From 1d41464ccb6ff66441947eef0305518e3ce79a77 Mon Sep 17 00:00:00 2001
From: kaliko <efrim@azylum.org>
Date: Fri, 1 Nov 2013 11:58:23 +0100
Subject: [PATCH] Add levenstein fuzzy method in SimaStr

---
 sima/client.py      | 14 +++-----------
 sima/lib/simastr.py | 42 +++++++++++++++++++++++++++++-------------
 2 files changed, 32 insertions(+), 24 deletions(-)

diff --git a/sima/client.py b/sima/client.py
index 929f662..fba50fd 100644
--- a/sima/client.py
+++ b/sima/client.py
@@ -22,7 +22,6 @@ except ImportError as err:
 from .lib.player import Player
 from .lib.track import Track
 from .lib.simastr import SimaStr
-from .utils.leven import levenshtein_ratio
 
 
 class PlayerError(Exception):
@@ -151,21 +150,14 @@ class PlayerClient(Player):
                 matching_artists.append(fuzz_art)
                 self.log.debug('"%s" matches "%s".' % (fuzz_art, artist))
                 return matching_artists
-            # Proceed with levenshtein and SimaStr
-            leven = levenshtein_ratio(artist.stripped.lower(),
-                    SimaStr(fuzz_art).stripped.lower())
-            # SimaStr string __eq__, not regular string comparison here
+            # SimaStr string __eq__ (not regular string comparison here)
             if artist == fuzz_art:
                 matching_artists.append(fuzz_art)
                 self.log.info('"%s" quite probably matches "%s" (SimaStr)' %
                               (fuzz_art, artist))
-            elif leven >= 0.82:  # PARAM
-                matching_artists.append(fuzz_art)
-                self.log.debug('FZZZ: "%s" should match "%s" (lr=%1.3f)' %
-                               (fuzz_art, artist, leven))
             else:
-                self.log.debug('FZZZ: "%s" does not match "%s" (lr=%1.3f)' %
-                               (fuzz_art, artist, leven))
+                self.log.debug('FZZZ: "%s" does not match "%s"' %
+                               (fuzz_art, artist))
         return matching_artists
 
     def find_album(self, artist, album):
diff --git a/sima/lib/simastr.py b/sima/lib/simastr.py
index 7e7668c..56cd242 100644
--- a/sima/lib/simastr.py
+++ b/sima/lib/simastr.py
@@ -24,12 +24,11 @@ SimaStr
 Special unicode() subclass to perform fuzzy match on specific strings with
 known noise.
 
-Artist names often contain a leading 'The ' which might, or might not be
-present. Some other noise sources in artist name are 'and' words :
-    'and'/'&'/'n'/'N'.
-
-The SimaStr() object removes these words and compute equality on "stripped"
-strings.
+ * SimaStr() object removes specific patterns from the string
+ * Diacritic are removed
+ * Equality test is done on lower-cased string
+ * Equality test is not an exact comparison, the levenshtein edition distance
+   between stripped and filtered strings is used
 
 >>> from simastr import SimaStr
 >>> art0 = SimaStr('The Desert Sessions & PJ Harvey')
@@ -38,9 +37,14 @@ strings.
 >>> True
 >>> art0 == 'Desert Sessions And PJ Harvey'
 >>> True
+>>> # diacritic filter + levenshtein  example
+>>> art0 = sima.lib.simastr.SimaStr('Hubert Félix Thiéphaine')
+>>> art1 = sima.lib.simastr.SimaStr('Hubert-Felix Thiephaine')
+>>> art0 == art1
+>>> True
 >>>
 
-Current stripped word patterns (usually English followed by French andx
+Current stripped word patterns (usually English followed by French and
 Spanish alternatives)
     leading (case-insensitive):
             "the","le","la","les","el","los"
@@ -50,10 +54,9 @@ Spanish alternatives)
             combination of "[- !?\.]+" "\(? ?[Ll]ive ?\)?"
 
 
-Possibility to access to stripped string :
+Possibility to access to stripped string:
 
 >>> art0 = SimaStr('The Desert Sessions & PJ Harvey')
->>> art.stripped
 >>> print (art0, art0.stripped)
 >>> ('The Desert Sessions & PJ Harvey', 'Desert Sessions PJ Harvey')
 
@@ -63,11 +66,14 @@ TODO:
 """
 
 __author__ = 'Jack Kaliko'
-__version__ = '0.3'
+__version__ = '0.4'
 
 # IMPORTS
+import unicodedata
 from re import (compile, U, I)
 
+from ..utils.leven import levenshtein_ratio
+
 
 class SimaStr(str):
     """
@@ -95,11 +101,12 @@ class SimaStr(str):
     def __init__(self, fuzzstr):
         """
         """
-        str().__init__(fuzzstr)
+        super().__init__(fuzzstr)
         self.orig = str(fuzzstr)
         self.stripped = str(fuzzstr.strip())
         # fuzzy computation
         self._get_root()
+        self.remove_diacritics()
 
     def _get_root(self):
         """
@@ -121,13 +128,22 @@ class SimaStr(str):
             #print sea.groupdict()
             self.stripped = sea.group('root0')
 
+    def remove_diacritics(self):
+        self.stripped = ''.join(x for x in
+                                unicodedata.normalize('NFKD', self.stripped)
+                                if unicodedata.category(x) != 'Mn')
+
     def __hash__(self):
         return hash(self.stripped)
 
     def __eq__(self, other):
         if not isinstance(other, SimaStr):
-            return hash(self) == hash(SimaStr(other))
-        return hash(self) == hash(other)
+            other = SimaStr(other)
+        levenr = levenshtein_ratio(self.stripped.lower(),
+                                   other.stripped.lower())
+        if hash(self) == hash(other):
+            return True
+        return levenr >= 0.82
 
     def __ne__(self, other):
         if not isinstance(other, SimaStr):
-- 
2.39.2