def test_scale_and_offset_raw_value_iterable_for_set_cache(values,
offsets,
scales):
p = Parameter(name='test_scale_and_offset_raw_value', set_cmd=None)
# test that scale and offset does not change the default behaviour
p.cache.set(values)
assert p.raw_value == values
# test setting scale and offset does not change anything
p.scale = scales
p.offset = offsets
assert p.raw_value == values
np_values = np.array(values)
np_offsets = np.array(offsets)
np_scales = np.array(scales)
np_get_latest_values = np.array(p.get_latest())
# Without a call to ``get``, ``get_latest`` will just return old
# cached values without applying the set scale and offset
np.testing.assert_allclose(np_get_latest_values, np_values)
np_get_values = np.array(p.get())
# Now that ``get`` is called, the returned values are the result of
# application of the scale and offset. Obviously, calling
# ``get_latest`` now will also return the values with the applied
# scale and offset
np.testing.assert_allclose(np_get_values,
(np_values - np_offsets) / np_scales)
np_get_latest_values_after_get = np.array(p.get_latest())
np.testing.assert_allclose(np_get_latest_values_after_get,
(np_values - np_offsets) / np_scales)
# test ``cache.set`` for scalar values
if not isinstance(values, Iterable):
p.cache.set(values)
np.testing.assert_allclose(np.array(p.raw_value),
np_values * np_scales + np_offsets)
# No set/get cmd performed
# testing conversion back and forth
p.cache.set(values)
np_get_latest_values = np.array(p.get_latest())
# No set/get cmd performed
np.testing.assert_allclose(np_get_latest_values, np_values)
# adding statistics
if isinstance(offsets, Iterable):
event('Offset is array')
if isinstance(scales, Iterable):
event('Scale is array')
if isinstance(values, Iterable):
event('Value is array')
if isinstance(scales, Iterable) and isinstance(offsets, Iterable):
event('Scale is array and also offset')
if isinstance(scales, Iterable) and not isinstance(offsets, Iterable):
event('Scale is array but not offset')
def test_setting_numpy_array_valued_param_if_scale_and_offset_are_not_none():
param = Parameter(name='test_param', set_cmd=None, get_cmd=None)
values = np.array([1, 2, 3, 4, 5])
param.scale = 100
param.offset = 10
param(values)
assert isinstance(param.raw_value, np.ndarray)
def test_numpy_array_valued_parameter_preserves_type_if_scale_and_offset_are_set(
):
def rands():
return np.random.randn(5)
param = Parameter(name='test_param', set_cmd=None, get_cmd=rands)
param.scale = 10
param.offset = 7
values = param()
assert isinstance(values, np.ndarray)
def test_scale_and_offset_raw_value_iterable(self, values, offsets,
scales):
p = Parameter(name='test_scale_and_offset_raw_value', set_cmd=None)
# test that scale and offset does not change the default behaviour
p(values)
self.assertEqual(p.raw_value, values)
# test setting scale and offset does not change anything
p.scale = scales
p.offset = offsets
self.assertEqual(p.raw_value, values)
np_values = np.array(values)
np_offsets = np.array(offsets)
np_scales = np.array(scales)
np_get_values = np.array(p())
np.testing.assert_allclose(np_get_values, (np_values - np_offsets) /
np_scales) # No set/get cmd performed
# test set, only for scalar values
if not isinstance(values, Iterable):
p(values)
np.testing.assert_allclose(np.array(p.raw_value),
np_values * np_scales +
np_offsets) # No set/get cmd performed
# testing conversion back and forth
p(values)
np_get_values = np.array(p())
np.testing.assert_allclose(np_get_values,
np_values) # No set/get cmd performed
# adding statistics
if isinstance(offsets, Iterable):
event('Offset is array')
if isinstance(scales, Iterable):
event('Scale is array')
if isinstance(values, Iterable):
event('Value is array')
if isinstance(scales, Iterable) and isinstance(offsets, Iterable):
event('Scale is array and also offset')
if isinstance(scales, Iterable) and not isinstance(offsets, Iterable):
event('Scale is array but not offset')
