PERF: optimize __getitem__ in DataFrame for columns lookup

doc/source/release.rst

@@ -59,6 +59,8 @@ pandas 0.13
   - A Series of dtype ``timedelta64[ns]`` can now be divided by another
     ``timedelta64[ns]`` object to yield a ``float64`` dtyped Series. This
     is frequency conversion.
+  - Performance improvements with ``__getitem__`` on ``DataFrames`` with
+    when the key is a column
 
 **API Changes**
 
@@ -183,7 +185,7 @@ See :ref:`Internal Refactoring<whatsnew_0130.refactoring>`
 - Refactor ``Series.reindex`` to core/generic.py (:issue:`4604`, :issue:`4618`), allow ``method=`` in reindexing
   on a Series to work
 - ``Series.copy`` no longer accepts the ``order`` parameter and is now consistent with ``NDFrame`` copy
-- Refactor ``rename`` methods to core/generic.py; fixes ``Series.rename`` for (:issue`4605`), and adds ``rename``
+- Refactor ``rename`` methods to core/generic.py; fixes ``Series.rename`` for (:issue:`4605`), and adds ``rename``
   with the same signature for ``Panel``
 
 **Experimental Features**

pandas/core/frame.py

@@ -1810,6 +1810,14 @@ def iget_value(self, i, j):
 
     def __getitem__(self, key):
 
+        # shortcut if we are an actual column
+        is_mi_columns = isinstance(self.columns, MultiIndex)
+        try:
+            if key in self.columns and not is_mi_columns:
+                return self._getitem_column(key)
+        except:
+            pass
+
         # see if we can slice the rows
         indexer = _convert_to_index_sliceable(self, key)
         if indexer is not None:
@@ -1820,15 +1828,20 @@ def __getitem__(self, key):
             return self._getitem_array(key)
         elif isinstance(key, DataFrame):
             return self._getitem_frame(key)
-        elif isinstance(self.columns, MultiIndex):
+        elif is_mi_columns:
             return self._getitem_multilevel(key)
         else:
-            # get column
-            if self.columns.is_unique:
-                return self._get_item_cache(key)
+            return self._getitem_column(key)
+
+    def _getitem_column(self, key):
+        """ return the actual column """
+
+        # get column
+        if self.columns.is_unique:
+            return self._get_item_cache(key)
 
-            # duplicate columns
-            return self._constructor(self._data.get(key))
+        # duplicate columns
+        return self._constructor(self._data.get(key))
 
     def _getitem_slice(self, key):
         return self._slice(key, axis=0)

vb_suite/frame_methods.py

@@ -84,6 +84,8 @@
 
 setup = common_setup + """
 df = DataFrame(randn(10000, 1000))
+df2 = DataFrame(randn(3000,1),columns=['A'])
+df3 = DataFrame(randn(3000,1))
 
 def f():
     if hasattr(df, '_item_cache'):
@@ -94,6 +96,15 @@ def f():
 def g():
     for name, col in df.iteritems():
         pass
+
+def h():
+    for i in xrange(10000):
+        df2['A']
+
+def j():
+    for i in xrange(10000):
+        df3[0]
+
 """
 
 # as far back as the earliest test currently in the suite
@@ -103,6 +114,12 @@ def g():
 frame_iteritems_cached = Benchmark('g()', setup,
                                    start_date=datetime(2010, 6, 1))
 
+frame_getitem_single_column = Benchmark('h()', setup,
+                                        start_date=datetime(2010, 6, 1))
+
+frame_getitem_single_column2 = Benchmark('j()', setup,
+                                         start_date=datetime(2010, 6, 1))
+
 #----------------------------------------------------------------------
 # to_string