from_numpy_tensor

def from_numpy_tensor[mut: Bool, //, dtype: DType, rank: Int, origin: Origin[mut=mut]](ref[origin] array: PythonObject) -> NumPyView[dtype, rank, origin]

Borrows an N-D C-contiguous NumPy array as a NumPyView.

This is the N-D counterpart of from_numpy_array: the view aliases the array's whole buffer in C order (last axis varying fastest) and carries the extents needed to index it. No bytes are copied, and the same aliasing rules as from_numpy_array apply.

Example:

from std.python import Python
from std.python.numpy import from_numpy_tensor

var np = Python.import_module("numpy")
var array = np.arange(6, dtype="float64").reshape(2, 3)
var view = from_numpy_tensor[DType.float64, 2](array)
var value = view[1, 2]          # array[1, 2]
var flat = view.data            # the buffer as a `Span`, in C order

Constraints:

dtype must be one of the fixed-width numeric dtypes supported by NumPy. rank must be positive.

Parameters:

• ​mut (Bool): The mutability of the borrow, inferred from array.
• ​dtype (DType): The expected element dtype of the array.
• ​rank (Int): The expected number of dimensions.
• ​origin (Origin[mut=mut]): The origin of the borrow, inferred from array.

Args:

• ​array (PythonObject): A C-contiguous NumPy ndarray of rank dimensions whose dtype matches dtype.

Returns:

NumPyView[dtype, rank, origin]: A NumPyView of the array's buffer and shape, with the same mutability and origin as the array binding.

Raises:

If array.ndim is not rank, is not C-contiguous, has a dtype that does not match dtype, or is not writable when borrowed mutably.