def test_death_z_test(self, reset_herbivore_params, omega_dict):
"""
Based on H.E. Plesser: biolab/test_bacteria.py
Probabilistic test of death function. Testing on herbivores.
Given that the sample size is large. Under the Z-test the distribution under the null
hypothesis can be estimated approximately by the normal distribution.
See https://en.wikipedia.org/wiki/Z-test.
Assuming low fitness of animals such that omega can be interpreted as an approximation of
the death probability. Testing with different values of omega
We compute the number of dead animals returned by our death function from class Animal.
Then we compare this value to the mean of dead animals derived from a fixed probability.
Null hypothesis: The number of dead animals returned by the death function is
statistically significant with a p-value greater than the alpha parameter.
Alternative hypothesis: The number of dead animals returned is not statistically
significant and we reject the null hypothesis.
"""
random.seed(123)
# High age ensures low fitness
herb = Herbivore(age=100, weight=10)
# with low fitness we assume that the death probability is the same as omega
herb.set_params(omega_dict)
# death probability set equal to omega
p = Herbivore.p["omega"]
# Number of animals
N = 1000
# Number of dead animals
n = sum(herb.death() for _ in range(N))
# Performing the z-test
assert phi_z_test(N, p, n) > TestAnimal.alpha
def test_set_params(self):
"""
Test that parameters can be set
"""
my_params = {"w_birth": 10, "sigma_birth": 2.5}
Herbivore.set_params(my_params)
assert Herbivore.p["w_birth"] == 10
assert Herbivore.p["sigma_birth"] == 2.5
def test_aging(self):
"""
Test that the animal age increases
"""
herb = Herbivore(weight=10, age=0)
carn = Carnivore(weight=10)
herb.aging()
carn.aging()
assert herb.age > 0, carn.age > 0
def test_add_remove_animals(self, highland_cell):
"""
:method: LandscapeCell.add_animals
:method: LandscapeCell.remove_animals
:property: LandscapeCell.animal_count
Test that animals can be added and removed correctly
"""
animals = [Herbivore(), Carnivore(), Herbivore()]
highland_cell.add_animals(animals)
highland_cell.remove_animals([animals[0], animals[1]])
assert highland_cell.herb_count == 1
def test_eat_fodder(self):
"""
Weight of animal shall increase after eating fodder
"""
herb = Herbivore(weight=10, age=0)
herb_weight = herb.weight
herb.eat_fodder(cell=Lowland())
# new weight
herb_weight_after = herb.weight
assert herb_weight < herb_weight_after
def set_animal_parameters(species, params):
"""Set parameters for animal species.
:param species: String, name of animal species
:param params: Dict with valid parameter specification for species
"""
if species == "Herbivore":
Herbivore.set_params(params)
elif species == "Carnivore":
Carnivore.set_params(params)
else:
raise ValueError("species needs to be either Herbivore or Carnivore!")
def test_certain_birth(self, mocker, reset_herbivore_params):
"""
test give birth function
Mock the random number generator to always return one.
Then as long as weight is not zero. give_birth function shall return True.
"""
herb = Herbivore(weight=800, age=5)
num_herbs = 10
mocker.patch("random.random", return_value=0)
give_birth, _ = herb.give_birth(num_herbs)
assert give_birth
def test_lose_weight(self, reset_herbivore_params, reset_carnivore_params):
"""
Test that animals lose weight
"""
herb, carn = Herbivore(weight=20), Carnivore(weight=20)
# Decreasing parameters
herb.p['eta'] = 0.1
carn.p['eta'] = 0.2
herb_initial_weight, carn_initial_weight = herb.weight, carn.weight
herb.lose_weight(), carn.lose_weight()
# New weight of animal must be less than before
assert herb.weight < herb_initial_weight
assert carn.weight < carn_initial_weight
def test_sorted_herbivores_and_carnivores(self, highland_cell):
"""
:method: LandscapeCell.add_animals
:method: LandscapeCell.sorted_herbivores
:method: LandscapeCell.sorted_carnivores
Check that sorting algorithms sort the lists by fitness
"""
highland_cell.add_animals([Herbivore(weight=50), Herbivore(weight=20)])
highland_cell.add_animals([Carnivore(weight=25), Carnivore(weight=40)])
sorted_herbivores = list(
[herb[0] for herb in highland_cell.sorted_herbivores])
sorted_carnivores = highland_cell.sorted_carnivores
assert sorted_herbivores == highland_cell.herbivores[::-1]
assert sorted_carnivores == highland_cell.carnivores[::-1]
def test_constructor(self):
"""
Animals can be created
"""
herb = Herbivore(weight=10, age=0)
carn = Carnivore()
assert isinstance(herb, Herbivore), isinstance(carn, Carnivore)
def procreation(self, cell):
"""Iterates through each animal in the cell and procreates.
:param cell: Current cell object
:type cell: object
"""
new_herbs = []
new_carns = []
n_herbs, n_carns = cell.herb_count, cell.carn_count
for herb in cell.herbivores: # Herbivores give birth)
give_birth, birth_weight = herb.give_birth(n_herbs)
if give_birth:
new_herbs.append(Herbivore(weight=birth_weight, age=0))
for carn in cell.carnivores: # Carnivores give birth
give_birth, birth_weight = carn.give_birth(n_carns)
if give_birth:
new_carns.append(Carnivore(weight=birth_weight, age=0))
cell.add_animals(new_herbs + new_carns) # Add new animals to cell
self._island.count_animals(num_herbs=len(new_herbs), num_carns=len(new_carns))
def add_population(self, population):
"""Add a population to specific `island` cells by providing a list of dictionaries.
:param population: List of dictionaries specifying population
:Example:
.. code-block:: python
example_pop = {
'loc': (4,4),
'pop': [
{'species': 'Herbivore', 'age': 2, 'weight': 60},
{'species': 'Herbivore', 'age': 9, 'weight': 30},
{'species': 'Herbivore', 'age': 16, 'weight': 14}
]
}
"""
if type(population) == list:
for loc_dict in population: # This loop will be replaced with a more elegant iteration
new_animals = [
Herbivore.from_dict(animal_dict)
if animal_dict["species"] == "Herbivore"
else Carnivore.from_dict(animal_dict)
for animal_dict in loc_dict["pop"]
]
self._island.landscape[loc_dict["loc"]].add_animals(new_animals)
self._island.count_animals(animal_list=new_animals)
else:
raise ValueError(
f"Pop list needs to be a list of dicts! Was of type " f"{type(population)}."
)
def animals(self):
"""
create animals of different type, age and weight to use in test of fitness
"""
animals = [
Herbivore(age=0, weight=5),
Herbivore(age=0, weight=1000),
Herbivore(age=100, weight=5),
Herbivore(age=100, weight=1000),
Herbivore(age=0, weight=5),
Carnivore(age=0, weight=5),
Carnivore(age=0, weight=1000),
Carnivore(age=100, weight=5),
Carnivore(age=100, weight=1000)
]
return animals
def test_set_invalid_params(self, reset_herbivore_params):
"""
Test errors with illegal keys and values
"""
with pytest.raises(KeyError):
assert Herbivore.p["w_death"]
with pytest.raises(ValueError):
assert Herbivore.set_params({"sigma_birth": -5})
def test_mean_birth_weight(self, birth_dict, reset_herbivore_params):
""" Test that the birth weight of animals are normal distributed using the normaltest
from scipy.
https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.normaltest.html
Null hypothesis: The birth weight of the animal is normally distributed
Alternative hypothesis: The birth weight are not normally distributed.
We keep the null hypothesis if the p-value is larger than the significance level alpha
"""
random.seed(123)
N = 1000
Herbivore.set_params(birth_dict)
herb_birth_weights = [
Herbivore(age=5, weight=20).birth_weight for _ in range(N)
]
k2, phi = stats.normaltest(herb_birth_weights)
assert phi > TestAnimal.alpha
def test_animal_fitness(self, biosim):
"""
:property: Island.animal_fitness
Test that animal fitness property is of correct shape and value
"""
fixed_herb_fitness = Herbivore(weight=20, age=5).fitness
fixed_carn_fitness = Carnivore(weight=25, age=4).fitness
assert biosim._island.animal_fitness == [[
fixed_herb_fitness, fixed_herb_fitness, fixed_herb_fitness
], [fixed_carn_fitness, fixed_carn_fitness]]
def test_shuffle_herbs(self, highland_cell):
"""
:method: LandscapeCell.add_animals
:method: LandscapeCell.randomize_herbs
Test that shuffle method shuffles herbivores list
"""
highland_cell.add_animals([Herbivore() for _ in range(1000)])
original_herbs = [animal for animal in highland_cell.herbivores]
highland_cell.randomize_herbs()
assert highland_cell.herbivores != original_herbs
def test_kill_prey(self):
"""
Test kill prey. With a high fitness diff the carnivore will always kill the herbivore.
"""
carn = Carnivore(age=5, weight=30)
herb_list = [Herbivore(age=100, weight=50) for _ in range(100)]
mock_sorted_list = [(herb, herb.fitness) for herb in herb_list]
kill_count = 0
for _ in mock_sorted_list:
if carn.kill_prey(mock_sorted_list):
kill_count += 1
assert kill_count == 100
def test_reset_animals(self, highland_cell):
"""
:method: LandscapeCell.add_animal
:method: LandscapeCell.reset_animals
:property: LandscapeCell.carnivores
:property: LandscapeCell.has_moved
Test that has_moved property is correctly set and reset
"""
highland_cell.add_animals([Carnivore(), Herbivore()])
if highland_cell.carnivores[0].has_moved:
highland_cell.reset_animals()
assert not highland_cell.carnivores[0].has_moved
def test_count_del_animals(self, island):
"""
:method: Island.count_animals
:method: Island.del_animals
Test counting and removing animals with lists passed and arguments only
"""
island.count_animals(num_herbs=10, num_carns=10)
island.del_animals(animal_list=[Herbivore()])
island.del_animals(num_herbs=5, num_carns=5)
assert island.num_animals == 9
assert island.num_herbs == 4
assert island.num_carns == 5
def test_give_birth(self, gamma_dict, reset_herbivore_params):
"""Test that for animals with fitness close to one, and two animals of same type one specie
in a cell. The give_birth function should be well approximated by the parameter gamma.
An we test this against our function under the significance level alpha.
Null hypothesis: The give_birth function returns correct with fixed gamma
Alternative hypothesis: The give_birth function does not return correct. We reject our
null hypothesis.
"""
random.seed(123)
N = 1000
Herbivore.set_params(gamma_dict)
num_herbs = 2
p = gamma_dict["gamma"]
list_birth = [
Herbivore(weight=200, age=5).give_birth(num_herbs)
for _ in range(N)
]
# number when births return True
n = sum([item[0] for item in list_birth])
mean = N * p
assert phi_z_test(N, p, n) > TestAnimal.alpha
def test_single_procreation(self):
"""
test that the initial herbivore population will not reproduce a newborn population of
greater numbers during a year cycle. Each mother can at most give birth to one animal.
A high fitness and gamma parameter ensures highly fertile animals.
"""
num_newborns = 0
adult_herbs = [Herbivore(age=5, weight=40) for _ in range(100)]
num_adults = len(adult_herbs)
for herb in adult_herbs:
herb.set_params(({"gamma": 0.99}))
if herb.give_birth(num_adults):
num_newborns += 1
assert num_newborns <= num_adults
def test_weight_gain(self, reset_carnivore_params, reset_herbivore_params,
params):
"""
Testing weight gain of carnivores. Assuming they have access to more fodder than they
will eat. Making old heavy herbivores with low fitness. Carnivores should add beta * F
weight
Assuming fitness diff is larger than DeltaPhiMax such that the carnivore always kills
the herbivore.
"""
carn = Carnivore(age=5, weight=40)
# number of herbivores
N = 1000
herb_list = [Herbivore(age=100, weight=200) for _ in range(N)]
mock_sorted_list = [(herb, herb.fitness) for herb in herb_list]
initial_weight = carn.weight
_ = carn.kill_prey(mock_sorted_list)
new_weight = carn.weight
assert new_weight == initial_weight + carn.p["beta"] * carn.p["F"]
def herbivore(self):
return Herbivore()
def test_constructor(self):
"""
Test herbivores can be constructed
"""
herb = Herbivore()
assert isinstance(herb, Herbivore)
def test_parameters(self, reset_herbivore_params, reset_carnivore_params):
"""
Test parameters of herbs and carns
"""
herb, carn = Herbivore(), Carnivore()
assert herb.p != carn.p
def reset_herbivore_params():
"""
Based on test_dish.py
set parameters of herbivores back to defaults
"""
yield Herbivore.set_params(Herbivore.p)