def test_with_numeric_policies():
sim = run_sim(
[v('a', 't*x + 1')],
policies=pols(x=['100', '1000']),
num_steps=2,
)
sim.assert_values_match('a', [1, 101, 201], policy=0)
sim.assert_values_match('a', [1, 1001, 2001], policy=1)
def test_broadcasting_dataset_var_in_policy_expression():
sim = run_sim(
[v('x', 'attr')],
policies=pols(attr=['EXAMPLE_LABEL.workers == 1', '5']),
dataset=WORKERS_DATASET_DEF,
allow_errs=True,
)
sim.assert_errors_match([['attr', 'equation', 'extra dimension']])
def test_join_fn_multiple_policies():
sim = run_sim(
[
v('a_var', '3'),
v('b_var',
"join(EXAMPLE_LABEL.factor_4 == a_var, EXAMPLE_LABEL.factor_2)"),
],
policies=pols(attribute1=['1', '3']),
dataset=ASSETS_DATASET_DEF,
num_sims=2,
)
assert (sim.results.b_var == 20).all()
def test_with_policy_formulas():
sim = run_sim(
[
v('a', 'where(t>1, x[t-1], 1)', '1'),
v('b', 'x[t-1]', '0'),
],
policies=pols(x=['a*2', 'a*3']),
)
sim.assert_values_match('a', [1, 1, 2, 4], policy=0)
sim.assert_values_match('a', [1, 1, 3, 9], policy=1)
sim.assert_values_match('b', [0, 2, 2, 4], policy=0)
sim.assert_values_match('b', [0, 3, 3, 9], policy=1)
def test_with_numeric_policies_and_two_sims():
# As above, but with 2 sims
sim = run_sim(
[v('a', 't * x + 1')],
policies=pols(x=['100', '1000']),
num_steps=2,
num_sims=2,
)
for sim_id in range(2):
# Values should be identical across simulations.
sim.assert_values_match('a', [1, 101, 201], policy=0, sim_id=sim_id)
sim.assert_values_match('a', [1, 1001, 2001], policy=1, sim_id=sim_id)
def test_join_fn_policy_dependent_mismatch():
# For policy 1, this will give a valid 1:1 join. But for policy 2, it will
# overmatch (due to there being 2 rows having factor_1=3)
sim = run_sim(
[
v('b_var',
"join(EXAMPLE_LABEL.factor_1 == att1, EXAMPLE_LABEL.factor_2)"),
],
policies=pols(att1=["1", "3"], ),
dataset=ASSETS_DATASET_DEF,
allow_errs=True,
)
sim.assert_errors_match([['b_var', 'equation', 'Join']])
def test_dummy_classifier_multiple_policies():
sim = run_sim(
[
v('pred', 'CLF.predict(t)'),
v('prob', 'CLF.predict_proba(t)'),
],
policies=pols(x=['1', '2']),
models=clf_factory('dummy.pickle'),
)
df = sim.results
# All predicted labels should be 0 or 1
assert df.pred.isin([0, 1]).all()
# All probabilities should be .5
assert (df.prob == .5).all()
assert len(df) == 4 * 2 # 4 timesteps, 2 policies, 1 sim
def test_sum_constant_user_dataset_var():
"""Calling sum on a DatasetAdditionVar with a constant value
(and also using multiple sims and policies)"""
db = deepcopy(WORKERS_DATASET_DEF)
db._add_var(v('seven', '7'))
sim = run_sim(
[
v('x', 'sum(EXAMPLE_LABEL.seven)'),
],
policies=pols(attribute1=['1', '2']),
dataset=db,
num_sims=5,
)
x = sim.results.x
assert (x == 14).all()
assert len(x) == 40 # 5 sims * 2 policies * 4 timesteps
def test_time_indexing_dataset_vars():
db = deepcopy(ASSETS_DATASET_DEF)
db._add_var(v('ds_var1', 'EXAMPLE_LABEL.factor_1'))
db._add_var(v('ds_var2', 'ds_var1[t-1]', '0'))
sim = run_sim(
[
v('lagged_var',
'sum(EXAMPLE_LABEL.ds_var1[t-1] + EXAMPLE_LABEL.ds_var2)', '0')
],
policies=pols(attribute1=['1']),
dataset=db,
)
sim.assert_values_match(
'lagged_var',
[0, 48, 48, 48],
)
def test_with_dataset_formulas_accessing_raw_data_and_attributes_and_vars():
db = deepcopy(WORKERS_DATASET_DEF)
db._add_var(
v(
'rev_per_worker',
"EXAMPLE_LABEL.max_production / EXAMPLE_LABEL.workers * price_per_unit * a_var",
))
sim = run_sim(
[v('a_var', 't/10')],
policies=pols(price_per_unit=['10', '5']),
dataset=db,
num_sims=4,
)
sim.assert_values_match(
'EXAMPLE_LABEL.rev_per_worker',
[50, 10 / 3],
t=1,
policy=1,
)
def test_with_dataset_formulas_accessing_raw_data():
db = deepcopy(WORKERS_DATASET_DEF)
db._add_var(v('t_again', 't'))
sim = run_sim(
policies=pols(price_per_unit=['10', '1']),
dataset=db,
)
sim.assert_values_match(
'EXAMPLE_LABEL.t_again',
[
1,
1,
2,
2,
3,
3,
],
policy=0,
)
def test_with_dataset_formulas_accessing_attributes():
db = deepcopy(WORKERS_DATASET_DEF)
db._add_var(
v(
'rev_per_worker',
"EXAMPLE_LABEL.max_production / EXAMPLE_LABEL.workers * price_per_unit"
))
sim = run_sim(
[],
policies=pols(price_per_unit=['10', '5']),
dataset=db,
num_sims=4,
)
sim.assert_values_match_across_time(
'EXAMPLE_LABEL.rev_per_worker',
[1000, 200 / 3],
policy=0,
)
sim.assert_values_match_across_time(
'EXAMPLE_LABEL.rev_per_worker',
[500, 100 / 3],
policy=1,
)
import pytest
from sample_datasets import *
from helpers import run_sim, pols
from decisionai.variable import Variable as v
from copy import deepcopy
# TODO: All through this file we're reusing some mocks, copying and mutating
# that's unlike actual usage patterns. Fix that.
# Used when we want to test a sim with multiple policies, but we don't really
# care about the contents of their attributes.
DUMMY_POLS = pols(unused_attr=['0', '5'])
def test_join_fn_simple():
sim = run_sim(
[
v('a_var', '3'),
v('b_var',
"join(EXAMPLE_LABEL.factor_4 == a_var, EXAMPLE_LABEL.factor_2)"),
],
dataset=ASSETS_DATASET_DEF,
num_sims=2,
)
sim.assert_values_match(
'b_var',
[20, 20, 20, 20],
sim_id=1,
)
def test_attribute_syntax_error():
sim = run_sim(
policies=pols(x=['1', 't-']),
)
sim.assert_errors_match([ ['x', 'equation'] ])
