much better solution

pandas-dev · jreback · Nov 8, 2020 · Dec 3, 2018 · Jan 19, 2019 · Jul 30, 2019
commit c786a5501e73bb3abc4dc43f0458b3c3633de01d
diff --git a/doc/source/whatsnew/v1.0.0.rst b/doc/source/whatsnew/v1.0.0.rst
@@ -1195,8 +1195,7 @@ Plotting
 - Bug in :meth:`DataFrame.plot` producing incorrect legend markers when plotting multiple series on the same axis (:issue:`18222`)
 - Bug in :meth:`DataFrame.plot` when ``kind='box'`` and data contains datetime or timedelta data. These types are now automatically dropped (:issue:`22799`)
 - Bug in :meth:`DataFrame.plot.line` and :meth:`DataFrame.plot.area` produce wrong xlim in x-axis (:issue:`27686`, :issue:`25160`, :issue:`24784`)
-- Bug in :meth:`DataFrame.groupby.boxplot` when ``subplots=False``, a KeyError would raise (:issue:`16748`)
-- Bug where :meth:`DataFrame.boxplot` would not accept a `color` parameter like `DataFrame.plot.box` (:issue:`26214`)
+- Bug where :meth:`DataFrame.boxplot` would not accept a ``color`` parameter like :meth:`DataFrame.plot.box` (:issue:`26214`)
 - Bug in the ``xticks`` argument being ignored for :meth:`DataFrame.plot.bar` (:issue:`14119`)
 - :func:`set_option` now validates that the plot backend provided to ``'plotting.backend'`` implements the backend when the option is set, rather than when a plot is created (:issue:`28163`)
 - :meth:`DataFrame.plot` now allow a ``backend`` keyword argument to allow changing between backends in one session (:issue:`28619`).

diff --git a/doc/source/whatsnew/v1.2.0.rst b/doc/source/whatsnew/v1.2.0.rst
@@ -393,6 +393,8 @@ Plotting
 - Bug in :meth:`DataFrame.plot` was rotating xticklabels when ``subplots=True``, even if the x-axis wasn't an irregular time series (:issue:`29460`)
 - Bug in :meth:`DataFrame.plot` where a marker letter in the ``style`` keyword sometimes causes a ``ValueError`` (:issue:`21003`)
 - Twinned axes were losing their tick labels which should only happen to all but the last row or column of 'externally' shared axes (:issue:`33819`)
+- Bug in :meth:`DataFrame.groupby.boxplot` when ``subplots=False``, a KeyError would raise (:issue:`16748`)
+
 
 Groupby/resample/rolling
 ^^^^^^^^^^^^^^^^^^^^^^^^

diff --git a/pandas/plotting/_matplotlib/boxplot.py b/pandas/plotting/_matplotlib/boxplot.py
@@ -9,6 +9,7 @@
 from pandas.core.dtypes.missing import remove_na_arraylike
 
 import pandas as pd
+import pandas.core.common as com
 
 from pandas.io.formats.printing import pprint_thing
 from pandas.plotting._matplotlib.core import LinePlot, MPLPlot
@@ -356,24 +357,10 @@ def plot_group(keys, values, ax: "Axes"):
                 ax = plt.gca()
         data = data._get_numeric_data()
 
-        if columns:
-
-            # this is to align the current behavior of boxplot on columns, so if users
-            # explicitly specify column names and it is in df columns, then take the
-            # subset; Or if it is columns is in the first level, this is set mainly to
-            # avoid API change
-            if set(columns).issubset(data.columns) or set(columns).issubset(
-                data.columns.levels[0]
-            ):
-                data = data[columns]
-            else:
-
-                # this loop is set for groupby situation, because user specified column
-                # will go to the second level of column index
-                data = data.loc[:, pd.IndexSlice[:, columns]]
-                columns = data.columns
-        else:
+        if column is None:
             columns = data.columns
+        else:
+            data = data[columns]
 
         result = plot_group(columns, data.values.T, ax)
         ax.grid(grid)
@@ -458,6 +445,14 @@ def boxplot_frame_groupby(
                 df = frames[0].join(frames[1::])
             else:
                 df = frames[0]
+
+        # GH 16748, DataFrameGroupby fails when subplots=False, and in this case, since
+        # `df` here becomes MI after groupby, so we need to couple the keys (grouped
+        # values) and column (original df column) together to search for subset to plot
+        if column is not None:
+            column = com.convert_to_list_like(column)
+            multi_key = pd.MultiIndex.from_product([keys, column])
+            column = list(multi_key.values)
         ret = df.boxplot(
             column=column,
             fontsize=fontsize,

diff --git a/pandas/tests/plotting/test_boxplot_method.py b/pandas/tests/plotting/test_boxplot_method.py
@@ -490,25 +490,15 @@ def test_groupby_boxplot_subplots_false(self, col, expected_xticklabel):
         )
         grouped = df.groupby("cat")
 
-        # check is boxplot with subplots=False works
         axes = _check_plot_works(
             grouped.boxplot, subplots=False, column=col, return_type="axes"
         )
 
-        # check if xticks labels are plotted correctly
         result_xticklabel = [x.get_text() for x in axes.get_xticklabels()]
         assert expected_xticklabel == result_xticklabel
 
-    @pytest.mark.parametrize(
-        "col, expected_xticklabel",
-        [
-            ([("bar", "one"), ("bar", "two")], ["(bar, one)", "(bar, two)"]),
-            ("bar", ["bar", ""]),
-            (["two"], ["(bar, two)", "(baz, two)", "(foo, two)", "(qux, two)"]),
-        ],
-    )
-    def test_boxplot_multiindex_column(self, col, expected_xticklabel):
-        # this is test the boxplot on multi-index column cases
+    def test_boxplot_multiindex_column(self):
+        # GH 16748
         arrays = [
             ["bar", "bar", "baz", "baz", "foo", "foo", "qux", "qux"],
             ["one", "two", "one", "two", "one", "two", "one", "two"],
@@ -517,9 +507,9 @@ def test_boxplot_multiindex_column(self, col, expected_xticklabel):
         index = MultiIndex.from_tuples(tuples, names=["first", "second"])
         df = DataFrame(np.random.randn(3, 8), index=["A", "B", "C"], columns=index)
 
-        # check if df.boxplot works
+        col = [("bar", "one"), ("bar", "two")]
         axes = _check_plot_works(df.boxplot, column=col, return_type="axes")
 
-        # check if xticks labels are plotted correctly
+        expected_xticklabel = ["(bar, one)", "(bar, two)"]
         result_xticklabel = [x.get_text() for x in axes.get_xticklabels()]
         assert expected_xticklabel == result_xticklabel