Encoding fixes

talon/signature/learning/classifier.py

@@ -36,10 +36,10 @@ def load(saved_classifier_filename, train_data_filename):
         return joblib.load(saved_classifier_filename)
     except ValueError:
         import sys
-        pickle_options = {}
+        kwargs = {}
         if sys.version_info > (3, 0):
-            pickle_options["encoding"] = "bytes"
+            kwargs["encoding"] = "latin1"
 
-        loaded = pickle.load(open(saved_classifier_filename, 'rb'), **pickle_options)
+        loaded = pickle.load(open(saved_classifier_filename, 'rb'), **kwargs)
         joblib.dump(loaded, saved_classifier_filename, compress=True)
-        return loaded
+        return joblib.load(saved_classifier_filename)

talon/signature/learning/dataset.py

@@ -61,7 +61,7 @@ def parse_msg_sender(filename, sender_known=True):
     import sys
     kwargs = {}
     if sys.version_info > (3, 0):
-        kwargs["encoding"] = "bytes"
+        kwargs["encoding"] = "utf8"
 
     sender, msg = None, None
     if os.path.isfile(filename) and not is_sender_filename(filename):

tests/html_quotations_test.py

@@ -306,7 +306,7 @@ def extract_reply_and_check(filename):
     import sys
     kwargs = {}
     if sys.version_info > (3, 0):
-        kwargs["encoding"] = "bytes"
+        kwargs["encoding"] = "utf8"
 
     f = open(filename, **kwargs)
 

tests/signature/extraction_test.py

@@ -27,7 +27,7 @@ def test_messages_longer_SIGNATURE_MAX_LINES():
     import sys
     kwargs = {}
     if sys.version_info > (3, 0):
-        kwargs["encoding"] = "bytes"
+        kwargs["encoding"] = "utf8"
 
     for filename in os.listdir(STRIPPED):
         filename = os.path.join(STRIPPED, filename)