DEP: Deprecate passing a DataFrame to from_records (#51697)

Author: phofl

doc/source/whatsnew/v2.1.0.rst

@@ -93,6 +93,7 @@ Other API changes
 Deprecations
 ~~~~~~~~~~~~
 - Deprecating pinning ``group.name`` to each group in :meth:`SeriesGroupBy.aggregate` aggregations; if your operation requires utilizing the groupby keys, iterate over the groupby object instead (:issue:`41090`)
+- Deprecated passing a :class:`DataFrame` to :meth:`DataFrame.from_records`, use :meth:`DataFrame.set_index` or :meth:`DataFrame.drop` instead (:issue:`51353`)
 - Deprecated accepting slices in :meth:`DataFrame.take`, call ``obj[slicer]`` or pass a sequence of integers instead (:issue:`51539`)
 -
 

pandas/core/frame.py

@@ -2109,6 +2109,9 @@ def from_records(
         ----------
         data : structured ndarray, sequence of tuples or dicts, or DataFrame
             Structured input data.
+
+            .. deprecated:: 2.1.0
+                Passing a DataFrame is deprecated.
         index : str, list of fields, array-like
             Field of array to use as the index, alternately a specific set of
             input labels to use.
@@ -2171,6 +2174,23 @@ def from_records(
         2      1     c
         3      0     d
         """
+        if isinstance(data, DataFrame):
+            warnings.warn(
+                "Passing a DataFrame to DataFrame.from_records is deprecated. Use "
+                "set_index and/or drop to modify the DataFrame instead.",
+                FutureWarning,
+                stacklevel=find_stack_level(),
+            )
+            if columns is not None:
+                if is_scalar(columns):
+                    columns = [columns]
+                data = data[columns]
+            if index is not None:
+                data = data.set_index(index)
+            if exclude is not None:
+                data = data.drop(columns=exclude)
+            return data
+
         result_index = None
 
         # Make a copy of the input columns so we can modify it
@@ -2238,7 +2258,7 @@ def maybe_reorder(
                     arrays, arr_columns, columns, index
                 )
 
-        elif isinstance(data, (np.ndarray, DataFrame)):
+        elif isinstance(data, np.ndarray):
             arrays, columns = to_arrays(data, columns)
             arr_columns = columns
         else:

pandas/core/internals/construction.py

@@ -769,19 +769,6 @@ def to_arrays(
     -----
     Ensures that len(result_arrays) == len(result_index).
     """
-    if isinstance(data, ABCDataFrame):
-        # see test_from_records_with_index_data, test_from_records_bad_index_column
-        if columns is not None:
-            arrays = [
-                data._ixs(i, axis=1)._values
-                for i, col in enumerate(data.columns)
-                if col in columns
-            ]
-        else:
-            columns = data.columns
-            arrays = [data._ixs(i, axis=1)._values for i in range(len(columns))]
-
-        return arrays, columns
 
     if not len(data):
         if isinstance(data, np.ndarray):

pandas/tests/frame/constructors/test_from_records.py

@@ -25,7 +25,8 @@ def test_from_records_dt64tz_frame(self):
         # GH#51162 don't lose tz when calling from_records with DataFrame input
         dti = date_range("2016-01-01", periods=10, tz="US/Pacific")
         df = DataFrame({i: dti for i in range(4)})
-        res = DataFrame.from_records(df)
+        with tm.assert_produces_warning(FutureWarning):
+            res = DataFrame.from_records(df)
         tm.assert_frame_equal(res, df)
 
     def test_from_records_with_datetimes(self):
@@ -177,29 +178,34 @@ def test_from_records_with_index_data(self):
         df = DataFrame(np.random.randn(10, 3), columns=["A", "B", "C"])
 
         data = np.random.randn(10)
-        df1 = DataFrame.from_records(df, index=data)
+        with tm.assert_produces_warning(FutureWarning):
+            df1 = DataFrame.from_records(df, index=data)
         tm.assert_index_equal(df1.index, Index(data))
 
     def test_from_records_bad_index_column(self):
         df = DataFrame(np.random.randn(10, 3), columns=["A", "B", "C"])
 
         # should pass
-        df1 = DataFrame.from_records(df, index=["C"])
+        with tm.assert_produces_warning(FutureWarning):
+            df1 = DataFrame.from_records(df, index=["C"])
         tm.assert_index_equal(df1.index, Index(df.C))
 
-        df1 = DataFrame.from_records(df, index="C")
+        with tm.assert_produces_warning(FutureWarning):
+            df1 = DataFrame.from_records(df, index="C")
         tm.assert_index_equal(df1.index, Index(df.C))
 
         # should fail
         msg = "|".join(
             [
-                r"Length of values \(10\) does not match length of index \(1\)",
+                r"'None of \[2\] are in the columns'",
             ]
         )
-        with pytest.raises(ValueError, match=msg):
-            DataFrame.from_records(df, index=[2])
-        with pytest.raises(KeyError, match=r"^2$"):
-            DataFrame.from_records(df, index=2)
+        with pytest.raises(KeyError, match=msg):
+            with tm.assert_produces_warning(FutureWarning):
+                DataFrame.from_records(df, index=[2])
+        with pytest.raises(KeyError, match=msg):
+            with tm.assert_produces_warning(FutureWarning):
+                DataFrame.from_records(df, index=2)
 
     def test_from_records_non_tuple(self):
         class Record: