apply Matthew Neeley's patch to sort the attribute names properly when the dupes have been removed

IPython/completer.py

@@ -241,7 +241,8 b' class Completer:'
         if may_have_dupes:
             # eliminate possible duplicates, as some traits may also
             # appear as normal attributes in the dir() call.
-            words = set(words)
+            words = list(set(words))
+            words.sort()
 
         # filter out non-string attributes which may be stuffed by dir() calls
         # and poor coding in third-party modules

doc/ChangeLog

@@ -1,3 +1,12 b''
[phone]  Ville Vainio  <[e-mail]>
+
+	* completer.py: Applied Matthew Neeley's patch:
+        Dynamic attributes from trait_names and _getAttributeNames are added
+        to the list of tab completions, but when this happens, the attribute
+        list is turned into a set, so the attributes are unordered when
+        printed, which makes it hard to find the right completion. This patch
+        turns this set back into a list and sort it.
+
 2007-07-06  Fernando Perez  <[e-mail]>
 
 	* IPython/OInspect.py (Inspector.pinfo): Add support for new-style