Fix #856 by ignoring combining diacritics when computing line length

pycodestyle.py

@@ -49,15 +49,16 @@
 """
 from __future__ import with_statement
 
+import bisect
 import inspect
 import keyword
 import os
 import re
 import sys
 import time
 import tokenize
+import unicodedata
 import warnings
-import bisect
 
 try:
     from functools import lru_cache
@@ -296,12 +297,15 @@ def maximum_line_length(physical_line, max_line_length, multiline,
             (len(chunks) == 2 and chunks[0] == '#')) and \
                 len(line) - len(chunks[-1]) < max_line_length - 7:
             return
-        if hasattr(line, 'decode'):   # Python 2
-            # The line could contain multi-byte characters
+        # Special case: multi-byte chars and combining diacritics
+        if sys.version_info >= (3,):
+            line_text = line
+        else:
             try:
-                length = len(line.decode('utf-8'))
-            except UnicodeError:
-                pass
+                line_text = line.decode('UTF-8')
+            except UnicodeDecodeError:
+                line_text = u''
+        length = sum(not unicodedata.combining(c) for c in line_text)
         if length > max_line_length:
             return (max_line_length, "E501 line too long "
                     "(%d > %d characters)" % (length, max_line_length))

testsuite/E50.py

@@ -59,6 +59,10 @@
 ('''
     ''' + ' \
 ')
+
+#
+#: Okay
+this = 'é́́́́xá́́́́mplé́́́́ há́́́́s ló́́́́ts ó́́́́f có́́́́mbí́́́́ní́́́́ng dí́́́́á́́́́crí́́́́tí́́́́cs, á́́́́nd thá́́́́t í́́́́s ó́́́́ká́́́́y.'
 #: E501 E225 E226
 very_long_identifiers=and_terrible_whitespace_habits(are_no_excuse+for_long_lines)
 #