Evaluate adding non-inherited ak.DataFrame methods to the pandas DataFrame accessor #5200
Description
Some functionality available on legacy ak.DataFrame is not inherited by pd.DataFrame when one or more columns are backed by ArkoudaExtensionArray. A DataFrame accessor exists (or is planned) and could be used to expose this additional functionality in a pandas-friendly way.
This ticket is to evaluate whether selected ak.DataFrame methods should be implemented or surfaced via the accessor, based on:
usefulness in pandas-style workflows
API clarity / discoverability (vs keeping legacy-only)
performance or Arkouda-native advantages
The goal is to identify which non-inherited ak.DataFrame methods (if any) belong in the accessor and which should remain legacy-only.
- GroupBy
- GroupBy_class
-
_MutableMapping__marker -
__abstractmethods__ -
__class_getitem__ -
__reversed__ -
__slots__ -
_abc_impl -
_drop_column -
_drop_row -
_get_head_tail -
_get_head_tail_server -
_ipython_key_completions_ -
_ncols -
_parse_col_names -
_prep_data -
_reindex -
_rename_column -
_rename_index -
_set_index -
_shape_str -
_to_hdf_snapshot - append
- apply_permutation
- argsort
- clear
- coargsort
- concat
- filter_by_range
- from_pandas
- from_return_msg
- fromkeys
- is_registered
- load
- memory_usage_info
- objType
- popitem
- read_csv
- register
- setdefault
- to_pandas
- transfer
- unregister
- update_hdf
- update_nrows