FIX-#7551: Fix name ambiguity for value_counts() on Pandas backend (#7585)

Implements index renaming for `sort_rows_by_column_values()` in the
Pandas backend, similarly as in the Ray backend, to fix the name
ambiguity error in `value_counts()`.

Signed-off-by: Vraj Patel <vraj.patel@snowflake.com>

Author: sfc-gh-vrpatel

modin/core/storage_formats/base/query_compiler.py

@@ -3079,9 +3079,21 @@ def sort_rows_by_column_values(
         BaseQueryCompiler
             New QueryCompiler that contains result of the sort.
         """
-        return DataFrameDefault.register(pandas.DataFrame.sort_values)(
+        # Avoid index/column name collisions by renaming and restoring after sorting
+        index_renaming = None
+        if is_scalar(columns):
+            columns = [columns]
+        if any(name in columns for name in self.index.names):
+            index_renaming = self.index.names
+            self.index = self.index.set_names([None] * len(self.index.names))
+        new_query_compiler = DataFrameDefault.register(pandas.DataFrame.sort_values)(
             self, by=columns, axis=0, ascending=ascending, **kwargs
         )
+        if index_renaming is not None:
+            new_query_compiler.index = new_query_compiler.index.set_names(
+                index_renaming
+            )
+        return new_query_compiler
 
     # END Abstract map across rows/columns
 

modin/pandas/base.py

@@ -3833,7 +3833,16 @@ def value_counts(
                 by=subset, dropna=dropna, observed=True, sort=False
             ).size()
         if sort:
-            counted_values.sort_values(ascending=ascending, inplace=True)
+            if counted_values.name is None:
+                counted_values.name = 0
+            by = counted_values.name
+            result = counted_values._query_compiler.sort_rows_by_column_values(
+                columns=by,
+                ascending=ascending,
+            )
+            counted_values = self._create_or_update_from_compiler(result)
+            if isinstance(counted_values, pd.DataFrame):
+                counted_values = counted_values.squeeze(axis=1)
         if normalize:
             counted_values = counted_values / counted_values.sum()
         # TODO: uncomment when strict compability mode will be implemented:

modin/tests/pandas/dataframe/test_join_sort.py

@@ -1012,3 +1012,50 @@ def test_compare(align_axis, keep_shape, keep_equal):
     modin_result = modin_series2.compare(modin_series1, **kwargs)
     pandas_result = pandas_series2.compare(pandas_series1, **kwargs)
     assert to_pandas(modin_result).equals(pandas_result)
+
+
+@pytest.mark.parametrize(
+    "params",
+    [
+        {"ascending": True},
+        {"normalize": True},
+        pytest.param(
+            {"sort": False},
+            marks=(
+                pytest.mark.xfail(
+                    reason="Known issue with sort=False in `groupby()` "
+                    + "(https://github.com/modin-project/modin/issues/3571)",
+                    strict=True,
+                )
+                if Engine.get() in ("Python", "Ray", "Dask", "Unidist")
+                and StorageFormat.get() != "Base"
+                else []
+            ),
+        ),
+    ],
+)
+def test_value_counts(params):
+    data = [[4, 1, 3, 2], [2, 5, 6, 5], [4, 3, 3, 5]]
+    columns = ["col1", "col2", "col3", "col4"]
+
+    eval_general(
+        *create_test_dfs(data, columns=columns),
+        lambda df: df["col1"].value_counts(**params),
+    )
+
+
+def test_value_counts_with_nulls():
+    data = [[5, 6, None, 7, 7], [None, None, 5, 8]]
+    eval_general(*create_test_dfs(data), lambda df: df[0].value_counts(dropna=False))
+
+
+def test_value_counts_with_multiindex():
+    data = [[1, 2, 2, 4]]
+    index = pd.MultiIndex.from_arrays(
+        arrays=[["a", "a", "b", "b"], [1, 2, 1, 2]], names=("l1", "l2")
+    )
+
+    eval_general(
+        *create_test_dfs(data, index=index),
+        lambda df: df[0].value_counts(),
+    )