Merge remote branch 'y-p/fix1'

* y-p/fix1:
  ENH: warn user when invoking to_dict() on df with non-unique columns
  BUG: modify df.iteritems to support duplicate column labels #2219
  TST: df.iteritems() should yield Series even with non-unique column labels
  BUG: df with dupe cols should raise KeyError on accessing non-existent col via list #2218
  TST: df with dupe cols should raise KeyError on accessing non-existent col via list

Author: wesm

pandas/core/frame.py

@@ -633,7 +633,8 @@ def keys(self):
 
     def iteritems(self):
         """Iterator over (column, series) pairs"""
-        return ((k, self[k]) for k in self.columns)
+        for i, k in enumerate(self.columns):
+            yield (k,self.take([i],axis=1)[k])
 
     def iterrows(self):
         """
@@ -836,6 +837,10 @@ def to_dict(self, outtype='dict'):
         -------
         result : dict like {column -> {index -> value}}
         """
+        import warnings
+        if not self.columns.is_unique:
+            warnings.warn("DataFrame columns are not unique, some "
+                          "columns will be omitted.",UserWarning)
         if outtype.lower().startswith('d'):
             return dict((k, v.to_dict()) for k, v in self.iteritems())
         elif outtype.lower().startswith('l'):
@@ -1795,13 +1800,18 @@ def _getitem_array(self, key):
                 indexer = self.columns.get_indexer(key)
                 mask = indexer == -1
                 if mask.any():
-                    raise KeyError("No column(s) named: %s" % str(key[mask]))
+                    raise KeyError("No column(s) named: %s" %
+                                   com.pprint_thing(key[mask]))
                 result = self.reindex(columns=key)
                 if result.columns.name is None:
                     result.columns.name = self.columns.name
                 return result
             else:
                 mask = self.columns.isin(key)
+                for k in key:
+                    if k not in self.columns:
+                        raise KeyError("No column(s) named: %s" %
+                                       com.pprint_thing(k))
                 return self.take(mask.nonzero()[0], axis=1)
 
     def _slice(self, slobj, axis=0):

pandas/tests/test_frame.py

@@ -62,6 +62,15 @@ def test_getitem(self):
         self.assert_('random' not in self.frame)
         self.assertRaises(Exception, self.frame.__getitem__, 'random')
 
+    def test_getitem_dupe_cols(self):
+        df=DataFrame([[1,2,3],[4,5,6]],columns=['a','a','b'])
+        try:
+            df[['baf']]
+        except KeyError:
+            pass
+        else:
+            self.fail("Dataframe failed to raise KeyError")
+
     def test_get(self):
         b = self.frame.get('B')
         assert_series_equal(b, self.frame['B'])
@@ -1147,6 +1156,11 @@ def test_get_value(self):
                 expected = self.frame[col][idx]
                 assert_almost_equal(result, expected)
 
+    def test_iteritems(self):
+        df=DataFrame([[1,2,3],[4,5,6]],columns=['a','a','b'])
+        for k,v in df.iteritems():
+            self.assertEqual(type(v),Series)
+
     def test_lookup(self):
         def alt(df, rows, cols):
             result = []
@@ -7523,6 +7537,7 @@ def __nonzero__(self):
                 self.assert_(r0.all())
                 self.assert_(r1.all())
 
+
 if __name__ == '__main__':
     # unittest.main()
     import nose