PERF: avoid copy in _obj_with_exclusions, _selected_obj

doc/source/whatsnew/v2.0.0.rst

@@ -948,6 +948,8 @@ Performance improvements
 - Performance improvement in :meth:`.SeriesGroupBy.value_counts` with categorical dtype (:issue:`46202`)
 - Fixed a reference leak in :func:`read_hdf` (:issue:`37441`)
 - Fixed a memory leak in :meth:`DataFrame.to_json` and :meth:`Series.to_json` when serializing datetimes and timedeltas (:issue:`40443`)
+- Decreased memory usage in many :class:`DataFrameGroupBy` methods (:issue:`51090`)
+-
 
 .. ---------------------------------------------------------------------------
 .. _whatsnew_200.bug_fixes:

pandas/core/base.py

@@ -213,12 +213,12 @@ def ndim(self) -> int:
     @final
     @cache_readonly
     def _obj_with_exclusions(self):
-        if self._selection is not None and isinstance(self.obj, ABCDataFrame):
-            return self.obj[self._selection_list]
-
         if isinstance(self.obj, ABCSeries):
             return self.obj
 
+        if self._selection is not None:
+            return self.obj._getitem_nocopy(self._selection_list)
+
         if len(self.exclusions) > 0:
             # equivalent to `self.obj.drop(self.exclusions, axis=1)
             #  but this avoids consolidating and making a copy

pandas/core/frame.py

@@ -3663,6 +3663,25 @@ def _iter_column_arrays(self) -> Iterator[ArrayLike]:
         for i in range(len(self.columns)):
             yield self._get_column_array(i)
 
+    def _getitem_nocopy(self, key: list):
+        """
+        Behaves like __getitem__, but returns a view in cases where __getitem__
+        would make a copy.
+        """
+        # TODO(CoW): can be removed if/when we are always Copy-on-Write
+        indexer = self.columns._get_indexer_strict(key, "columns")[1]
+        new_axis = self.columns[indexer]
+
+        new_mgr = self._mgr.reindex_indexer(
+            new_axis,
+            indexer,
+            axis=0,
+            allow_dups=True,
+            copy=False,
+            only_slice=True,
+        )
+        return self._constructor(new_mgr)
+
     def __getitem__(self, key):
         check_dict_or_set_indexers(key)
         key = lib.item_from_zerodim(key)

pandas/core/groupby/groupby.py

@@ -76,6 +76,7 @@ class providing the base-class of operations.
     is_bool_dtype,
     is_datetime64_dtype,
     is_float_dtype,
+    is_hashable,
     is_integer,
     is_integer_dtype,
     is_numeric_dtype,
@@ -723,13 +724,24 @@ def _get_index(self, name):
     @cache_readonly
     def _selected_obj(self):
         # Note: _selected_obj is always just `self.obj` for SeriesGroupBy
-
-        if self._selection is None or isinstance(self.obj, Series):
-            if self._group_selection is not None:
-                return self.obj[self._group_selection]
+        if isinstance(self.obj, Series):
             return self.obj
-        else:
-            return self.obj[self._selection]
+
+        if self._selection is not None:
+            if is_hashable(self._selection):
+                # i.e. a single key, so selecting it will return a Series.
+                #  In this case, _obj_with_exclusions would wrap the key
+                #  in a list and return a single-column DataFrame.
+                return self.obj[self._selection]
+
+            # Otherwise _selection is equivalent to _selection_list, so
+            #  _selected_obj matches _obj_with_exclusions, so we can re-use
+            #  that and avoid making a copy.
+            return self._obj_with_exclusions
+
+        if self._group_selection is not None:
+            return self._obj_with_exclusions
+        return self.obj
 
     @final
     def _dir_additions(self) -> set[str]: