(Aside: my question is equally applicable to numpy structured arrays and non-structured arrays.)

Suppose I have a numpy structured array with the dtype:

EXAMPLE_DTYPE = np.dtype([("alpha", np.str_), ("beta", np.int64)])

I have a wrapper around the numpy data array of this dtype, and then I want to implement a special __getitem__ for it:

class ExampleArray:
  data: np.ndarray = np.array([("hello", 0), ("world", 1)], dtype=EXAMPLE_DTYPE)

  def __getitem__(self, index: int|str) -> SpecializedArray:
    return SpecializedArray(index, self.data[index])

SpecializedArray is a class which keeps track of the indices used to specialize from a parent array into a derived array:

class SpecializedArray:
  specialization: int|str 
  data: np.ndarray

  def __init__(self, specialization: int|str, data: np.ndarray) -> None:
    self.specialization = specialization
    self.data = data

  def __repr__(self) -> str:
    return f"{self.specialization} -> {self.data}"

I would like SpecializedArray to ideally have a “reference” to the parent array it was derived from. What is the best way to provide such a reference? Should I provide the parent by slicing it, since slicing creates a view?

# method in ExampleArray
  def __getitem__(self, index: int|str) -> SpecializedArray:
    # suppose that SpecializedArray took the parent as a third argument
    return SpecializedArray(index, self.data[index], self.data[:])