1
1
mirror of https://github.com/ytdl-org/youtube-dl synced 2024-11-27 19:36:50 +01:00
youtube-dl/youtube_dl/downloader/http.py

299 lines
13 KiB
Python
Raw Normal View History

2014-11-26 12:27:36 +01:00
from __future__ import unicode_literals
import errno
import os
import socket
import time
2015-08-02 04:58:02 +02:00
import re
from .common import FileDownloader
Switch codebase to use sanitized_Request instead of compat_urllib_request.Request [downloader/dash] Use sanitized_Request [downloader/http] Use sanitized_Request [atresplayer] Use sanitized_Request [bambuser] Use sanitized_Request [bliptv] Use sanitized_Request [brightcove] Use sanitized_Request [cbs] Use sanitized_Request [ceskatelevize] Use sanitized_Request [collegerama] Use sanitized_Request [extractor/common] Use sanitized_Request [crunchyroll] Use sanitized_Request [dailymotion] Use sanitized_Request [dcn] Use sanitized_Request [dramafever] Use sanitized_Request [dumpert] Use sanitized_Request [eitb] Use sanitized_Request [escapist] Use sanitized_Request [everyonesmixtape] Use sanitized_Request [extremetube] Use sanitized_Request [facebook] Use sanitized_Request [fc2] Use sanitized_Request [flickr] Use sanitized_Request [4tube] Use sanitized_Request [gdcvault] Use sanitized_Request [extractor/generic] Use sanitized_Request [hearthisat] Use sanitized_Request [hotnewhiphop] Use sanitized_Request [hypem] Use sanitized_Request [iprima] Use sanitized_Request [ivi] Use sanitized_Request [keezmovies] Use sanitized_Request [letv] Use sanitized_Request [lynda] Use sanitized_Request [metacafe] Use sanitized_Request [minhateca] Use sanitized_Request [miomio] Use sanitized_Request [meovideo] Use sanitized_Request [mofosex] Use sanitized_Request [moniker] Use sanitized_Request [mooshare] Use sanitized_Request [movieclips] Use sanitized_Request [mtv] Use sanitized_Request [myvideo] Use sanitized_Request [neteasemusic] Use sanitized_Request [nfb] Use sanitized_Request [niconico] Use sanitized_Request [noco] Use sanitized_Request [nosvideo] Use sanitized_Request [novamov] Use sanitized_Request [nowness] Use sanitized_Request [nuvid] Use sanitized_Request [played] Use sanitized_Request [pluralsight] Use sanitized_Request [pornhub] Use sanitized_Request [pornotube] Use sanitized_Request [primesharetv] Use sanitized_Request [promptfile] Use sanitized_Request [qqmusic] Use sanitized_Request [rtve] Use sanitized_Request [safari] Use sanitized_Request [sandia] Use sanitized_Request [shared] Use sanitized_Request [sharesix] Use sanitized_Request [sina] Use sanitized_Request [smotri] Use sanitized_Request [sohu] Use sanitized_Request [spankwire] Use sanitized_Request [sportdeutschland] Use sanitized_Request [streamcloud] Use sanitized_Request [streamcz] Use sanitized_Request [tapely] Use sanitized_Request [tube8] Use sanitized_Request [tubitv] Use sanitized_Request [twitch] Use sanitized_Request [twitter] Use sanitized_Request [udemy] Use sanitized_Request [vbox7] Use sanitized_Request [veoh] Use sanitized_Request [vessel] Use sanitized_Request [vevo] Use sanitized_Request [viddler] Use sanitized_Request [videomega] Use sanitized_Request [viewvster] Use sanitized_Request [viki] Use sanitized_Request [vk] Use sanitized_Request [vodlocker] Use sanitized_Request [voicerepublic] Use sanitized_Request [wistia] Use sanitized_Request [xfileshare] Use sanitized_Request [xtube] Use sanitized_Request [xvideos] Use sanitized_Request [yandexmusic] Use sanitized_Request [youku] Use sanitized_Request [youporn] Use sanitized_Request [youtube] Use sanitized_Request [patreon] Use sanitized_Request [extractor/common] Remove unused import [nfb] PEP 8
2015-11-21 17:18:17 +01:00
from ..compat import compat_urllib_error
from ..utils import (
ContentTooShortError,
encodeFilename,
sanitize_open,
Switch codebase to use sanitized_Request instead of compat_urllib_request.Request [downloader/dash] Use sanitized_Request [downloader/http] Use sanitized_Request [atresplayer] Use sanitized_Request [bambuser] Use sanitized_Request [bliptv] Use sanitized_Request [brightcove] Use sanitized_Request [cbs] Use sanitized_Request [ceskatelevize] Use sanitized_Request [collegerama] Use sanitized_Request [extractor/common] Use sanitized_Request [crunchyroll] Use sanitized_Request [dailymotion] Use sanitized_Request [dcn] Use sanitized_Request [dramafever] Use sanitized_Request [dumpert] Use sanitized_Request [eitb] Use sanitized_Request [escapist] Use sanitized_Request [everyonesmixtape] Use sanitized_Request [extremetube] Use sanitized_Request [facebook] Use sanitized_Request [fc2] Use sanitized_Request [flickr] Use sanitized_Request [4tube] Use sanitized_Request [gdcvault] Use sanitized_Request [extractor/generic] Use sanitized_Request [hearthisat] Use sanitized_Request [hotnewhiphop] Use sanitized_Request [hypem] Use sanitized_Request [iprima] Use sanitized_Request [ivi] Use sanitized_Request [keezmovies] Use sanitized_Request [letv] Use sanitized_Request [lynda] Use sanitized_Request [metacafe] Use sanitized_Request [minhateca] Use sanitized_Request [miomio] Use sanitized_Request [meovideo] Use sanitized_Request [mofosex] Use sanitized_Request [moniker] Use sanitized_Request [mooshare] Use sanitized_Request [movieclips] Use sanitized_Request [mtv] Use sanitized_Request [myvideo] Use sanitized_Request [neteasemusic] Use sanitized_Request [nfb] Use sanitized_Request [niconico] Use sanitized_Request [noco] Use sanitized_Request [nosvideo] Use sanitized_Request [novamov] Use sanitized_Request [nowness] Use sanitized_Request [nuvid] Use sanitized_Request [played] Use sanitized_Request [pluralsight] Use sanitized_Request [pornhub] Use sanitized_Request [pornotube] Use sanitized_Request [primesharetv] Use sanitized_Request [promptfile] Use sanitized_Request [qqmusic] Use sanitized_Request [rtve] Use sanitized_Request [safari] Use sanitized_Request [sandia] Use sanitized_Request [shared] Use sanitized_Request [sharesix] Use sanitized_Request [sina] Use sanitized_Request [smotri] Use sanitized_Request [sohu] Use sanitized_Request [spankwire] Use sanitized_Request [sportdeutschland] Use sanitized_Request [streamcloud] Use sanitized_Request [streamcz] Use sanitized_Request [tapely] Use sanitized_Request [tube8] Use sanitized_Request [tubitv] Use sanitized_Request [twitch] Use sanitized_Request [twitter] Use sanitized_Request [udemy] Use sanitized_Request [vbox7] Use sanitized_Request [veoh] Use sanitized_Request [vessel] Use sanitized_Request [vevo] Use sanitized_Request [viddler] Use sanitized_Request [videomega] Use sanitized_Request [viewvster] Use sanitized_Request [viki] Use sanitized_Request [vk] Use sanitized_Request [vodlocker] Use sanitized_Request [voicerepublic] Use sanitized_Request [wistia] Use sanitized_Request [xfileshare] Use sanitized_Request [xtube] Use sanitized_Request [xvideos] Use sanitized_Request [yandexmusic] Use sanitized_Request [youku] Use sanitized_Request [youporn] Use sanitized_Request [youtube] Use sanitized_Request [patreon] Use sanitized_Request [extractor/common] Remove unused import [nfb] PEP 8
2015-11-21 17:18:17 +01:00
sanitized_Request,
write_xattr,
XAttrMetadataError,
XAttrUnavailableError,
)
class HttpFD(FileDownloader):
def real_download(self, filename, info_dict):
url = info_dict['url']
class DownloadContext(dict):
__getattr__ = dict.get
__setattr__ = dict.__setitem__
__delattr__ = dict.__delitem__
ctx = DownloadContext()
ctx.filename = filename
ctx.tmpfilename = self.temp_name(filename)
ctx.stream = None
# Do not include the Accept-Encoding header
headers = {'Youtubedl-no-compression': 'True'}
add_headers = info_dict.get('http_headers')
if add_headers:
headers.update(add_headers)
Switch codebase to use sanitized_Request instead of compat_urllib_request.Request [downloader/dash] Use sanitized_Request [downloader/http] Use sanitized_Request [atresplayer] Use sanitized_Request [bambuser] Use sanitized_Request [bliptv] Use sanitized_Request [brightcove] Use sanitized_Request [cbs] Use sanitized_Request [ceskatelevize] Use sanitized_Request [collegerama] Use sanitized_Request [extractor/common] Use sanitized_Request [crunchyroll] Use sanitized_Request [dailymotion] Use sanitized_Request [dcn] Use sanitized_Request [dramafever] Use sanitized_Request [dumpert] Use sanitized_Request [eitb] Use sanitized_Request [escapist] Use sanitized_Request [everyonesmixtape] Use sanitized_Request [extremetube] Use sanitized_Request [facebook] Use sanitized_Request [fc2] Use sanitized_Request [flickr] Use sanitized_Request [4tube] Use sanitized_Request [gdcvault] Use sanitized_Request [extractor/generic] Use sanitized_Request [hearthisat] Use sanitized_Request [hotnewhiphop] Use sanitized_Request [hypem] Use sanitized_Request [iprima] Use sanitized_Request [ivi] Use sanitized_Request [keezmovies] Use sanitized_Request [letv] Use sanitized_Request [lynda] Use sanitized_Request [metacafe] Use sanitized_Request [minhateca] Use sanitized_Request [miomio] Use sanitized_Request [meovideo] Use sanitized_Request [mofosex] Use sanitized_Request [moniker] Use sanitized_Request [mooshare] Use sanitized_Request [movieclips] Use sanitized_Request [mtv] Use sanitized_Request [myvideo] Use sanitized_Request [neteasemusic] Use sanitized_Request [nfb] Use sanitized_Request [niconico] Use sanitized_Request [noco] Use sanitized_Request [nosvideo] Use sanitized_Request [novamov] Use sanitized_Request [nowness] Use sanitized_Request [nuvid] Use sanitized_Request [played] Use sanitized_Request [pluralsight] Use sanitized_Request [pornhub] Use sanitized_Request [pornotube] Use sanitized_Request [primesharetv] Use sanitized_Request [promptfile] Use sanitized_Request [qqmusic] Use sanitized_Request [rtve] Use sanitized_Request [safari] Use sanitized_Request [sandia] Use sanitized_Request [shared] Use sanitized_Request [sharesix] Use sanitized_Request [sina] Use sanitized_Request [smotri] Use sanitized_Request [sohu] Use sanitized_Request [spankwire] Use sanitized_Request [sportdeutschland] Use sanitized_Request [streamcloud] Use sanitized_Request [streamcz] Use sanitized_Request [tapely] Use sanitized_Request [tube8] Use sanitized_Request [tubitv] Use sanitized_Request [twitch] Use sanitized_Request [twitter] Use sanitized_Request [udemy] Use sanitized_Request [vbox7] Use sanitized_Request [veoh] Use sanitized_Request [vessel] Use sanitized_Request [vevo] Use sanitized_Request [viddler] Use sanitized_Request [videomega] Use sanitized_Request [viewvster] Use sanitized_Request [viki] Use sanitized_Request [vk] Use sanitized_Request [vodlocker] Use sanitized_Request [voicerepublic] Use sanitized_Request [wistia] Use sanitized_Request [xfileshare] Use sanitized_Request [xtube] Use sanitized_Request [xvideos] Use sanitized_Request [yandexmusic] Use sanitized_Request [youku] Use sanitized_Request [youporn] Use sanitized_Request [youtube] Use sanitized_Request [patreon] Use sanitized_Request [extractor/common] Remove unused import [nfb] PEP 8
2015-11-21 17:18:17 +01:00
basic_request = sanitized_Request(url, None, headers)
request = sanitized_Request(url, None, headers)
is_test = self.params.get('test', False)
if is_test:
request.add_header('Range', 'bytes=0-%s' % str(self._TEST_FILE_SIZE - 1))
ctx.open_mode = 'wb'
ctx.resume_len = 0
if self.params.get('continuedl', True):
# Establish possible resume length
if os.path.isfile(encodeFilename(ctx.tmpfilename)):
ctx.resume_len = os.path.getsize(encodeFilename(ctx.tmpfilename))
count = 0
retries = self.params.get('retries', 0)
class SucceedDownload(Exception):
pass
class RetryDownload(Exception):
def __init__(self, source_error):
self.source_error = source_error
def establish_connection():
if ctx.resume_len != 0:
self.report_resuming_byte(ctx.resume_len)
request.add_header('Range', 'bytes=%d-' % ctx.resume_len)
ctx.open_mode = 'ab'
# Establish connection
try:
ctx.data = self.ydl.urlopen(request)
# When trying to resume, Content-Range HTTP header of response has to be checked
# to match the value of requested Range HTTP header. This is due to a webservers
# that don't support resuming and serve a whole file with no Content-Range
# set in response despite of requested Range (see
# https://github.com/rg3/youtube-dl/issues/6057#issuecomment-126129799)
if ctx.resume_len > 0:
content_range = ctx.data.headers.get('Content-Range')
if content_range:
content_range_m = re.search(r'bytes (\d+)-', content_range)
# Content-Range is present and matches requested Range, resume is possible
if content_range_m and ctx.resume_len == int(content_range_m.group(1)):
return
# Content-Range is either not present or invalid. Assuming remote webserver is
# trying to send the whole file, resume is not possible, so wiping the local file
# and performing entire redownload
self.report_unable_to_resume()
ctx.resume_len = 0
ctx.open_mode = 'wb'
return
except (compat_urllib_error.HTTPError, ) as err:
if (err.code < 500 or err.code >= 600) and err.code != 416:
# Unexpected HTTP error
raise
elif err.code == 416:
# Unable to resume (requested range not satisfiable)
try:
# Open the connection again without the range header
ctx.data = self.ydl.urlopen(basic_request)
content_length = ctx.data.info()['Content-Length']
except (compat_urllib_error.HTTPError, ) as err:
if err.code < 500 or err.code >= 600:
raise
else:
# Examine the reported length
if (content_length is not None and
(ctx.resume_len - 100 < int(content_length) < ctx.resume_len + 100)):
# The file had already been fully downloaded.
# Explanation to the above condition: in issue #175 it was revealed that
# YouTube sometimes adds or removes a few bytes from the end of the file,
# changing the file size slightly and causing problems for some users. So
# I decided to implement a suggested change and consider the file
# completely downloaded if the file size differs less than 100 bytes from
# the one in the hard drive.
self.report_file_already_downloaded(ctx.filename)
self.try_rename(ctx.tmpfilename, ctx.filename)
self._hook_progress({
'filename': ctx.filename,
'status': 'finished',
'downloaded_bytes': ctx.resume_len,
'total_bytes': ctx.resume_len,
})
raise SucceedDownload()
else:
# The length does not match, we start the download over
self.report_unable_to_resume()
ctx.resume_len = 0
ctx.open_mode = 'wb'
return
raise RetryDownload(err)
except socket.error as err:
if err.errno != errno.ECONNRESET:
2015-02-02 12:05:39 +01:00
# Connection reset is no problem, just retry
raise
raise RetryDownload(err)
def download():
data_len = ctx.data.info().get('Content-length', None)
# Range HTTP header may be ignored/unsupported by a webserver
# (e.g. extractor/scivee.py, extractor/bambuser.py).
# However, for a test we still would like to download just a piece of a file.
# To achieve this we limit data_len to _TEST_FILE_SIZE and manually control
# block size when downloading a file.
if is_test and (data_len is None or int(data_len) > self._TEST_FILE_SIZE):
data_len = self._TEST_FILE_SIZE
if data_len is not None:
data_len = int(data_len) + ctx.resume_len
min_data_len = self.params.get('min_filesize')
max_data_len = self.params.get('max_filesize')
if min_data_len is not None and data_len < min_data_len:
self.to_screen('\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len, min_data_len))
return False
if max_data_len is not None and data_len > max_data_len:
self.to_screen('\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len, max_data_len))
return False
2015-02-02 12:05:39 +01:00
byte_counter = 0 + ctx.resume_len
block_size = self.params.get('buffersize', 1024)
start = time.time()
# measure time over whole while-loop, so slow_down() and best_block_size() work together properly
now = None # needed for slow_down() in the first loop run
before = start # start measuring
def retry(e):
if ctx.tmpfilename != '-':
ctx.stream.close()
ctx.stream = None
ctx.resume_len = os.path.getsize(encodeFilename(ctx.tmpfilename))
raise RetryDownload(e)
while True:
try:
# Download and write
data_block = ctx.data.read(block_size if not is_test else min(block_size, data_len - byte_counter))
# socket.timeout is a subclass of socket.error but may not have
# errno set
except socket.timeout as e:
retry(e)
except socket.error as e:
if e.errno not in (errno.ECONNRESET, errno.ETIMEDOUT):
raise
retry(e)
byte_counter += len(data_block)
# exit loop when download is finished
if len(data_block) == 0:
break
# Open destination file just in time
if ctx.stream is None:
try:
ctx.stream, ctx.tmpfilename = sanitize_open(
ctx.tmpfilename, ctx.open_mode)
assert ctx.stream is not None
ctx.filename = self.undo_temp_name(ctx.tmpfilename)
self.report_destination(ctx.filename)
except (OSError, IOError) as err:
self.report_error('unable to open for writing: %s' % str(err))
return False
if self.params.get('xattr_set_filesize', False) and data_len is not None:
try:
write_xattr(ctx.tmpfilename, 'user.ytdl.filesize', str(data_len).encode('utf-8'))
except (XAttrUnavailableError, XAttrMetadataError) as err:
self.report_error('unable to set filesize xattr: %s' % str(err))
try:
ctx.stream.write(data_block)
except (IOError, OSError) as err:
self.to_stderr('\n')
self.report_error('unable to write data: %s' % str(err))
return False
# Apply rate limit
self.slow_down(start, now, byte_counter - ctx.resume_len)
# end measuring of one loop run
now = time.time()
after = now
# Adjust block size
if not self.params.get('noresizebuffer', False):
block_size = self.best_block_size(after - before, len(data_block))
before = after
# Progress message
speed = self.calc_speed(start, now, byte_counter - ctx.resume_len)
if data_len is None:
eta = None
else:
eta = self.calc_eta(start, time.time(), data_len - ctx.resume_len, byte_counter - ctx.resume_len)
self._hook_progress({
'status': 'downloading',
'downloaded_bytes': byte_counter,
'total_bytes': data_len,
'tmpfilename': ctx.tmpfilename,
'filename': ctx.filename,
'eta': eta,
'speed': speed,
'elapsed': now - start,
})
if is_test and byte_counter == data_len:
break
if ctx.stream is None:
2014-11-26 12:27:36 +01:00
self.to_stderr('\n')
self.report_error('Did not get any data blocks')
return False
if ctx.tmpfilename != '-':
ctx.stream.close()
if data_len is not None and byte_counter != data_len:
err = ContentTooShortError(byte_counter, int(data_len))
if count <= retries:
retry(err)
raise err
self.try_rename(ctx.tmpfilename, ctx.filename)
# Update file modification time
if self.params.get('updatetime', True):
info_dict['filetime'] = self.try_utime(ctx.filename, ctx.data.info().get('last-modified', None))
self._hook_progress({
'downloaded_bytes': byte_counter,
'total_bytes': byte_counter,
'filename': ctx.filename,
'status': 'finished',
'elapsed': time.time() - start,
})
return True
while count <= retries:
try:
establish_connection()
download()
return True
except RetryDownload as e:
count += 1
if count <= retries:
self.report_retry(e.source_error, count, retries)
continue
except SucceedDownload:
return True
self.report_error('giving up after %s retries' % retries)
return False