfilename': ctx
.tmpfilename
,
310 'filename': ctx
.filename
,
313 'elapsed': now
- ctx
.start_time
,
314 'ctx_id': info_dict
.get('ctx_id'),
317 if data_len
is not None and byte_counter
== data_len
:
320 if speed
and speed
< (self
.params
.get('throttledratelimit') or 0):
321 # The speed must stay below the limit for 3 seconds
322 # This prevents raising error when the speed temporarily goes down
323 if ctx
.throttle_start
is None:
324 ctx
.throttle_start
= now
325 elif now
- ctx
.throttle_start
> 3:
326 if ctx
.stream
is not None and ctx
.tmpfilename
!= '-':
328 raise ThrottledDownload()
330 ctx
.throttle_start
= None
332 if not is_test
and ctx
.chunk_size
and ctx
.content_len
is not None and byte_counter
< ctx
.content_len
:
333 ctx
.resume_len
= byte_counter
334 # ctx.block_size = block_size
337 if ctx
.stream
is None:
339 self
.report_error('Did not get any data blocks')
341 if ctx
.tmpfilename
!= '-':
344 if data_len
is not None and byte_counter
!= data_len
:
345 err
= ContentTooShortError(byte_counter
, int(data_len
))
350 self
.try_rename(ctx
.tmpfilename
, ctx
.filename
)
352 # Update file modification time
353 if self
.params
.get('updatetime', True):
354 info_dict
['filetime'] = self
.try_utime(ctx
.filename
, ctx
.data
.info().get('last-modified', None))
356 self
._hook
_progress
({
357 'downloaded_bytes': byte_counter
,
358 'total_bytes': byte_counter
,
359 'filename': ctx
.filename
,
360 'status': 'finished',
361 'elapsed': time
.time() - ctx
.start_time
,
362 'ctx_id': info_dict
.get('ctx_id'),
367 while count
<= retries
:
369 establish_connection()
371 except RetryDownload
as e
:
374 self
.report_retry(e
.source_error
, count
, retries
)
376 self
.to_screen(f
'[download] Got server HTTP error: {e.source_error}')
380 except SucceedDownload
:
383 self
.report_error('giving up after %s retries' % retries
)