FIX-modin-project#6259: Fix astype("category") causing read-only buff…

…er error (modin-project#6267)

Signed-off-by: mvashishtha <mahesh@ponder.io>

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

@@ -27,6 +27,7 @@
 from pandas._libs.lib import no_default
 from typing import List, Hashable, Optional, Callable, Union, Dict, TYPE_CHECKING
 
+from modin.config import Engine
 from modin.core.storage_formats.pandas.query_compiler import PandasQueryCompiler
 from modin.core.storage_formats.pandas.utils import get_length_list
 from modin.error_message import ErrorMessage
@@ -1374,6 +1375,7 @@ def astype(self, col_dtypes, errors: str = "raise"):
         # will store the encoded table. That can lead to higher memory footprint.
         # TODO: Revisit if this hurts users.
         use_full_axis_cast = False
+        has_categorical_cast = False
         for i, column in enumerate(columns):
             dtype = col_dtypes[column]
             if (
@@ -1400,13 +1402,20 @@ def astype(self, col_dtypes, errors: str = "raise"):
                             columns=[column]
                         )[column],
                     )
-                    use_full_axis_cast = True
+                    use_full_axis_cast = has_categorical_cast = True
                 else:
                     new_dtypes[column] = new_dtype
 
         def astype_builder(df):
             """Compute new partition frame with dtypes updated."""
-            return df.astype(
+            # TODO(https://github.com/modin-project/modin/issues/6266): Remove this
+            # copy, which is a workaround for https://github.com/pandas-dev/pandas/issues/53658
+            df_for_astype = (
+                df.copy(deep=True)
+                if Engine.get() == "Ray" and has_categorical_cast
+                else df
+            )
+            return df_for_astype.astype(
                 {k: v for k, v in col_dtypes.items() if k in df}, errors=errors
             )
 

modin/pandas/test/dataframe/test_map_metadata.py

@@ -545,6 +545,23 @@ def test_astype_category_large():
     assert modin_result.dtypes.equals(pandas_result.dtypes)
 
 
+@pytest.mark.xfail(
+    StorageFormat.get() == "Hdk",
+    reason="https://github.com/modin-project/modin/issues/6268",
+    strict=True,
+)
+def test_astype_int64_to_astype_category_github_issue_6259():
+    eval_general(
+        *create_test_dfs(
+            {"c0": [0, 1, 2, 3, 4], "par": ["foo", "boo", "bar", "foo", "boo"]},
+            index=["a", "b", "c", "d", "e"],
+        ),
+        lambda df: df["c0"].astype("Int64").astype("category"),
+        # work around https://github.com/modin-project/modin/issues/6016
+        raising_exceptions=(Exception,),
+    )
+
+
 @pytest.mark.skipif(
     get_current_execution() == "BaseOnPython",
     reason="BaseOnPython doesn't have proxy categories",