def test_filter_mutable_dimension(f):
# Check that changing the order of dimensions doesn't change
# the outcome up to numeric precision.
xr_assert_allclose(
f(dims=('y', 'x')).apply(ds),
f(dims=('x', 'y')).apply(ds)
)
def test_apply_with_vars_keep_vars():
ds = generate_test_dataset()
ref = ds.mean('time')
result = utils.apply(ds,
lambda a: a.mean(axis=0),
signature='(time,var)->(var)')
xr_assert_allclose(result.transpose(*ref.nd.dims),
ref.transpose(*ref.nd.dims))
def test_apply_with_vars():
ds = generate_test_dataset()
ref = ds.to_array(dim='var').mean('var')
result = utils.apply(ds,
lambda a: a.mean(axis=1),
signature='(time,var)->(time)')
xr_assert_allclose(result.transpose(*ref.nd.dims),
ref.transpose(*ref.nd.dims))
def test_apply(dims):
ds = generate_test_dataset()
ref = ds.mean(dims)
result = utils.apply(ds,
np.mean,
signature='({})->()'.format(",".join(dims)))
xr_assert_allclose(result.transpose(*ref.nd.dims),
ref.transpose(*ref.nd.dims))
def test_parallel_merge():
dims = dict(x=50, y=50, time=50)
ds = generate_test_dataset(dims=dims)
def _fn(ds):
return (ds - ds.mean('time')) / ds.std('time')
result = _fn(ds)
result_1 = utils.parallel(_fn, 'x')(ds)
result_2 = utils.xr_merge(utils.parallel(_fn, 'x', merge=False)(ds),
dim='x')
xr_assert_allclose(result, result_1)
xr_assert_allclose(result_1, result_2)
def test_generate_covariance_matrix():
ds = generate_test_dataset(var=['VV__im', 'VV__re', 'VH__im', 'VH__re'])
assemble_complex(ds, inplace=True)
cov = generate_covariance_matrix(ds)
xr_assert_allclose(np.abs(ds.VV)**2, cov.C11)
xr_assert_allclose(np.abs(ds.VH)**2, cov.C22)
xr_assert_allclose(ds.VV * np.conj(ds.VH), cov.C12)
def test_parallelized_filter(f, dims):
xr_assert_allclose(
f(dims=dims).apply(ds),
f(dims=dims).apply(ds, njobs=2)
)