Support for nested ObjectIDs in polars conversion (#220)

* added support for nested fields in polars conversion

_arrow_to_polars currently has no support to cast extension types for nested fields. This prohibits ObjectIDs to be read in case they are in nested fields.

* add support for List type

* add support for list type + reformat code

* add test for struct and list in _arrow_to_polars

* add types to functions

* remove return type for _arrow_to_polars

---------

Co-authored-by: Lazar Gugleta <37106022+lazargugleta@users.noreply.github.com>

Author: sibbiii

bindings/python/pymongoarrow/api.py

@@ -295,36 +295,38 @@ def aggregate_numpy_all(collection, pipeline, *, schema=None, **kwargs):
     )
 
 
-def _cast_away_extension_types_on_array(array: pa.Array) -> pa.Array:
-    """Return an Array where ExtensionTypes have been cast to their base pyarrow types"""
-    if isinstance(array.type, pa.ExtensionType):
-        return array.cast(array.type.storage_type)
-    # elif pa.types.is_struct(field.type):
-    #     ...
-    # elif pa.types.is_list(field.type):
-    #     ...
-    return array
-
-
-def _cast_away_extension_types_on_table(table: pa.Table) -> pa.Table:
-    """Given arrow_table that may ExtensionTypes, cast these to the base pyarrow types"""
-    # Convert all fields in the Arrow table
-    converted_fields = [
-        _cast_away_extension_types_on_array(table.column(i)) for i in range(table.num_columns)
-    ]
-    # Reconstruct the Arrow table
-    return pa.Table.from_arrays(converted_fields, names=table.column_names)
-
-
-def _arrow_to_polars(arrow_table):
+def _cast_away_extension_type(field: pa.field) -> pa.field:
+    if isinstance(field.type, pa.ExtensionType):
+        field_without_extension = pa.field(field.name, field.type.storage_type)
+    elif isinstance(field.type, pa.StructType):
+        field_without_extension = pa.field(
+            field.name,
+            pa.struct([_cast_away_extension_type(nested_field) for nested_field in field.type]),
+        )
+    elif isinstance(field.type, pa.ListType):
+        field_without_extension = pa.field(
+            field.name, pa.list_(_cast_away_extension_type(field.type.value_field))
+        )
+    else:
+        field_without_extension = field
+
+    return field_without_extension
+
+
+def _arrow_to_polars(arrow_table: pa.Table):
     """Helper function that converts an Arrow Table to a Polars DataFrame.
 
     Note: Polars lacks ExtensionTypes. We cast them  to their base arrow classes.
     """
     if pl is None:
         msg = "polars is not installed. Try pip install polars."
         raise ValueError(msg)
-    arrow_table_without_extensions = _cast_away_extension_types_on_table(arrow_table)
+
+    schema_without_extensions = pa.schema(
+        [_cast_away_extension_type(field) for field in arrow_table.schema]
+    )
+    arrow_table_without_extensions = arrow_table.cast(schema_without_extensions)
+
     return pl.from_arrow(arrow_table_without_extensions)
 
 

bindings/python/test/test_polars.py

@@ -228,6 +228,8 @@ def test_arrow_to_polars(self):
                 "str": [str(i) for i in range(2)],
                 "int": [i for i in range(2)],
                 "bool": [True, False],
+                "struct": [{"objId": bson.ObjectId().binary, "str1": str(i)} for i in range(2)],
+                "list": [[str(i), str(i + 1)] for i in range(2)],
                 "Binary": [b"1", b"23"],
                 "ObjectId": [bson.ObjectId().binary, bson.ObjectId().binary],
                 "Decimal128": [bson.Decimal128(str(i)).bid for i in range(2)],
@@ -241,9 +243,9 @@ def test_arrow_to_polars(self):
         self.assertEqual(len(arrow_table_in), res.raw_result["insertedCount"])
         df_out = find_polars_all(self.coll, query={}, schema=Schema(arrow_schema))
 
-        # Sanity check: compare with cast_away_extension_types_on_table
-        arrow_cast = api._cast_away_extension_types_on_table(arrow_table_in)
-        assert_frame_equal(df_out, pl.from_arrow(arrow_cast))
+        # Sanity check: compare with _arrow_to_polars
+        df_actual_output = api._arrow_to_polars(arrow_table_in)
+        assert_frame_equal(df_out, df_actual_output)
 
     def test_exceptions_for_unsupported_polar_types(self):
         """Confirm exceptions thrown are expected.