fields.buffers.base.BufferBase.from_array#
- abstract classmethod BufferBase.from_array(obj: Any, *args, dtype: Any | None = None, **kwargs) BufferBase[source]#
Attempt to construct a new buffer instance from an array-like object.
This method is the canonical entry point for converting arbitrary array-like inputs into a buffer of this type. It behaves similarly to a cast operation, and will coerce the input as needed to match the expected backend format (e.g.,
ndarray, class:~unyt.unyt_array, etc.).The method should be overridden in subclasses to handle type conversion, unit attachment, memory layout, or any other backend-specific behavior.
- Parameters:
obj (
array-like) – Input data to be wrapped. This can be any object that is compatible with the backend’s array casting rules—such as lists, tuples, NumPy arrays, unyt arrays, or backend-native types (e.g., HDF5 datasets). The input will be coerced into a backend-compatible array before being wrapped in a buffer instance. If coercion fails, a TypeError will be raised.dtype (
data-type, optional) – Desired data type of the resulting array. If not specified, the type is inferred from obj.*args –
Additional arguments to customize the construction. These may include:
units for unit-aware buffers
order, copy, or device for backend-specific configuration
Any arguments accepted by the backend constructor
**kwargs –
Additional arguments to customize the construction. These may include:
units for unit-aware buffers
order, copy, or device for backend-specific configuration
Any arguments accepted by the backend constructor
- Returns:
A new buffer instance wrapping the coerced array.
- Return type:
- Raises:
TypeError – If the input cannot be coerced into a valid array for this backend.