def test_plot_knee():
"""Test that plotting is functional"""
with np.errstate(divide="ignore"):
x = np.linspace(0.0, 1, 10)
y = np.true_divide(-1, x + 0.1) + 5
kl = KneeLocator(x, y, S=1.0, curve_nature="concave")
num_figures_before = plt.gcf().number
kl.plot_knee()
num_figures_after = plt.gcf().number
assert num_figures_before < num_figures_after
def test_convex_decreasing():
"""Tests that a correct knee point is detected in
curve having convex and decreasing nature."""
kn = KneeLocator(x,
y_convex_dec,
curve_nature="convex",
curve_direction="decreasing")
assert kn.knee == 2
def test_concave_increasing_truncated():
"""Tests that a correct knee point is detected in
curve having truncated concave increasing nature"""
kn = KneeLocator(
x[:-3] / 10,
y_concave_inc[:-3] / 10,
curve_nature="concave",
curve_direction="increasing",
)
assert kn.knee == 0.2
def test_properties():
"""Tests that elbow and knee can be used interchangeably."""
kn = KneeLocator(x,
y_concave_inc,
curve_nature="concave",
curve_direction="increasing")
assert kn.knee == kn.elbow
assert kn.norm_knee == kn.norm_elbow
# pytest compares all elements in each list.
assert kn.all_knees == kn.all_elbows
assert kn.all_norm_knees == kn.all_norm_elbows
def test_gamma_online_offline(online, expected):
"""Tests online and offline knee detection.
Notable that a large number of samples are highly sensitive to S parameter
"""
np.random.seed(23)
n = 1000
x = range(1, n + 1)
y = sorted(np.random.gamma(0.5, 1.0, n), reverse=True)
kl = KneeLocator(x,
y,
curve_nature="convex",
curve_direction="decreasing",
online=online)
assert kl.knee == expected
def test_y():
"""Test the y value"""
with np.errstate(divide="ignore"):
x = np.linspace(0.0, 1, 10)
y = np.true_divide(-1, x + 0.1) + 5
kl = KneeLocator(x, y, S=1.0, curve_nature="concave")
assert kl.knee_y == pytest.approx(1.897, 0.03)
assert kl.all_knees_y[0] == pytest.approx(1.897, 0.03)
assert kl.norm_knee_y == pytest.approx(0.758, 0.03)
assert kl.all_norm_knees_y[0] == pytest.approx(0.758, 0.03)
assert kl.elbow_y == pytest.approx(1.897, 0.03)
assert kl.all_elbows_y[0] == pytest.approx(1.897, 0.03)
assert kl.norm_elbow_y == pytest.approx(0.758, 0.03)
assert kl.all_norm_elbows_y[0] == pytest.approx(0.758, 0.03)
def test_x_equals_y():
"""Test that a runtime warning is raised when no maxima are found"""
x = range(10)
y = [1] * len(x)
with pytest.warns(RuntimeWarning):
KneeLocator(x, y)
