[YoutubeDL] format spec: correctly handle dashes and other unused operators

[youtube-dl] / youtube_dl / YoutubeDL.py

diff --git a/youtube_dl/YoutubeDL.py b/youtube_dl/YoutubeDL.py
index da7c510083820353d64db31a84dec7ccfac85c3a..1446b3254f0085cf933d368b7109723af3720fba 100755 (executable)
--- a/youtube_dl/YoutubeDL.py
+++ b/youtube_dl/YoutubeDL.py
@@ -264,6 +264,8 @@ class YoutubeDL(object):
     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.
+    postprocessor_args: A list of additional command-line arguments for the
+                        postprocessor.
     """
 
     params = None
@@ -931,6 +933,37 @@ class YoutubeDL(object):
                 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
@@ -958,6 +991,8 @@ class YoutubeDL(object):
                         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), [])
@@ -1107,7 +1142,7 @@ class YoutubeDL(object):
 
         stream = io.BytesIO(format_spec.encode('utf-8'))
         try:
-            tokens = list(compat_tokenize_tokenize(stream.readline))
+            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)))
 
@@ -1175,7 +1210,7 @@ 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
@@ -1269,7 +1304,8 @@ class YoutubeDL(object):
         if req_format is None:
             req_format_list = []
             if (self.params.get('outtmpl', DEFAULT_OUTTMPL) != '-' and
-                    info_dict['extractor'] in ['youtube', 'ted']):
+                    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')