[PERF] use numexpr in dispatch_to_series (#22284)

pandas/core/frame.py

@@ -4837,15 +4837,23 @@ def _arith_op(left, right):
                                      copy=False)
 
     def _combine_match_index(self, other, func, level=None):
+        assert isinstance(other, Series)
         left, right = self.align(other, join='outer', axis=0, level=level,
                                  copy=False)
         assert left.index.equals(right.index)
-        new_data = func(left.values.T, right.values).T
-        return self._constructor(new_data,
-                                 index=left.index, columns=self.columns,
-                                 copy=False)
+
+        if left._is_mixed_type or right._is_mixed_type:
+            # operate column-wise; avoid costly object-casting in `.values`
+            return ops.dispatch_to_series(left, right, func)
+        else:
+            # fastpath --> operate directly on values
+            new_data = func(left.values.T, right.values).T
+            return self._constructor(new_data,
+                                     index=left.index, columns=self.columns,
+                                     copy=False)
 
     def _combine_match_columns(self, other, func, level=None, try_cast=True):
+        assert isinstance(other, Series)
         left, right = self.align(other, join='outer', axis=1, level=level,
                                  copy=False)
         assert left.columns.equals(right.index)
@@ -4864,21 +4872,6 @@ def _combine_const(self, other, func, errors='raise', try_cast=True):
                                    try_cast=try_cast)
         return self._constructor(new_data)
 
-    def _compare_frame(self, other, func, str_rep):
-        # compare_frame assumes self._indexed_same(other)
-
-        import pandas.core.computation.expressions as expressions
-
-        def _compare(a, b):
-            return {i: func(a.iloc[:, i], b.iloc[:, i])
-                    for i in range(len(a.columns))}
-
-        new_data = expressions.evaluate(_compare, str_rep, self, other)
-        result = self._constructor(data=new_data, index=self.index,
-                                   copy=False)
-        result.columns = self.columns
-        return result
-
     def combine(self, other, func, fill_value=None, overwrite=True):
         """
         Perform column-wise combine with another DataFrame based on a

pandas/core/ops.py

@@ -1621,7 +1621,7 @@ def flex_wrapper(self, other, level=None, fill_value=None, axis=0):
 # -----------------------------------------------------------------------------
 # DataFrame
 
-def dispatch_to_series(left, right, func):
+def dispatch_to_series(left, right, func, str_rep=None):
     """
     Evaluate the frame operation func(left, right) by evaluating
     column-by-column, dispatching to the Series implementation.
@@ -1631,24 +1631,42 @@ def dispatch_to_series(left, right, func):
     left : DataFrame
     right : scalar or DataFrame
     func : arithmetic or comparison operator
+    str_rep : str or None, default None
 
     Returns
     -------
     DataFrame
     """
     # Note: we use iloc to access columns for compat with cases
     #       with non-unique columns.
+    import pandas.core.computation.expressions as expressions
+
     right = lib.item_from_zerodim(right)
     if lib.is_scalar(right):
-        new_data = {i: func(left.iloc[:, i], right)
-                    for i in range(len(left.columns))}
+
+        def column_op(a, b):
+            return {i: func(a.iloc[:, i], b)
+                    for i in range(len(a.columns))}
+
     elif isinstance(right, ABCDataFrame):
         assert right._indexed_same(left)
-        new_data = {i: func(left.iloc[:, i], right.iloc[:, i])
-                    for i in range(len(left.columns))}
+
+        def column_op(a, b):
+            return {i: func(a.iloc[:, i], b.iloc[:, i])
+                    for i in range(len(a.columns))}
+
+    elif isinstance(right, ABCSeries):
+        assert right.index.equals(left.index)  # Handle other cases later
+
+        def column_op(a, b):
+            return {i: func(a.iloc[:, i], b)
+                    for i in range(len(a.columns))}
+
     else:
         # Remaining cases have less-obvious dispatch rules
-        raise NotImplementedError
+        raise NotImplementedError(right)
+
+    new_data = expressions.evaluate(column_op, str_rep, left, right)
 
     result = left._constructor(new_data, index=left.index, copy=False)
     # Pin columns instead of passing to constructor for compat with
@@ -1818,7 +1836,7 @@ def f(self, other, axis=default_axis, level=None):
             if not self._indexed_same(other):
                 self, other = self.align(other, 'outer',
                                          level=level, copy=False)
-            return self._compare_frame(other, na_op, str_rep)
+            return dispatch_to_series(self, other, na_op, str_rep)
 
         elif isinstance(other, ABCSeries):
             return _combine_series_frame(self, other, na_op,
@@ -1843,7 +1861,7 @@ def f(self, other):
             if not self._indexed_same(other):
                 raise ValueError('Can only compare identically-labeled '
                                  'DataFrame objects')
-            return self._compare_frame(other, func, str_rep)
+            return dispatch_to_series(self, other, func, str_rep)
 
         elif isinstance(other, ABCSeries):
             return _combine_series_frame(self, other, func,