collate/icu/__init__.py

   1 """ICU-based collation.
   2
   3 This collation backend uses the International Components for Unicode
   4 library to provide accurate and high-performance collation. It
   5 supports multiple locales and advanced sorting capabilities.
   6
   7 Use this collation backend if possible; it's by far the best.
   8
   9 Avoid this backend if...
  10  - ICU is not available for your system.
  11
  12 """
  13
  14 __all__ = ["Collator"]
  15
  16 import collate._abcollator
  17 import collate._locale
  18 import collate.errors
  19
  20 from collate.icu import _icu
  21
  22 class Collator(collate._abcollator.Collator):
  23     """ICU-based collation."""
  24
  25     def __init__(self, locale, encoding=None):
  26         super(Collator, self).__init__(locale, encoding)
  27         locale, encoding = collate._locale.getpair(locale, encoding)
  28         icu_locale = "root" if locale == "C" else locale
  29         self._collator = _icu.Collator(icu_locale)
  30         self.locale = self._collator.locale
  31         self.encoding = collate._locale.encoding(encoding)
  32         if self._collator.used_default_information and locale != "C":
  33             raise collate.errors.InvalidLocaleError(locale)
  34
  35         try:
  36             self._breaker = _icu.WordBreaker(icu_locale)
  37         except ValueError:
  38             # Thai is the only language with a special break locale,
  39             # so this is a harmless error.
  40             self._breaker = _icu.WordBreaker("root")
  41
  42     def words(self, string):
  43         """Split the string along word boundries."""
  44         if isinstance(string, str):
  45             string = string.decode(self.encoding)
  46         words = self._breaker.words(string)
  47         return [w for w in words if not w.isspace()]
  48
  49     def key(self, string):
  50         """Sort key for a string.
  51
  52         If the string is a str instance, it is decoded to a unicode
  53         instance according to the 'encoding' attribute of the
  54         Collator.
  55         """
  56         if isinstance(string, str):
  57             string = string.decode(self.encoding, 'replace')
  58         return self._collator.key(string)
  59