def test_select_remove_empty(self):
survey = surveys.Survey(
sources=surveys.txrx_coordinates_to_dict(emg3d.TxElectricDipole,
([0, 1, 2], 0, 0, 0, 0)),
receivers=surveys.txrx_coordinates_to_dict(
emg3d.RxElectricPoint, ([100, 101, 102], 0, 0, 0, 0)),
frequencies=np.array([1, 2, 3]),
)
selection = {
'sources': ['TxED-2', 'TxED-3'],
'receivers': ['RxEP-2', 'RxEP-3'],
'frequencies': ['f-2', 'f-3']
}
# Without 'observed', it should have no effect.
new = survey.select(**selection)
assert new.shape == (2, 2, 2)
new = survey.select(**selection, remove_empty=False)
assert new.shape == (2, 2, 2)
# Create and add 'observed' data.
data = np.arange(27.).reshape(3, 3, 3)
data = data + 1j * data
data[1:, 1:, 1:] = np.nan
data[2, 2, 2] = 26.0 + 26.0j
survey.data.observed[...] = data
# With observed, it should remove if True.
new = survey.select(**selection)
assert new.shape == (1, 1, 1)
new = survey.select(**selection, remove_empty=False)
assert new.shape == (2, 2, 2)
def test_add_noise(self):
offs = np.linspace(0, 10000, 21)
rec = surveys.txrx_coordinates_to_dict(
emg3d.electrodes.RxElectricPoint, (offs, 0, 0, 0, 0))
data = np.logspace(0, -20,
offs.size) + 1j * np.logspace(0, -20, offs.size)
survey = surveys.Survey(sources=emg3d.electrodes.TxElectricDipole(
(0, 0, 0, 0, 0)),
receivers=rec,
frequencies=1.0,
data=data,
relative_error=0.01,
noise_floor=1e-15)
# Defined cutting
survey.add_noise(min_offset=1000, min_amplitude=1e-19, add_to='test1')
# Ensure short offsets are NaN
assert np.all(np.isnan(survey.data.test1.data[:, :2, :]))
# Ensure low amplitudes are NaN
assert np.all(np.isnan(survey.data.test1.data[:, -1:, :]))
# Ensure no others are none
assert np.sum(np.isnan(survey.data.test1.data)) == 3
# No cutting
survey.add_noise(min_offset=0, min_amplitude=10e-50, add_to='test2')
assert np.sum(np.isnan(survey.data.test2.data)) == 0
# Defaults
survey.add_noise()
# Ensure low amplitudes are NaN
assert np.all(np.isnan(survey.data.observed.data[:, -5:, :]))
assert np.sum(np.isnan(survey.data.observed.data)) == 5
def test_txrx_coordinates_to_dict():
sources = surveys.txrx_coordinates_to_dict(emg3d.TxElectricDipole, ([
-1,
1,
], 0, 0, [-10, 10], 0),
strength=[100, 1])
assert sources['TxED-1'].strength == 100
assert sources['TxED-1'].azimuth == -10
assert_allclose(sources['TxED-1'].center, (-1, 0, 0))
assert sources['TxED-2'].strength == 1
assert sources['TxED-2'].azimuth == 10
assert_allclose(sources['TxED-2'].center, (1, 0, 0))
def test_txrx_lists_to_dict():
electric = [emg3d.RxElectricPoint((x, 0, 0, 0, 0)) for x in [1000, 1100]]
magnetic = surveys.txrx_coordinates_to_dict(
emg3d.RxMagneticPoint, ([950, 1050, 1150], 0, 0, 0, 90))
streamer = emg3d.RxElectricPoint((5, 0, 0, 0, 0), relative=True)
# If instance, it should give the instance in a dict.
rec1 = surveys.txrx_lists_to_dict(streamer)
assert streamer == rec1['RxEP-1']
# If dict, it should yield the same.
rec2 = surveys.txrx_lists_to_dict(magnetic)
assert magnetic['RxMP-1'] == rec2['RxMP-1']
rec3 = surveys.txrx_lists_to_dict([streamer, electric, magnetic])
assert rec3['RxEP-1'] == streamer
assert rec3['RxEP-2'] == electric[0]
assert rec3['RxEP-3'] == electric[1]
assert rec3['RxMP-4'] == magnetic['RxMP-1']
assert rec3['RxMP-5'] == magnetic['RxMP-2']
assert rec3['RxMP-6'] == magnetic['RxMP-3']
rec4 = surveys.txrx_lists_to_dict((streamer, tuple(electric), magnetic))
assert rec3 == rec4
class TestSurvey():
sources = surveys.txrx_coordinates_to_dict(
emg3d.TxElectricDipole,
(0, [1000, 2000, 3000, 4000, 5000], -950, 0, 0))
receivers = surveys.txrx_coordinates_to_dict(
emg3d.RxElectricPoint, ([1000, 2000, 3000, 4000], 2000, -1000, 0, 0))
frequencies = (1, 0.1, 2, 3)
shape = (len(sources), len(receivers), len(frequencies))
def test_defaults(self):
srvy = surveys.Survey(self.sources, self.receivers, self.frequencies)
assert isinstance(srvy.sources, dict)
assert srvy.sources['TxED-1'].center[0] == 0
assert srvy.count == 0
assert srvy.size == 80
assert srvy.shape == self.shape
assert 'Coordinates' in srvy._repr_html_()
assert 'Coordinates' in srvy.__repr__()
# Check defaults
assert_allclose(srvy.data.observed.data,
np.ones(self.shape) * (np.nan + 1j * np.nan))
assert srvy.noise_floor is None
assert srvy.relative_error is None
assert srvy.name is None
assert srvy.date is None
assert srvy.info is None
with pytest.raises(TypeError, match="Unexpected "):
surveys.Survey(self.sources,
self.receivers,
self.frequencies,
bla='a')
def test_basics(self):
data = np.arange(np.prod(self.shape)).reshape(self.shape)
srvy = surveys.Survey(
self.sources,
self.receivers,
self.frequencies,
data={'test': data},
relative_error=0.05,
noise_floor=1e-15,
name='MySurvey',
info='RainyDay',
date='today',
)
assert isinstance(srvy.sources, dict)
assert srvy.sources['TxED-1'].center[0] == 0
assert srvy.count == 0
assert srvy.size == 80
assert srvy.shape == self.shape
assert 'MySurvey' in srvy._repr_html_()
assert 'MySurvey' in srvy.__repr__()
assert 'RainyDay' in srvy._repr_html_()
assert 'RainyDay' in srvy.__repr__()
assert 'today' in srvy._repr_html_()
assert 'today' in srvy.__repr__()
assert_allclose(srvy.data.test.data, data)
assert_allclose(srvy.data.observed.data,
np.ones(self.shape) * (np.nan + 1j * np.nan))
assert srvy.noise_floor == 1e-15
assert srvy.relative_error == 0.05
assert srvy.name == 'MySurvey'
assert srvy.date == 'today'
assert srvy.info == 'RainyDay'
def test_no_receiver(self):
srvy = surveys.Survey(self.sources, None, self.frequencies)
assert isinstance(srvy.sources, dict)
assert srvy.sources['TxED-1'].center[0] == 0
assert srvy.receivers == {}
assert srvy.shape == (self.shape[0], 0, self.shape[2])
def test_standard_deviation(self):
srvy0 = surveys.Survey(self.sources,
self.receivers,
self.frequencies,
data=np.ones(self.shape))
assert srvy0.standard_deviation is None
srvy1 = surveys.Survey(self.sources,
self.receivers,
self.frequencies,
relative_error=0.5,
noise_floor=0.1,
data=np.ones(self.shape))
std = np.sqrt(
np.ones(self.shape) * 0.1**2 + np.ones(self.shape) * 0.5**2)
assert_allclose(srvy1.standard_deviation.data, std)
# Test f-dependent noise-floor and source-dependent rel. error.
nf = np.arange(1, 5)[None, None, :] * 1e-15 # No noise floor for f=1.0
re = np.arange(1, 6)[:, None, None] / 100 # No rel. error for Tx1
srvy2 = surveys.Survey(self.sources,
self.receivers,
self.frequencies,
relative_error=re,
noise_floor=nf,
data=np.ones(self.shape))
assert_allclose(srvy2.noise_floor, np.ones(srvy2.shape) * nf)
assert_allclose(srvy2.relative_error, np.ones(srvy2.shape) * re)
assert_allclose(srvy2.data._noise_floor[0, 0, :], np.squeeze(nf))
assert_allclose(srvy2.data._relative_error[:, 0, 0], np.squeeze(re))
# As data are ones, we can check standard_deviation without it.
std = np.sqrt(
np.ones(srvy2.shape) * nf**2 + np.ones(srvy2.shape) * re**2)
assert_allclose(srvy2.standard_deviation.data, std)
# Set the standard deviations
test_std = np.arange(1, srvy2.size + 1).reshape(srvy2.shape)
srvy2.standard_deviation = test_std
assert_allclose(srvy2.data._noise_floor[0, 0, :], np.squeeze(nf))
assert_allclose(srvy2.data._relative_error[:, 0, 0], np.squeeze(re))
assert_allclose(srvy2.standard_deviation.data, test_std)
srvy2.standard_deviation = None # Delete again
assert_allclose(srvy2.standard_deviation.data, std)
with pytest.raises(ValueError, match='All values of `standard_dev'):
srvy2.standard_deviation = np.zeros(srvy2.shape)
with pytest.raises(ValueError, match='All values of `noise_floor`'):
srvy2.noise_floor = 0.0
with pytest.raises(ValueError, match='All values of `relative_error'):
srvy2.relative_error = 0.0
with pytest.raises(ValueError, match='operands could not be '):
srvy2.noise_floor = np.ones(srvy2.shape)[:, :2, :]
with pytest.raises(ValueError, match='operands could not be '):
srvy2.relative_error = np.ones(srvy2.shape)[2:6, :, :]
def test_copy(self, tmpdir):
sources = emg3d.TxElectricDipole((0, 0, 0, 0, 0))
receivers = emg3d.RxElectricPoint((1000, 0, 0, 0, 0))
# This also checks to_dict()/from_dict().
srvy1 = surveys.Survey(sources, receivers, 1.0)
# Set observed and standard deviation.
srvy1.observed = [[[3 + 3j]]]
srvy1.standard_deviation = np.array([[[1.1]]])
srvy2 = srvy1.copy()
assert srvy1.sources == srvy2.sources
srvy3 = surveys.Survey(sources, receivers, 1.0, [[[3 + 3j]]])
srvy4 = srvy3.copy()
srvy4.standard_deviation = np.array([[[1.1]]])
assert srvy3.sources == srvy4.sources
# Also check to_file()/from_file().
srvy4.to_file(tmpdir + '/test.npz')
srvy5 = surveys.Survey.from_file(tmpdir + '/test.npz')
assert srvy4.name == srvy5.name
assert srvy4.sources == srvy5.sources
assert srvy4.receivers == srvy5.receivers
assert srvy4.frequencies == srvy5.frequencies
assert_allclose(srvy4.data.observed, srvy5.data.observed)
assert_allclose(srvy4.standard_deviation, srvy5.standard_deviation)
srvy7 = surveys.Survey.from_file(tmpdir + '/test.npz', verb=-1)
assert srvy5.name == srvy7[0].name
assert 'Data loaded from' in srvy7[1]
def test_select(self):
sources = [
emg3d.TxElectricDipole((x, 0, 0, 0, 0)) for x in [0, 50, 100]
]
receivers = [
emg3d.RxElectricPoint((x, 0, 0, 0, 0))
for x in [1000, 1100, 1200, 1300, 1400]
]
survey = surveys.Survey(
sources,
receivers,
frequencies=(1.0, 2.0, 3.4, 4.0),
data=np.arange(3 * 5 * 4).reshape((3, 5, 4)),
noise_floor=np.ones((3, 5, 4)),
relative_error=np.ones((3, 5, 4)),
name='Test',
)
t1 = survey.select('TxED-1', ['RxEP-1', 'RxEP-5'], 'f-1')
assert t1.shape == (1, 2, 1)
t2 = survey.select(frequencies=[], receivers='RxEP-1')
assert t2.shape == (3, 1, 0)
def test_select_remove_empty(self):
survey = surveys.Survey(
sources=surveys.txrx_coordinates_to_dict(emg3d.TxElectricDipole,
([0, 1, 2], 0, 0, 0, 0)),
receivers=surveys.txrx_coordinates_to_dict(
emg3d.RxElectricPoint, ([100, 101, 102], 0, 0, 0, 0)),
frequencies=np.array([1, 2, 3]),
)
selection = {
'sources': ['TxED-2', 'TxED-3'],
'receivers': ['RxEP-2', 'RxEP-3'],
'frequencies': ['f-2', 'f-3']
}
# Without 'observed', it should have no effect.
new = survey.select(**selection)
assert new.shape == (2, 2, 2)
new = survey.select(**selection, remove_empty=False)
assert new.shape == (2, 2, 2)
# Create and add 'observed' data.
data = np.arange(27.).reshape(3, 3, 3)
data = data + 1j * data
data[1:, 1:, 1:] = np.nan
data[2, 2, 2] = 26.0 + 26.0j
survey.data.observed[...] = data
# With observed, it should remove if True.
new = survey.select(**selection)
assert new.shape == (1, 1, 1)
new = survey.select(**selection, remove_empty=False)
assert new.shape == (2, 2, 2)
def test_src_rec_coordinates(self):
survey = surveys.Survey(
sources=[
emg3d.TxElectricDipole((0, 0, 0, 0, 0)),
emg3d.TxElectricDipole((100, 200, 300, 400, 500, 600))
],
receivers=[
emg3d.RxElectricPoint((1000, 0, 2000, 10, 0)),
emg3d.RxElectricPoint((1000, 0, 2000, 0, 0), relative=True),
emg3d.RxMagneticPoint((3000, 0, 2000, 0, 20)),
],
frequencies=1,
)
assert_allclose(survey.source_coordinates(),
[[0, 150], [0, 350], [0, 550]])
assert_allclose(survey.receiver_coordinates(),
[[1000, 1000, 1150, 3000], [0, 0, 350, 0],
[2000, 2000, 2550, 2000]])
assert_allclose(survey.receiver_coordinates('TxED-2'),
[[1000, 1150, 3000], [0, 350, 0], [2000, 2550, 2000]])
erec, _ = survey._irec_types
assert_allclose(erec, [0, 1])
assert_allclose(survey._imrec, [2])
ecoo, mcoo = survey._rec_types_coord('TxED-1')
assert_allclose(
ecoo,
([1000., 1000.], [0., 0.], [2000., 2000.], [10., 0.], [0., 0.]))
assert_allclose(mcoo, ([3000.], [0.], [2000.], [0.], [20.]))
def test_add_noise(self):
offs = np.linspace(0, 10000, 21)
rec = surveys.txrx_coordinates_to_dict(
emg3d.electrodes.RxElectricPoint, (offs, 0, 0, 0, 0))
data = np.logspace(0, -20,
offs.size) + 1j * np.logspace(0, -20, offs.size)
survey = surveys.Survey(sources=emg3d.electrodes.TxElectricDipole(
(0, 0, 0, 0, 0)),
receivers=rec,
frequencies=1.0,
data=data,
relative_error=0.01,
noise_floor=1e-15)
# Defined cutting
survey.add_noise(min_offset=1000, min_amplitude=1e-19, add_to='test1')
# Ensure short offsets are NaN
assert np.all(np.isnan(survey.data.test1.data[:, :2, :]))
# Ensure low amplitudes are NaN
assert np.all(np.isnan(survey.data.test1.data[:, -1:, :]))
# Ensure no others are none
assert np.sum(np.isnan(survey.data.test1.data)) == 3
# No cutting
survey.add_noise(min_offset=0, min_amplitude=10e-50, add_to='test2')
assert np.sum(np.isnan(survey.data.test2.data)) == 0
# Defaults
survey.add_noise()
# Ensure low amplitudes are NaN
assert np.all(np.isnan(survey.data.observed.data[:, -5:, :]))
assert np.sum(np.isnan(survey.data.observed.data)) == 5