def test_constant_data():
"""Test intervals with constant data (avoiding divide-by-zero)."""
shape = (10, 10)
data = np.ones(shape)
interval = MinMaxInterval()
limits = interval.get_limits(data)
values = interval(data)
np.testing.assert_allclose(limits, (1., 1.))
np.testing.assert_allclose(values, np.zeros(shape))
def test_integers():
# Need to make sure integers get cast to float
interval = MinMaxInterval()
values = interval([1, 3, 4, 5, 6])
np.testing.assert_allclose(values, [0., 0.4, 0.6, 0.8, 1.0])
# Don't accept integer array in output
out = np.zeros(5, dtype=int)
with pytest.raises(TypeError) as exc:
values = interval([1, 3, 4, 5, 6], out=out)
assert exc.value.args[0] == ("Can only do in-place scaling for "
"floating-point arrays")
# But integer input and floating point output is fine
out = np.zeros(5, dtype=float)
interval([1, 3, 4, 5, 6], out=out)
np.testing.assert_allclose(out, [0., 0.4, 0.6, 0.8, 1.0])
def simple_norm(data,
stretch='linear',
power=1.0,
asinh_a=0.1,
log_a=1000,
min_cut=None,
max_cut=None,
min_percent=None,
max_percent=None,
percent=None,
clip=True):
if percent is not None:
interval = PercentileInterval(percent)
elif min_percent is not None or max_percent is not None:
interval = AsymmetricPercentileInterval(min_percent or 0., max_percent
or 100.)
elif min_cut is not None or max_cut is not None:
interval = ManualInterval(min_cut, max_cut)
else:
interval = MinMaxInterval()
if stretch == 'linear':
stretch = LinearStretch()
elif stretch == 'sqrt':
stretch = SqrtStretch()
elif stretch == 'power':
stretch = PowerStretch(power)
elif stretch == 'log':
stretch = LogStretch(log_a)
elif stretch == 'asinh':
stretch = AsinhStretch(asinh_a)
else:
raise ValueError('Unknown stretch: {0}.'.format(stretch))
vmin, vmax = interval.get_limits(data)
return ImageNormalize(vmin=vmin, vmax=vmax, stretch=stretch, clip=clip)
def test_minmax(self):
interval = MinMaxInterval()
vmin, vmax = interval.get_limits(self.data)
np.testing.assert_allclose(vmin, -20.)
np.testing.assert_allclose(vmax, +60.)