+
+def _no_diacritics_regexp(query):
+ """ returns a regexp for searching for a query without diacritics
+
+ should be locale-aware """
+ names = {
+ u'a':u'aąĄ', u'c':u'cćĆ', u'e':u'eęĘ', u'l': u'lłŁ', u'n':u'nńŃ', u'o':u'oóÓ', u's':u'sśŚ', u'z':u'zźżŹŻ',
+ u'ą':u'ąĄ', u'ć':u'ćĆ', u'ę':u'ęĘ', u'ł': u'łŁ', u'ń':u'ńŃ', u'ó':u'óÓ', u'ś':u'śŚ', u'ź':u'źŹ', u'ż':u'żŻ'
+ }
+ def repl(m):
+ l = m.group()
+ return u"(%s)" % '|'.join(names[l])
+ return re.sub(u'[%s]'%(u''.join(names.keys())), repl, query)
+
+def unicode_re_escape(query):
+ """ Unicode-friendly version of re.escape """
+ return re.sub('(?u)(\W)', r'\\\1', query)
+
+def _word_starts_with(name, prefix):
+ """returns a Q object getting models having `name` contain a word
+ starting with `prefix`
+
+ We define word characters as alphanumeric and underscore, like in JS.
+
+ Works for MySQL, PostgreSQL, Oracle.
+ For SQLite, _sqlite* version is substituted for this.
+ """
+ kwargs = {}
+
+ prefix = _no_diacritics_regexp(unicode_re_escape(prefix))
+ # can't use [[:<:]] (word start),
+ # but we want both `xy` and `(xy` to catch `(xyz)`
+ kwargs['%s__iregex' % name] = u"(^|[^[:alnum:]_])%s" % prefix
+ print kwargs['%s__iregex' % name]
+
+ return Q(**kwargs)
+
+
+def _sqlite_word_starts_with(name, prefix):
+ """ version of _word_starts_with for SQLite
+
+ SQLite in Django uses Python re module
+ """
+ kwargs = {}
+ prefix = _no_diacritics_regexp(unicode_re_escape(prefix))
+ kwargs['%s__iregex' % name] = ur"(^|(?<=[^\wąćęłńóśźżĄĆĘŁŃÓŚŹŻ]))%s" % prefix
+ return Q(**kwargs)
+
+
+if settings.DATABASE_ENGINE == 'sqlite3':
+ _word_starts_with = _sqlite_word_starts_with
+
+
fnp / wolnelektury.git / blobdiff