def test_correlation_with_anxcor_data_scheme(self):
tau=None
starttime = UTCDateTime("2017-10-01 06:00:00").timestamp
starttime_utc = UTCDateTime("2017-10-01 06:00:00")
endtime_utc = UTCDateTime("2017-10-01 06:10:00")
anxcor_main = build_anxcor(tau)
result = anxcor_main.process([starttime])
source_np_array = convert_xarray_to_np_array(result,variable='src:BB rec:Nodal')
target_np_array = get_obspy_correlation(starttime_utc,endtime_utc)
source_stream = numpy_2_stream(source_np_array)
target_stream = numpy_2_stream(target_np_array)
source_stream.normalize()
target_stream.normalize()
np.assert_allclose(source_stream[0].data, target_stream[0].data, atol=1e-5)
def test_matrix_norms(self):
# Not all of these are matrix norms in the most technical sense.
np.random.seed(1234)
for n, m in (1, 1), (1, 3), (3, 1), (4, 4), (4, 5), (5, 4):
for t in np.single, np.double, np.csingle, np.cdouble, np.int64:
A = 10 * np.random.randn(n, m).astype(t)
if np.issubdtype(A.dtype, np.complexfloating):
A = (A + 10j * np.random.randn(n, m)).astype(t)
t_high = np.cdouble
else:
t_high = np.double
for order in (None, 'fro', 1, -1, 2, -2, np.inf, -np.inf):
actual = norm(A, ord=order)
desired = np.linalg.norm(A, ord=order)
# SciPy may return higher precision matrix norms.
# This is a consequence of using LAPACK.
if not np.allclose(actual, desired):
desired = np.linalg.norm(A.astype(t_high), ord=order)
np.assert_allclose(actual, desired)
def test_spherical_shell_three_dim_adaptive_discret(): # pylint: disable=too-many-locals
"Compare numerical result with analytical solution for 3D adaptive discretization"
# Define computation point located on the equator at 1km above mean Earth radius
ellipsoid = get_ellipsoid()
radius = ellipsoid.mean_radius + 1e3
coordinates = [0, 0, radius]
# Define shape of spherical shell model made of tesseroids
shape = (6, 6)
# Define a density for the shell
density = 1000
# Define different values for the spherical shell thickness
thicknesses = np.logspace(1, 5, 5)
for thickness in thicknesses:
# Create list of tesseroids for the spherical shell model
tesseroids = []
# Define boundary coordinates of each tesseroid
top = ellipsoid.mean_radius
bottom = top - thickness
longitude = np.linspace(0, 360, shape[0] + 1)
latitude = np.linspace(-90, 90, shape[1] + 1)
west, east = longitude[:-1], longitude[1:]
south, north = latitude[:-1], latitude[1:]
for w, e in zip(west, east):
for s, n in zip(south, north):
tesseroids.append([w, e, s, n, bottom, top])
# Compute gravitational fields of the spherical shell
numerical = {"potential": 0, "g_r": 0}
for tesseroid in tesseroids:
for field in numerical:
numerical[field] += tesseroid_gravity(
coordinates,
tesseroid,
density,
field=field,
radial_adaptive_discretization=True,
)
# Get analytical solutions
analytical = spherical_shell_analytical(top, bottom, density, radius)
# Assert analytical and numerical solution are bellow the accuracy threshold
for field in numerical:
np.assert_allclose(
analytical[field], numerical[field], rtol=accuracy_threshold
)
def test_matrix_norms(self):
# Not all of these are matrix norms in the most technical sense.
np.random.seed(1234)
for n, m in (1, 1), (1, 3), (3, 1), (4, 4), (4, 5), (5, 4):
for t in np.single, np.double, np.csingle, np.cdouble, np.int64:
A = 10 * np.random.randn(n, m).astype(t)
if np.issubdtype(A.dtype, np.complexfloating):
A = (A + 10j * np.random.randn(n, m)).astype(t)
t_high = np.cdouble
else:
t_high = np.double
for order in (None, 'fro', 1, -1, 2, -2, np.inf, -np.inf):
actual = norm(A, ord=order)
desired = np.linalg.norm(A, ord=order)
# SciPy may return higher precision matrix norms.
# This is a consequence of using LAPACK.
if not np.allclose(actual, desired):
desired = np.linalg.norm(A.astype(t_high), ord=order)
np.assert_allclose(actual, desired)