Merge pull request #48 from isotoma/fix-setup-py3k

Fix reading README & CHANGES RST files when the system encoding is not UTF-8

setup.py

@@ -16,13 +16,15 @@ def find_version(*file_paths):
     raise RuntimeError("Unable to find version string.")
 
 
-if sys.version_info[0] < 3:
-    def read(*parts):
-        return codecs.open(os.path.join(os.path.dirname(__file__), *parts)).read()
+def read(*parts):
+    return codecs.open(os.path.join(os.path.dirname(__file__), *parts),
+                       encoding='utf8').read()
 
-else:
-    def read(*parts):
-        return open(os.path.join(os.path.dirname(__file__), *parts), 'r').read()
+
+try:
+    bytes
+except NameError:
+    bytes = str
 
 
 class UltraMagicString(object):
@@ -31,19 +33,26 @@ class UltraMagicString(object):
     http://stackoverflow.com/questions/1162338/whats-the-right-way-to-use-unicode-metadata-in-setup-py
     '''
     def __init__(self, value):
+        if not isinstance(value, bytes):
+            value = value.encode('utf8')
         self.value = value
 
-    def __str__(self):
+    def __bytes__(self):
         return self.value
 
     def __unicode__(self):
         return self.value.decode('UTF-8')
 
+    if sys.version_info[0] < 3:
+        __str__ = __bytes__
+    else:
+        __str__ = __unicode__
+
     def __add__(self, other):
-        return UltraMagicString(self.value + str(other))
+        return UltraMagicString(self.value + bytes(other))
 
     def split(self, *args, **kw):
-        return self.value.split(*args, **kw)
+        return str(self).split(*args, **kw)
 
 
 long_description = UltraMagicString('\n\n'.join((