Fix normalizing string subclasses with incompatible __new__.

pekkaklarck · pekkaklarck · commit 6d597531424d · 2018-03-16T11:38:23.000+02:00
Fixes robotframework#2739.
diff --git a/src/robot/utils/normalizing.py b/src/robot/utils/normalizing.py
@@ -17,7 +17,7 @@
 from collections import MutableMapping
 
 from .platform import PY3, IRONPYTHON
-from .robottypes import is_dict_like
+from .robottypes import is_dict_like, is_unicode
 
 
 def normalize(string, ignore=(), caseless=True, spaceless=True):
@@ -26,7 +26,7 @@ def normalize(string, ignore=(), caseless=True, spaceless=True):
     By default string is turned to lower case and all whitespace is removed.
     Additional characters can be removed by giving them in ``ignore`` list.
     """
-    empty = type(string)()
+    empty = u'' if is_unicode(string) else b''
     if PY3 and isinstance(ignore, bytes):
         # Iterating bytes in Python3 yields integers.
         ignore = [bytes([i]) for i in ignore]
diff --git a/utest/utils/test_normalizing.py b/utest/utils/test_normalizing.py
@@ -56,6 +56,14 @@ def test_bytes(self):
         self._verify(b'FOO & Bar', b'bar', ignore=b'F&o')
         self._verify(b'FOO & Bar', b'OOBar', ignore=b'F&o', caseless=False)
 
+    def test_string_subclass_without_compatible_init(self):
+        class BrokenLikeSudsText(str):
+            def __new__(cls, value):
+                return str.__new__(cls, value)
+        self._verify(BrokenLikeSudsText('suds.sax.text.Text is BROKEN'),
+                     'suds.sax.text.textisbroken')
+        self._verify(BrokenLikeSudsText(''), '')
+
 
 class TestNormalizedDict(unittest.TestCase):
 
