deeptrack.types Module#

Type declarations for internal use.

This module defines type aliases and utility types to standardize the type annotations used throughout the codebase. It enhances code readability, maintainability, and reduces redundancy in type annotations. These types are particularly useful for properties and array-like structures used within the library.

Defined Types#

  • PropertyLike

    A type alias representing a value of type T or a callable returning T.

  • ArrayLike

    A type alias for array-like structures, namely, tuples, lists, NumPy arrays, and PyTorch tensors.

  • NumberLike

    A type alias for numeric types, including scalars and arrays, namely, NumPy arrays, PyTorch tensors, bool, int, float, and complex.

Examples#

>>> import deeptrack as dt

Using `PropertyLike`

>>> def scale(value: PropertyLike[float]) -> float:
...     if callable(value):
...         return value()
...     return value

It works for a given type (in this case, a float):

>>> scale(3.14)

It also works for function returning the same type (in this case, a function returning a float):

>>> scale(lambda: 2.71)

PropertyLike[Type] is generally used for typing arguments passed to a feature that are then passed to the constructor of the feature parent, because these can be intrinsically either Type or Callable[…, Type].

Using `ArrayLike`

>>> def print_arraylike(array: dt.types.ArrayLike[float]) -> None:
...     print(array)

It works for:

  • Lists:

>>> print_arraylike([1.0, 2.0, 3.0])

  • Tuples:

>>> print_arraylike((4.0, 5.0, 6.0))

  • NumPy arrays:

>>> import numpy as np
>>>
>>> print_arraylike(np.array([7.0, 8.0, 9.0]))

  • PyTorch tensors:

>>> import torch
>>>
>>> print_arraylike(torch.tensor([1.0, 2.0, 3.0]))

Using `NumberLike`

>>> def add_numbers(a: NumberLike, b: NumberLike) -> NumberLike:
...     return a + b

It works for:

  • Scalars (bool, int, float, complex):

>>> add_numbers(5, 3.2)

  • NumPy arrays:

>>> import numpy as np
>>>
>>> add_numbers(np.array([1, 2, 3]), 4)

  • PyTorch tensors:

>>> import torch
>>>
>>> add_numbers(torch.tensor([1, 2, 3]), 4)