Embedded Subtitles provider: improve detection of potentially incomplete subtitles

libs/subliminal_patch/providers/embeddedsubtitles.py

@@ -117,6 +117,8 @@ class EmbeddedSubtitlesProvider(Provider):
             self._blacklist.add(path)
             streams = []
 
+        streams = _discard_possible_incomplete_subtitles(list(streams))
+
         if not streams:
             logger.debug("No subtitles found for container: %s", video)
 
@@ -260,6 +262,39 @@ def _check_hi_fallback(streams, languages):
             logger.debug("HI fallback not needed: %s", compatible_streams)
 
 
+def _discard_possible_incomplete_subtitles(streams):
+    """Check number_of_frames attributes from subtitle streams in order to find
+    supposedly incomplete subtitles"""
+    try:
+        max_frames = max(stream.number_of_frames for stream in streams)
+    except ValueError:
+        return []
+
+    # Blatantly assume there's nothing to discard as some ffprobe streams don't
+    # have number_of_frames tags
+    if not max_frames:
+        return streams
+
+    logger.debug("Checking possible incomplete subtitles (max frames: %d)", max_frames)
+
+    valid_streams = []
+
+    for stream in streams:
+        # 500 < 1200
+        if stream.number_of_frames < max_frames // 2:
+            logger.debug(
+                "Possible bad subtitle found: %s (%s frames - %s frames)",
+                stream,
+                stream.number_of_frames,
+                max_frames,
+            )
+            continue
+
+        valid_streams.append(stream)
+
+    return valid_streams
+
+
 def _is_fuse_rclone_mount(path: str):
     # Experimental!
 

tests/subliminal_patch/test_embeddedsubtitles.py

@@ -9,10 +9,11 @@ import subliminal_patch
 from subliminal_patch.core import Episode
 from subliminal_patch.core import Movie
 from subliminal_patch.exceptions import MustGetBlacklisted
-from subliminal_patch.providers.embeddedsubtitles import \
-    _MemoizedFFprobeVideoContainer
-from subliminal_patch.providers.embeddedsubtitles import \
-    EmbeddedSubtitlesProvider
+from subliminal_patch.providers.embeddedsubtitles import _MemoizedFFprobeVideoContainer
+from subliminal_patch.providers.embeddedsubtitles import EmbeddedSubtitlesProvider
+from subliminal_patch.providers.embeddedsubtitles import (
+    _discard_possible_incomplete_subtitles,
+)
 from subzero.language import Language
 
 
@@ -249,3 +250,30 @@ def test_memoized(video_single_language, mocker):
             ]
             is not None
         )
+
+
+@pytest.mark.parametrize(
+    "number_of_frames,expected_len",
+    [((34, 811), 1), ((0, 0), 2), ((811, 34), 1), ((900, 1000), 2), ((0, 900), 1)],
+)
+def test_discard_possible_incomplete_subtitles(number_of_frames, expected_len):
+    subtitle_1 = FFprobeSubtitleStream(
+        {
+            "index": 1,
+            "codec_name": "subrip",
+            "codec_long_name": "SubRip subtitle",
+            "disposition": {},
+            "tags": {"language": "eng", "NUMBER_OF_FRAMES": number_of_frames[0]},
+        }
+    )
+    subtitle_2 = FFprobeSubtitleStream(
+        {
+            "index": 2,
+            "codec_name": "subrip",
+            "codec_long_name": "SubRip subtitle",
+            "disposition": {},
+            "tags": {"language": "eng", "NUMBER_OF_FRAMES": number_of_frames[1]},
+        }
+    )
+    new_list = _discard_possible_incomplete_subtitles([subtitle_1, subtitle_2])
+    assert len(new_list) == expected_len