1
0
Fork 0
mirror of https://gitlab.com/dstftw/youtube-dl.git synced 2020-11-16 09:42:26 +00:00
youtube-dl/youtube_dl/utils.py

1573 lines
47 KiB
Python
Raw Normal View History

#!/usr/bin/env python
# -*- coding: utf-8 -*-
2014-11-02 10:37:49 +00:00
from __future__ import unicode_literals
2014-03-24 00:40:09 +00:00
import calendar
2014-04-04 21:00:51 +00:00
import codecs
import contextlib
import ctypes
import datetime
import email.utils
import errno
import gzip
import itertools
2012-11-27 23:09:17 +00:00
import io
2012-12-20 12:13:24 +00:00
import json
import locale
import math
import os
import pipes
import platform
import re
import ssl
import socket
2014-02-15 15:24:43 +00:00
import struct
2013-12-09 17:29:07 +00:00
import subprocess
import sys
import tempfile
import traceback
import xml.etree.ElementTree
import zlib
from .compat import (
compat_chr,
compat_getenv,
compat_html_entities,
compat_parse_qs,
compat_str,
compat_urllib_error,
compat_urllib_parse,
compat_urllib_parse_urlparse,
compat_urllib_request,
compat_urlparse,
shlex_quote,
)
# This is not clearly defined otherwise
compiled_regex_type = type(re.compile(''))
std_headers = {
'User-Agent': 'Mozilla/5.0 (X11; Linux x86_64; rv:10.0) Gecko/20100101 Firefox/10.0 (Chrome)',
2012-11-28 01:04:46 +00:00
'Accept-Charset': 'ISO-8859-1,utf-8;q=0.7,*;q=0.7',
'Accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8',
'Accept-Encoding': 'gzip, deflate',
'Accept-Language': 'en-us,en;q=0.5',
}
2014-11-23 19:41:03 +00:00
def preferredencoding():
2012-11-28 01:04:46 +00:00
"""Get preferred encoding.
2012-11-28 01:04:46 +00:00
Returns the best encoding scheme for the system, based on
locale.getpreferredencoding() and some further tweaks.
"""
try:
pref = locale.getpreferredencoding()
2014-11-17 06:16:12 +00:00
'TEST'.encode(pref)
2012-11-28 01:04:46 +00:00
except:
pref = 'UTF-8'
2012-11-28 01:04:46 +00:00
return pref
2012-12-20 12:13:24 +00:00
def write_json_file(obj, fn):
""" Encode obj as JSON and write it to fn, atomically if possible """
fn = encodeFilename(fn)
if sys.version_info < (3, 0) and sys.platform != 'win32':
encoding = get_filesystem_encoding()
# os.path.basename returns a bytes object, but NamedTemporaryFile
# will fail if the filename contains non ascii characters unless we
# use a unicode object
path_basename = lambda f: os.path.basename(fn).decode(encoding)
# the same for os.path.dirname
path_dirname = lambda f: os.path.dirname(fn).decode(encoding)
else:
path_basename = os.path.basename
path_dirname = os.path.dirname
args = {
'suffix': '.tmp',
'prefix': path_basename(fn) + '.',
'dir': path_dirname(fn),
'delete': False,
}
# In Python 2.x, json.dump expects a bytestream.
# In Python 3.x, it writes to a character stream
if sys.version_info < (3, 0):
args['mode'] = 'wb'
else:
args.update({
'mode': 'w',
'encoding': 'utf-8',
})
tf = tempfile.NamedTemporaryFile(**args)
try:
with tf:
json.dump(obj, tf)
if sys.platform == 'win32':
# Need to remove existing file on Windows, else os.rename raises
# WindowsError or FileExistsError.
try:
os.unlink(fn)
except OSError:
pass
os.rename(tf.name, fn)
except:
try:
os.remove(tf.name)
except OSError:
pass
raise
if sys.version_info >= (2, 7):
2013-07-11 14:12:08 +00:00
def find_xpath_attr(node, xpath, key, val):
""" Find the xpath xpath[@key=val] """
2014-07-25 09:39:17 +00:00
assert re.match(r'^[a-zA-Z-]+$', key)
assert re.match(r'^[a-zA-Z0-9@\s:._-]*$', val)
2014-11-26 10:50:22 +00:00
expr = xpath + "[@%s='%s']" % (key, val)
2013-07-11 14:12:08 +00:00
return node.find(expr)
else:
def find_xpath_attr(node, xpath, key, val):
2014-09-13 06:34:15 +00:00
# Here comes the crazy part: In 2.6, if the xpath is a unicode,
# .//node does not match if a node is a direct child of . !
if isinstance(xpath, unicode):
xpath = xpath.encode('ascii')
2013-07-11 14:12:08 +00:00
for f in node.findall(xpath):
if f.attrib.get(key) == val:
return f
return None
# On python2.6 the xml.etree.ElementTree.Element methods don't support
# the namespace parameter
2014-11-23 19:41:03 +00:00
def xpath_with_ns(path, ns_map):
components = [c.split(':') for c in path.split('/')]
replaced = []
for c in components:
if len(c) == 1:
replaced.append(c[0])
else:
ns, tag = c
replaced.append('{%s}%s' % (ns_map[ns], tag))
return '/'.join(replaced)
def xpath_text(node, xpath, name=None, fatal=False):
if sys.version_info < (2, 7): # Crazy 2.6
xpath = xpath.encode('ascii')
n = node.find(xpath)
if n is None or n.text is None:
if fatal:
name = xpath if name is None else name
raise ExtractorError('Could not find XML element %s' % name)
else:
return None
return n.text
def get_element_by_id(id, html):
"""Return the content of the tag with the specified ID in the passed HTML document"""
return get_element_by_attribute("id", id, html)
def get_element_by_attribute(attribute, value, html):
"""Return the content of the tag with the specified attribute in the passed HTML document"""
m = re.search(r'''(?xs)
<([a-zA-Z0-9:._-]+)
(?:\s+[a-zA-Z0-9:._-]+(?:=[a-zA-Z0-9:._-]+|="[^"]+"|='[^']+'))*?
\s+%s=['"]?%s['"]?
(?:\s+[a-zA-Z0-9:._-]+(?:=[a-zA-Z0-9:._-]+|="[^"]+"|='[^']+'))*?
\s*>
(?P<content>.*?)
</\1>
''' % (re.escape(attribute), re.escape(value)), html)
if not m:
return None
res = m.group('content')
if res.startswith('"') or res.startswith("'"):
res = res[1:-1]
return unescapeHTML(res)
def clean_html(html):
2012-11-28 01:04:46 +00:00
"""Clean an HTML snippet into a readable string"""
# Newline vs <br />
html = html.replace('\n', ' ')
html = re.sub(r'\s*<\s*br\s*/?\s*>\s*', '\n', html)
html = re.sub(r'<\s*/\s*p\s*>\s*<\s*p[^>]*>', '\n', html)
2012-11-28 01:04:46 +00:00
# Strip html tags
html = re.sub('<.*?>', '', html)
# Replace html entities
html = unescapeHTML(html)
2013-03-29 14:59:13 +00:00
return html.strip()
def sanitize_open(filename, open_mode):
2012-11-28 01:04:46 +00:00
"""Try to open the given filename, and slightly tweak it if this fails.
Attempts to open the given filename. If this fails, it tries to change
the filename slightly, step by step, until it's either able to open it
or it fails and raises a final exception, like the standard open()
function.
It returns the tuple (stream, definitive_file_name).
"""
try:
2014-11-17 06:16:12 +00:00
if filename == '-':
2012-11-28 01:04:46 +00:00
if sys.platform == 'win32':
import msvcrt
msvcrt.setmode(sys.stdout.fileno(), os.O_BINARY)
2013-03-28 12:13:03 +00:00
return (sys.stdout.buffer if hasattr(sys.stdout, 'buffer') else sys.stdout, filename)
2012-11-28 01:04:46 +00:00
stream = open(encodeFilename(filename), open_mode)
return (stream, filename)
except (IOError, OSError) as err:
if err.errno in (errno.EACCES,):
raise
2012-11-28 01:04:46 +00:00
# In case of error, try to remove win32 forbidden chars
alt_filename = os.path.join(
2014-11-23 21:21:46 +00:00
re.sub('[/<>:"\\|\\\\?\\*]', '#', path_part)
for path_part in os.path.split(filename)
)
if alt_filename == filename:
raise
else:
# An exception here should be caught in the caller
stream = open(encodeFilename(filename), open_mode)
return (stream, alt_filename)
def timeconvert(timestr):
2012-11-28 01:04:46 +00:00
"""Convert RFC 2822 defined time string into system timestamp"""
timestamp = None
timetuple = email.utils.parsedate_tz(timestr)
if timetuple is not None:
timestamp = email.utils.mktime_tz(timetuple)
return timestamp
2012-11-26 22:58:46 +00:00
2014-11-23 19:41:03 +00:00
def sanitize_filename(s, restricted=False, is_id=False):
2012-11-28 01:04:46 +00:00
"""Sanitizes a string so it could be used as part of a filename.
If restricted is set, use a stricter subset of allowed characters.
Set is_id if this is not an arbitrary string, but an ID that should be kept if possible
2012-11-28 01:04:46 +00:00
"""
def replace_insane(char):
if char == '?' or ord(char) < 32 or ord(char) == 127:
return ''
elif char == '"':
return '' if restricted else '\''
elif char == ':':
return '_-' if restricted else ' -'
elif char in '\\/|*<>':
return '_'
2012-11-28 11:59:27 +00:00
if restricted and (char in '!&\'()[]{}$;`^,#' or char.isspace()):
2012-11-28 01:04:46 +00:00
return '_'
if restricted and ord(char) > 127:
return '_'
return char
2014-11-17 06:16:12 +00:00
result = ''.join(map(replace_insane, s))
if not is_id:
while '__' in result:
result = result.replace('__', '_')
result = result.strip('_')
# Common case of "Foreign band name - English song title"
if restricted and result.startswith('-_'):
result = result[2:]
if not result:
result = '_'
2012-11-28 01:04:46 +00:00
return result
2014-11-23 19:41:03 +00:00
def orderedSet(iterable):
2012-11-28 01:04:46 +00:00
""" Remove all duplicates from the input iterable """
res = []
for el in iterable:
if el not in res:
res.append(el)
return res
2014-03-24 00:40:09 +00:00
2014-08-27 17:11:45 +00:00
def _htmlentity_transform(entity):
"""Transforms an HTML entity to a character."""
# Known non-numeric HTML entity
if entity in compat_html_entities.name2codepoint:
return compat_chr(compat_html_entities.name2codepoint[entity])
mobj = re.match(r'#(x?[0-9]+)', entity)
if mobj is not None:
numstr = mobj.group(1)
2014-11-17 06:16:12 +00:00
if numstr.startswith('x'):
2014-08-27 17:11:45 +00:00
base = 16
2014-11-17 06:16:12 +00:00
numstr = '0%s' % numstr
2014-08-27 17:11:45 +00:00
else:
base = 10
return compat_chr(int(numstr, base))
# Unknown entity in name, return its literal representation
2014-11-17 06:16:12 +00:00
return ('&%s;' % entity)
2014-08-27 17:11:45 +00:00
def unescapeHTML(s):
2014-03-24 00:40:09 +00:00
if s is None:
return None
assert type(s) == compat_str
2014-08-27 17:11:45 +00:00
return re.sub(
r'&([^;]+);', lambda m: _htmlentity_transform(m.group(1)), s)
def encodeFilename(s, for_subprocess=False):
2012-11-28 01:04:46 +00:00
"""
@param s The name of the file
"""
assert type(s) == compat_str
2012-11-28 01:04:46 +00:00
# Python 3 has a Unicode API
if sys.version_info >= (3, 0):
return s
2012-11-28 01:04:46 +00:00
if sys.platform == 'win32' and sys.getwindowsversion()[0] >= 5:
2014-11-17 06:16:12 +00:00
# Pass '' directly to use Unicode APIs on Windows 2000 and up
2012-11-28 01:04:46 +00:00
# (Detecting Windows NT 4 is tricky because 'major >= 4' would
# match Windows 9x series as well. Besides, NT 4 is obsolete.)
if not for_subprocess:
return s
else:
# For subprocess calls, encode with locale encoding
# Refer to http://stackoverflow.com/a/9951851/35070
encoding = preferredencoding()
2012-11-28 01:04:46 +00:00
else:
encoding = sys.getfilesystemencoding()
if encoding is None:
encoding = 'utf-8'
return s.encode(encoding, 'ignore')
def encodeArgument(s):
if not isinstance(s, compat_str):
# Legacy code that uses byte strings
# Uncomment the following line after fixing all post processors
2014-12-16 23:06:41 +00:00
# assert False, 'Internal error: %r should be of type %r, is %r' % (s, compat_str, type(s))
s = s.decode('ascii')
return encodeFilename(s, True)
def decodeOption(optval):
if optval is None:
return optval
if isinstance(optval, bytes):
optval = optval.decode(preferredencoding())
assert isinstance(optval, compat_str)
return optval
2014-11-23 19:41:03 +00:00
def formatSeconds(secs):
if secs > 3600:
return '%d:%02d:%02d' % (secs // 3600, (secs % 3600) // 60, secs % 60)
elif secs > 60:
return '%d:%02d' % (secs // 60, secs % 60)
else:
return '%d' % secs
2013-12-29 14:28:32 +00:00
def make_HTTPS_handler(opts_no_check_certificate, **kwargs):
if hasattr(ssl, 'create_default_context'): # Python >= 3.4 or 2.7.9
context = ssl.create_default_context(ssl.Purpose.CLIENT_AUTH)
if opts_no_check_certificate:
context.verify_mode = ssl.CERT_NONE
2014-12-13 22:27:21 +00:00
try:
return compat_urllib_request.HTTPSHandler(context=context, **kwargs)
except TypeError:
# Python 2.7.8
# (create_default_context present but HTTPSHandler has no context=)
pass
if sys.version_info < (3, 2):
import httplib
class HTTPSConnectionV3(httplib.HTTPSConnection):
def __init__(self, *args, **kwargs):
httplib.HTTPSConnection.__init__(self, *args, **kwargs)
def connect(self):
sock = socket.create_connection((self.host, self.port), self.timeout)
if getattr(self, '_tunnel_host', False):
self.sock = sock
self._tunnel()
try:
self.sock = ssl.wrap_socket(sock, self.key_file, self.cert_file, ssl_version=ssl.PROTOCOL_TLSv1)
2013-11-25 05:06:18 +00:00
except ssl.SSLError:
self.sock = ssl.wrap_socket(sock, self.key_file, self.cert_file, ssl_version=ssl.PROTOCOL_SSLv23)
class HTTPSHandlerV3(compat_urllib_request.HTTPSHandler):
def https_open(self, req):
return self.do_open(HTTPSConnectionV3, req)
2013-12-29 14:28:32 +00:00
return HTTPSHandlerV3(**kwargs)
else: # Python < 3.4
context = ssl.SSLContext(ssl.PROTOCOL_SSLv23)
2013-05-04 10:19:02 +00:00
context.verify_mode = (ssl.CERT_NONE
if opts_no_check_certificate
2013-05-04 10:19:02 +00:00
else ssl.CERT_REQUIRED)
context.set_default_verify_paths()
2013-12-29 14:28:32 +00:00
return compat_urllib_request.HTTPSHandler(context=context, **kwargs)
2013-05-04 10:19:02 +00:00
class ExtractorError(Exception):
"""Error during info extraction."""
2014-11-23 19:41:03 +00:00
def __init__(self, msg, tb=None, expected=False, cause=None, video_id=None):
""" tb, if given, is the original traceback (so that it can be printed out).
If expected is set, this is a normal error message and most likely not a bug in youtube-dl.
"""
if sys.exc_info()[0] in (compat_urllib_error.URLError, socket.timeout, UnavailableVideoError):
expected = True
if video_id is not None:
msg = video_id + ': ' + msg
2014-09-30 05:56:24 +00:00
if cause:
2014-11-17 06:16:12 +00:00
msg += ' (caused by %r)' % cause
if not expected:
if ytdl_is_updateable():
update_cmd = 'type youtube-dl -U to update'
else:
update_cmd = 'see https://yt-dl.org/update on how to update'
msg += '; please report this issue on https://yt-dl.org/bug .'
msg += ' Make sure you are using the latest version; %s.' % update_cmd
msg += ' Be sure to call youtube-dl with the --verbose flag and include its complete output.'
super(ExtractorError, self).__init__(msg)
self.traceback = tb
self.exc_info = sys.exc_info() # preserve original exception
2013-08-28 02:25:38 +00:00
self.cause = cause
self.video_id = video_id
def format_traceback(self):
if self.traceback is None:
return None
2014-11-17 06:16:12 +00:00
return ''.join(traceback.format_tb(self.traceback))
class UnsupportedError(ExtractorError):
def __init__(self, url):
super(UnsupportedError, self).__init__(
'Unsupported URL: %s' % url, expected=True)
self.url = url
class RegexNotFoundError(ExtractorError):
"""Error when a regex didn't match"""
pass
class DownloadError(Exception):
2012-11-28 01:04:46 +00:00
"""Download Error exception.
2012-11-28 01:04:46 +00:00
This exception may be thrown by FileDownloader objects if they are not
configured to continue on errors. They will contain the appropriate
error message.
"""
2014-11-23 19:41:03 +00:00
def __init__(self, msg, exc_info=None):
""" exc_info, if given, is the original exception that caused the trouble (as returned by sys.exc_info()). """
super(DownloadError, self).__init__(msg)
self.exc_info = exc_info
class SameFileError(Exception):
2012-11-28 01:04:46 +00:00
"""Same File exception.
2012-11-28 01:04:46 +00:00
This exception will be thrown by FileDownloader objects if they detect
multiple files would have to be downloaded to the same file on disk.
"""
pass
class PostProcessingError(Exception):
2012-11-28 01:04:46 +00:00
"""Post Processing exception.
2012-11-28 01:04:46 +00:00
This exception may be raised by PostProcessor's .run() method to
indicate an error in the postprocessing task.
"""
2014-11-23 19:41:03 +00:00
2013-01-12 14:07:59 +00:00
def __init__(self, msg):
self.msg = msg
2014-11-23 19:41:03 +00:00
class MaxDownloadsReached(Exception):
2012-11-28 01:04:46 +00:00
""" --max-downloads limit has been reached. """
pass
class UnavailableVideoError(Exception):
2012-11-28 01:04:46 +00:00
"""Unavailable Format exception.
2012-11-28 01:04:46 +00:00
This exception will be thrown when a video is requested
in a format that is not available for that video.
"""
pass
class ContentTooShortError(Exception):
2012-11-28 01:04:46 +00:00
"""Content Too Short exception.
2012-11-28 01:04:46 +00:00
This exception may be raised by FileDownloader objects when a file they
download is too small for what the server announced first, indicating
the connection was probably interrupted.
"""
# Both in bytes
downloaded = None
expected = None
2012-11-28 01:04:46 +00:00
def __init__(self, downloaded, expected):
self.downloaded = downloaded
self.expected = expected
2014-11-23 19:41:03 +00:00
class YoutubeDLHandler(compat_urllib_request.HTTPHandler):
2012-11-28 01:04:46 +00:00
"""Handler for HTTP requests and responses.
This class, when installed with an OpenerDirector, automatically adds
the standard headers to every HTTP request and handles gzipped and
deflated responses from web servers. If compression is to be avoided in
a particular request, the original request in the program code only has
to include the HTTP header "Youtubedl-No-Compression", which will be
removed before making the real request.
Part of this code was copied from:
http://techknack.net/python-urllib2-handlers/
Andrew Rowls, the author of that code, agreed to release it to the
public domain.
"""
@staticmethod
def deflate(data):
try:
return zlib.decompress(data, -zlib.MAX_WBITS)
except zlib.error:
return zlib.decompress(data)
@staticmethod
def addinfourl_wrapper(stream, headers, url, code):
if hasattr(compat_urllib_request.addinfourl, 'getcode'):
return compat_urllib_request.addinfourl(stream, headers, url, code)
ret = compat_urllib_request.addinfourl(stream, headers, url)
ret.code = code
return ret
def http_request(self, req):
for h, v in std_headers.items():
if h not in req.headers:
req.add_header(h, v)
2012-11-28 01:04:46 +00:00
if 'Youtubedl-no-compression' in req.headers:
if 'Accept-encoding' in req.headers:
del req.headers['Accept-encoding']
del req.headers['Youtubedl-no-compression']
if 'Youtubedl-user-agent' in req.headers:
if 'User-agent' in req.headers:
del req.headers['User-agent']
req.headers['User-agent'] = req.headers['Youtubedl-user-agent']
del req.headers['Youtubedl-user-agent']
if sys.version_info < (2, 7) and '#' in req.get_full_url():
# Python 2.6 is brain-dead when it comes to fragments
req._Request__original = req._Request__original.partition('#')[0]
req._Request__r_type = req._Request__r_type.partition('#')[0]
2012-11-28 01:04:46 +00:00
return req
def http_response(self, req, resp):
2012-11-28 01:04:46 +00:00
old_resp = resp
# gzip
if resp.headers.get('Content-encoding', '') == 'gzip':
content = resp.read()
gz = gzip.GzipFile(fileobj=io.BytesIO(content), mode='rb')
try:
uncompressed = io.BytesIO(gz.read())
except IOError as original_ioerror:
# There may be junk add the end of the file
# See http://stackoverflow.com/q/4928560/35070 for details
for i in range(1, 1024):
try:
gz = gzip.GzipFile(fileobj=io.BytesIO(content[:-i]), mode='rb')
uncompressed = io.BytesIO(gz.read())
except IOError:
continue
break
else:
raise original_ioerror
resp = self.addinfourl_wrapper(uncompressed, old_resp.headers, old_resp.url, old_resp.code)
2012-11-28 01:04:46 +00:00
resp.msg = old_resp.msg
# deflate
if resp.headers.get('Content-encoding', '') == 'deflate':
gz = io.BytesIO(self.deflate(resp.read()))
resp = self.addinfourl_wrapper(gz, old_resp.headers, old_resp.url, old_resp.code)
resp.msg = old_resp.msg
return resp
https_request = http_request
https_response = http_response
2014-02-06 10:29:46 +00:00
2014-05-17 17:04:02 +00:00
def parse_iso8601(date_str, delimiter='T'):
2014-03-24 00:40:09 +00:00
""" Return a UNIX timestamp from the given date """
if date_str is None:
return None
m = re.search(
r'(\.[0-9]+)?(?:Z$| ?(?P<sign>\+|-)(?P<hours>[0-9]{2}):?(?P<minutes>[0-9]{2})$)',
2014-03-24 00:40:09 +00:00
date_str)
if not m:
timezone = datetime.timedelta()
else:
date_str = date_str[:-len(m.group(0))]
if not m.group('sign'):
timezone = datetime.timedelta()
else:
sign = 1 if m.group('sign') == '+' else -1
timezone = datetime.timedelta(
hours=sign * int(m.group('hours')),
minutes=sign * int(m.group('minutes')))
date_format = '%Y-%m-%d{0}%H:%M:%S'.format(delimiter)
2014-05-17 17:04:02 +00:00
dt = datetime.datetime.strptime(date_str, date_format) - timezone
2014-03-24 00:40:09 +00:00
return calendar.timegm(dt.timetuple())
def unified_strdate(date_str, day_first=True):
"""Return a string with the date in the format YYYYMMDD"""
2014-03-21 13:38:37 +00:00
if date_str is None:
return None
upload_date = None
2014-11-23 19:41:03 +00:00
# Replace commas
2014-02-09 17:09:57 +00:00
date_str = date_str.replace(',', ' ')
# %z (UTC offset) is only supported in python>=3.2
2014-02-09 17:09:57 +00:00
date_str = re.sub(r' ?(\+|-)[0-9]{2}:?[0-9]{2}$', '', date_str)
# Remove AM/PM + timezone
date_str = re.sub(r'(?i)\s*(?:AM|PM)\s+[A-Z]+', '', date_str)
format_expressions = [
'%d %B %Y',
2014-02-16 20:47:03 +00:00
'%d %b %Y',
'%B %d %Y',
'%b %d %Y',
2014-06-28 18:02:02 +00:00
'%b %dst %Y %I:%M%p',
'%b %dnd %Y %I:%M%p',
'%b %dth %Y %I:%M%p',
'%Y-%m-%d',
'%Y/%m/%d',
2014-02-27 10:44:05 +00:00
'%d.%m.%Y',
'%d/%m/%Y',
'%d/%m/%y',
'%Y/%m/%d %H:%M:%S',
'%Y-%m-%d %H:%M:%S',
'%Y-%m-%d %H:%M:%S.%f',
'%d.%m.%Y %H:%M',
2014-03-11 21:18:43 +00:00
'%d.%m.%Y %H.%M',
'%Y-%m-%dT%H:%M:%SZ',
2013-11-20 05:13:19 +00:00
'%Y-%m-%dT%H:%M:%S.%fZ',
'%Y-%m-%dT%H:%M:%S.%f0Z',
'%Y-%m-%dT%H:%M:%S',
2014-02-23 12:00:51 +00:00
'%Y-%m-%dT%H:%M:%S.%f',
2014-02-06 10:29:46 +00:00
'%Y-%m-%dT%H:%M',
]
if day_first:
format_expressions.extend([
'%d/%m/%Y %H:%M:%S',
])
else:
format_expressions.extend([
'%m/%d/%Y %H:%M:%S',
])
for expression in format_expressions:
try:
upload_date = datetime.datetime.strptime(date_str, expression).strftime('%Y%m%d')
2014-02-06 10:29:46 +00:00
except ValueError:
pass
if upload_date is None:
timetuple = email.utils.parsedate_tz(date_str)
if timetuple:
upload_date = datetime.datetime(*timetuple[:6]).strftime('%Y%m%d')
return upload_date
2014-11-23 19:41:03 +00:00
2014-11-17 06:16:12 +00:00
def determine_ext(url, default_ext='unknown_video'):
2014-08-01 12:08:09 +00:00
if url is None:
return default_ext
2014-11-17 06:16:12 +00:00
guess = url.partition('?')[0].rpartition('.')[2]
2013-07-07 23:13:55 +00:00
if re.match(r'^[A-Za-z0-9]+$', guess):
return guess
else:
return default_ext
2013-07-07 23:13:55 +00:00
2014-11-23 19:41:03 +00:00
def subtitles_filename(filename, sub_lang, sub_format):
2014-11-17 06:16:12 +00:00
return filename.rsplit('.', 1)[0] + '.' + sub_lang + '.' + sub_format
2014-11-23 19:41:03 +00:00
def date_from_str(date_str):
"""
Return a datetime object from a string in the format YYYYMMDD or
(now|today)[+-][0-9](day|week|month|year)(s)?"""
today = datetime.date.today()
if date_str in ('now', 'today'):
return today
if date_str == 'yesterday':
return today - datetime.timedelta(days=1)
match = re.match('(now|today)(?P<sign>[+-])(?P<time>\d+)(?P<unit>day|week|month|year)(s)?', date_str)
if match is not None:
sign = match.group('sign')
time = int(match.group('time'))
if sign == '-':
time = -time
unit = match.group('unit')
2014-11-23 19:41:03 +00:00
# A bad aproximation?
if unit == 'month':
unit = 'day'
time *= 30
elif unit == 'year':
unit = 'day'
time *= 365
unit += 's'
delta = datetime.timedelta(**{unit: time})
return today + delta
return datetime.datetime.strptime(date_str, "%Y%m%d").date()
2014-11-23 19:41:03 +00:00
def hyphenate_date(date_str):
"""
Convert a date in 'YYYYMMDD' format to 'YYYY-MM-DD' format"""
match = re.match(r'^(\d\d\d\d)(\d\d)(\d\d)$', date_str)
if match is not None:
return '-'.join(match.groups())
else:
return date_str
2014-11-23 19:41:03 +00:00
class DateRange(object):
"""Represents a time interval between two dates"""
2014-11-23 19:41:03 +00:00
def __init__(self, start=None, end=None):
"""start and end must be strings in the format accepted by date"""
if start is not None:
self.start = date_from_str(start)
else:
self.start = datetime.datetime.min.date()
if end is not None:
self.end = date_from_str(end)
else:
self.end = datetime.datetime.max.date()
if self.start > self.end:
raise ValueError('Date range: "%s" , the start date must be before the end date' % self)
2014-11-23 19:41:03 +00:00
@classmethod
def day(cls, day):
"""Returns a range that only contains the given day"""
2014-11-23 19:41:03 +00:00
return cls(day, day)
def __contains__(self, date):
"""Check if the date is in the range"""
if not isinstance(date, datetime.date):
date = date_from_str(date)
return self.start <= date <= self.end
2014-11-23 19:41:03 +00:00
def __str__(self):
2014-11-23 19:41:03 +00:00
return '%s - %s' % (self.start.isoformat(), self.end.isoformat())
def platform_name():
""" Returns the platform name as a compat_str """
res = platform.platform()
if isinstance(res, bytes):
res = res.decode(preferredencoding())
assert isinstance(res, compat_str)
return res
def _windows_write_string(s, out):
""" Returns True if the string was written using special methods,
False if it has yet to be written out."""
# Adapted from http://stackoverflow.com/a/3259271/35070
import ctypes
import ctypes.wintypes
WIN_OUTPUT_IDS = {
1: -11,
2: -12,
}
try:
fileno = out.fileno()
except AttributeError:
# If the output stream doesn't have a fileno, it's virtual
return False
if fileno not in WIN_OUTPUT_IDS:
return False
GetStdHandle = ctypes.WINFUNCTYPE(
ctypes.wintypes.HANDLE, ctypes.wintypes.DWORD)(
(b"GetStdHandle", ctypes.windll.kernel32))
h = GetStdHandle(WIN_OUTPUT_IDS[fileno])
WriteConsoleW = ctypes.WINFUNCTYPE(
ctypes.wintypes.BOOL, ctypes.wintypes.HANDLE, ctypes.wintypes.LPWSTR,
ctypes.wintypes.DWORD, ctypes.POINTER(ctypes.wintypes.DWORD),
ctypes.wintypes.LPVOID)((b"WriteConsoleW", ctypes.windll.kernel32))
written = ctypes.wintypes.DWORD(0)
GetFileType = ctypes.WINFUNCTYPE(ctypes.wintypes.DWORD, ctypes.wintypes.DWORD)((b"GetFileType", ctypes.windll.kernel32))
FILE_TYPE_CHAR = 0x0002
FILE_TYPE_REMOTE = 0x8000
GetConsoleMode = ctypes.WINFUNCTYPE(
ctypes.wintypes.BOOL, ctypes.wintypes.HANDLE,
ctypes.POINTER(ctypes.wintypes.DWORD))(
(b"GetConsoleMode", ctypes.windll.kernel32))
INVALID_HANDLE_VALUE = ctypes.wintypes.DWORD(-1).value
def not_a_console(handle):
if handle == INVALID_HANDLE_VALUE or handle is None:
return True
return ((GetFileType(handle) & ~FILE_TYPE_REMOTE) != FILE_TYPE_CHAR
or GetConsoleMode(handle, ctypes.byref(ctypes.wintypes.DWORD())) == 0)
if not_a_console(h):
return False
def next_nonbmp_pos(s):
try:
return next(i for i, c in enumerate(s) if ord(c) > 0xffff)
except StopIteration:
return len(s)
while s:
count = min(next_nonbmp_pos(s), 1024)
ret = WriteConsoleW(
h, s, count if count else 2, ctypes.byref(written), None)
if ret == 0:
raise OSError('Failed to write string')
if not count: # We just wrote a non-BMP character
assert written.value == 2
s = s[1:]
else:
assert written.value > 0
s = s[written.value:]
return True
2014-04-07 17:57:42 +00:00
def write_string(s, out=None, encoding=None):
if out is None:
out = sys.stderr
assert type(s) == compat_str
if sys.platform == 'win32' and encoding is None and hasattr(out, 'fileno'):
if _windows_write_string(s, out):
return
if ('b' in getattr(out, 'mode', '') or
sys.version_info[0] < 3): # Python 2 lies about mode of sys.stderr
byt = s.encode(encoding or preferredencoding(), 'ignore')
out.write(byt)
elif hasattr(out, 'buffer'):
enc = encoding or getattr(out, 'encoding', None) or preferredencoding()
byt = s.encode(enc, 'ignore')
out.buffer.write(byt)
else:
out.write(s)
out.flush()
2013-08-28 12:28:55 +00:00
def bytes_to_intlist(bs):
if not bs:
return []
if isinstance(bs[0], int): # Python 3
return list(bs)
else:
return [ord(c) for c in bs]
2013-08-28 13:59:07 +00:00
def intlist_to_bytes(xs):
if not xs:
return b''
return struct_pack('%dB' % len(xs), *xs)
# Cross-platform file locking
if sys.platform == 'win32':
import ctypes.wintypes
import msvcrt
class OVERLAPPED(ctypes.Structure):
_fields_ = [
('Internal', ctypes.wintypes.LPVOID),
('InternalHigh', ctypes.wintypes.LPVOID),
('Offset', ctypes.wintypes.DWORD),
('OffsetHigh', ctypes.wintypes.DWORD),
('hEvent', ctypes.wintypes.HANDLE),
]
kernel32 = ctypes.windll.kernel32
LockFileEx = kernel32.LockFileEx
LockFileEx.argtypes = [
ctypes.wintypes.HANDLE, # hFile
ctypes.wintypes.DWORD, # dwFlags
ctypes.wintypes.DWORD, # dwReserved
ctypes.wintypes.DWORD, # nNumberOfBytesToLockLow
ctypes.wintypes.DWORD, # nNumberOfBytesToLockHigh
ctypes.POINTER(OVERLAPPED) # Overlapped
]
LockFileEx.restype = ctypes.wintypes.BOOL
UnlockFileEx = kernel32.UnlockFileEx
UnlockFileEx.argtypes = [
ctypes.wintypes.HANDLE, # hFile
ctypes.wintypes.DWORD, # dwReserved
ctypes.wintypes.DWORD, # nNumberOfBytesToLockLow
ctypes.wintypes.DWORD, # nNumberOfBytesToLockHigh
ctypes.POINTER(OVERLAPPED) # Overlapped
]
UnlockFileEx.restype = ctypes.wintypes.BOOL
whole_low = 0xffffffff
whole_high = 0x7fffffff
def _lock_file(f, exclusive):
overlapped = OVERLAPPED()
overlapped.Offset = 0
overlapped.OffsetHigh = 0
overlapped.hEvent = 0
f._lock_file_overlapped_p = ctypes.pointer(overlapped)
handle = msvcrt.get_osfhandle(f.fileno())
if not LockFileEx(handle, 0x2 if exclusive else 0x0, 0,
whole_low, whole_high, f._lock_file_overlapped_p):
raise OSError('Locking file failed: %r' % ctypes.FormatError())
def _unlock_file(f):
assert f._lock_file_overlapped_p
handle = msvcrt.get_osfhandle(f.fileno())
if not UnlockFileEx(handle, 0,
whole_low, whole_high, f._lock_file_overlapped_p):
raise OSError('Unlocking file failed: %r' % ctypes.FormatError())
else:
import fcntl
def _lock_file(f, exclusive):
fcntl.flock(f, fcntl.LOCK_EX if exclusive else fcntl.LOCK_SH)
def _unlock_file(f):
fcntl.flock(f, fcntl.LOCK_UN)
class locked_file(object):
def __init__(self, filename, mode, encoding=None):
assert mode in ['r', 'a', 'w']
self.f = io.open(filename, mode, encoding=encoding)
self.mode = mode
def __enter__(self):
exclusive = self.mode != 'r'
try:
_lock_file(self.f, exclusive)
except IOError:
self.f.close()
raise
return self
def __exit__(self, etype, value, traceback):
try:
_unlock_file(self.f)
finally:
self.f.close()
def __iter__(self):
return iter(self.f)
def write(self, *args):
return self.f.write(*args)
def read(self, *args):
return self.f.read(*args)
def get_filesystem_encoding():
encoding = sys.getfilesystemencoding()
return encoding if encoding is not None else 'utf-8'
def shell_quote(args):
quoted_args = []
encoding = get_filesystem_encoding()
for a in args:
if isinstance(a, bytes):
# We may get a filename encoded with 'encodeFilename'
a = a.decode(encoding)
quoted_args.append(pipes.quote(a))
2014-11-17 06:16:12 +00:00
return ' '.join(quoted_args)
2013-10-17 22:46:35 +00:00
def takewhile_inclusive(pred, seq):
""" Like itertools.takewhile, but include the latest evaluated element
(the first element so that Not pred(e)) """
for e in seq:
yield e
if not pred(e):
return
def smuggle_url(url, data):
""" Pass additional data in a URL for internal use. """
sdata = compat_urllib_parse.urlencode(
2014-11-17 06:16:12 +00:00
{'__youtubedl_smuggle': json.dumps(data)})
return url + '#' + sdata
def unsmuggle_url(smug_url, default=None):
2014-12-09 22:11:26 +00:00
if '#__youtubedl_smuggle' not in smug_url:
return smug_url, default
2014-11-17 06:16:12 +00:00
url, _, sdata = smug_url.rpartition('#')
jsond = compat_parse_qs(sdata)['__youtubedl_smuggle'][0]
data = json.loads(jsond)
return url, data
def format_bytes(bytes):
if bytes is None:
2014-11-17 06:16:12 +00:00
return 'N/A'
if type(bytes) is str:
bytes = float(bytes)
if bytes == 0.0:
exponent = 0
else:
exponent = int(math.log(bytes, 1024.0))
2014-11-17 06:16:12 +00:00
suffix = ['B', 'KiB', 'MiB', 'GiB', 'TiB', 'PiB', 'EiB', 'ZiB', 'YiB'][exponent]
converted = float(bytes) / float(1024 ** exponent)
2014-11-17 06:16:12 +00:00
return '%.2f%s' % (converted, suffix)
2013-12-09 17:29:07 +00:00
2014-11-25 08:54:54 +00:00
def parse_filesize(s):
if s is None:
return None
# The lower-case forms are of course incorrect and inofficial,
# but we support those too
_UNIT_TABLE = {
'B': 1,
'b': 1,
'KiB': 1024,
'KB': 1000,
'kB': 1024,
'Kb': 1000,
'MiB': 1024 ** 2,
'MB': 1000 ** 2,
'mB': 1024 ** 2,
'Mb': 1000 ** 2,
'GiB': 1024 ** 3,
'GB': 1000 ** 3,
'gB': 1024 ** 3,
'Gb': 1000 ** 3,
'TiB': 1024 ** 4,
'TB': 1000 ** 4,
'tB': 1024 ** 4,
'Tb': 1000 ** 4,
'PiB': 1024 ** 5,
'PB': 1000 ** 5,
'pB': 1024 ** 5,
'Pb': 1000 ** 5,
'EiB': 1024 ** 6,
'EB': 1000 ** 6,
'eB': 1024 ** 6,
'Eb': 1000 ** 6,
'ZiB': 1024 ** 7,
'ZB': 1000 ** 7,
'zB': 1024 ** 7,
'Zb': 1000 ** 7,
'YiB': 1024 ** 8,
'YB': 1000 ** 8,
'yB': 1024 ** 8,
'Yb': 1000 ** 8,
}
units_re = '|'.join(re.escape(u) for u in _UNIT_TABLE)
m = re.match(
r'(?P<num>[0-9]+(?:[,.][0-9]*)?)\s*(?P<unit>%s)' % units_re, s)
2014-11-25 08:54:54 +00:00
if not m:
return None
num_str = m.group('num').replace(',', '.')
mult = _UNIT_TABLE[m.group('unit')]
return int(float(num_str) * mult)
2014-11-25 08:54:54 +00:00
2013-12-09 17:29:07 +00:00
def get_term_width():
columns = compat_getenv('COLUMNS', None)
2013-12-09 17:29:07 +00:00
if columns:
return int(columns)
try:
sp = subprocess.Popen(
['stty', 'size'],
stdout=subprocess.PIPE, stderr=subprocess.PIPE)
out, err = sp.communicate()
return int(out.split()[1])
except:
pass
return None
2013-12-09 18:39:41 +00:00
def month_by_name(name):
""" Return the number of a month by (locale-independently) English name """
ENGLISH_NAMES = [
2014-11-17 06:16:12 +00:00
'January', 'February', 'March', 'April', 'May', 'June',
'July', 'August', 'September', 'October', 'November', 'December']
2013-12-09 18:39:41 +00:00
try:
return ENGLISH_NAMES.index(name) + 1
except ValueError:
return None
2013-12-10 20:03:53 +00:00
2014-01-20 21:11:34 +00:00
def fix_xml_ampersands(xml_str):
2013-12-10 20:03:53 +00:00
"""Replace all the '&' by '&amp;' in XML"""
2014-01-20 21:11:34 +00:00
return re.sub(
r'&(?!amp;|lt;|gt;|apos;|quot;|#x[0-9a-fA-F]{,4};|#[0-9]{,4};)',
2014-11-17 06:16:12 +00:00
'&amp;',
2014-01-20 21:11:34 +00:00
xml_str)
def setproctitle(title):
assert isinstance(title, compat_str)
try:
libc = ctypes.cdll.LoadLibrary("libc.so.6")
except OSError:
return
2014-03-23 13:28:22 +00:00
title_bytes = title.encode('utf-8')
buf = ctypes.create_string_buffer(len(title_bytes))
buf.value = title_bytes
try:
2014-03-23 13:28:22 +00:00
libc.prctl(15, buf, 0, 0, 0)
except AttributeError:
return # Strange libc, just skip this
2013-12-16 12:56:13 +00:00
def remove_start(s, start):
if s.startswith(start):
return s[len(start):]
return s
def remove_end(s, end):
if s.endswith(end):
return s[:-len(end)]
return s
def url_basename(url):
path = compat_urlparse.urlparse(url).path
2014-11-17 06:16:12 +00:00
return path.strip('/').split('/')[-1]
2013-12-20 16:05:28 +00:00
class HEADRequest(compat_urllib_request.Request):
def get_method(self):
return "HEAD"
2014-07-21 10:02:44 +00:00
def int_or_none(v, scale=1, default=None, get_attr=None, invscale=1):
if get_attr:
if v is not None:
v = getattr(v, get_attr, None)
if v == '':
v = None
2014-07-21 10:02:44 +00:00
return default if v is None else (int(v) * invscale // scale)
def str_or_none(v, default=None):
return default if v is None else compat_str(v)
2014-07-21 10:02:44 +00:00
def str_to_int(int_str):
2014-08-31 21:51:36 +00:00
""" A more relaxed version of int_or_none """
2014-07-21 10:02:44 +00:00
if int_str is None:
return None
2014-11-17 06:16:12 +00:00
int_str = re.sub(r'[,\.\+]', '', int_str)
2014-07-21 10:02:44 +00:00
return int(int_str)
2014-07-21 10:02:44 +00:00
def float_or_none(v, scale=1, invscale=1, default=None):
return default if v is None else (float(v) * invscale / scale)
def parse_duration(s):
if s is None:
return None
2014-08-30 23:41:30 +00:00
s = s.strip()
m = re.match(
2014-11-16 13:55:22 +00:00
r'''(?ix)T?
2014-12-04 16:35:40 +00:00
(?:
(?P<only_mins>[0-9.]+)\s*(?:mins?|minutes?)\s*|
(?P<only_hours>[0-9.]+)\s*(?:hours?)|
2014-11-16 13:55:22 +00:00
(?:
(?:(?P<hours>[0-9]+)\s*(?:[:h]|hours?)\s*)?
(?P<mins>[0-9]+)\s*(?:[:m]|mins?|minutes?)\s*
)?
2014-12-04 16:35:40 +00:00
(?P<secs>[0-9]+)(?P<ms>\.[0-9]+)?\s*(?:s|secs?|seconds?)?
)$''', s)
if not m:
return None
2014-12-04 16:35:40 +00:00
res = 0
if m.group('only_mins'):
return float_or_none(m.group('only_mins'), invscale=60)
if m.group('only_hours'):
return float_or_none(m.group('only_hours'), invscale=60 * 60)
if m.group('secs'):
res += int(m.group('secs'))
if m.group('mins'):
res += int(m.group('mins')) * 60
2014-12-04 16:35:40 +00:00
if m.group('hours'):
res += int(m.group('hours')) * 60 * 60
2014-08-25 10:59:53 +00:00
if m.group('ms'):
res += float(m.group('ms'))
return res
def prepend_extension(filename, ext):
2014-11-23 19:41:03 +00:00
name, real_ext = os.path.splitext(filename)
2014-11-17 06:16:12 +00:00
return '{0}.{1}{2}'.format(name, ext, real_ext)
def check_executable(exe, args=[]):
""" Checks if the given binary is installed somewhere in PATH, and returns its name.
args can be a list of arguments for a short output (like -version) """
try:
subprocess.Popen([exe] + args, stdout=subprocess.PIPE, stderr=subprocess.PIPE).communicate()
except OSError:
return False
return exe
def get_exe_version(exe, args=['--version'],
version_re=None, unrecognized='present'):
""" Returns the version of the specified executable,
or False if the executable is not present """
try:
out, _ = subprocess.Popen(
[exe] + args,
stdout=subprocess.PIPE, stderr=subprocess.STDOUT).communicate()
except OSError:
return False
if isinstance(out, bytes): # Python 2.x
out = out.decode('ascii', 'ignore')
return detect_exe_version(out, version_re, unrecognized)
def detect_exe_version(output, version_re=None, unrecognized='present'):
assert isinstance(output, compat_str)
if version_re is None:
version_re = r'version\s+([-0-9._a-zA-Z]+)'
m = re.search(version_re, output)
if m:
return m.group(1)
else:
return unrecognized
class PagedList(object):
2014-01-22 20:43:33 +00:00
def __len__(self):
# This is only useful for tests
return len(self.getslice())
class OnDemandPagedList(PagedList):
def __init__(self, pagefunc, pagesize):
self._pagefunc = pagefunc
self._pagesize = pagesize
def getslice(self, start=0, end=None):
res = []
for pagenum in itertools.count(start // self._pagesize):
firstid = pagenum * self._pagesize
nextfirstid = pagenum * self._pagesize + self._pagesize
if start >= nextfirstid:
continue
page_results = list(self._pagefunc(pagenum))
startv = (
start % self._pagesize
if firstid <= start < nextfirstid
else 0)
endv = (
((end - 1) % self._pagesize) + 1
if (end is not None and firstid <= end <= nextfirstid)
else None)
if startv != 0 or endv is not None:
page_results = page_results[startv:endv]
res.extend(page_results)
# A little optimization - if current page is not "full", ie. does
# not contain page_size videos then we can assume that this page
# is the last one - there are no more ids on further pages -
# i.e. no need to query again.
if len(page_results) + startv < self._pagesize:
break
# If we got the whole page, but the next page is not interesting,
# break out early as well
if end == nextfirstid:
break
return res
class InAdvancePagedList(PagedList):
def __init__(self, pagefunc, pagecount, pagesize):
self._pagefunc = pagefunc
self._pagecount = pagecount
self._pagesize = pagesize
def getslice(self, start=0, end=None):
res = []
start_page = start // self._pagesize
end_page = (
self._pagecount if end is None else (end // self._pagesize + 1))
skip_elems = start - start_page * self._pagesize
only_more = None if end is None else end - start
for pagenum in range(start_page, end_page):
page = list(self._pagefunc(pagenum))
if skip_elems:
page = page[skip_elems:]
skip_elems = None
if only_more is not None:
if len(page) < only_more:
only_more -= len(page)
else:
page = page[:only_more]
res.extend(page)
break
res.extend(page)
return res
def uppercase_escape(s):
2014-04-04 21:00:51 +00:00
unicode_escape = codecs.getdecoder('unicode_escape')
return re.sub(
r'\\U[0-9a-fA-F]{8}',
2014-04-04 21:00:51 +00:00
lambda m: unicode_escape(m.group(0))[0],
s)
2014-02-15 15:24:43 +00:00
def escape_rfc3986(s):
"""Escape non-ASCII characters as suggested by RFC 3986"""
if sys.version_info < (3, 0) and isinstance(s, unicode):
s = s.encode('utf-8')
2014-11-02 10:37:49 +00:00
return compat_urllib_parse.quote(s, b"%/;:@&=+$,!~*'()?#[]")
def escape_url(url):
"""Escape URL as suggested by RFC 3986"""
url_parsed = compat_urllib_parse_urlparse(url)
return url_parsed._replace(
path=escape_rfc3986(url_parsed.path),
params=escape_rfc3986(url_parsed.params),
query=escape_rfc3986(url_parsed.query),
fragment=escape_rfc3986(url_parsed.fragment)
).geturl()
2014-02-15 15:24:43 +00:00
try:
2014-11-17 06:16:12 +00:00
struct.pack('!I', 0)
2014-02-15 15:24:43 +00:00
except TypeError:
# In Python 2.6 (and some 2.7 versions), struct requires a bytes argument
def struct_pack(spec, *args):
if isinstance(spec, compat_str):
spec = spec.encode('ascii')
return struct.pack(spec, *args)
def struct_unpack(spec, *args):
if isinstance(spec, compat_str):
spec = spec.encode('ascii')
return struct.unpack(spec, *args)
else:
struct_pack = struct.pack
struct_unpack = struct.unpack
def read_batch_urls(batch_fd):
def fixup(url):
if not isinstance(url, compat_str):
url = url.decode('utf-8', 'replace')
2014-11-17 06:16:12 +00:00
BOM_UTF8 = '\xef\xbb\xbf'
if url.startswith(BOM_UTF8):
url = url[len(BOM_UTF8):]
url = url.strip()
if url.startswith(('#', ';', ']')):
return False
return url
with contextlib.closing(batch_fd) as fd:
return [url for url in map(fixup, fd) if url]
def urlencode_postdata(*args, **kargs):
return compat_urllib_parse.urlencode(*args, **kargs).encode('ascii')
try:
etree_iter = xml.etree.ElementTree.Element.iter
except AttributeError: # Python <=2.6
etree_iter = lambda n: n.findall('.//*')
def parse_xml(s):
class TreeBuilder(xml.etree.ElementTree.TreeBuilder):
def doctype(self, name, pubid, system):
pass # Ignore doctypes
parser = xml.etree.ElementTree.XMLParser(target=TreeBuilder())
kwargs = {'parser': parser} if sys.version_info >= (2, 7) else {}
tree = xml.etree.ElementTree.XML(s.encode('utf-8'), **kwargs)
# Fix up XML parser in Python 2.x
if sys.version_info < (3, 0):
for n in etree_iter(tree):
if n.text is not None:
if not isinstance(n.text, compat_str):
n.text = n.text.decode('utf-8')
return tree
2014-03-18 13:27:42 +00:00
2014-03-20 23:59:51 +00:00
US_RATINGS = {
'G': 0,
'PG': 10,
'PG-13': 13,
'R': 16,
'NC': 18,
}
2014-10-03 12:37:25 +00:00
def parse_age_limit(s):
if s is None:
return None
2014-10-03 12:37:25 +00:00
m = re.match(r'^(?P<age>\d{1,2})\+?$', s)
return int(m.group('age')) if m else US_RATINGS.get(s, None)
2014-10-03 12:37:25 +00:00
def strip_jsonp(code):
2014-11-13 15:28:05 +00:00
return re.sub(
r'(?s)^[a-zA-Z0-9_]+\s*\(\s*(.*)\);?\s*?(?://[^\n]*)*$', r'\1', code)
2014-04-21 05:12:02 +00:00
2014-08-22 00:33:29 +00:00
def js_to_json(code):
def fix_kv(m):
2014-09-30 09:12:59 +00:00
v = m.group(0)
if v in ('true', 'false', 'null'):
return v
if v.startswith('"'):
return v
if v.startswith("'"):
v = v[1:-1]
v = re.sub(r"\\\\|\\'|\"", lambda m: {
'\\\\': '\\\\',
"\\'": "'",
'"': '\\"',
}[m.group(0)], v)
return '"%s"' % v
2014-08-22 00:33:29 +00:00
res = re.sub(r'''(?x)
2014-09-30 09:12:59 +00:00
"(?:[^"\\]*(?:\\\\|\\")?)*"|
'(?:[^'\\]*(?:\\\\|\\')?)*'|
[a-zA-Z_][a-zA-Z_0-9]*
2014-08-22 00:33:29 +00:00
''', fix_kv, code)
res = re.sub(r',(\s*\])', lambda m: m.group(1), res)
return res
2014-04-21 05:12:02 +00:00
def qualities(quality_ids):
""" Get a numeric quality value out of a list of possible values """
def q(qid):
try:
return quality_ids.index(qid)
except ValueError:
return -1
return q
DEFAULT_OUTTMPL = '%(title)s-%(id)s.%(ext)s'
def limit_length(s, length):
""" Add ellipses to overly long strings """
if s is None:
return None
ELLIPSES = '...'
if len(s) > length:
return s[:length - len(ELLIPSES)] + ELLIPSES
return s
def version_tuple(v):
return tuple(int(e) for e in re.split(r'[-.]', v))
def is_outdated_version(version, limit, assume_new=True):
if not version:
return not assume_new
try:
return version_tuple(version) < version_tuple(limit)
except ValueError:
return not assume_new
def ytdl_is_updateable():
""" Returns if youtube-dl can be updated with -U """
from zipimport import zipimporter
return isinstance(globals().get('__loader__'), zipimporter) or hasattr(sys, 'frozen')
def args_to_str(args):
# Get a short string representation for a subprocess command
return ' '.join(shlex_quote(a) for a in args)
2015-01-04 01:20:45 +00:00
def urlhandle_detect_ext(url_handle):
try:
url_handle.headers
getheader = lambda h: url_handle.headers[h]
except AttributeError: # Python < 3
getheader = url_handle.info().getheader
return getheader('Content-Type').split("/")[1]
def age_restricted(content_limit, age_limit):
""" Returns True iff the content should be blocked """
if age_limit is None: # No limit set
return False
if content_limit is None:
return False # Content available for everyone
return age_limit < content_limit