improve sanitization

server/djrandom/scanner/metadata.py

@@ -6,9 +6,14 @@ import re
 # Compiling and including 'unicodedata' apparently makes it possible
 # to correctly interpret accented characters in unicode tag metadata.
 _nonalpha_pattern = re.compile(r'\W+', re.UNICODE)
+_spaces_pattern = re.compile(r'\s+', re.UNICODE)
 
 def normalize_string(s):
-    return _nonalpha_pattern.sub(' ', s).lower()
+    s = s.replace('_', ' ')
+    s = _nonalpha_pattern.sub(' ', s)
+    s = _spaces_pattern.sub(' ', s)
+    s = s.lower().strip()
+    return s
 
 
 def analyze_mp3(path):