unicode strings should not raise an error if used in repr.

Python 2 doesn't allow unicode objects in __repr__ methods. Therefore we need to encode them as utf-8 bytes.

jedi/_compatibility.py

@@ -180,3 +180,21 @@ def no_unicode_pprint(dct):
     import pprint
     s = pprint.pformat(dct)
     print(re.sub("u'", "'", s))
+
+
+def utf8_repr(func):
+    """
+    ``__repr__`` methods in Python 2 don't allow unicode objects to be
+    returned. Therefore cast them to utf-8 bytes in this decorator.
+    """
+    def wrapper(self):
+        result = func(self)
+        if isinstance(result, unicode):
+            return result.encode('utf-8')
+        else:
+            return result
+
+    if is_py3:
+        return func
+    else:
+        return wrapper

jedi/debug.py

@@ -1,4 +1,4 @@
-from jedi._compatibility import encoding, is_py3
+from jedi._compatibility import encoding, is_py3, u
 import inspect
 import os
 import time
@@ -63,13 +63,13 @@ def dbg(message, *args):
         mod = inspect.getmodule(frm[0])
         if not (mod.__name__ in ignored_modules):
             i = ' ' * _debug_indent
-            debug_function(NOTICE, i + 'dbg: ' + message % args)
+            debug_function(NOTICE, i + 'dbg: ' + message % tuple(u(repr(a)) for a in args))
 
 
 def warning(message, *args):
     if debug_function and enable_warning:
         i = ' ' * _debug_indent
-        debug_function(WARNING, i + 'warning: ' + message % args)
+        debug_function(WARNING, i + 'warning: ' + message % tuple(u(repr(a)) for a in args))
 
 
 def speed(name):

jedi/parser/tree.py

@@ -34,7 +34,7 @@ from inspect import cleandoc
 from itertools import chain
 import textwrap
 
-from jedi._compatibility import (Python3Method, encoding, is_py3,
+from jedi._compatibility import (Python3Method, encoding, is_py3, utf8_repr,
                                  literal_eval, use_metaclass, unicode)
 from jedi import cache
 
@@ -223,8 +223,9 @@ class Leaf(Base):
                     return None
                 return self.parent.children[i - 1]
 
+    @utf8_repr
     def __repr__(self):
-        return "<%s: %s>" % (type(self).__name__, repr(self.value))
+        return "<%s: %s>" % (type(self).__name__, self.value)
 
 
 class LeafWithNewLines(Leaf):
@@ -336,16 +337,6 @@ class Literal(LeafWithNewLines):
     def eval(self):
         return literal_eval(self.value)
 
-    def __repr__(self):
-        # TODO remove?
-        """
-        if is_py3:
-            s = self.literal
-        else:
-            s = self.literal.encode('ascii', 'replace')
-        """
-        return "<%s: %s>" % (type(self).__name__, self.value)
-
 
 class Number(Literal):
     type = 'number'
@@ -472,6 +463,7 @@ class BaseNode(Base):
         except AttributeError:
             return self.children[0]
 
+    @utf8_repr
     def __repr__(self):
         code = self.get_code().replace('\n', ' ')
         if not is_py3:

test/test_parser/test_parser.py

@@ -202,3 +202,8 @@ def test_param_splitting():
 
     check('def x(a, (b, c)):\n pass', ['a'])
     check('def x((b, c)):\n pass', [])
+
+
+def test_unicode_string():
+    s = pt.String(None, u'bö', (0, 0))
+    assert repr(s)  # Should not raise an Error!