def test_pension_env_step_payout_action(do_pay_out):
env = pension_env.PensionEnv()
env.reset()
payout_action = 1
env.step(payout_action)
# do_pay_out.assert_called_once()
assert do_pay_out.call_count == 1
def test_pension_env_step_new_years(mocker):
# _leave_cdf, _death_cdf, _new_cdf, live_one_year, run_company):
"""Given N Clients, env.year will advance with every Nth step(),
InsuranceCompany.run_company will be called once per year and
Client.live_one_year will be called once for each step.
"""
# Patch Client object to never leave nor join nor die
mocker.patch.object(pension_env.Client, "_leave_cdf", return_value=1.0)
mocker.patch.object(pension_env.Client, "_death_cdf", return_value=0.0)
mocker.patch.object(pension_env.Client, "_new_cdf", return_value=0.0)
run_company = mocker.patch.object(pension_env.InsuranceCompany,
"run_company")
env = pension_env.PensionEnv()
env.reset()
first_client = env.humans[0]
second_client = pension_env.Client.maybe_make_client(env, force=True)
env.humans.append(second_client)
num_clients = len(env.humans) # There are 2 clients
print("num_clients", num_clients)
mocker.spy(first_client, "live_one_year")
mocker.spy(second_client, "live_one_year")
steps = 10
y0 = env.year # probably 0
debit_action = 0
for _ in range(steps):
env.step(debit_action)
y5 = env.year
expected_year_changes = floor(steps / num_clients)
# 0 + floor(10 / 2) == 5
assert y0 + expected_year_changes == y5
assert run_company.call_count == y0 + expected_year_changes
assert first_client.live_one_year.call_count == steps / num_clients
assert second_client.live_one_year.call_count == steps / num_clients
def test_pension_env_seed():
env = pension_env.PensionEnv()
env.seed(0)
r1 = pension_env.np_random.choice(range(1000))
env.seed(0)
r2 = pension_env.np_random.choice(range(1000))
assert r1 == r2
def test_pension_env_step_debit_action(do_debit_premium):
env = pension_env.PensionEnv()
env.reset()
debit_action = 0
env.step(debit_action)
# do_debit_premium.assert_called_once()
assert do_debit_premium.call_count == 1
def test_pension_env_reset():
env = pension_env.PensionEnv()
env.seed(0) # PensionEnv start state is currently deterministic, but still
s1 = env.reset()
for _ in range(100):
env.step(env.action_space.sample())
env.seed(0)
s2 = env.reset()
assert (s1 == s2).all()
def test_pension_env_step_before_reset():
env = pension_env.PensionEnv()
with pytest.raises(pension_env.PensionEnvError, match="before reset"):
env.step(env.action_space.sample())
def test_pension_env_close():
env = pension_env.PensionEnv()
env.close()
def test_pension_env_render():
env = pension_env.PensionEnv()
env.render()
def test_pension_env_observation_shape():
env = pension_env.PensionEnv()
s = env.reset()
assert s.shape[0] == env.observation_space.low.shape[0]
def test_pension_env_companies():
env = pension_env.PensionEnv()
assert len(env.companies) == 0
env.reset()
assert len(env.companies) > 0
def test_pension_env_init():
assert pension_env.PensionEnv()