7 from .common
import FileDownloader
8 from ..compat
import compat_http_client
, compat_urllib_error
13 XAttrUnavailableError
,
22 RESPONSE_READ_EXCEPTIONS
= (
24 socket
.timeout
, # compat: py < 3.10
27 compat_http_client
.HTTPException
31 class HttpFD(FileDownloader
):
32 def real_download(self
, filename
, info_dict
):
33 url
= info_dict
['url']
34 request_data
= info_dict
.get('request_data', None)
36 class DownloadContext(dict):
37 __getattr__
= dict.get
38 __setattr__
= dict.__setitem
__
39 __delattr__
= dict.__delitem
__
41 ctx
= DownloadContext()
42 ctx
.filename
= filename
43 ctx
.tmpfilename
= self
.temp_name(filename
)
46 # Do not include the Accept-Encoding header
47 headers
= {'Youtubedl-no-compression': 'True'}
48 add_headers
= info_dict
.get('http_headers')
50 headers
.update(add_headers
)
52 is_test
= self
.params
.get('test', False)
53 chunk_size
= self
._TEST
_FILE
_SIZE
if is_test
else (
54 self
.params
.get('http_chunk_size')
55 or info_dict
.get('downloader_options', {}).get('http_chunk_size')
60 ctx
.block_size
= self
.params
.get('buffersize', 1024)
61 ctx
.start_time
= time
.time()
64 req_start
, req_end
, _
= parse_http_range(headers
.get('Range'))
66 if self
.params
.get('continuedl', True):
67 # Establish possible resume length
68 if os
.path
.isfile(encodeFilename(ctx
.tmpfilename
)):
69 ctx
.resume_len
= os
.path
.getsize(
70 encodeFilename(ctx
.tmpfilename
))
72 ctx
.is_resume
= ctx
.resume_len
> 0
75 retries
= self
.params
.get('retries', 0)
77 class SucceedDownload(Exception):
80 class RetryDownload(Exception):
81 def __init__(self
, source_error
):
82 self
.source_error
= source_error
84 class NextFragment(Exception):
87 def establish_connection():
88 ctx
.chunk_size
= (random
.randint(int(chunk_size
* 0.95), chunk_size
)
89 if not is_test
and chunk_size
else chunk_size
)
90 if ctx
.resume_len
> 0:
91 range_start
= ctx
.resume_len
92 if req_start
is not None:
93 # offset the beginning of Range to be within request
94 range_start
+= req_start
96 self
.report_resuming_byte(ctx
.resume_len
)
98 elif req_start
is not None:
99 range_start
= req_start
100 elif ctx
.chunk_size
> 0:
104 ctx
.is_resume
= False
107 chunk_aware_end
= range_start
+ ctx
.chunk_size
- 1
108 # we're not allowed to download outside Range
109 range_end
= chunk_aware_end
if req_end
is None else min(chunk_aware_end
, req_end
)
110 elif req_end
is not None:
111 # there's no need for chunked downloads, so download until the end of Range
116 if try_call(lambda: range_start
> range_end
):
119 raise RetryDownload(Exception(f
'Conflicting range. (start={range_start} > end={range_end})'))
121 if try_call(lambda: range_end
>= ctx
.content_len
):
122 range_end
= ctx
.content_len
- 1
124 request
= sanitized_Request(url
, request_data
, headers
)
125 has_range
= range_start
is not None
127 request
.add_header('Range', f
'bytes={int(range_start)}-{int_or_none(range_end) or ""}')
128 # Establish connection
130 ctx
.data
= self
.ydl
.urlopen(request
)
131 # When trying to resume, Content-Range HTTP header of response has to be checked
132 # to match the value of requested Range HTTP header. This is due to a webservers
133 # that don't support resuming and serve a whole file with no Content-Range
134 # set in response despite of requested Range (see
135 # https://github.com/ytdl-org/youtube-dl/issues/6057#issuecomment-126129799)
137 content_range
= ctx
.data
.headers
.get('Content-Range')
138 content_range_start
, content_range_end
, content_len
= parse_http_range(content_range
)
139 if content_range_start
is not None and range_start
== content_range_start
:
140 # Content-Range is present and matches requested Range, resume is possible
141 accept_content_len
= (
142 # Non-chunked download
144 # Chunked download and requested piece or
145 # its part is promised to be served
146 or content_range_end
== range_end
147 or content_len
< range_end
)
148 if accept_content_len
:
149 ctx
.content_len
= content_len
150 if content_len
or req_end
:
151 ctx
.data_len
= min(content_len
or req_end
, req_end
or content_len
) - (req_start
or 0)
153 # Content-Range is either not present or invalid. Assuming remote webserver is
154 # trying to send the whole file, resume is not possible, so wiping the local file
155 # and performing entire redownload
156 self
.report_unable_to_resume()
159 ctx
.data_len
= ctx
.content_len
= int_or_none(ctx
.data
.info().get('Content-length', None))
160 except compat_urllib_error
.HTTPError
as err
:
162 # Unable to resume (requested range not satisfiable)
164 # Open the connection again without the range header
165 ctx
.data
= self
.ydl
.urlopen(
166 sanitized_Request(url
, request_data
, headers
))
167 content_length
= ctx
.data
.info()['Content-Length']
168 except compat_urllib_error
.HTTPError
as err
:
169 if err
.code
< 500 or err
.code
>= 600:
172 # Examine the reported length
173 if (content_length
is not None
174 and (ctx
.resume_len
- 100 < int(content_length
) < ctx
.resume_len
+ 100)):
175 # The file had already been fully downloaded.
176 # Explanation to the above condition: in issue #175 it was revealed that
177 # YouTube sometimes adds or removes a few bytes from the end of the file,
178 # changing the file size slightly and causing problems for some users. So
179 # I decided to implement a suggested change and consider the file
180 # completely downloaded if the file size differs less than 100 bytes from
181 # the one in the hard drive.
182 self
.report_file_already_downloaded(ctx
.filename
)
183 self
.try_rename(ctx
.tmpfilename
, ctx
.filename
)
184 self
._hook
_progress
({
185 'filename': ctx
.filename
,
186 'status': 'finished',
187 'downloaded_bytes': ctx
.resume_len
,
188 'total_bytes': ctx
.resume_len
,
190 raise SucceedDownload()
192 # The length does not match, we start the download over
193 self
.report_unable_to_resume()
197 elif err
.code
< 500 or err
.code
>= 600:
198 # Unexpected HTTP error
200 raise RetryDownload(err
)
201 except compat_urllib_error
.URLError
as err
:
202 if isinstance(err
.reason
, ssl
.CertificateError
):
204 raise RetryDownload(err
)
205 # In urllib.request.AbstractHTTPHandler, the response is partially read on request.
206 # Any errors that occur during this will not be wrapped by URLError
207 except RESPONSE_READ_EXCEPTIONS
as err
:
208 raise RetryDownload(err
)
211 data_len
= ctx
.data
.info().get('Content-length', None)
213 # Range HTTP header may be ignored/unsupported by a webserver
214 # (e.g. extractor/scivee.py, extractor/bambuser.py).
215 # However, for a test we still would like to download just a piece of a file.
216 # To achieve this we limit data_len to _TEST_FILE_SIZE and manually control
217 # block size when downloading a file.
218 if is_test
and (data_len
is None or int(data_len
) > self
._TEST
_FILE
_SIZE
):
219 data_len
= self
._TEST
_FILE
_SIZE
221 if data_len
is not None:
222 data_len
= int(data_len
) + ctx
.resume_len
223 min_data_len
= self
.params
.get('min_filesize')
224 max_data_len
= self
.params
.get('max_filesize')
225 if min_data_len
is not None and data_len
< min_data_len
:
227 f
'\r[download] File is smaller than min-filesize ({data_len} bytes < {min_data_len} bytes). Aborting.')
229 if max_data_len
is not None and data_len
> max_data_len
:
231 f
'\r[download] File is larger than max-filesize ({data_len} bytes > {max_data_len} bytes). Aborting.')
234 byte_counter
= 0 + ctx
.resume_len
235 block_size
= ctx
.block_size
238 # measure time over whole while-loop, so slow_down() and best_block_size() work together properly
239 now
= None # needed for slow_down() in the first loop run
240 before
= start
# start measuring
243 to_stdout
= ctx
.tmpfilename
== '-'
244 if ctx
.stream
is not None:
248 ctx
.resume_len
= byte_counter
if to_stdout
else os
.path
.getsize(encodeFilename(ctx
.tmpfilename
))
249 raise RetryDownload(e
)
254 data_block
= ctx
.data
.read(block_size
if not is_test
else min(block_size
, data_len
- byte_counter
))
255 except RESPONSE_READ_EXCEPTIONS
as err
:
258 byte_counter
+= len(data_block
)
260 # exit loop when download is finished
261 if len(data_block
) == 0:
264 # Open destination file just in time
265 if ctx
.stream
is None:
267 ctx
.stream
, ctx
.tmpfilename
= self
.sanitize_open(
268 ctx
.tmpfilename
, ctx
.open_mode
)
269 assert ctx
.stream
is not None
270 ctx
.filename
= self
.undo_temp_name(ctx
.tmpfilename
)
271 self
.report_destination(ctx
.filename
)
272 except OSError as err
:
273 self
.report_error('unable to open for writing: %s' % str(err
))
276 if self
.params
.get('xattr_set_filesize', False) and data_len
is not None:
278 write_xattr(ctx
.tmpfilename
, 'user.ytdl.filesize', str(data_len
).encode())
279 except (XAttrUnavailableError
, XAttrMetadataError
) as err
:
280 self
.report_error('unable to set filesize xattr: %s' % str(err
))
283 ctx
.stream
.write(data_block
)
284 except OSError as err
:
286 self
.report_error('unable to write data: %s' % str(err
))
290 self
.slow_down(start
, now
, byte_counter
- ctx
.resume_len
)
292 # end measuring of one loop run
297 if not self
.params
.get('noresizebuffer', False):
298 block_size
= self
.best_block_size(after
- before
, len(data_block
))
303 speed
= self
.calc_speed(start
, now
, byte_counter
- ctx
.resume_len
)
304 if ctx
.data_len
is None:
307 eta
= self
.calc_eta(start
, time
.time(), ctx
.data_len
- ctx
.resume_len
, byte_counter
- ctx
.resume_len
)
309 self
._hook
_progress
({
310 'status': 'downloading',
311 'downloaded_bytes': byte_counter
,
312 'total_bytes': ctx
.data_len
,
313 'tmpfilename': ctx
.tmpfilename
,
314 'filename': ctx
.filename
,
317 'elapsed': now
- ctx
.start_time
,
318 'ctx_id': info_dict
.get('ctx_id'),
321 if data_len
is not None and byte_counter
== data_len
:
324 if speed
and speed
< (self
.params
.get('throttledratelimit') or 0):
325 # The speed must stay below the limit for 3 seconds
326 # This prevents raising error when the speed temporarily goes down
327 if ctx
.throttle_start
is None:
328 ctx
.throttle_start
= now
329 elif now
- ctx
.throttle_start
> 3:
330 if ctx
.stream
is not None and ctx
.tmpfilename
!= '-':
332 raise ThrottledDownload()
334 ctx
.throttle_start
= None
336 if not is_test
and ctx
.chunk_size
and ctx
.content_len
is not None and byte_counter
< ctx
.content_len
:
337 ctx
.resume_len
= byte_counter
338 # ctx.block_size = block_size
341 if ctx
.stream
is None:
343 self
.report_error('Did not get any data blocks')
345 if ctx
.tmpfilename
!= '-':
348 if data_len
is not None and byte_counter
!= data_len
:
349 err
= ContentTooShortError(byte_counter
, int(data_len
))
354 self
.try_rename(ctx
.tmpfilename
, ctx
.filename
)
356 # Update file modification time
357 if self
.params
.get('updatetime', True):
358 info_dict
['filetime'] = self
.try_utime(ctx
.filename
, ctx
.data
.info().get('last-modified', None))
360 self
._hook
_progress
({
361 'downloaded_bytes': byte_counter
,
362 'total_bytes': byte_counter
,
363 'filename': ctx
.filename
,
364 'status': 'finished',
365 'elapsed': time
.time() - ctx
.start_time
,
366 'ctx_id': info_dict
.get('ctx_id'),
371 while count
<= retries
:
373 establish_connection()
375 except RetryDownload
as e
:
378 self
.report_retry(e
.source_error
, count
, retries
)
380 self
.to_screen(f
'[download] Got server HTTP error: {e.source_error}')
384 except SucceedDownload
:
387 self
.report_error('giving up after %s retries' % retries
)