Fix xapian indexer for unicode

Make regular expression pattern match words as unicode.

Author: rouilj

roundup/backends/indexer_xapian.py

@@ -95,7 +95,7 @@ def add_text(self, identifier, text, mime_type='text/plain'):
 
         for match in re.finditer(r'\b\w{%d,%d}\b'
                                  % (self.minlength, self.maxlength),
-                                 text.upper()):
+                                 text.upper(), re.UNICODE):
             word = match.group(0)
             if self.is_stopword(word):
                 continue

test/test_indexer.py

@@ -50,14 +50,6 @@
     skip_whoosh = mark_class(pytest.mark.skip(
         "Skipping Whoosh indexer tests: 'whoosh' not installed"))
 
-
-import sys
-if sys.version_info[0] > 2:
-    unicode_fails_py2 = lambda func, *args, **kwargs: func
-else:
-    unicode_fails_py2 = pytest.mark.xfail(
-       reason="Unicode indexing expected to fail under python 2")
-    
 class db:
     class config(dict):
         DATABASE = 'test-index'
@@ -165,7 +157,6 @@ def test_manyresults(self):
             self.dex.add_text(('test', str(i), 'many'), 'many')
         self.assertEqual(len(self.dex.find(['many'])), 123)
 
-    @unicode_fails_py2
     def test_unicode(self):
         """Test with unicode words. see:
            https://issues.roundup-tracker.org/issue1344046"""