Documentation, Typing and Refactors

[Jeadie]

Changeset

• Add typing to functions
• Add docstrings to most functions
• Change Definition from a collections.namedtuple to a dataclass
• Simplify ann_benchmarks.definitions.create_definitions_from_algorithm by decomposing into separate functions.
• Create a named tuple for Metric

  class Metric(NamedTuple):
      distance: Callable[[np.ndarray, np.ndarray], float]
      distance_valid: Callable[[float], bool]

• Simplify main function in ann_benchmarks/main.py by moving filter functions (e.g. filter_disabled_algorithms) to separate python functions.
• NOTE: The above refactors should not alter the underlying logic of ann-benchmarks.

Motivation

I started working on specific changes to ann-benchmarks (e.g. adding new ANN algorithms, extending ann-benchmarks to be used as a performance testing library), and found the codebase hard to understand as a new user. This PR attempts to make the codebase easier to understand and well documented.

[maumueller]

Thanks @Jeadie, never got around to do this super useful work!

Any reason not to merge this @erikbern?

[erikbern]

Hi, sorry, let's ship this. Thank you so much for the contribution @Jeadie !!!