Word-splitting.

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

diff --git a/collate/syslocale.py b/collate/syslocale.py
index b9f9b92..e2aeed9 100644 (file)
--- a/collate/syslocale.py
+++ b/collate/syslocale.py
@@ -12,16 +12,18 @@ locale of all previous collators and anything else using the system
 locale information.
 
 Use this collation backend if...
- - You are on a system without ICU or UCA datafiles for the locale,
-   and DUCET results are not acceptable.
+ - You are on a system without ICU.
 
 Avoid this backend if...
- - ICU or UCA support is available for the current locale.
+ - ICU is available for the current locale.
  - You are sorting strings from alphabets outside the primary locale.
  - You need to support collating multiple locales at once.
+ - You need the same results across multiple platforms.
+ 
 """
 
 import locale
+import re
 
 import collate.errors
 import collate._abcollator
@@ -31,10 +33,12 @@ class Collator(collate._abcollator.Collator):
     """C library locale-based collation."""
 
     def __init__(self, locale_code, encoding=None):
+        locale_code, encoding = collate._locale.getpair(locale_code, encoding)
         try:
-            locale.setlocale(locale.LC_COLLATE, locale_code)
+            setlocale = locale_code + "." + encoding
+            locale.setlocale(locale.LC_COLLATE, setlocale)
         except locale.Error:
-            raise collate.errors.InvalidLocaleError(locale_code)
+            raise collate.errors.InvalidLocaleError(setlocale)
         self.locale = locale.getlocale(locale.LC_COLLATE)[0]
         self.encoding = collate._locale.encoding(encoding)
 
@@ -56,10 +60,16 @@ class Collator(collate._abcollator.Collator):
         If strs rather than unicodes are passed in, they are first
         decoded according to the 'encoding' attribute of the Collator.
         """
-
         if isinstance(a, str):
             a = a.decode(self.encoding, "replace")
         if isinstance(b, str):
             b = b.decode(self.encoding, "replace")
         return locale.strcoll(a, b)
                                   
+    def words(self, string, sep=re.compile(r"\W+", re.UNICODE)):
+        """Split the string into separate words.
+
+        This split is done using the locale's notion of a word boundry.
+        """
+        return re.split(sep, string)
+