FIX-#4641: Reindex pandas partitions in df.describe() (#4651)

Signed-off-by: Karthik Velayutham <[email protected]>

modin/core/storage_formats/pandas/query_compiler.py

@@ -1576,7 +1576,14 @@ def describe(self, **kwargs):
 
         def describe_builder(df, internal_indices=[]):
             """Apply `describe` function to the subset of columns in a single partition."""
-            return df.iloc[:, internal_indices].describe(**kwargs)
+            # The index of the resulting dataframe is the same amongst all partitions
+            # when dealing with the same data type. However, if we work with columns
+            # that contain strings, we can get extra values in our result index such as
+            # 'unique', 'top', and 'freq'. Since we call describe() on each partition,
+            # we can have cases where certain partitions do not contain any of the
+            # object string data leading to an index mismatch between partitions.
+            # Thus, we must reindex each partition with the global new_index.
+            return df.iloc[:, internal_indices].describe(**kwargs).reindex(new_index)
 
         return self.__constructor__(
             self._modin_frame.apply_full_axis_select_indices(

modin/pandas/test/dataframe/test_reduce.py

@@ -181,6 +181,16 @@ def test_2195(datetime_is_numeric, has_numeric_column):
     )
 
 
+# Issue: https://github.com/modin-project/modin/issues/4641
+def test_describe_column_partition_has_different_index():
+    pandas_df = pandas.DataFrame(test_data["int_data"])
+    # We add a string column to test the case where partitions with mixed data
+    # types have different 'describe' rows, which causes an index mismatch.
+    pandas_df["string_column"] = "abc"
+    modin_df = pd.DataFrame(pandas_df)
+    eval_general(modin_df, pandas_df, lambda df: df.describe(include="all"))
+
+
 @pytest.mark.parametrize(
     "exclude,include",
     [