Use word boundaries for alnum seps and no boundaries for symbols

Author: JOJ0

beetsplug/lastgenre/__init__.py

@@ -44,15 +44,15 @@
 )
 
 DEFAULT_ARTIST_SEPARATORS = [
-    "feat\\.",
+    "feat.",
     "featuring",
     "&",
-    "vs\\.",
-    "\\bx\\b",  # Match "x" only as whole word
+    "vs.",
+    "x",  # Match "x" only as whole word
     "/",
     "+",
     "and",
-    "\\|",
+    "|",
 ]
 
 
@@ -100,10 +100,18 @@ def split_on_separators(text, separators):
     if not seps:
         return [text]
 
-    # One regex: separators at token boundaries (no letters/digits touching)
-    # (?<!\S) == start or whitespace; (?!\S) == end or whitespace
-    alt = "|".join(re.escape(s) for s in seps)  # escape special chars
-    pattern = rf"(?<!\S)(?:{alt})(?!\S)"
+    # Build patterns: word boundaries for pure alphanumeric, no boundaries for others
+    patterns = []
+    for s in seps:
+        escaped = re.escape(s)
+        if s.replace(" ", "").isalnum():  # treat spaced separators like symbols
+            # Alphanumeric needs word boundaries (like "x", "and")
+            patterns.append(rf"\b{escaped}\b")
+        else:
+            # Symbols like "/", " / " need no boundaries
+            patterns.append(escaped)
+
+    pattern = "|".join(patterns)
 
     if not re.search(pattern, text, flags=re.IGNORECASE):
         return [text]