New approach - find split points based on Unicode categories.

[python-collate.git] / collate / _abcollator.py

diff --git a/collate/_abcollator.py b/collate/_abcollator.py
index bc43dc3..fdd7783 100644 (file)
--- a/collate/_abcollator.py
+++ b/collate/_abcollator.py
@@ -5,20 +5,14 @@ class Collator(object):
         """Return negative if a < b, zero if a == b, positive if a > b."""
         return cmp(self.key(string1), self.key(string2))
 
-    def words(self, string):
-        """Split the string into separate words.
-
-        This split is done using Unicode's definition of whitespace.
-        """
-        return string.split()
-
-    def sortemes(self, string):
-        words = []
-        for word in self.words(string):
-            words.extend(collate._strings.alnumsplit(word))
-        return filter(collate._strings.wordlike, words)
-
-    def sortemekey(self, string):
-        words = map(collate._strings.numeric, self.sortemes(string))
-        words = [(i, self.key(word)) for (i, word) in words]
-        return words
+    def sortemekey(self, string, invalid=float('inf')):
+        keys = []
+        for sorteme in collate._strings.sortemes(string):
+            num, alpha = collate._strings.numeric(sorteme, invalid)
+            if num == invalid:
+                keys.append(self.key(alpha))
+            else:
+                keys.append(num)
+        # Shove the sortkeyed original string on the end to resolve
+        # ties intelligently.
+        return (keys, self.key(string))