def test_add_and_remove_independent_variable():
mg = ModelGetter()
m = mg.model
# Create an independent variable
independent_variable = IndependentVariable("time", 1.0, u.s)
# Try to add it
m.add_independent_variable(independent_variable)
# Try to remove it
m.remove_independent_variable("time")
with pytest.raises(AssertionError):
m.add_independent_variable(Parameter("time", 1.0))
# Try to add it twice, which shouldn't fail
m.add_independent_variable(independent_variable)
m.add_independent_variable(independent_variable)
# Try to display it just to make sure it works
m.display()
# Now try to use it
link_law = Powerlaw()
link_law.K.value = 1.0
link_law.index.value = -1.0
n_free_before_link = len(m.free_parameters)
m.link(m.one.spectrum.main.Powerlaw.K, independent_variable, link_law)
# The power law adds two parameters, but the link removes one, so
assert len(m.free_parameters) - 1 == n_free_before_link
# Now see if it works
for t in np.linspace(0, 10, 100):
independent_variable.value = t
assert m.one.spectrum.main.Powerlaw.K.value == link_law(t)
def test_time_domain_integration():
po = Powerlaw()
default_powerlaw = Powerlaw()
src = PointSource("test", ra=0.0, dec=0.0, spectral_shape=po)
m = Model(src) # type: model.Model
# Add time independent variable
time = IndependentVariable("time", 0.0, u.s)
m.add_independent_variable(time)
# Now link one of the parameters with a simple line law
line = Line()
line.a = 0.0
m.link(po.index, time, line)
# Test the display just to make sure it doesn't crash
m.display()
# Now test the average with the integral
energies = np.linspace(1, 10, 10)
results = m.get_point_source_fluxes(0, energies,
tag=(time, 0, 10)) # type: np.ndarray
assert np.all(results == 1.0)
# Now test the linking of the normalization, first with a constant then with a line with a certain
# angular coefficient
m.unlink(po.index)
po.index.value = default_powerlaw.index.value
line2 = Line()
line2.a = 0.0
line2.b = 1.0
m.link(po.K, time, line2)
time.value = 1.0
results = m.get_point_source_fluxes(0, energies, tag=(time, 0, 10))
assert np.allclose(results, default_powerlaw(energies))
# Now make something actually happen
line2.a = 1.0
line2.b = 1.0
results = m.get_point_source_fluxes(0, energies,
tag=(time, 0, 10)) # type: np.ndarray
# Compare with analytical result
def F(x):
return line2.a.value / 2.0 * x**2 + line2.b.value * x
effective_norm = (F(10) - F(0)) / 10.0
expected_results = default_powerlaw(
energies) * effective_norm # type: np.ndarray
assert np.allclose(expected_results, results)