"""Test stuff on bit strings."""
from typing import Any, Callable, Final, Iterable, cast
import numpy as np
from numpy.random import Generator, default_rng
from pycommons.types import check_int_range, type_error, type_name_of
from moptipy.algorithms.so.fitness import Fitness
from moptipy.api.algorithm import Algorithm, check_algorithm
from moptipy.api.execution import Execution
from moptipy.api.mo_algorithm import MOAlgorithm
from moptipy.api.mo_problem import MOProblem
from moptipy.api.objective import Objective
from moptipy.api.operators import Op0, Op1, Op1WithStepSize, Op2
from moptipy.examples.bitstrings.ising1d import Ising1d
from moptipy.examples.bitstrings.leadingones import LeadingOnes
from moptipy.examples.bitstrings.onemax import OneMax
from moptipy.examples.bitstrings.zeromax import ZeroMax
from moptipy.mo.problem.weighted_sum import WeightedSum
from moptipy.operators.bitstrings.op0_random import Op0Random
from moptipy.spaces.bitstrings import BitStrings
from moptipy.tests.algorithm import validate_algorithm
from moptipy.tests.fitness import validate_fitness
from moptipy.tests.mo_algorithm import validate_mo_algorithm
from moptipy.tests.op0 import validate_op0
from moptipy.tests.op1 import validate_op1
from moptipy.tests.op1_with_step_size import validate_op1_with_step_size
from moptipy.tests.op2 import validate_op2
from moptipy.utils.nputils import array_to_str
[docs]
def dimensions_for_tests() -> Iterable[int]:
"""
Get a sequence of dimensions for tests.
:returns: the sequence of integers
"""
r = default_rng()
bs: list[int] = [1, 2, 3, 4, 5, 10, 16, 100,
int(r.integers(20, 50)), int(r.integers(200, 300))]
r.shuffle(cast(list, bs))
return bs
[docs]
def bitstrings_for_tests() -> Iterable[BitStrings]:
"""
Get a sequence of bit strings for tests.
:returns: the sequence of BitStrings
"""
return [BitStrings(i) for i in dimensions_for_tests()]
[docs]
def random_bit_string(random: Generator, x: np.ndarray) -> np.ndarray:
"""
Randomize a bit string.
:param random: the random number generator
:param x: the bit string
:returns: the array
"""
ri = random.integers
for i in range(len(x)): # pylint: disable=C0200
x[i] = ri(2) <= 0
return x
[docs]
def validate_op0_on_1_bitstrings(
op0: Op0 | Callable[[BitStrings], Op0],
search_space: BitStrings,
number_of_samples: int | None = None,
min_unique_samples:
int | Callable[[int, BitStrings], int] | None
= None) -> None:
"""
Validate the unary operator on one bit strings instance.
:param op0: the operator or operator factory
:param search_space: the search space
:param number_of_samples: the optional number of samples
:param min_unique_samples: the optional unique samples
"""
args: dict[str, Any] = {
"op0": op0(search_space) if callable(op0) else op0,
"search_space": search_space,
"make_search_space_element_valid": random_bit_string,
}
if number_of_samples is not None:
args["number_of_samples"] = number_of_samples
if min_unique_samples is not None:
args["min_unique_samples"] = min_unique_samples
validate_op0(**args)
[docs]
def validate_op0_on_bitstrings(
op0: Op0 | Callable[[BitStrings], Op0],
number_of_samples: int | None = None,
min_unique_samples:
int | Callable[[int, BitStrings], int] | None
= None) -> None:
"""
Validate the unary operator on several BitStrings instances.
:param op0: the operator or operator factory
:param number_of_samples: the optional number of samples
:param min_unique_samples: the optional unique samples
"""
for bst in bitstrings_for_tests():
validate_op0_on_1_bitstrings(op0, bst,
number_of_samples, min_unique_samples)
[docs]
def validate_op1_on_1_bitstrings(
op1: Op1 | Callable[[BitStrings], Op1],
search_space: BitStrings,
number_of_samples: int | None = None,
min_unique_samples:
int | Callable[[int, BitStrings], int] | None
= None) -> None:
"""
Validate the unary operator on one BitStrings instance.
:param op1: the operator or operator factory
:param search_space: the search space
:param number_of_samples: the optional number of samples
:param min_unique_samples: the optional unique samples
"""
args: dict[str, Any] = {
"op1": op1(search_space) if callable(op1) else op1,
"search_space": search_space,
"make_search_space_element_valid": random_bit_string,
}
if number_of_samples is not None:
args["number_of_samples"] = number_of_samples
if min_unique_samples is not None:
args["min_unique_samples"] = min_unique_samples
validate_op1(**args)
[docs]
def validate_op1_on_bitstrings(
op1: Op1 | Callable[[BitStrings], Op1],
number_of_samples: int | None = None,
min_unique_samples:
int | Callable[[int, BitStrings], int] | None
= None) -> None:
"""
Validate the unary operator on several BitStrings instances.
:param op1: the operator or operator factory
:param number_of_samples: the optional number of samples
:param min_unique_samples: the optional unique samples
"""
for bst in bitstrings_for_tests():
validate_op1_on_1_bitstrings(op1, bst,
number_of_samples, min_unique_samples)
[docs]
def validate_op1_with_step_size_on_1_bitstrings(
op1: Op1WithStepSize | Callable[[BitStrings], Op1WithStepSize],
search_space: BitStrings,
number_of_samples: int | None = None,
min_unique_samples: int | Callable[[
int, BitStrings], int] | None = None,
step_sizes: Iterable[float] | Callable[
[BitStrings], Iterable[float]] = (),
get_step_size: Callable[[
BitStrings, np.ndarray, np.ndarray,
], float | None] | None = None) -> None:
"""
Validate the step-sized unary operator on one `BitStrings` instance.
:param op1: the operator or operator factory
:param search_space: the search space
:param number_of_samples: the optional number of samples
:param min_unique_samples: the optional unique samples
:param step_sizes: the step sizes to test
:param get_step_size: try to get the step size from two space elements
"""
args: dict[str, Any] = {
"op1": op1(search_space) if callable(op1) else op1,
"search_space": search_space,
"make_search_space_element_valid": random_bit_string,
"step_sizes": step_sizes(search_space) if callable(step_sizes)
else step_sizes,
"get_step_size": get_step_size,
}
if number_of_samples is not None:
args["number_of_samples"] = number_of_samples
if min_unique_samples is not None:
args["min_unique_samples"] = min_unique_samples
validate_op1_with_step_size(**args)
[docs]
def validate_op1_with_step_size_on_bitstrings(
op1: Op1WithStepSize | Callable[[BitStrings], Op1WithStepSize],
number_of_samples: int | None = None,
min_unique_samples: int | Callable[[
int, BitStrings], int] | None = None,
step_sizes: Iterable[float] | Callable[
[BitStrings], Iterable[float]] = (),
get_step_size: Callable[[
BitStrings, np.ndarray, np.ndarray,
], float | None] | None = None) -> None:
"""
Validate the unary operator on several `BitStrings` instances.
:param op1: the operator or operator factory
:param number_of_samples: the optional number of samples
:param min_unique_samples: the optional unique samples
:param step_sizes: the step sizes to test
:param get_step_size: try to get the step size from two space elements
"""
for bs in bitstrings_for_tests():
validate_op1_with_step_size_on_1_bitstrings(
op1, bs, number_of_samples, min_unique_samples, step_sizes,
get_step_size)
[docs]
def validate_op2_on_1_bitstrings(
op2: Op2 | Callable[[BitStrings], Op2],
search_space: BitStrings,
number_of_samples: int | None = None,
min_unique_samples:
int | Callable[[int, BitStrings], int] | None
= None) -> None:
"""
Validate the binary operator on one BitStrings instance.
:param op2: the operator or operator factory
:param search_space: the search space
:param number_of_samples: the optional number of samples
:param min_unique_samples: the optional unique samples
"""
args: dict[str, Any] = {
"op2": op2(search_space) if callable(op2) else op2,
"search_space": search_space,
"make_search_space_element_valid": random_bit_string,
}
if number_of_samples is not None:
args["number_of_samples"] = number_of_samples
if min_unique_samples is not None:
args["min_unique_samples"] = min_unique_samples
validate_op2(**args)
[docs]
def validate_op2_on_bitstrings(
op2: Op2 | Callable[[BitStrings], Op2],
number_of_samples: int | None = None,
min_unique_samples:
int | Callable[[int, BitStrings], int] | None
= None) -> None:
"""
Validate the binary operator on several BitStrings instances.
:param op2: the operator or operator factory
:param number_of_samples: the optional number of samples
:param min_unique_samples: the optional unique samples
"""
for bst in bitstrings_for_tests():
validate_op2_on_1_bitstrings(op2, bst,
number_of_samples, min_unique_samples)
[docs]
def validate_algorithm_on_bitstrings(
objective: Objective | Callable[[int], Objective],
algorithm: Algorithm | Callable[
[BitStrings, Objective], Algorithm],
dimension: int = 5, max_fes: int = 100,
required_result: int | Callable[
[int, int], int] | None = None,
uses_all_fes_if_goal_not_reached: bool = True,
post: Callable[[Algorithm, int], Any] | None = None) -> None:
"""
Check the validity of a black-box algorithm on a bit strings problem.
:param algorithm: the algorithm or algorithm factory
:param objective: the objective function or function factory
:param dimension: the dimension of the problem
:param max_fes: the maximum number of FEs
:param required_result: the optional required result quality
:param uses_all_fes_if_goal_not_reached: will the algorithm use all FEs
unless it reaches the goal?
:param post: a check to run after each execution of the algorithm,
receiving the algorithm and the number of consumed FEs as parameter
"""
if not (isinstance(algorithm, Algorithm) or callable(algorithm)):
raise type_error(algorithm, "algorithm", Algorithm, True)
if not (isinstance(objective, Objective) or callable(objective)):
raise type_error(objective, "objective", Objective, True)
check_int_range(dimension, "dimension", 1, 1_000_000)
if (post is not None) and (not callable(post)):
raise type_error(post, "post", None, call=True)
if callable(objective):
objective = objective(dimension)
if not isinstance(objective, Objective):
raise type_error(objective, "result of callable 'objective'",
Objective)
bs: Final[BitStrings] = BitStrings(dimension)
if callable(algorithm):
algorithm = algorithm(bs, objective)
if not isinstance(algorithm, Algorithm):
raise type_error(algorithm, "result of callable 'algorithm'",
Algorithm)
goal: Final[int | None] = required_result(max_fes, dimension) \
if callable(required_result) else required_result
validate_algorithm(
algorithm=algorithm, solution_space=bs, objective=objective,
max_fes=max_fes, required_result=goal,
uses_all_fes_if_goal_not_reached=uses_all_fes_if_goal_not_reached,
post=post)
[docs]
def validate_algorithm_on_onemax(
algorithm: Algorithm | Callable[
[BitStrings, Objective], Algorithm],
post: Callable[[Algorithm, int], Any] | None = None) -> None:
"""
Check the validity of a black-box algorithm on OneMax.
:param algorithm: the algorithm or algorithm factory
:param post: a check to run after each execution of the algorithm,
receiving the algorithm and the number of consumed FEs as parameter
"""
max_fes: Final[int] = 100
for i in dimensions_for_tests():
rr: int = 1 if i < 3 else (1 + max(1, i // 2, i - int(max_fes ** 0.5)))
validate_algorithm_on_bitstrings(
objective=OneMax,
algorithm=algorithm,
dimension=i,
max_fes=max_fes,
required_result=rr,
post=post)
[docs]
def validate_algorithm_on_leadingones(
algorithm: Algorithm | Callable[
[BitStrings, Objective], Algorithm],
post: Callable[[Algorithm, int], Any] | None = None) -> None:
"""
Check the validity of a black-box algorithm on LeadingOnes.
:param algorithm: the algorithm or algorithm factory
:param post: a check to run after each execution of the algorithm,
receiving the algorithm and the number of consumed FEs as parameter
"""
max_fes: Final[int] = 100
for i in dimensions_for_tests():
rr: int
if i < 3:
rr = 1
elif max_fes > (10 * (i ** 1.5)):
rr = i - 1
else:
rr = i
validate_algorithm_on_bitstrings(
objective=LeadingOnes,
algorithm=algorithm,
dimension=i,
max_fes=int(1.25 * max_fes),
required_result=rr,
post=post)
[docs]
def validate_mo_algorithm_on_bitstrings(
problem: MOProblem | Callable[[int], MOProblem],
algorithm: MOAlgorithm | Callable[
[BitStrings, MOProblem], MOAlgorithm],
dimension: int = 5,
max_fes: int = 100) -> None:
"""
Check a black-box multi-objective algorithm on a bit strings problem.
:param algorithm: the algorithm or algorithm factory
:param problem: the multi-objective optimization problem or factory
:param dimension: the dimension of the problem
:param max_fes: the maximum number of FEs
"""
if not (isinstance(algorithm, MOAlgorithm) or callable(algorithm)):
raise type_error(algorithm, "algorithm", MOAlgorithm, True)
if not (isinstance(problem, MOProblem) or callable(problem)):
raise type_error(problem, "problem", MOProblem, True)
check_int_range(dimension, "dimension", 1, 1_000_000)
if callable(problem):
problem = problem(dimension)
if not isinstance(problem, MOProblem):
raise type_error(problem, "result of callable 'problem'",
MOProblem)
bs: Final[BitStrings] = BitStrings(dimension)
if callable(algorithm):
algorithm = algorithm(bs, problem)
if not isinstance(algorithm, MOAlgorithm):
raise type_error(algorithm, "result of callable 'algorithm'",
MOAlgorithm)
validate_mo_algorithm(algorithm=algorithm,
solution_space=bs,
problem=problem,
max_fes=max_fes)
[docs]
def validate_mo_algorithm_on_2_bitstring_problems(
algorithm: MOAlgorithm | Callable[
[BitStrings, MOProblem], MOAlgorithm]) -> None:
"""
Check the validity of a black-box algorithm on OneMax and ZeroMax.
:param algorithm: the algorithm or algorithm factory
"""
max_fes: Final[int] = 100
random: Final[Generator] = default_rng()
for i in dimensions_for_tests():
weights: list[int | float] = [float(random.uniform(0.01, 10)),
float(random.uniform(0.01, 10))] \
if random.integers(2) <= 0 else \
[1 + int(random.integers(1 << random.integers(40))),
1 + int(random.integers(1 << random.integers(40)))]
validate_mo_algorithm_on_bitstrings(
problem=WeightedSum([OneMax(i), ZeroMax(i)], weights),
algorithm=algorithm,
dimension=i,
max_fes=max_fes)
[docs]
def validate_mo_algorithm_on_3_bitstring_problems(
algorithm: MOAlgorithm | Callable[
[BitStrings, MOProblem], MOAlgorithm]) -> None:
"""
Check the validity of an algorithm on OneMax, ZeroMax, and Ising1d.
:param algorithm: the algorithm or algorithm factory
"""
max_fes: Final[int] = 100
random: Final[Generator] = default_rng()
for i in dimensions_for_tests():
weights: list[int | float] = [float(random.uniform(0.01, 10)),
float(random.uniform(0.01, 10)),
float(random.uniform(0.01, 10))] \
if random.integers(2) <= 0 else \
[1 + int(random.integers(1 << random.integers(40))),
1 + int(random.integers(1 << random.integers(40))),
1 + int(random.integers(1 << random.integers(40)))]
validate_mo_algorithm_on_bitstrings(
problem=WeightedSum([OneMax(i), ZeroMax(i), Ising1d(i)],
weights),
algorithm=algorithm,
dimension=i,
max_fes=max_fes)
[docs]
def verify_algorithms_equivalent(
algorithms: Iterable[Callable[[BitStrings, Objective], Algorithm]],
max_fes: int | None = None) \
-> None:
"""
Verify that a set of algorithms performs identical steps.
:param algorithms: the sequence of algorithms
:param max_fes: the maximum number of FEs
"""
if not isinstance(algorithms, Iterable):
raise type_error(algorithms, "algorithms", Iterable)
if (max_fes is not None) and (not isinstance(max_fes, int)):
raise type_error(max_fes, "max_fes", int)
random: Final[Generator] = default_rng()
dim: Final[int] = int(random.integers(4, 16))
space: Final[BitStrings] = BitStrings(dim)
steps: Final[int] = int(random.integers(100, 1000)) \
if max_fes is None else max_fes
if steps <= 0:
raise ValueError(f"invalid maximum FEs={steps} for max_fes={max_fes}")
choice: Final[int] = int(random.integers(3))
f: Final[Objective] = \
LeadingOnes(dim) if choice <= 0 else \
OneMax(dim) if choice <= 1 else \
Ising1d(dim)
evaluate: Final[Callable] = f.evaluate # noqa
seed: Final[int] = int(random.integers(1 << 62))
result1: Final[list[bool]] = []
result2: Final[list[bool]] = []
first: bool = True
first_name: str = ""
do_fes: int = -1
do_res: int | float = -1
index: int = -1
for algo in algorithms:
index += 1
if not callable(algo):
raise type_error(algo, f"algorithms[{index}] for {f}", call=True)
algorithm: Algorithm = check_algorithm(algo(space, f))
current_name: str = str(algorithm)
if first:
first_name = current_name
result = result1
else:
result = result2
result.clear()
def ff(x, rr=(result, )) -> int:
nonlocal evaluate
rres = evaluate(x)
rr[0].extend(x)
return rres
ex = Execution()
ex.set_algorithm(algorithm)
ex.set_solution_space(space)
f.evaluate = ff # type: ignore
ex.set_objective(f)
ex.set_rand_seed(seed)
ex.set_max_fes(steps)
with ex.execute() as p:
cf = p.get_consumed_fes()
if not (0 < cf <= steps):
raise ValueError(f"{current_name} consumed {cf} FS for "
f"{steps} max FEs on {f} for seed {seed}.")
if first:
do_fes = cf
elif do_fes != cf:
raise ValueError(f"{current_name} consumed {cf} FEs but "
f"{first_name} consumed {do_fes} FEs on "
f"{f} for seed {seed}.")
res = p.get_best_f()
if not (0 <= res <= dim):
raise ValueError(f"{current_name} got {res} as objective "
f"value on {f} for seed {seed}.")
if first:
do_res = res
elif do_res != res:
raise ValueError(
f"{current_name} got {res} as objective value on {f} but "
f"{first_name} got {do_res} for seed {seed}.")
if len(result) != (cf * dim):
raise ValueError(
f"len(result) == {len(result)}, but should be {cf * dim} "
f"for {current_name} for seed {seed} on {f}.")
if (not first) and (result1 != result2):
raise ValueError(
f"{current_name} produced different steps than {first_name} "
f"on {f} for seed {seed}: is "
f"{array_to_str(np.array(result2))}"
f" but should be {array_to_str(np.array(result1))}.")
first = False
[docs]
def validate_fitness_on_bitstrings(
fitness: Fitness | Callable[[Objective], Fitness],
class_needed: str | type = Fitness,
prepare_objective: Callable[[Objective], Objective] = lambda x: x) \
-> None:
"""
Validate a fitness assignment process on bit strings.
:param fitness: the fitness assignment process, or a callable creating it
:param class_needed: the required class
:param prepare_objective: prepare the objective function
"""
if (not isinstance(fitness, Fitness)) and (not callable(fitness)):
raise type_error(fitness, "fitness", Fitness, call=True)
if not isinstance(class_needed, str | type):
raise type_error(class_needed, "class_needed", (str, type))
if not callable(prepare_objective):
raise type_error(prepare_objective, "prepare_objective", call=True)
random: Final[Generator] = default_rng()
sizes: set[int] = set()
while len(sizes) < 4:
sizes.add(int(random.integers(2, 10)))
op0: Op0Random = Op0Random()
for s in sizes:
space: BitStrings = BitStrings(s)
f: Objective = OneMax(s) if random.integers(2) <= 0 else LeadingOnes(s)
f2 = prepare_objective(f)
if not isinstance(f2, Objective):
raise type_error(f2, f"prepare_objective({f})", Objective)
del f
if callable(fitness):
ff = fitness(f2)
if not isinstance(ff, Fitness):
raise type_error(ff, f"fitness({f2})", Fitness)
if isinstance(class_needed, str):
name = type_name_of(ff)
if name != class_needed:
raise TypeError(f"fitness assignment process should be "
f"{class_needed!r}, but is {name!r}.")
elif not isinstance(ff, class_needed):
raise type_error(ff, f"fitness({f2})", class_needed)
else:
ff = fitness
validate_fitness(ff, f2, space, op0)