def test_linint2pts_non_monotonic_np(self):
import warnings
with self.assertWarns(Warning):
linint2pts(self._fi_np[:, :, ::-1, :],
self._xo,
self._yo,
0,
xi=self._xi,
yi=self._yi_reverse)
def test_linint2pts_non_monotonic_xr(self):
fi = xr.DataArray(self._fi_np[:, :, ::-1, :],
dims=['time', 'level', 'lat', 'lon'],
coords={
'lat': self._yi_reverse,
'lon': self._xi
}).chunk(self._chunks)
import warnings
with self.assertWarns(Warning):
linint2pts(fi, self._xo, self._yo, 0).compute()
def test_linint2pts_fi_np(self):
fo = linint2pts(self._fi_np,
self._xo,
self._yo,
0,
xi=self._xi,
yi=self._yi)
self.assertEqual((self._shape0, self._shape1), fo.shape[:-1])
self.assertEqual(np.float64, fo.dtype)
newshape = (self._shape0 * self._shape1 * self._no, )
fo_vals = fo.values.reshape(newshape).tolist()
print(
self._fi_np.reshape(
(self._shape0 * self._shape1 * self._ni * self._ni, )))
print(fo_vals)
print(self._xo)
# Use numpy.testing.assert_almost_equal() instead of ut.TestCase.assertAlmostEqual() because the former can
# handle NaNs but the latter cannot.
# Compare the function-generated fo array to NCL ground-truth up to 5 decimal points
np.testing.assert_almost_equal(self._ncl_truth, fo_vals, decimal=5)
def test_linint2pts_msg_99(self):
# fi_msg_99 = xr.DataArray(self._fi_np_msg_nan, dims=['time', 'level', 'lat', 'lon'], coords={'lat': self._yi, 'lon': self._xi}).chunk(self._chunks)
fi_msg_99 = xr.DataArray(self._fi_np_msg_99,
dims=['time', 'level', 'lat', 'lon'],
coords={
'lat': self._yi,
'lon': self._xi
}).chunk(self._chunks)
fo = linint2pts(fi_msg_99,
self._xo,
self._yo,
0,
msg_py=self._fi_np_msg_99[0, 0, 0, 0])
self.assertEqual((self._shape0, self._shape1), fo.shape[:-1])
self.assertEqual(np.float64, fo.dtype)
newshape = (self._shape0 * self._shape1 * self._no, )
fo_vals = fo.values.reshape(newshape).tolist()
# Use numpy.testing.assert_almost_equal() instead of ut.TestCase.assertAlmostEqual() because the former can
# handle NaNs but the latter cannot.
# Compare the function-generated fo array to NCL ground-truth up to 5 decimal points
np.testing.assert_almost_equal(self.groundtruth_msg_99,
fo_vals,
decimal=5)
def test_linint2pts_chunked_interp(self):
# use 1 for interpolated dimension chunk sizes -- this should throw a ChunkError
wrong_chunks = {'time': 1, 'level': 1, 'lat': 1, 'lon': 1}
fi = xr.DataArray(self._fi_np,
dims=['time', 'level', 'lat', 'lon'],
coords={
'lat': self._yi,
'lon': self._xi
}).chunk(wrong_chunks)
with self.assertRaises(ChunkError):
fo = linint2pts(fi, self._xo, self._yo, 0)
def test_linint2pts_chunked_leftmost(self):
# use 1 for time and level chunk sizes
custom_chunks = {
'time': 1,
'level': 1,
'lat': self._fi_np.shape[2],
'lon': self._fi_np.shape[3]
}
fi = xr.DataArray(self._fi_np,
dims=['time', 'level', 'lat', 'lon'],
coords={
'lat': self._yi,
'lon': self._xi
}).chunk(custom_chunks)
fo = linint2pts(fi, self._xo, self._yo, 0)
newshape = (self._shape0 * self._shape1 * self._no, )
fo_vals = fo.values.reshape(newshape).tolist()
# Use numpy.testing.assert_almost_equal() instead of ut.TestCase.assertAlmostEqual() because the former can
# handle NaNs but the latter cannot.
# Compare the function-generated fo array to NCL ground-truth up to 5 decimal points
np.testing.assert_almost_equal(self._ncl_truth, fo_vals, decimal=5)
def test_linint2pts_fi_np_no_xi(self):
with self.assertRaises(CoordinateError):
fo = linint2pts(self._fi_np, self._xo, self._yo, 0, yi=self._yi)