]> Raphaƫl G. Git Repositories - youtubedl/blobdiff - youtube_dl/YoutubeDL.py
Update changelog.
[youtubedl] / youtube_dl / YoutubeDL.py
index 74e4261680da06d9892b4c6f60a26481771c950b..1783ce01bf683ab0cfa349a1b040362f88cc1a47 100755 (executable)
@@ -4,8 +4,10 @@
 from __future__ import absolute_import, unicode_literals
 
 import collections
+import contextlib
 import datetime
 import errno
+import fileinput
 import io
 import itertools
 import json
@@ -19,23 +21,25 @@ import subprocess
 import socket
 import sys
 import time
+import tokenize
 import traceback
 
 if os.name == 'nt':
     import ctypes
 
 from .compat import (
-    compat_basestring,
     compat_cookiejar,
     compat_expanduser,
+    compat_get_terminal_size,
     compat_http_client,
     compat_kwargs,
     compat_str,
+    compat_tokenize_tokenize,
     compat_urllib_error,
     compat_urllib_request,
+    compat_urllib_request_DataHandler,
 )
 from .utils import (
-    escape_url,
     ContentTooShortError,
     date_from_str,
     DateRange,
@@ -46,28 +50,30 @@ from .utils import (
     ExtractorError,
     format_bytes,
     formatSeconds,
-    get_term_width,
     locked_file,
     make_HTTPS_handler,
     MaxDownloadsReached,
     PagedList,
     parse_filesize,
+    PerRequestProxyHandler,
     PostProcessingError,
     platform_name,
     preferredencoding,
     render_table,
     SameFileError,
     sanitize_filename,
+    sanitize_path,
     std_headers,
     subtitles_filename,
-    takewhile_inclusive,
     UnavailableVideoError,
     url_basename,
     version_tuple,
     write_json_file,
     write_string,
+    YoutubeDLCookieProcessor,
     YoutubeDLHandler,
     prepend_extension,
+    replace_extension,
     args_to_str,
     age_restricted,
 )
@@ -114,7 +120,7 @@ class YoutubeDL(object):
 
     username:          Username for authentication purposes.
     password:          Password for authentication purposes.
-    videopassword:     Password for acces a video.
+    videopassword:     Password for accessing a video.
     usenetrc:          Use netrc for authentication instead.
     verbose:           Print additional info to stdout.
     quiet:             Do not print messages to stdout.
@@ -131,10 +137,10 @@ class YoutubeDL(object):
                        (or video) as a single JSON line.
     simulate:          Do not download the video files.
     format:            Video format code. See options.py for more information.
-    format_limit:      Highest quality format to try.
     outtmpl:           Template for output names.
     restrictfilenames: Do not allow "&" and spaces in file names
     ignoreerrors:      Do not stop on download errors.
+    force_generic_extractor: Force downloader to use the generic extractor
     nooverwrites:      Prevent overwriting files.
     playliststart:     Playlist item to start at.
     playlistend:       Playlist item to end at.
@@ -181,6 +187,8 @@ class YoutubeDL(object):
     prefer_insecure:   Use HTTP instead of HTTPS to retrieve information.
                        At the moment, this is only supported by YouTube.
     proxy:             URL of the proxy server to use
+    cn_verification_proxy:  URL of the proxy to use for IP address verification
+                       on Chinese sites. (Experimental)
     socket_timeout:    Time to wait for unresponsive hosts, in seconds
     bidi_workaround:   Work around buggy terminals without bidirectional text
                        support, using fridibi
@@ -247,15 +255,16 @@ class YoutubeDL(object):
     hls_prefer_native: Use the native HLS downloader instead of ffmpeg/avconv.
 
     The following parameters are not used by YoutubeDL itself, they are used by
-    the FileDownloader:
+    the downloader (see youtube_dl/downloader/common.py):
     nopart, updatetime, buffersize, ratelimit, min_filesize, max_filesize, test,
     noresizebuffer, retries, continuedl, noprogress, consoletitle,
-    xattr_set_filesize.
+    xattr_set_filesize, external_downloader_args.
 
     The following options are used by the post processors:
     prefer_ffmpeg:     If True, use ffmpeg instead of avconv if both are available,
                        otherwise prefer avconv.
-    exec_cmd:          Arbitrary command to run after downloading
+    postprocessor_args: A list of additional command-line arguments for the
+                        postprocessor.
     """
 
     params = None
@@ -277,14 +286,18 @@ class YoutubeDL(object):
         self._num_downloads = 0
         self._screen_file = [sys.stdout, sys.stderr][params.get('logtostderr', False)]
         self._err_file = sys.stderr
-        self.params = params
+        self.params = {
+            # Default parameters
+            'nocheckcertificate': False,
+        }
+        self.params.update(params)
         self.cache = Cache(self)
 
         if params.get('bidi_workaround', False):
             try:
                 import pty
                 master, slave = pty.openpty()
-                width = get_term_width()
+                width = compat_get_terminal_size().columns
                 if width is None:
                     width_args = []
                 else:
@@ -317,8 +330,10 @@ class YoutubeDL(object):
                 'Set the LC_ALL environment variable to fix this.')
             self.params['restrictfilenames'] = True
 
-        if '%(stitle)s' in self.params.get('outtmpl', ''):
-            self.report_warning('%(stitle)s is deprecated. Use the %(title)s and the --restrict-filenames flag(which also secures %(uploader)s et al) instead.')
+        if isinstance(params.get('outtmpl'), bytes):
+            self.report_warning(
+                'Parameter outtmpl is bytes, but should be a unicode string. '
+                'Put  from __future__ import unicode_literals  at the top of your code file or consider switching to Python 3.x.')
 
         self._setup_opener()
 
@@ -565,7 +580,7 @@ class YoutubeDL(object):
             # to workaround encoding issues with subprocess on python2 @ Windows
             if sys.version_info < (3, 0) and sys.platform == 'win32':
                 filename = encodeFilename(filename, True).decode(preferredencoding())
-            return filename
+            return sanitize_path(filename)
         except ValueError as err:
             self.report_error('Error in output template: ' + str(err) + ' (encoding: ' + repr(preferredencoding()) + ')')
             return None
@@ -619,12 +634,15 @@ class YoutubeDL(object):
             info_dict.setdefault(key, value)
 
     def extract_info(self, url, download=True, ie_key=None, extra_info={},
-                     process=True):
+                     process=True, force_generic_extractor=False):
         '''
         Returns a list with a dictionary for each video we find.
         If 'download', also downloads the videos.
         extra_info is a dict containing the extra values to add to each result
-         '''
+        '''
+
+        if not ie_key and force_generic_extractor:
+            ie_key = 'Generic'
 
         if ie_key:
             ies = [self.get_info_extractor(ie_key)]
@@ -753,7 +771,9 @@ class YoutubeDL(object):
             if isinstance(ie_entries, list):
                 n_all_entries = len(ie_entries)
                 if playlistitems:
-                    entries = [ie_entries[i - 1] for i in playlistitems]
+                    entries = [
+                        ie_entries[i - 1] for i in playlistitems
+                        if -n_all_entries <= i - 1 < n_all_entries]
                 else:
                     entries = ie_entries[playliststart:playlistend]
                 n_entries = len(entries)
@@ -838,8 +858,8 @@ class YoutubeDL(object):
         else:
             raise Exception('Invalid result type: %s' % result_type)
 
-    def _apply_format_filter(self, format_spec, available_formats):
-        " Returns a tuple of the remaining format_spec and filtered formats "
+    def _build_format_filter(self, filter_spec):
+        " Returns a function to filter the formats according to the filter_spec "
 
         OPERATORS = {
             '<': operator.lt,
@@ -849,13 +869,13 @@ class YoutubeDL(object):
             '=': operator.eq,
             '!=': operator.ne,
         }
-        operator_rex = re.compile(r'''(?x)\s*\[
+        operator_rex = re.compile(r'''(?x)\s*
             (?P<key>width|height|tbr|abr|vbr|asr|filesize|fps)
             \s*(?P<op>%s)(?P<none_inclusive>\s*\?)?\s*
             (?P<value>[0-9.]+(?:[kKmMgGtTpPeEzZyY]i?[Bb]?)?)
-            \]$
+            $
             ''' % '|'.join(map(re.escape, OPERATORS.keys())))
-        m = operator_rex.search(format_spec)
+        m = operator_rex.search(filter_spec)
         if m:
             try:
                 comparison_value = int(m.group('value'))
@@ -866,7 +886,7 @@ class YoutubeDL(object):
                 if comparison_value is None:
                     raise ValueError(
                         'Invalid value %r in format specification %r' % (
-                            m.group('value'), format_spec))
+                            m.group('value'), filter_spec))
             op = OPERATORS[m.group('op')]
 
         if not m:
@@ -874,78 +894,283 @@ class YoutubeDL(object):
                 '=': operator.eq,
                 '!=': operator.ne,
             }
-            str_operator_rex = re.compile(r'''(?x)\s*\[
+            str_operator_rex = re.compile(r'''(?x)
                 \s*(?P<key>ext|acodec|vcodec|container|protocol)
                 \s*(?P<op>%s)(?P<none_inclusive>\s*\?)?
                 \s*(?P<value>[a-zA-Z0-9_-]+)
-                \s*\]$
+                \s*$
                 ''' % '|'.join(map(re.escape, STR_OPERATORS.keys())))
-            m = str_operator_rex.search(format_spec)
+            m = str_operator_rex.search(filter_spec)
             if m:
                 comparison_value = m.group('value')
                 op = STR_OPERATORS[m.group('op')]
 
         if not m:
-            raise ValueError('Invalid format specification %r' % format_spec)
+            raise ValueError('Invalid filter specification %r' % filter_spec)
 
         def _filter(f):
             actual_value = f.get(m.group('key'))
             if actual_value is None:
                 return m.group('none_inclusive')
             return op(actual_value, comparison_value)
-        new_formats = [f for f in available_formats if _filter(f)]
+        return _filter
+
+    def build_format_selector(self, format_spec):
+        def syntax_error(note, start):
+            message = (
+                'Invalid format specification: '
+                '{0}\n\t{1}\n\t{2}^'.format(note, format_spec, ' ' * start[1]))
+            return SyntaxError(message)
+
+        PICKFIRST = 'PICKFIRST'
+        MERGE = 'MERGE'
+        SINGLE = 'SINGLE'
+        GROUP = 'GROUP'
+        FormatSelector = collections.namedtuple('FormatSelector', ['type', 'selector', 'filters'])
+
+        def _parse_filter(tokens):
+            filter_parts = []
+            for type, string, start, _, _ in tokens:
+                if type == tokenize.OP and string == ']':
+                    return ''.join(filter_parts)
+                else:
+                    filter_parts.append(string)
+
+        def _remove_unused_ops(tokens):
+            # Remove operators that we don't use and join them with the sourrounding strings
+            # for example: 'mp4' '-' 'baseline' '-' '16x9' is converted to 'mp4-baseline-16x9'
+            ALLOWED_OPS = ('/', '+', ',', '(', ')')
+            last_string, last_start, last_end, last_line = None, None, None, None
+            for type, string, start, end, line in tokens:
+                if type == tokenize.OP and string == '[':
+                    if last_string:
+                        yield tokenize.NAME, last_string, last_start, last_end, last_line
+                        last_string = None
+                    yield type, string, start, end, line
+                    # everything inside brackets will be handled by _parse_filter
+                    for type, string, start, end, line in tokens:
+                        yield type, string, start, end, line
+                        if type == tokenize.OP and string == ']':
+                            break
+                elif type == tokenize.OP and string in ALLOWED_OPS:
+                    if last_string:
+                        yield tokenize.NAME, last_string, last_start, last_end, last_line
+                        last_string = None
+                    yield type, string, start, end, line
+                elif type in [tokenize.NAME, tokenize.NUMBER, tokenize.OP]:
+                    if not last_string:
+                        last_string = string
+                        last_start = start
+                        last_end = end
+                    else:
+                        last_string += string
+            if last_string:
+                yield tokenize.NAME, last_string, last_start, last_end, last_line
+
+        def _parse_format_selection(tokens, inside_merge=False, inside_choice=False, inside_group=False):
+            selectors = []
+            current_selector = None
+            for type, string, start, _, _ in tokens:
+                # ENCODING is only defined in python 3.x
+                if type == getattr(tokenize, 'ENCODING', None):
+                    continue
+                elif type in [tokenize.NAME, tokenize.NUMBER]:
+                    current_selector = FormatSelector(SINGLE, string, [])
+                elif type == tokenize.OP:
+                    if string == ')':
+                        if not inside_group:
+                            # ')' will be handled by the parentheses group
+                            tokens.restore_last_token()
+                        break
+                    elif inside_merge and string in ['/', ',']:
+                        tokens.restore_last_token()
+                        break
+                    elif inside_choice and string == ',':
+                        tokens.restore_last_token()
+                        break
+                    elif string == ',':
+                        if not current_selector:
+                            raise syntax_error('"," must follow a format selector', start)
+                        selectors.append(current_selector)
+                        current_selector = None
+                    elif string == '/':
+                        if not current_selector:
+                            raise syntax_error('"/" must follow a format selector', start)
+                        first_choice = current_selector
+                        second_choice = _parse_format_selection(tokens, inside_choice=True)
+                        current_selector = FormatSelector(PICKFIRST, (first_choice, second_choice), [])
+                    elif string == '[':
+                        if not current_selector:
+                            current_selector = FormatSelector(SINGLE, 'best', [])
+                        format_filter = _parse_filter(tokens)
+                        current_selector.filters.append(format_filter)
+                    elif string == '(':
+                        if current_selector:
+                            raise syntax_error('Unexpected "("', start)
+                        group = _parse_format_selection(tokens, inside_group=True)
+                        current_selector = FormatSelector(GROUP, group, [])
+                    elif string == '+':
+                        video_selector = current_selector
+                        audio_selector = _parse_format_selection(tokens, inside_merge=True)
+                        if not video_selector or not audio_selector:
+                            raise syntax_error('"+" must be between two format selectors', start)
+                        current_selector = FormatSelector(MERGE, (video_selector, audio_selector), [])
+                    else:
+                        raise syntax_error('Operator not recognized: "{0}"'.format(string), start)
+                elif type == tokenize.ENDMARKER:
+                    break
+            if current_selector:
+                selectors.append(current_selector)
+            return selectors
+
+        def _build_selector_function(selector):
+            if isinstance(selector, list):
+                fs = [_build_selector_function(s) for s in selector]
+
+                def selector_function(formats):
+                    for f in fs:
+                        for format in f(formats):
+                            yield format
+                return selector_function
+            elif selector.type == GROUP:
+                selector_function = _build_selector_function(selector.selector)
+            elif selector.type == PICKFIRST:
+                fs = [_build_selector_function(s) for s in selector.selector]
+
+                def selector_function(formats):
+                    for f in fs:
+                        picked_formats = list(f(formats))
+                        if picked_formats:
+                            return picked_formats
+                    return []
+            elif selector.type == SINGLE:
+                format_spec = selector.selector
+
+                def selector_function(formats):
+                    formats = list(formats)
+                    if not formats:
+                        return
+                    if format_spec == 'all':
+                        for f in formats:
+                            yield f
+                    elif format_spec in ['best', 'worst', None]:
+                        format_idx = 0 if format_spec == 'worst' else -1
+                        audiovideo_formats = [
+                            f for f in formats
+                            if f.get('vcodec') != 'none' and f.get('acodec') != 'none']
+                        if audiovideo_formats:
+                            yield audiovideo_formats[format_idx]
+                        # for audio only (soundcloud) or video only (imgur) urls, select the best/worst audio format
+                        elif (all(f.get('acodec') != 'none' for f in formats) or
+                              all(f.get('vcodec') != 'none' for f in formats)):
+                            yield formats[format_idx]
+                    elif format_spec == 'bestaudio':
+                        audio_formats = [
+                            f for f in formats
+                            if f.get('vcodec') == 'none']
+                        if audio_formats:
+                            yield audio_formats[-1]
+                    elif format_spec == 'worstaudio':
+                        audio_formats = [
+                            f for f in formats
+                            if f.get('vcodec') == 'none']
+                        if audio_formats:
+                            yield audio_formats[0]
+                    elif format_spec == 'bestvideo':
+                        video_formats = [
+                            f for f in formats
+                            if f.get('acodec') == 'none']
+                        if video_formats:
+                            yield video_formats[-1]
+                    elif format_spec == 'worstvideo':
+                        video_formats = [
+                            f for f in formats
+                            if f.get('acodec') == 'none']
+                        if video_formats:
+                            yield video_formats[0]
+                    else:
+                        extensions = ['mp4', 'flv', 'webm', '3gp', 'm4a', 'mp3', 'ogg', 'aac', 'wav']
+                        if format_spec in extensions:
+                            filter_f = lambda f: f['ext'] == format_spec
+                        else:
+                            filter_f = lambda f: f['format_id'] == format_spec
+                        matches = list(filter(filter_f, formats))
+                        if matches:
+                            yield matches[-1]
+            elif selector.type == MERGE:
+                def _merge(formats_info):
+                    format_1, format_2 = [f['format_id'] for f in formats_info]
+                    # The first format must contain the video and the
+                    # second the audio
+                    if formats_info[0].get('vcodec') == 'none':
+                        self.report_error('The first format must '
+                                          'contain the video, try using '
+                                          '"-f %s+%s"' % (format_2, format_1))
+                        return
+                    output_ext = (
+                        formats_info[0]['ext']
+                        if self.params.get('merge_output_format') is None
+                        else self.params['merge_output_format'])
+                    return {
+                        'requested_formats': formats_info,
+                        'format': '%s+%s' % (formats_info[0].get('format'),
+                                             formats_info[1].get('format')),
+                        'format_id': '%s+%s' % (formats_info[0].get('format_id'),
+                                                formats_info[1].get('format_id')),
+                        'width': formats_info[0].get('width'),
+                        'height': formats_info[0].get('height'),
+                        'resolution': formats_info[0].get('resolution'),
+                        'fps': formats_info[0].get('fps'),
+                        'vcodec': formats_info[0].get('vcodec'),
+                        'vbr': formats_info[0].get('vbr'),
+                        'stretched_ratio': formats_info[0].get('stretched_ratio'),
+                        'acodec': formats_info[1].get('acodec'),
+                        'abr': formats_info[1].get('abr'),
+                        'ext': output_ext,
+                    }
+                video_selector, audio_selector = map(_build_selector_function, selector.selector)
 
-        new_format_spec = format_spec[:-len(m.group(0))]
-        if not new_format_spec:
-            new_format_spec = 'best'
+                def selector_function(formats):
+                    formats = list(formats)
+                    for pair in itertools.product(video_selector(formats), audio_selector(formats)):
+                        yield _merge(pair)
 
-        return (new_format_spec, new_formats)
+            filters = [self._build_format_filter(f) for f in selector.filters]
 
-    def select_format(self, format_spec, available_formats):
-        while format_spec.endswith(']'):
-            format_spec, available_formats = self._apply_format_filter(
-                format_spec, available_formats)
-        if not available_formats:
-            return None
+            def final_selector(formats):
+                for _filter in filters:
+                    formats = list(filter(_filter, formats))
+                return selector_function(formats)
+            return final_selector
 
-        if format_spec == 'best' or format_spec is None:
-            return available_formats[-1]
-        elif format_spec == 'worst':
-            return available_formats[0]
-        elif format_spec == 'bestaudio':
-            audio_formats = [
-                f for f in available_formats
-                if f.get('vcodec') == 'none']
-            if audio_formats:
-                return audio_formats[-1]
-        elif format_spec == 'worstaudio':
-            audio_formats = [
-                f for f in available_formats
-                if f.get('vcodec') == 'none']
-            if audio_formats:
-                return audio_formats[0]
-        elif format_spec == 'bestvideo':
-            video_formats = [
-                f for f in available_formats
-                if f.get('acodec') == 'none']
-            if video_formats:
-                return video_formats[-1]
-        elif format_spec == 'worstvideo':
-            video_formats = [
-                f for f in available_formats
-                if f.get('acodec') == 'none']
-            if video_formats:
-                return video_formats[0]
-        else:
-            extensions = ['mp4', 'flv', 'webm', '3gp', 'm4a', 'mp3', 'ogg', 'aac', 'wav']
-            if format_spec in extensions:
-                filter_f = lambda f: f['ext'] == format_spec
-            else:
-                filter_f = lambda f: f['format_id'] == format_spec
-            matches = list(filter(filter_f, available_formats))
-            if matches:
-                return matches[-1]
-        return None
+        stream = io.BytesIO(format_spec.encode('utf-8'))
+        try:
+            tokens = list(_remove_unused_ops(compat_tokenize_tokenize(stream.readline)))
+        except tokenize.TokenError:
+            raise syntax_error('Missing closing/opening brackets or parenthesis', (0, len(format_spec)))
+
+        class TokenIterator(object):
+            def __init__(self, tokens):
+                self.tokens = tokens
+                self.counter = 0
+
+            def __iter__(self):
+                return self
+
+            def __next__(self):
+                if self.counter >= len(self.tokens):
+                    raise StopIteration()
+                value = self.tokens[self.counter]
+                self.counter += 1
+                return value
+
+            next = __next__
+
+            def restore_last_token(self):
+                self.counter -= 1
+
+        parsed_selector = _parse_format_selection(iter(TokenIterator(tokens)))
+        return _build_selector_function(parsed_selector)
 
     def _calc_headers(self, info_dict):
         res = std_headers.copy()
@@ -988,7 +1213,7 @@ class YoutubeDL(object):
                 t.get('preference'), t.get('width'), t.get('height'),
                 t.get('id'), t.get('url')))
             for i, t in enumerate(thumbnails):
-                if 'width' in t and 'height' in t:
+                if t.get('width') and t.get('height'):
                     t['resolution'] = '%dx%d' % (t['width'], t['height'])
                 if t.get('id') is None:
                     t['id'] = '%d' % i
@@ -1000,29 +1225,30 @@ class YoutubeDL(object):
             info_dict['display_id'] = info_dict['id']
 
         if info_dict.get('upload_date') is None and info_dict.get('timestamp') is not None:
-            # Working around negative timestamps in Windows
-            # (see http://bugs.python.org/issue1646728)
-            if info_dict['timestamp'] < 0 and os.name == 'nt':
-                info_dict['timestamp'] = 0
-            upload_date = datetime.datetime.utcfromtimestamp(
-                info_dict['timestamp'])
-            info_dict['upload_date'] = upload_date.strftime('%Y%m%d')
+            # Working around out-of-range timestamp values (e.g. negative ones on Windows,
+            # see http://bugs.python.org/issue1646728)
+            try:
+                upload_date = datetime.datetime.utcfromtimestamp(info_dict['timestamp'])
+                info_dict['upload_date'] = upload_date.strftime('%Y%m%d')
+            except (ValueError, OverflowError, OSError):
+                pass
+
+        subtitles = info_dict.get('subtitles')
+        if subtitles:
+            for _, subtitle in subtitles.items():
+                for subtitle_format in subtitle:
+                    if 'ext' not in subtitle_format:
+                        subtitle_format['ext'] = determine_ext(subtitle_format['url']).lower()
 
         if self.params.get('listsubtitles', False):
             if 'automatic_captions' in info_dict:
                 self.list_subtitles(info_dict['id'], info_dict.get('automatic_captions'), 'automatic captions')
-            self.list_subtitles(info_dict['id'], info_dict.get('subtitles'), 'subtitles')
+            self.list_subtitles(info_dict['id'], subtitles, 'subtitles')
             return
         info_dict['requested_subtitles'] = self.process_subtitles(
-            info_dict['id'], info_dict.get('subtitles'),
+            info_dict['id'], subtitles,
             info_dict.get('automatic_captions'))
 
-        # This extractors handle format selection themselves
-        if info_dict['extractor'] in ['Youku']:
-            if download:
-                self.process_info(info_dict)
-            return info_dict
-
         # We now pick which formats have to be downloaded
         if info_dict.get('formats') is None:
             # There's only one format available
@@ -1033,6 +1259,8 @@ class YoutubeDL(object):
         if not formats:
             raise ExtractorError('No video formats found!')
 
+        formats_dict = {}
+
         # We check that all the formats have the format and format_id fields
         for i, format in enumerate(formats):
             if 'url' not in format:
@@ -1040,6 +1268,18 @@ class YoutubeDL(object):
 
             if format.get('format_id') is None:
                 format['format_id'] = compat_str(i)
+            format_id = format['format_id']
+            if format_id not in formats_dict:
+                formats_dict[format_id] = []
+            formats_dict[format_id].append(format)
+
+        # Make sure all formats have unique format_id
+        for format_id, ambiguous_formats in formats_dict.items():
+            if len(ambiguous_formats) > 1:
+                for i, format in enumerate(ambiguous_formats):
+                    format['format_id'] = '%s-%d' % (format_id, i)
+
+        for i, format in enumerate(formats):
             if format.get('format') is None:
                 format['format'] = '{id} - {res}{note}'.format(
                     id=format['format_id'],
@@ -1055,12 +1295,6 @@ class YoutubeDL(object):
             full_format_info.update(format)
             format['http_headers'] = self._calc_headers(full_format_info)
 
-        format_limit = self.params.get('format_limit', None)
-        if format_limit:
-            formats = list(takewhile_inclusive(
-                lambda f: f['format_id'] != format_limit, formats
-            ))
-
         # TODO Central sorting goes here
 
         if formats[0] is not info_dict:
@@ -1078,58 +1312,17 @@ class YoutubeDL(object):
 
         req_format = self.params.get('format')
         if req_format is None:
-            req_format = 'best'
-        formats_to_download = []
-        # The -1 is for supporting YoutubeIE
-        if req_format in ('-1', 'all'):
-            formats_to_download = formats
-        else:
-            for rfstr in req_format.split(','):
-                # We can accept formats requested in the format: 34/5/best, we pick
-                # the first that is available, starting from left
-                req_formats = rfstr.split('/')
-                for rf in req_formats:
-                    if re.match(r'.+?\+.+?', rf) is not None:
-                        # Two formats have been requested like '137+139'
-                        format_1, format_2 = rf.split('+')
-                        formats_info = (self.select_format(format_1, formats),
-                                        self.select_format(format_2, formats))
-                        if all(formats_info):
-                            # The first format must contain the video and the
-                            # second the audio
-                            if formats_info[0].get('vcodec') == 'none':
-                                self.report_error('The first format must '
-                                                  'contain the video, try using '
-                                                  '"-f %s+%s"' % (format_2, format_1))
-                                return
-                            output_ext = (
-                                formats_info[0]['ext']
-                                if self.params.get('merge_output_format') is None
-                                else self.params['merge_output_format'])
-                            selected_format = {
-                                'requested_formats': formats_info,
-                                'format': '%s+%s' % (formats_info[0].get('format'),
-                                                     formats_info[1].get('format')),
-                                'format_id': '%s+%s' % (formats_info[0].get('format_id'),
-                                                        formats_info[1].get('format_id')),
-                                'width': formats_info[0].get('width'),
-                                'height': formats_info[0].get('height'),
-                                'resolution': formats_info[0].get('resolution'),
-                                'fps': formats_info[0].get('fps'),
-                                'vcodec': formats_info[0].get('vcodec'),
-                                'vbr': formats_info[0].get('vbr'),
-                                'stretched_ratio': formats_info[0].get('stretched_ratio'),
-                                'acodec': formats_info[1].get('acodec'),
-                                'abr': formats_info[1].get('abr'),
-                                'ext': output_ext,
-                            }
-                        else:
-                            selected_format = None
-                    else:
-                        selected_format = self.select_format(rf, formats)
-                    if selected_format is not None:
-                        formats_to_download.append(selected_format)
-                        break
+            req_format_list = []
+            if (self.params.get('outtmpl', DEFAULT_OUTTMPL) != '-' and
+                    info_dict['extractor'] in ['youtube', 'ted'] and
+                    not info_dict.get('is_live')):
+                merger = FFmpegMergerPP(self)
+                if merger.available and merger.can_merge():
+                    req_format_list.append('bestvideo+bestaudio')
+            req_format_list.append('best')
+            req_format = '/'.join(req_format_list)
+        format_selector = self.build_format_selector(req_format)
+        formats_to_download = list(format_selector(formats))
         if not formats_to_download:
             raise ExtractorError('requested format not available',
                                  expected=True)
@@ -1208,9 +1401,6 @@ class YoutubeDL(object):
         if len(info_dict['title']) > 200:
             info_dict['title'] = info_dict['title'][:197] + '...'
 
-        # Keep for backwards compatibility
-        info_dict['stitle'] = info_dict['title']
-
         if 'format' not in info_dict:
             info_dict['format'] = info_dict['ext']
 
@@ -1256,7 +1446,7 @@ class YoutubeDL(object):
             return
 
         try:
-            dn = os.path.dirname(encodeFilename(filename))
+            dn = os.path.dirname(sanitize_path(encodeFilename(filename)))
             if dn and not os.path.exists(dn):
                 os.makedirs(dn)
         except (OSError, IOError) as err:
@@ -1264,7 +1454,7 @@ class YoutubeDL(object):
             return
 
         if self.params.get('writedescription', False):
-            descfn = filename + '.description'
+            descfn = replace_extension(filename, 'description', info_dict.get('ext'))
             if self.params.get('nooverwrites', False) and os.path.exists(encodeFilename(descfn)):
                 self.to_screen('[info] Video description is already present')
             elif info_dict.get('description') is None:
@@ -1279,7 +1469,7 @@ class YoutubeDL(object):
                     return
 
         if self.params.get('writeannotations', False):
-            annofn = filename + '.annotations.xml'
+            annofn = replace_extension(filename, 'annotations.xml', info_dict.get('ext'))
             if self.params.get('nooverwrites', False) and os.path.exists(encodeFilename(annofn)):
                 self.to_screen('[info] Video annotations are already present')
             else:
@@ -1326,13 +1516,13 @@ class YoutubeDL(object):
                     return
 
         if self.params.get('writeinfojson', False):
-            infofn = os.path.splitext(filename)[0] + '.info.json'
+            infofn = replace_extension(filename, 'info.json', info_dict.get('ext'))
             if self.params.get('nooverwrites', False) and os.path.exists(encodeFilename(infofn)):
                 self.to_screen('[info] Video description metadata is already present')
             else:
                 self.to_screen('[info] Writing video description metadata as JSON to: ' + infofn)
                 try:
-                    write_json_file(info_dict, infofn)
+                    write_json_file(self.filter_requested_info(info_dict), infofn)
                 except (OSError, IOError):
                     self.report_error('Cannot write metadata to JSON file ' + infofn)
                     return
@@ -1352,24 +1542,57 @@ class YoutubeDL(object):
                 if info_dict.get('requested_formats') is not None:
                     downloaded = []
                     success = True
-                    merger = FFmpegMergerPP(self, not self.params.get('keepvideo'))
+                    merger = FFmpegMergerPP(self)
                     if not merger.available:
                         postprocessors = []
                         self.report_warning('You have requested multiple '
                                             'formats but ffmpeg or avconv are not installed.'
-                                            ' The formats won\'t be merged')
+                                            ' The formats won\'t be merged.')
                     else:
                         postprocessors = [merger]
-                    for f in info_dict['requested_formats']:
-                        new_info = dict(info_dict)
-                        new_info.update(f)
-                        fname = self.prepare_filename(new_info)
-                        fname = prepend_extension(fname, 'f%s' % f['format_id'])
-                        downloaded.append(fname)
-                        partial_success = dl(fname, new_info)
-                        success = success and partial_success
-                    info_dict['__postprocessors'] = postprocessors
-                    info_dict['__files_to_merge'] = downloaded
+
+                    def compatible_formats(formats):
+                        video, audio = formats
+                        # Check extension
+                        video_ext, audio_ext = audio.get('ext'), video.get('ext')
+                        if video_ext and audio_ext:
+                            COMPATIBLE_EXTS = (
+                                ('mp3', 'mp4', 'm4a', 'm4p', 'm4b', 'm4r', 'm4v'),
+                                ('webm')
+                            )
+                            for exts in COMPATIBLE_EXTS:
+                                if video_ext in exts and audio_ext in exts:
+                                    return True
+                        # TODO: Check acodec/vcodec
+                        return False
+
+                    filename_real_ext = os.path.splitext(filename)[1][1:]
+                    filename_wo_ext = (
+                        os.path.splitext(filename)[0]
+                        if filename_real_ext == info_dict['ext']
+                        else filename)
+                    requested_formats = info_dict['requested_formats']
+                    if self.params.get('merge_output_format') is None and not compatible_formats(requested_formats):
+                        info_dict['ext'] = 'mkv'
+                        self.report_warning(
+                            'Requested formats are incompatible for merge and will be merged into mkv.')
+                    # Ensure filename always has a correct extension for successful merge
+                    filename = '%s.%s' % (filename_wo_ext, info_dict['ext'])
+                    if os.path.exists(encodeFilename(filename)):
+                        self.to_screen(
+                            '[download] %s has already been downloaded and '
+                            'merged' % filename)
+                    else:
+                        for f in requested_formats:
+                            new_info = dict(info_dict)
+                            new_info.update(f)
+                            fname = self.prepare_filename(new_info)
+                            fname = prepend_extension(fname, 'f%s' % f['format_id'], new_info['ext'])
+                            downloaded.append(fname)
+                            partial_success = dl(fname, new_info)
+                            success = success and partial_success
+                        info_dict['__postprocessors'] = postprocessors
+                        info_dict['__files_to_merge'] = downloaded
                 else:
                     # Just a single file
                     success = dl(filename, info_dict)
@@ -1439,7 +1662,8 @@ class YoutubeDL(object):
         for url in url_list:
             try:
                 # It also downloads the videos
-                res = self.extract_info(url)
+                res = self.extract_info(
+                    url, force_generic_extractor=self.params.get('force_generic_extractor', False))
             except UnavailableVideoError:
                 self.report_error('unable to download video')
             except MaxDownloadsReached:
@@ -1452,8 +1676,11 @@ class YoutubeDL(object):
         return self._download_retcode
 
     def download_with_info_file(self, info_filename):
-        with io.open(info_filename, 'r', encoding='utf-8') as f:
-            info = json.load(f)
+        with contextlib.closing(fileinput.FileInput(
+                [info_filename], mode='r',
+                openhook=fileinput.hook_encoded('utf-8'))) as f:
+            # FileInput doesn't have a read method, we can't call json.load
+            info = self.filter_requested_info(json.loads('\n'.join(f)))
         try:
             self.process_ie_result(info, download=True)
         except DownloadError:
@@ -1465,6 +1692,12 @@ class YoutubeDL(object):
                 raise
         return self._download_retcode
 
+    @staticmethod
+    def filter_requested_info(info_dict):
+        return dict(
+            (k, v) for k, v in info_dict.items()
+            if k not in ['requested_formats', 'requested_subtitles'])
+
     def post_process(self, filename, ie_info):
         """Run all the postprocessors on the given file."""
         info = dict(ie_info)
@@ -1474,24 +1707,18 @@ class YoutubeDL(object):
             pps_chain.extend(ie_info['__postprocessors'])
         pps_chain.extend(self._pps)
         for pp in pps_chain:
-            keep_video = None
-            old_filename = info['filepath']
+            files_to_delete = []
             try:
-                keep_video_wish, info = pp.run(info)
-                if keep_video_wish is not None:
-                    if keep_video_wish:
-                        keep_video = keep_video_wish
-                    elif keep_video is None:
-                        # No clear decision yet, let IE decide
-                        keep_video = keep_video_wish
+                files_to_delete, info = pp.run(info)
             except PostProcessingError as e:
                 self.report_error(e.msg)
-            if keep_video is False and not self.params.get('keepvideo', False):
-                try:
+            if files_to_delete and not self.params.get('keepvideo', False):
+                for old_filename in files_to_delete:
                     self.to_screen('Deleting original file %s (pass -k to keep)' % old_filename)
-                    os.remove(encodeFilename(old_filename))
-                except (IOError, OSError):
-                    self.report_warning('Unable to remove downloaded video file')
+                    try:
+                        os.remove(encodeFilename(old_filename))
+                    except (IOError, OSError):
+                        self.report_warning('Unable to remove downloaded original file')
 
     def _make_archive_id(self, info_dict):
         # Future-proof against any change in case
@@ -1643,26 +1870,6 @@ class YoutubeDL(object):
 
     def urlopen(self, req):
         """ Start an HTTP download """
-
-        # According to RFC 3986, URLs can not contain non-ASCII characters, however this is not
-        # always respected by websites, some tend to give out URLs with non percent-encoded
-        # non-ASCII characters (see telemb.py, ard.py [#3412])
-        # urllib chokes on URLs with non-ASCII characters (see http://bugs.python.org/issue3991)
-        # To work around aforementioned issue we will replace request's original URL with
-        # percent-encoded one
-        req_is_string = isinstance(req, compat_basestring)
-        url = req if req_is_string else req.get_full_url()
-        url_escaped = escape_url(url)
-
-        # Substitute URL if any change after escaping
-        if url != url_escaped:
-            if req_is_string:
-                req = url_escaped
-            else:
-                req = compat_urllib_request.Request(
-                    url_escaped, data=req.data, headers=req.headers,
-                    origin_req_host=req.origin_req_host, unverifiable=req.unverifiable)
-
         return self._opener.open(req, timeout=self._socket_timeout)
 
     def print_debug_header(self):
@@ -1694,10 +1901,10 @@ class YoutubeDL(object):
             out = out.decode().strip()
             if re.match('[0-9a-f]+', out):
                 self._write_string('[debug] Git HEAD: ' + out + '\n')
-        except:
+        except Exception:
             try:
                 sys.exc_clear()
-            except:
+            except Exception:
                 pass
         self._write_string('[debug] Python version %s - %s\n' % (
             platform.python_version(), platform_name()))
@@ -1745,8 +1952,7 @@ class YoutubeDL(object):
             if os.access(opts_cookiefile, os.R_OK):
                 self.cookiejar.load()
 
-        cookie_processor = compat_urllib_request.HTTPCookieProcessor(
-            self.cookiejar)
+        cookie_processor = YoutubeDLCookieProcessor(self.cookiejar)
         if opts_proxy is not None:
             if opts_proxy == '':
                 proxies = {}
@@ -1757,13 +1963,15 @@ class YoutubeDL(object):
             # Set HTTPS proxy to HTTP one if given (https://github.com/rg3/youtube-dl/issues/805)
             if 'http' in proxies and 'https' not in proxies:
                 proxies['https'] = proxies['http']
-        proxy_handler = compat_urllib_request.ProxyHandler(proxies)
+        proxy_handler = PerRequestProxyHandler(proxies)
 
         debuglevel = 1 if self.params.get('debug_printtraffic') else 0
         https_handler = make_HTTPS_handler(self.params, debuglevel=debuglevel)
         ydlh = YoutubeDLHandler(self.params, debuglevel=debuglevel)
+        data_handler = compat_urllib_request_DataHandler()
         opener = compat_urllib_request.build_opener(
-            https_handler, proxy_handler, cookie_processor, ydlh)
+            proxy_handler, https_handler, cookie_processor, ydlh, data_handler)
+
         # Delete the default user-agent header, which would otherwise apply in
         # cases where our custom HTTP handler doesn't come into play
         # (See https://github.com/rg3/youtube-dl/issues/1309 for details)
@@ -1804,7 +2012,7 @@ class YoutubeDL(object):
             thumb_ext = determine_ext(t['url'], 'jpg')
             suffix = '_%s' % t['id'] if len(thumbnails) > 1 else ''
             thumb_display_id = '%s ' % t['id'] if len(thumbnails) > 1 else ''
-            thumb_filename = os.path.splitext(filename)[0] + suffix + '.' + thumb_ext
+            t['filename'] = thumb_filename = os.path.splitext(filename)[0] + suffix + '.' + thumb_ext
 
             if self.params.get('nooverwrites', False) and os.path.exists(encodeFilename(thumb_filename)):
                 self.to_screen('[%s] %s: Thumbnail %sis already present' %
@@ -1814,7 +2022,7 @@ class YoutubeDL(object):
                                (info_dict['extractor'], info_dict['id'], thumb_display_id))
                 try:
                     uf = self.urlopen(t['url'])
-                    with open(thumb_filename, 'wb') as thumbf:
+                    with open(encodeFilename(thumb_filename), 'wb') as thumbf:
                         shutil.copyfileobj(uf, thumbf)
                     self.to_screen('[%s] %s: Writing thumbnail %sto: %s' %
                                    (info_dict['extractor'], info_dict['id'], thumb_display_id, thumb_filename))