Static-frame

Enhanced support for ``fill_value`` as a ``FillValueAuto``, a mapping, or a sequence of fill values (per column) where appropriate.

Added ``Index.dropna()``, ``Index.dropfalsy()``, ``IndexHierarchy.dropna()``, ``IndexHierarchy.dropfalsy().

Added ``Index.fillfalsy()``, ``IndexHierarchy.fillfalsy()``.

Performance improvements for ``Frame.iter_group_labels()``, ``Frame.iter_group_labels_items()``, ``Frame.iter_group_labels_array()``, ``Frame.iter_group_labels_array_items()``.

Fixed usage of ``dtypes`` argument when encountering zero-sized data in ``Frame.from_records()`` and ``Frame.from_pandas()``.

Improved ``Frame.iter_tuple`` to not coerce types through arrays.

Added ``Frame.set_columns()``, ``Frame.set_columns_hierarchy()``, and ``Frame.unset_columns()``.

Added ``apply_element()`` and ``apply_element_items()`` methods to ``FrameAssign`` and ``SeriesAssign`` interfaces.

Added implementation of ``__array__()`` and ``__array_ufunc__()`` to all containers for better support with NumPy objects and binary operators.

Added ``Series.iter_group_array()``, ``Series.iter_group_array_items()``, ``Series.iter_group_labels_array()``, ``Series.iter_group_labels_array_items()``

Added ``Frame.iter_group_array()``, ``Frame.iter_group_array_items()``, ``Frame.iter_group_labels_array()``, ``Frame.iter_group_labels_array_items()``

Corrected issue when using binary operators with a `FrameGO` and a `Series`.

Corrected issue and performance of ``name`` assignment when extracting ``Series`` from ``Frame`` with an ``IndexHierarchy``.

Corrected more issues when calling ``IndexHierarchy.loc[]`` with another ``IndexHierarchy``, or when calling ``Frame.assign.apply`` when that frame has ``IndexHierarchy`` columns.

Corrected undesirable type coercion from happening in single-row selections from ``IndexHierarchy``.

Corrected issue when calling `IndexDatetime.loc[]` with an empty list.

Corrected issue when calling `IndexHierarchy.loc[]` with another `IndexHierarchy`

API change: ``Bus`` no longer accepts a ``Series`` on initialization; use ``Bus.from_series()``.

API change: ``Batch`` no longer normalizes containers after each step in processing; use ``Batch.via_container`` to force elements or arrays to ``Frame`` or ``Series``.

API change: ``Index`` objects can no longer be created with ``np.datetime64`` arrays; such labels must use an ``IndexDatetime`` subclass instead. If this is happening implicitly with an operation, that operation should expose a parameter for ``index_constructor`` or ``index_constructors``.

API change: ``IndexAutoFactory`` is no longer accepted as an ``index_constructor`` argument in ``Series.from_pandas()`` and ``Frame.from_pandas()``; ``IndexAutoFactory`` should be passed as an ``index`` or ``columns`` argument instead.

Minimum Python version is now 3.7

New implementation of ``IndexHierarchy``, offering significantly improved performance and removal of the requirement of tree hierarchies.

Added ``Batch.to_series()``.

Fixed issue when using ``Frame.from_npz`` with an NPZ created with a ``FrameGO``.

Fixed issue when supplying overspecified mappings to ``Frame.astype``.