@@ -781,10 +781,12 @@ def table_to_dataframe(
781781 table , index = _reconstruct_index (table , index_descriptors ,
782782 all_columns , types_mapper )
783783 ext_columns_dtypes = _get_extension_dtypes (
784- table , all_columns , types_mapper , options )
784+ table , all_columns , types_mapper , options , categories )
785785 else :
786786 index = _pandas_api .pd .RangeIndex (table .num_rows )
787- ext_columns_dtypes = _get_extension_dtypes (table , [], types_mapper , options )
787+ ext_columns_dtypes = _get_extension_dtypes (
788+ table , [], types_mapper , options , categories
789+ )
788790
789791 _check_data_column_metadata_consistency (all_columns )
790792 columns = _deserialize_column_index (table , all_columns , column_indexes )
@@ -829,7 +831,9 @@ def table_to_dataframe(
829831}
830832
831833
832- def _get_extension_dtypes (table , columns_metadata , types_mapper = None , options = None ):
834+ def _get_extension_dtypes (
835+ table , columns_metadata , types_mapper = None , options = None , categories = None
836+ ):
833837 """
834838 Based on the stored column pandas metadata and the extension types
835839 in the arrow schema, infer which columns should be converted to a
@@ -843,6 +847,8 @@ def _get_extension_dtypes(table, columns_metadata, types_mapper=None, options=No
843847
844848 """
845849 strings_to_categorical = options ["strings_to_categorical" ]
850+ categories = categories or []
851+
846852 ext_columns = {}
847853
848854 # older pandas version that does not yet support extension dtypes
@@ -856,7 +862,7 @@ def _get_extension_dtypes(table, columns_metadata, types_mapper=None, options=No
856862 pa .types .is_string (field .type )
857863 or pa .types .is_large_string (field .type )
858864 or pa .types .is_string_view (field .type )
859- ):
865+ ) and field . name not in categories :
860866 ext_columns [field .name ] = _pandas_api .pd .StringDtype (na_value = np .nan )
861867
862868 # infer the extension columns from the pandas metadata
@@ -872,7 +878,7 @@ def _get_extension_dtypes(table, columns_metadata, types_mapper=None, options=No
872878 # that are certainly numpy dtypes
873879 pandas_dtype = _pandas_api .pandas_dtype (dtype )
874880 if isinstance (pandas_dtype , _pandas_api .extension_dtype ):
875- if strings_to_categorical and isinstance (
881+ if ( strings_to_categorical or name in categories ) and isinstance (
876882 pandas_dtype , _pandas_api .pd .StringDtype
877883 ):
878884 continue
0 commit comments