Centralize some unicode and exception helpers into pyversion.

While we're at it, teach format_exception() to use the new
exc_to_unicode() helper derived from work done in capture.py.

3 files changed, 59 insertions, 51 deletions

diff --git a/nose/plugins/capture.py b/nose/plugins/capture.py
index d9041d8..fa4e5dc 100644
--- a/nose/plugins/capture.py
+++ b/nose/plugins/capture.py
@@ -13,7 +13,7 @@ import logging
 import os
 import sys
 from nose.plugins.base import Plugin
-from nose.pyversion import is_base_exception
+from nose.pyversion import exc_to_unicode, force_unicode
 from nose.util import ln
 from StringIO import StringIO
 
@@ -87,30 +87,8 @@ class Capture(Plugin):
         return self.formatError(test, err)
 
     def addCaptureToErr(self, ev, output):
-        if is_base_exception(ev):
-            if hasattr(ev, '__unicode__'):
-                # 2.6+
-                try:
-                    ev = unicode(ev)
-                except UnicodeDecodeError:
-                    # We need a unicode string... take our best shot at getting,
-                    # since we don't know what the original encoding is in.
-                    ev = str(ev).decode('utf8', 'replace')
-            else:
-                # 2.5-
-                if not hasattr(ev, 'message'):
-                    # 2.4
-                    msg = len(ev.args) and ev.args[0] or ''
-                else:
-                    msg = ev.message
-                if (isinstance(msg, basestring) and
-                    not isinstance(msg, unicode)):
-                    msg = msg.decode('utf8', 'replace')
-                ev = u'%s: %s' % (ev.__class__.__name__, msg)
-        elif not isinstance(ev, basestring):
-            ev = repr(ev)
-        if not isinstance(output, unicode):
-            output = output.decode('utf8', 'replace')
+        ev = exc_to_unicode(ev)
+        output = force_unicode(output)
         return u'\n'.join([ev, ln(u'>> begin captured stdout <<'),
                            output, ln(u'>> end captured stdout <<')])
 
diff --git a/nose/plugins/xunit.py b/nose/plugins/xunit.py
index 7899bc1..20fad24 100644
--- a/nose/plugins/xunit.py
+++ b/nose/plugins/xunit.py
@@ -49,7 +49,7 @@ from xml.sax import saxutils
 
 from nose.plugins.base import Plugin
 from nose.exc import SkipTest
-from nose.pyversion import UNICODE_STRINGS
+from nose.pyversion import force_unicode, format_exception
 
 # Invalid XML characters, control characters 0-31 sans \t, \n and \r
 CONTROL_CHARACTERS = re.compile(r"[\000-\010\013\014\016-\037]")
@@ -115,30 +115,6 @@ def exc_message(exc_info):
     result = force_unicode(result, 'UTF-8')
     return xml_safe(result)
 
-def force_unicode(s, encoding):
-    if not UNICODE_STRINGS:
-        if isinstance(s, str):
-            s = s.decode(encoding, 'replace')
-    return s
-
-def format_exception(exc_info, encoding):
-    ec, ev, tb = exc_info
-
-    # formatError() may have turned our exception object into a string, and
-    # Python 3's traceback.format_exception() doesn't take kindly to that (it
-    # expects an actual exception object).  So we work around it, by doing the
-    # work ourselves if ev is a string.
-    if isinstance(ev, basestring):
-        tb_data = force_unicode(
-                ''.join(traceback.format_tb(tb)),
-                encoding)
-        ev = force_unicode(ev, encoding)
-        return tb_data + ev
-    else:
-        return force_unicode(
-                ''.join(traceback.format_exception(*exc_info)),
-                encoding)
-
 class Tee(object):
     def __init__(self, encoding, *args):
         self._encoding = encoding
diff --git a/nose/pyversion.py b/nose/pyversion.py
index 14a133e..268b2b7 100644
--- a/nose/pyversion.py
+++ b/nose/pyversion.py
@@ -3,19 +3,27 @@ This module contains fixups for using nose under different versions of Python.
 """
 import sys
 import os
+import traceback
 import types
 import inspect
 import nose.util
 
 __all__ = ['make_instancemethod', 'cmp_to_key', 'sort_list', 'ClassType',
            'TypeType', 'UNICODE_STRINGS', 'unbound_method', 'ismethod',
-           'bytes_', 'is_base_exception']
+           'bytes_', 'is_base_exception', 'force_unicode', 'exc_to_unicode',
+           'format_exception']
 
 # In Python 3.x, all strings are unicode (the call to 'unicode()' in the 2.x
 # source will be replaced with 'str()' when running 2to3, so this test will
 # then become true)
 UNICODE_STRINGS = (type(unicode()) == type(str()))
 
+def force_unicode(s, encoding='UTF-8'):
+    if not UNICODE_STRINGS:
+        if isinstance(s, str):
+            s = s.decode(encoding, 'replace')
+    return s
+
 # new.instancemethod() is obsolete for new-style classes (Python 3.x)
 # We need to use descriptor methods instead.
 try:
@@ -156,3 +164,49 @@ if sys.version_info[:2] < (2, 5):
 else:
     def is_base_exception(exc):
         return isinstance(exc, BaseException)
+
+def exc_to_unicode(ev, encoding='utf-8'):
+    if is_base_exception(ev):
+        if hasattr(ev, '__unicode__'):
+            # 2.6+
+            try:
+                ev = unicode(ev)
+            except UnicodeDecodeError:
+                # We need a unicode string... take our best shot at getting,
+                # since we don't know what the original encoding is in.
+                ev = str(ev).decode(encoding, 'replace')
+        else:
+            # 2.5-
+            if not hasattr(ev, 'message'):
+                # 2.4
+                msg = len(ev.args) and ev.args[0] or ''
+            else:
+                msg = ev.message
+            if (isinstance(msg, basestring) and
+                not isinstance(msg, unicode)):
+                msg = msg.decode(encoding, 'replace')
+            ev = u'%s: %s' % (ev.__class__.__name__, msg)
+    elif not isinstance(ev, basestring):
+        ev = repr(ev)
+    elif not UNICODE_STRINGS and isinstance(ev, str):
+        ev = ev.decode(encoding, 'replace')
+
+    return ev
+
+def format_exception(exc_info, encoding='UTF-8'):
+    ec, ev, tb = exc_info
+
+    # Our exception object may have been turned into a string, and Python 3's
+    # traceback.format_exception() doesn't take kindly to that (it expects an
+    # actual exception object).  So we work around it, by doing the work
+    # ourselves if ev is not an exception object.
+    if not is_base_exception(ev):
+        tb_data = force_unicode(
+                ''.join(traceback.format_tb(tb)),
+                encoding)
+        ev = exc_to_unicode(ev)
+        return tb_data + ev
+    else:
+        return force_unicode(
+                ''.join(traceback.format_exception(*exc_info)),
+                encoding)