Revert to using regex based sample detection

Closes #1964

sabnzbd/misc.py

@@ -43,6 +43,7 @@ from sabnzbd.filesystem import userxbit
 TAB_UNITS = ("", "K", "M", "G", "T", "P")
 RE_UNITS = re.compile(r"(\d+\.*\d*)\s*([KMGTP]?)", re.I)
 RE_VERSION = re.compile(r"(\d+)\.(\d+)\.(\d+)([a-zA-Z]*)(\d*)")
+RE_SAMPLE = re.compile(r"((^|[\W_])(sample|proof))", re.I)  # something-sample or something-proof
 RE_IP4 = re.compile(r"inet\s+(addr:\s*)?(\d+\.\d+\.\d+\.\d+)")
 RE_IP6 = re.compile(r"inet6\s+(addr:\s*)?([0-9a-f:]+)", re.I)
 
@@ -808,6 +809,11 @@ def get_all_passwords(nzo) -> List[str]:
     return unique_passwords
 
 
+def is_sample(filename: str) -> bool:
+    """Try to determine if filename is (most likely) a sample"""
+    return bool(re.search(RE_SAMPLE, filename))
+
+
 def find_on_path(targets):
     """Search the PATH for a program and return full path"""
     if sabnzbd.WIN32:

sabnzbd/postproc.py

@@ -39,7 +39,7 @@ from sabnzbd.newsunpack import (
     is_sfv_file,
 )
 from threading import Thread
-from sabnzbd.misc import on_cleanup_list
+from sabnzbd.misc import on_cleanup_list, is_sample
 from sabnzbd.filesystem import (
     real_path,
     get_unique_path,
@@ -65,7 +65,7 @@ from sabnzbd.filesystem import (
     get_filename,
 )
 from sabnzbd.nzbstuff import NzbObject
-from sabnzbd.sorting import Sorter, is_sample
+from sabnzbd.sorting import Sorter
 from sabnzbd.constants import (
     REPAIR_PRIORITY,
     FORCE_PRIORITY,

sabnzbd/sorting.py

@@ -39,6 +39,7 @@ from sabnzbd.filesystem import (
 )
 import sabnzbd.cfg as cfg
 from sabnzbd.constants import EXCLUDED_GUESSIT_PROPERTIES, IGNORED_MOVIE_FOLDERS
+from sabnzbd.misc import is_sample
 from sabnzbd.nzbstuff import NzbObject, scan_password
 
 # Do not rename .vob files as they are usually DVD's
@@ -630,40 +631,9 @@ def guess_what(name: str, sort_type: Optional[str] = None) -> MatchesDict:
         ):
             guess["type"] = "unknown"
 
-    # Remove sample indicators from groupnames, e.g. 'sample-groupname' or 'groupname-proof'
-    group = guess.get("release_group", "")
-    if group.lower().startswith(("sample-", "proof-")) or group.lower().endswith(("-sample", "-proof")):
-        # Set clean groupname
-        guess["release_group"] = re.sub("^(sample|proof)-|-(sample|proof)$", "", group, re.I)
-        # Add 'Sample' property to the guess
-        other = guess.get("other")
-        if not other:
-            guess.setdefault("other", "Sample")
-        else:
-            if "Sample" not in guess["other"]:
-                # Pre-existing 'other' may be a string or a list
-                try:
-                    guess["other"].append("Sample")
-                except AttributeError:
-                    guess["other"] = [other, "Sample"]
-
     return guess
 
 
-def is_sample(filename: str) -> bool:
-    """Try to determine if filename belongs to a sample"""
-    if os.path.splitext(filename)[0].lower().strip() in ("sample", "proof"):
-        # The entire filename is just 'sample.ext' or similar
-        return True
-
-    # If that didn't work, start guessing
-    guess = guess_what(filename).get("other", "")
-    if isinstance(guess, list):
-        return any(item in ("Sample", "Proof") for item in guess)
-    else:
-        return guess in ("Sample", "Proof")
-
-
 def path_subst(path: str, mapping: List[Tuple[str, str]]) -> str:
     """Replace the sort string elements in the path with the real values provided by the mapping;
     non-elements are copied verbatim."""

tests/test_misc.py

@@ -215,6 +215,44 @@ class TestMisc:
         os.unlink("test.key")
 
     @pytest.mark.parametrize(
+        "name, result",
+        [
+            ("Free.Open.Source.Movie.2001.1080p.WEB-DL.DD5.1.H264-FOSS", False),  # Not samples
+            ("Setup.exe", False),
+            ("23.123.hdtv-rofl", False),
+            ("Something.1080p.WEB-DL.DD5.1.H264-EMRG-sample", True),  # Samples
+            ("Something.1080p.WEB-DL.DD5.1.H264-EMRG-sample.ogg", True),
+            ("Sumtin_Else_1080p_WEB-DL_DD5.1_H264_proof-EMRG", True),
+            ("Wot.Eva.540i.WEB-DL.aac.H264-Groupie sample.mp4", True),
+            ("file-sample.mkv", True),
+            ("PROOF.JPG", True),
+            ("Bla.s01e02.title.1080p.aac-sample proof.mkv", True),
+            ("Bla.s01e02.title.1080p.aac-proof.mkv", True),
+            ("Bla.s01e02.title.1080p.aac sample proof.mkv", True),
+            ("Bla.s01e02.title.1080p.aac proof.mkv", True),
+            ("Lwtn.s08e26.1080p.web.h264-glhf-sample.par2", True),
+            ("Lwtn.s08e26.1080p.web.h264-glhf-sample.vol001-002.par2", True),
+            ("Look at That 2011 540i WEB-DL.H265-NoSample", False),
+        ],
+    )
+    def test_is_sample(self, name, result):
+        assert misc.is_sample(name) == result
+
+    @pytest.mark.parametrize(
+        "name, result",
+        [
+            ("Not Death Proof (2022) 1080p x264 (DD5.1) BE Subs", False),  # Try to trigger some false positives
+            ("Proof.of.Everything.(2042).4320p.x266-4U", False),
+            ("Crime_Scene_S01E13_Free_Sample_For_Sale_480p-OhDear", False),
+            ("Sample That 2011 480p WEB-DL.H265-aMiGo", False),
+            ("NOT A SAMPLE.JPG", False),
+        ],
+    )
+    def test_is_sample_known_false_positives(self, name, result):
+        """We know these fail, but don't have a better solution for them at the moment."""
+        assert misc.is_sample(name) != result
+
+    @pytest.mark.parametrize(
         "test_input, expected_output",
         [
             (["cmd1", 9, "cmd3"], '"cmd1" "9" "cmd3"'),  # sending all commands as valid string

tests/test_sorting.py

@@ -66,7 +66,7 @@ class TestSortingFunctions:
                     "country": "US",
                 },
             ),
-            ("Test Movie 720p HDTV AAC x265 sample-MYgroup", {"release_group": "MYgroup", "other": "Sample"}),
+            ("Test Movie 720p HDTV AAC x265 MYgroup-Sample", {"release_group": "MYgroup", "other": "Sample"}),
             (None, None),  # Jobname missing
             ("", None),
         ],
@@ -86,33 +86,6 @@ class TestSortingFunctions:
                 else:
                     assert guess[key] == value
 
-    @pytest.mark.parametrize(
-        "name, result",
-        [
-            ("Free.Open.Source.Movie.2001.1080p.WEB-DL.DD5.1.H264-FOSS", False),  # Not samples
-            ("Setup.exe", False),
-            ("23.123.hdtv-rofl", False),
-            ("Something.1080p.WEB-DL.DD5.1.H264-EMRG-sample", True),  # Samples
-            ("Something.1080p.WEB-DL.DD5.1.H264-EMRG-sample.ogg", True),
-            ("Sumtin_Else_1080p_WEB-DL_DD5.1_H264_proof-EMRG", True),
-            ("Wot.Eva.540i.WEB-DL.aac.H264-Groupie sample.mp4", True),
-            ("file-sample.mkv", True),
-            ("PROOF.JPG", True),
-            ("Bla.s01e02.title.1080p.aac-sample proof.mkv", True),
-            ("Bla.s01e02.title.1080p.aac-proof.mkv", True),
-            ("Bla.s01e02.title.1080p.aac sample proof.mkv", True),
-            ("Bla.s01e02.title.1080p.aac proof.mkv", True),
-            ("Not Death Proof (2022) 1080p x264 (DD5.1) BE Subs", False),  # Try to trigger some false positives
-            ("Proof.of.Everything.(2042).4320p.x266-4U", False),
-            ("Crime_Scene_S01E13_Free_Sample_For_Sale_480p-OhDear", False),
-            ("Sample That 2011 480p WEB-DL.H265-aMiGo", False),
-            ("Look at That 2011 540i WEB-DL.H265-NoSample", False),
-            ("NOT A SAMPLE.JPG", False),
-        ],
-    )
-    def test_is_sample(self, name, result):
-        assert sorting.is_sample(name) == result
-
     @pytest.mark.parametrize("platform", ["linux", "darwin", "win32"])
     @pytest.mark.parametrize(
         "path, result_unix, result_win",