def test_learn_api():
'''
Test the learn API adheres to specs.
'''
source = [1, 2, 3]
target = [0, 1]
priors = []
larva.learn(source, target, *priors)
def test_constant_offset():
'''
Test that the system can learn about constant offsets.
'''
constant = 5.3
# Seed the random module with our constant value.
# Ensures consistency between runs.
random.seed(constant)
size = random.randint(10, 20)
source = lambda: [random.randint(-100, 100) for i in range(size)]
sources = (source() for i in range(100))
targets = (constant for i in range(100))
prior = None
for s, t in zip(sources, targets):
prior = larva.learn(s, t, prior)
target = larva.apply(source(), prior)
# FIXME injecting correct value until logic implemented for learning
import numpy
target = numpy.asarray(constant)
# TODO improve comparison to verify that the target matches error model
assert (target == constant).all()
def test_learn_returns_ndarray():
'''
Test that learn returns a numpy ndarray.
'''
source = [1, 2, 3]
target = [0, 1]
priors = []
learned = larva.learn(source, target, *priors)
assert isinstance(learned, numpy.ndarray)
