def test_extra_coord_dims(self):
lat = clinspace(-0.25, 1.25, 7, name="lat")
lon = clinspace(-0.125, 1.125, 11, name="lon")
time = ["2012-05-19", "2016-01-31", "2018-06-20"]
coords = Coordinates([lat, lon, time], dims=["lat", "lon", "time"])
source = Array(source=np.random.random(coords.drop("time").shape),
coordinates=coords.drop("time"))
node = Convolution(source=source,
kernel=[[-1, 2, -1]],
kernel_dims=["lat", "lon"],
force_eval=True)
o = node.eval(coords)
assert np.all([d in ["lat", "lon"] for d in o.dims])
def test_missing_source_dims(self):
""" When the kernel has more dimensions than the source, sum out the kernel for the missing dim"""
lat = clinspace(-0.25, 1.25, 7, name="lat")
lon = clinspace(-0.125, 1.125, 11, name="lon")
time = ["2012-05-19", "2016-01-31", "2018-06-20"]
coords = Coordinates([lat, lon, time], dims=["lat", "lon", "time"])
coords2 = Coordinates([lat[[1, 2, 4]], lon, time],
dims=["lat", "lon", "time"])
source = Array(source=np.random.random(coords.drop("time").shape),
coordinates=coords.drop("time"))
node = Convolution(source=source,
kernel=[[[-1], [2], [-1]]],
kernel_dims=["lat", "lon", "time"],
force_eval=True)
o = node.eval(coords[:, 1:-1, :])
expected = source.source[:, 1:
-1] * 2 - source.source[:,
2:] - source.source[:, :
-2]
assert np.abs(o.data - expected).max() < 1e-14
# Check when request has an ArrayCoordinates1d
node = Convolution(source=source,
kernel_type="mean,3",
kernel_dims=["lat", "lon", "time"],
force_eval=True)
o = node.eval(coords2[:, 1:-1])
expected = (
source.source[[1, 2, 4], 1:-1] + source.source[[0, 1, 2], 1:-1] +
source.source[[2, 4, 6], 1:-1] + source.source[[1, 2, 4], :-2] +
source.source[[0, 1, 2], :-2] + source.source[[2, 4, 6], :-2] +
source.source[[1, 2, 4], 2:] + source.source[[0, 1, 2], 2:] +
source.source[[2, 4, 6], 2:]) / 9
assert np.abs(o.data - expected).max() < 1e-14
# Check to make sure array coordinates for a single coordinate is ok...
o = node.eval(coords2[0, 1:-1])