Add pre-processor stage after_filter

* Move `_match_entry` and `post_extract` to `process_video_result`. It is also left in `process_info` for API compat
* `--list-...` options and `--force-write-archive` now obey filtering options
* Move `SponsorBlockPP` to `after_filter`. Closes https://github.com/yt-dlp/yt-dlp/issues/2536
* Reverts 4ec82a72bb since this commit addresses the issue it was solving

README.md

@@ -982,15 +982,17 @@ ## Post-Processing Options:
                                      semicolon ";" delimited list of NAME=VALUE.
                                      The "when" argument determines when the
                                      postprocessor is invoked. It can be one of
-                                     "pre_process" (after extraction),
+                                     "pre_process" (after video extraction),
-                                     "before_dl" (before video download),
+                                     "after_filter" (after video passes filter),
-                                     "post_process" (after video download;
+                                     "before_dl" (before each video download),
-                                     default), "after_move" (after moving file
+                                     "post_process" (after each video download;
-                                     to their final locations), "after_video"
+                                     default), "after_move" (after moving video
-                                     (after downloading and processing all
+                                     file to it's final locations),
-                                     formats of a video), or "playlist" (end of
+                                     "after_video" (after downloading and
-                                     playlist). This option can be used multiple
+                                     processing all formats of a video), or
-                                     times to add different postprocessors
+                                     "playlist" (at end of playlist). This
+                                     option can be used multiple times to add
+                                     different postprocessors
 
 ## SponsorBlock Options:
 Make chapter entries for, or remove various segments (sponsor,

test/test_YoutubeDL.py

@@ -30,9 +30,7 @@ def __init__(self, *args, **kwargs):
         self.msgs = []
 
     def process_info(self, info_dict):
-        info_dict = info_dict.copy()
+        self.downloaded_info_dicts.append(info_dict.copy())
-        info_dict.pop('__original_infodict', None)
-        self.downloaded_info_dicts.append(info_dict)
 
     def to_screen(self, msg):
         self.msgs.append(msg)
@@ -898,20 +896,6 @@ def run(self, info):
         os.unlink(filename)
 
     def test_match_filter(self):
-        class FilterYDL(YDL):
-            def __init__(self, *args, **kwargs):
-                super(FilterYDL, self).__init__(*args, **kwargs)
-                self.params['simulate'] = True
-
-            def process_info(self, info_dict):
-                super(YDL, self).process_info(info_dict)
-
-            def _match_entry(self, info_dict, incomplete=False):
-                res = super(FilterYDL, self)._match_entry(info_dict, incomplete)
-                if res is None:
-                    self.downloaded_info_dicts.append(info_dict.copy())
-                return res
-
         first = {
             'id': '1',
             'url': TEST_URL,
@@ -939,7 +923,7 @@ def _match_entry(self, info_dict, incomplete=False):
         videos = [first, second]
 
         def get_videos(filter_=None):
-            ydl = FilterYDL({'match_filter': filter_})
+            ydl = YDL({'match_filter': filter_, 'simulate': True})
             for v in videos:
                 ydl.process_ie_result(v, download=True)
             return [v['id'] for v in ydl.downloaded_info_dicts]

yt_dlp/YoutubeDL.py

@@ -1037,8 +1037,7 @@ def validate_outtmpl(cls, outtmpl):
     @staticmethod
     def _copy_infodict(info_dict):
         info_dict = dict(info_dict)
-        for key in ('__original_infodict', '__postprocessors'):
+        info_dict.pop('__postprocessors', None)
-            info_dict.pop(key, None)
         return info_dict
 
     def prepare_outtmpl(self, outtmpl, info_dict, sanitize=False):
@@ -2512,8 +2511,6 @@ def is_wellformed(f):
         if '__x_forwarded_for_ip' in info_dict:
             del info_dict['__x_forwarded_for_ip']
 
-        # TODO Central sorting goes here
-
         if self.params.get('check_formats') is True:
             formats = LazyList(self._check_formats(formats[::-1]), reverse=True)
 
@@ -2526,6 +2523,12 @@ def is_wellformed(f):
 
         info_dict, _ = self.pre_process(info_dict)
 
+        if self._match_entry(info_dict) is not None:
+            return info_dict
+
+        self.post_extract(info_dict)
+        info_dict, _ = self.pre_process(info_dict, 'after_filter')
+
         # The pre-processors may have modified the formats
         formats = info_dict.get('formats', [info_dict])
 
@@ -2610,15 +2613,12 @@ def is_wellformed(f):
                     + ', '.join([f['format_id'] for f in formats_to_download]))
             max_downloads_reached = False
             for i, fmt in enumerate(formats_to_download):
-                formats_to_download[i] = new_info = dict(info_dict)
+                formats_to_download[i] = new_info = self._copy_infodict(info_dict)
-                # Save a reference to the original info_dict so that it can be modified in process_info if needed
                 new_info.update(fmt)
-                new_info['__original_infodict'] = info_dict
                 try:
                     self.process_info(new_info)
                 except MaxDownloadsReached:
                     max_downloads_reached = True
-                new_info.pop('__original_infodict')
                 # Remove copied info
                 for key, val in tuple(new_info.items()):
                     if info_dict.get(key) == val:
@@ -2826,7 +2826,7 @@ def existing_file(self, filepaths, *, default_overwrite=True):
         return None
 
     def process_info(self, info_dict):
-        """Process a single resolved IE result. (Modified it in-place)"""
+        """Process a single resolved IE result. (Modifies it in-place)"""
 
         assert info_dict.get('_type', 'video') == 'video'
         original_infodict = info_dict
@@ -2834,18 +2834,22 @@ def process_info(self, info_dict):
         if 'format' not in info_dict and 'ext' in info_dict:
             info_dict['format'] = info_dict['ext']
 
+        # This is mostly just for backward compatibility of process_info
+        # As a side-effect, this allows for format-specific filters
         if self._match_entry(info_dict) is not None:
             info_dict['__write_download_archive'] = 'ignore'
             return
 
+        # Does nothing under normal operation - for backward compatibility of process_info
         self.post_extract(info_dict)
-        self._num_downloads += 1
 
         # info_dict['_filename'] needs to be set for backward compatibility
         info_dict['_filename'] = full_filename = self.prepare_filename(info_dict, warn=True)
         temp_filename = self.prepare_filename(info_dict, 'temp')
         files_to_move = {}
 
+        self._num_downloads += 1
+
         # Forced printings
         self.__forced_printings(info_dict, full_filename, incomplete=('format' not in info_dict))
 
@@ -3259,17 +3263,14 @@ def sanitize_info(info_dict, remove_private_keys=False):
             return info_dict
         info_dict.setdefault('epoch', int(time.time()))
         info_dict.setdefault('_type', 'video')
-        remove_keys = {'__original_infodict'}  # Always remove this since this may contain a copy of the entire dict
+
-        keep_keys = ['_type']  # Always keep this to facilitate load-info-json
         if remove_private_keys:
-            remove_keys |= {
+            reject = lambda k, v: v is None or (k.startswith('_') and k != '_type') or k in {
                 'requested_downloads', 'requested_formats', 'requested_subtitles', 'requested_entries',
                 'entries', 'filepath', 'infojson_filename', 'original_url', 'playlist_autonumber',
             }
-            reject = lambda k, v: k not in keep_keys and (
-                k.startswith('_') or k in remove_keys or v is None)
         else:
-            reject = lambda k, v: k in remove_keys
+            reject = lambda k, v: False
 
         def filter_fn(obj):
             if isinstance(obj, dict):
@@ -3296,14 +3297,8 @@ def actual_post_extract(info_dict):
                     actual_post_extract(video_dict or {})
                 return
 
-            post_extractor = info_dict.get('__post_extractor') or (lambda: {})
+            post_extractor = info_dict.pop('__post_extractor', None) or (lambda: {})
-            extra = post_extractor().items()
+            info_dict.update(post_extractor())
-            info_dict.update(extra)
-            info_dict.pop('__post_extractor', None)
-
-            original_infodict = info_dict.get('__original_infodict') or {}
-            original_infodict.update(extra)
-            original_infodict.pop('__post_extractor', None)
 
         actual_post_extract(info_dict or {})
 

yt_dlp/__init__.py

@@ -474,8 +474,8 @@ def report_unplayable_conflict(opt_name, arg, default=False, allowed=None):
             'key': 'SponsorBlock',
             'categories': sponsorblock_query,
             'api': opts.sponsorblock_api,
-            # Run this immediately after extraction is complete
+            # Run this after filtering videos
-            'when': 'pre_process'
+            'when': 'after_filter'
         })
     if opts.parse_metadata:
         postprocessors.append({

yt_dlp/options.py

@@ -1550,11 +1550,11 @@ def _dict_from_options_callback(
             'and (optionally) arguments to be passed to it, separated by a colon ":". '
             'ARGS are a semicolon ";" delimited list of NAME=VALUE. '
             'The "when" argument determines when the postprocessor is invoked. '
-            'It can be one of "pre_process" (after extraction), '
+            'It can be one of "pre_process" (after video extraction), "after_filter" (after video passes filter), '
-            '"before_dl" (before video download), "post_process" (after video download; default), '
+            '"before_dl" (before each video download), "post_process" (after each video download; default), '
-            '"after_move" (after moving file to their final locations), '
+            '"after_move" (after moving video file to it\'s final locations), '
             '"after_video" (after downloading and processing all formats of a video), '
-            'or "playlist" (end of playlist). '
+            'or "playlist" (at end of playlist). '
             'This option can be used multiple times to add different postprocessors'))
 
     sponsorblock = optparse.OptionGroup(parser, 'SponsorBlock Options', description=(

yt_dlp/utils.py

@@ -3166,7 +3166,7 @@ def q(qid):
     return q
 
 
-POSTPROCESS_WHEN = {'pre_process', 'before_dl', 'after_move', 'post_process', 'after_video', 'playlist'}
+POSTPROCESS_WHEN = {'pre_process', 'after_filter', 'before_dl', 'after_move', 'post_process', 'after_video', 'playlist'}
 
 
 DEFAULT_OUTTMPL = {