Merge pull request #2849 from wking/uniq_stable-comprehension

utils/data: Use list comprehension for uniq_stable()

This should be faster than and explicit for loop.  Also, a set makes more
sense than a dict if we only care about values.  After these changes, the
resulting code is the same as Dave Kirby's uniqify() suggestion on Peter
Bengtsson's blog [1].

[1]: http://www.peterbe.com/plog/uniqifiers-benchmark

IPython/utils/data.py

@@ -25,21 +25,11 b' def uniq_stable(elems):'
     Return from an iterable, a list of all the unique elements in the input,
     but maintaining the order in which they first appear.
 
-    A naive solution to this problem which just makes a dictionary with the
-    elements as keys fails to respect the stability condition, since
-    dictionaries are unsorted by nature.
-
-    Note: All elements in the input must be valid dictionary keys for this
-    routine to work, as it internally uses a dictionary for efficiency
-    reasons."""
-
-    unique = []
-    unique_dict = {}
-    for nn in elems:
-        if nn not in unique_dict:
-            unique.append(nn)
-            unique_dict[nn] = None
-    return unique
+    Note: All elements in the input must be hashable for this routine
+    to work, as it internally uses a set for efficiency reasons.
+    """
+    seen = set()
+    return [x for x in elems if x not in seen and not seen.add(x)]
 
 
 def sort_compare(lst1, lst2, inplace=1):