def setup_method(self, method):
# Create three signals with dimensions:
# s1 : <BaseSignal, title: , dimensions: (4, 3, 2|2, 3)>
# s2 : <BaseSignal, title: , dimensions: (2, 3|4, 3, 2)>
# s12 : <BaseSignal, title: , dimensions: (2, 3|4, 3, 2)>
# Where s12 data is transposed in respect to s2
dc1 = np.random.random((2, 3, 4, 3, 2))
dc2 = np.rollaxis(np.rollaxis(dc1, -1), -1)
s1 = signals.BaseSignal(dc1.copy())
s2 = signals.BaseSignal(dc2)
s12 = signals.BaseSignal(dc1.copy())
for i, axis in enumerate(s1.axes_manager._axes):
if i < 3:
axis.navigate = True
else:
axis.navigate = False
for i, axis in enumerate(s2.axes_manager._axes):
if i < 2:
axis.navigate = True
else:
axis.navigate = False
for i, axis in enumerate(s12.axes_manager._axes):
if i < 3:
axis.navigate = False
else:
axis.navigate = True
self.s1 = s1
self.s2 = s2
self.s12 = s12
def test_error_axes():
rng = np.random.RandomState(123)
s = signals.BaseSignal(rng.random((20, 100)))
with pytest.raises(AttributeError,
match="not possible to decompose a dataset"):
s.decomposition()
def setup_method(self, method):
s = signals.BaseSignal(np.zeros(1))
self.factors = np.ones([2, 3])
self.loadings = np.ones([2, 3])
s.learning_results.factors = self.factors.copy()
s.learning_results.loadings = self.loadings.copy()
self.s = s
def _create_signal(shape, dim, dtype, metadata):
data = np.arange(np.product(shape)).reshape(shape).astype(dtype)
if dim == 1:
if len(shape) > 2:
s = signals.EELSSpectrum(data)
if metadata:
s.set_microscope_parameters(beam_energy=100.,
convergence_angle=1.,
collection_angle=10.)
else:
s = signals.EDSTEMSpectrum(data)
if metadata:
s.set_microscope_parameters(beam_energy=100.,
live_time=1.,
tilt_stage=2.,
azimuth_angle=3.,
elevation_angle=4.,
energy_resolution_MnKa=5.)
else:
s = signals.BaseSignal(data).transpose(signal_axes=dim)
if metadata:
s.metadata.General.date = "2016-08-06"
s.metadata.General.time = "10:55:00"
s.metadata.General.title = "Test title"
for i, axis in enumerate(s.axes_manager._axes):
i += 1
axis.offset = i * 0.5
axis.scale = i * 100
axis.name = "%i" % i
if axis.navigate:
axis.units = "m"
else:
axis.units = "eV"
return s
def _read_signal_from_group(self, name, group, load_to_memory=True):
self._log.debug('Calling _read_signal_from_group')
from hyperspy import signals
# Extract essential data:
data = group.get('data')
if load_to_memory:
data = np.asanyarray(data)
# EMD does not have a standard way to describe the signal axis.
# Therefore we return a BaseSignal
signal = signals.BaseSignal(data)
# Set signal properties:
signal.set_signal_origin = group.attrs.get('signal_origin', '')
signal.set_signal_type = group.attrs.get('signal_type', '')
# Iterate over all dimensions:
for i in range(len(data.shape)):
dim = group.get('dim{}'.format(i + 1))
axis = signal.axes_manager._axes[i]
axis.name = dim.attrs.get('name', '')
units = re.findall('[^_\W]+', dim.attrs.get('units', ''))
axis.units = ''.join(units)
try:
axis.scale = dim[1] - dim[0]
axis.offset = dim[0]
# Hyperspy then uses defaults (1.0 and 0.0)!
except (IndexError, TypeError) as e:
self._log.warning(
'Could not calculate scale/offset of axis {}: {}'.format(
i, e))
# Extract metadata:
metadata = {}
for key, value in group.attrs.items():
metadata[key] = value
# Add signal:
self.add_signal(signal, name, metadata)
def setup_method(self, method):
self.s = signals.BaseSignal(np.random.rand(1, 2, 3, 4, 5, 6))
for ax, name in zip(self.s.axes_manager._axes, 'abcdef'):
ax.name = name
# just to make sure in case default changes
self.s.axes_manager.set_signal_dimension(6)
self.s.estimate_poissonian_noise_variance()
def setUp(self):
s = signals.BaseSignal(np.zeros(1))
self.factors = np.ones([2, 3])
self.loadings = np.ones([2, 3])
s.learning_results.factors = self.factors.copy()
s.learning_results.loadings = self.loadings.copy()
s.learning_results.bss_factors = self.factors.copy()
s.learning_results.bss_loadings = self.loadings.copy()
self.s = s
def test_ragged_slicing(slice_):
array_ragged = np.empty((2, 3), dtype=object)
for iy, ix in np.ndindex(array_ragged.shape):
array_ragged[iy, ix] = np.random.randint(0,
20,
size=np.random.randint(2, 10))
s = signals.BaseSignal(array_ragged, ragged=True)
s_lazy = s.as_lazy()
np.testing.assert_allclose(s.inav[slice_].data[0],
s_lazy.inav[slice_].data[0])
def test_broadcast_in_place(self):
s1 = self.s1
s1.axes_manager.set_signal_dimension(1) # (3|2)
s2 = signals.BaseSignal(np.ones((4, 2, 4, 3)))
s2c = s2
s2.axes_manager.set_signal_dimension(2) # (3, 4| 2, 4)
print(s2)
print(s1)
s2 += s1
assert_array_equal(s2.data, 2 * np.ones((4, 2, 4, 3)))
nt.assert_is(s2, s2c)
def test_broadcast_in_place(self):
s1 = self.s1
s1 = s1.transpose(signal_axes=1)
s2 = signals.BaseSignal(np.ones((4, 2, 4, 3)))
s2 = s2.transpose(signal_axes=2)
s2c = s2
print(s2)
print(s1)
s2 += s1
assert_array_equal(s2.data, 2 * np.ones((4, 2, 4, 3)))
nt.assert_is(s2, s2c)
def _read_signal_from_group(self, name, group, lazy=False):
self._log.debug('Calling _read_signal_from_group')
from hyperspy import signals
# Extract essential data:
data = group.get('data')
if lazy:
data = da.from_array(data, chunks=data.chunks)
else:
data = np.asanyarray(data)
# EMD does not have a standard way to describe the signal axis.
# Therefore we return a BaseSignal
signal = signals.BaseSignal(data)
# Set signal properties:
signal.set_signal_origin = group.attrs.get('signal_origin', '')
signal.set_signal_type = group.attrs.get('signal_type', '')
# Iterate over all dimensions:
for i in range(len(data.shape)):
dim = group.get('dim{}'.format(i + 1))
axis = signal.axes_manager._axes[i]
axis_name = dim.attrs.get('name', '')
if isinstance(axis_name, bytes):
axis_name = axis_name.decode('utf-8')
axis.name = axis_name
axis_units = dim.attrs.get('units', '')
if isinstance(axis_units, bytes):
axis_units = axis_units.decode('utf-8')
units = re.findall(r'[^_\W]+', axis_units)
axis.units = ''.join(units)
try:
if len(dim) == 1:
axis.scale = 1.
self._log.warning(
'Could not calculate scale of axis {}. '
'Setting scale to 1'.format(i))
else:
axis.scale = dim[1] - dim[0]
axis.offset = dim[0]
# HyperSpy then uses defaults (1.0 and 0.0)!
except (IndexError, TypeError) as e:
self._log.warning(
'Could not calculate scale/offset of '
'axis {}: {}'.format(i, e))
# Extract metadata:
metadata = {}
for key, value in group.attrs.items():
metadata[key] = value
if signal.data.dtype == np.object:
self._log.warning('HyperSpy could not load the data in {}, '
'skipping it'.format(name))
else:
# Add signal:
self.add_signal(signal, name, metadata)
def test_broadcast_in_place(self):
if self.s1._lazy:
pytest.skip("Inplace not supported by LazySignals")
s1 = self.s1
s1 = s1.transpose(signal_axes=1)
s2 = signals.BaseSignal(np.ones((4, 2, 4, 3)))
s2 = s2.transpose(signal_axes=2)
s2c = s2
print(s2)
print(s1)
s2 += s1
assert_array_equal(s2.data, 2 * np.ones((4, 2, 4, 3)))
assert s2 is s2c
def setup_method(self, method):
s = signals.BaseSignal(np.empty(1))
s.learning_results.explained_variance_ratio = np.asarray([
10e-1,
5e-2,
9e-3,
1e-3,
9e-5,
5e-5,
3.0e-5,
2.2e-5,
1.9e-5,
1.8e-5,
1.7e-5,
1.6e-5,
])
self.s = s
def setup_method(self, method):
self.signal = signals.BaseSignal(np.arange(10))
self.signal.axes_manager.set_signal_dimension(1)
self.signal.axes_manager[0].scale = 0.5
self.signal.axes_manager[0].offset = 0.25
self.data = self.signal.data.copy()
def setup_method(self, method):
self.signal = signals.BaseSignal(np.arange(24).reshape((2, 3, 4)))
self.data = self.signal.data.copy()
self.signal.axes_manager._axes[0].navigate = False
self.signal.axes_manager._axes[1].navigate = True
self.signal.axes_manager._axes[2].navigate = False
def denoised_data_to_signal(self):
signal = signals.BaseSignal(self.Y)
if self.signal_type == "spectrum":
return signal.as_signal1D(2)
if self.signal_type == "image":
return signal.as_signal2D((1,2))
def setUp(self):
s = signals.BaseSignal(np.empty(1))
self.s = s
def setup_method(self, method):
s = signals.BaseSignal(np.empty(1))
self.s = s
def setUp(self):
self.signal = signals.BaseSignal(np.arange(24).reshape((2, 3, 4)))
self.data = self.signal.data.copy()
self.signal.axes_manager.set_signal_dimension(0)
def test_error_axes():
s = signals.BaseSignal(generate_low_rank_matrix())
with pytest.raises(AttributeError, match="not possible to decompose a dataset"):
s.decomposition()
def test_lazy_to_device():
s = signals.BaseSignal(np.arange(10)).as_lazy()
with pytest.raises(BaseException):
s.to_device()
def setup_method(self, method):
self.signal = signals.BaseSignal(np.arange(24).reshape((2, 3, 4))).T
self.data = self.signal.data.copy()
def setup_method(self, method):
self.signal = signals.BaseSignal(
np.arange(2**5).reshape((2, 2, 2, 2, 2)))
self.signal.axes_manager.set_signal_dimension(1)
self.data = self.signal.data.copy()
def test_squeeze(shape):
s = signals.BaseSignal(np.random.random(shape))
s2 = s.transpose(signal_axes=[0, 1, 2])
s3 = s2.squeeze()
assert s2.data.squeeze().shape == s3.data.shape
assert s3.axes_manager.shape == (8, 6, 2, 7)
def test_to_host():
data = np.arange(10)
s = signals.BaseSignal(data)
s.to_host()
s.data is data
def estimate_parameters(self,
signal,
x1,
x2,
only_current=False,
out=False):
"""Estimate the parameters by the two area method
Parameters
----------
signal : Signal1D instance
x1 : float
Defines the left limit of the spectral range to use for the
estimation.
x2 : float
Defines the right limit of the spectral range to use for the
estimation.
only_current : bool
If False, estimates the parameters for the full dataset.
out : bool
If True, returns the result arrays directly without storing in the
parameter maps/values. The returned order is (A, r).
Returns
-------
{bool, tuple of values}
"""
super(PowerLaw, self)._estimate_parameters(signal)
axis = signal.axes_manager.signal_axes[0]
i1, i2 = axis.value_range_to_indices(x1, x2)
if not (i2 + i1) % 2 == 0:
i2 -= 1
if i2 == i1:
i2 += 2
i3 = (i2 + i1) // 2
x1 = axis.index2value(i1)
x2 = axis.index2value(i2)
x3 = axis.index2value(i3)
if only_current is True:
s = signal.get_current_signal()
else:
s = signal
if s._lazy:
import dask.array as da
log = da.log
I1 = s.isig[i1:i3].integrate1D(2j).data
I2 = s.isig[i3:i2].integrate1D(2j).data
else:
shape = s.data.shape[:-1]
I1_s = signals.BaseSignal(np.empty(shape, dtype='float'))
I2_s = signals.BaseSignal(np.empty(shape, dtype='float'))
# Use the `out` parameters to avoid doing the deepcopy
s.isig[i1:i3].integrate1D(2j, out=I1_s)
s.isig[i3:i2].integrate1D(2j, out=I2_s)
I1 = I1_s.data
I2 = I2_s.data
log = np.log
with np.errstate(divide='raise'):
try:
r = 2 * log(I1 / I2) / log(x2 / x1)
k = 1 - r
A = k * I2 / (x2**k - x3**k)
if s._lazy:
r = r.map_blocks(np.nan_to_num)
A = A.map_blocks(np.nan_to_num)
else:
r = np.nan_to_num(r)
A = np.nan_to_num(A)
except (RuntimeWarning, FloatingPointError):
_logger.warning('Power law paramaters estimation failed '
'because of a "divide by zero" error.')
return False
if only_current is True:
self.r.value = r
self.A.value = A
return True
if out:
return A, r
else:
if self.A.map is None:
self._create_arrays()
self.A.map['values'][:] = A
self.A.map['is_set'][:] = True
self.r.map['values'][:] = r
self.r.map['is_set'][:] = True
self.fetch_stored_values()
return True
