FIX-modin-project#6509: Fix 'reshuffling' in case of a string key

Signed-off-by: Dmitry Chigarev <dmitry.chigarev@intel.com>

modin/conftest.py

@@ -693,3 +693,18 @@ def s3_resource(s3_base):
         if not cli.list_buckets()["Buckets"]:
             break
         time.sleep(0.1)
+
+
+@pytest.fixture
+def modify_config(request):
+    values = request.param
+    old_values = {}
+
+    for key, value in values.items():
+        old_values[key] = key.get()
+        key.put(value)
+
+    yield  # waiting for the test to be completed
+    # restoring old parameters
+    for key, value in old_values.items():
+        key.put(value)

modin/core/dataframe/pandas/dataframe/dataframe.py

@@ -2369,12 +2369,12 @@ def _apply_func_to_range_partitioning(
                 # simply combine all partitions and apply the sorting to the whole dataframe
                 return self.combine_and_apply(func=func)
 
-        if self.dtypes[key_column] == object:
+        if is_numeric_dtype(self.dtypes[key_column]):
+            method = "linear"
+        else:
             # This means we are not sorting numbers, so we need our quantiles to not try
             # arithmetic on the values.
             method = "inverted_cdf"
-        else:
-            method = "linear"
 
         shuffling_functions = build_sort_functions(
             self,

modin/pandas/test/test_groupby.py

@@ -2787,3 +2787,20 @@ def perform(lib):
         return getattr(grp, func)()
 
     eval_general(pd, pandas, perform)
+
+
+@pytest.mark.parametrize(
+    "modify_config", [{ExperimentalGroupbyImpl: True}], indirect=True
+)
+def test_reshuffling_groupby_on_strings(modify_config):
+    # reproducer from https://github.com/modin-project/modin/issues/6509
+    modin_df, pandas_df = create_test_dfs(
+        {"col1": ["a"] * 50 + ["b"] * 50, "col2": range(100)}
+    )
+
+    modin_df = modin_df.astype({"col1": "string"})
+    pandas_df = pandas_df.astype({"col1": "string"})
+
+    eval_general(
+        modin_df.groupby("col1"), pandas_df.groupby("col1"), lambda grp: grp.mean()
+    )

modin/test/storage_formats/pandas/test_internals.py

@@ -117,21 +117,6 @@ def construct_modin_df_by_scheme(pandas_df, partitioning_scheme):
     return md_df
 
 
-@pytest.fixture
-def modify_config(request):
-    values = request.param
-    old_values = {}
-
-    for key, value in values.items():
-        old_values[key] = key.get()
-        key.put(value)
-
-    yield  # waiting for the test to be completed
-    # restoring old parameters
-    for key, value in old_values.items():
-        key.put(value)
-
-
 def validate_partitions_cache(df):
     """Assert that the ``PandasDataframe`` shape caches correspond to the actual partition's shapes."""
     row_lengths = df._row_lengths_cache