def test_argmin_max(dim_num, dtype, contains_nan, dask, func, skipna, aggdim):
# pandas-dev/pandas#16830, we do not check consistency with pandas but
# just make sure da[da.argmin()] == da.min()
if aggdim == 'y' and dim_num < 2:
pytest.skip('dim not in this test')
if dask and not has_dask:
pytest.skip('requires dask')
if contains_nan:
if not skipna:
pytest.skip("numpy's argmin (not nanargmin) does not handle "
"object-dtype")
if skipna and np.dtype(dtype).kind in 'iufc':
pytest.skip("numpy's nanargmin raises ValueError for all nan axis")
da = construct_dataarray(dim_num, dtype, contains_nan=contains_nan,
dask=dask)
with warnings.catch_warnings():
warnings.filterwarnings('ignore', 'All-NaN slice')
if aggdim == 'y' and contains_nan and skipna:
with pytest.raises(ValueError):
actual = da.isel(**{
aggdim: getattr(da, 'arg' + func)(
dim=aggdim, skipna=skipna).compute()})
return
actual = da.isel(**{aggdim: getattr(da, 'arg' + func)
(dim=aggdim, skipna=skipna).compute()})
expected = getattr(da, func)(dim=aggdim, skipna=skipna)
assert_allclose(actual.drop(actual.coords),
expected.drop(expected.coords))
def test_multiple_dims(dtype, dask, func):
if dask and not has_dask:
pytest.skip('requires dask')
da = construct_dataarray(3, dtype, contains_nan=True, dask=dask)
actual = getattr(da, func)(('x', 'y'))
expected = getattr(getattr(da, func)('x'), func)('y')
assert_allclose(actual, expected)
def test_notransform():
# regression test for https://github.com/pydata/xarray/issues/1686
# Create a geotiff file
with warnings.catch_warnings():
# rasterio throws a NotGeoreferencedWarning here, which is
# expected since we test rasterio's defaults in this case.
warnings.filterwarnings(
"ignore",
category=UserWarning,
message="Dataset has no geotransform set",
)
with create_tmp_file(suffix=".tif") as tmp_file:
# data
nx, ny, nz = 4, 3, 3
data = np.arange(nx * ny * nz, dtype=rasterio.float32).reshape(nz, ny, nx)
with rasterio.open(
tmp_file,
"w",
driver="GTiff",
height=ny,
width=nx,
count=nz,
dtype=rasterio.float32,
) as s:
s.descriptions = ("nx", "ny", "nz")
s.units = ("cm", "m", "km")
s.write(data)
# Tests
expected = DataArray(
data,
dims=("band", "y", "x"),
coords={
"band": [1, 2, 3],
"y": [0.5, 1.5, 2.5],
"x": [0.5, 1.5, 2.5, 3.5],
},
)
expected.coords[DEFAULT_GRID_MAP] = xr.Variable((), 0)
expected.coords[DEFAULT_GRID_MAP].attrs[
"GeoTransform"
] = "0.0 1.0 0.0 0.0 0.0 1.0"
with rioxarray.open_rasterio(tmp_file) as rioda:
assert_allclose(rioda, expected)
assert rioda.attrs["scale_factor"] == 1.0
assert rioda.attrs["add_offset"] == 0.0
assert rioda.attrs["long_name"] == ("nx", "ny", "nz")
assert rioda.attrs["units"] == ("cm", "m", "km")
assert isinstance(rioda.rio.resolution(), tuple)
assert isinstance(rioda.rio._cached_transform(), Affine)
def test_no_mftime(self):
# rasterio can accept "filename" urguments that are actually urls,
# including paths to remote files.
# In issue #1816, we found that these caused dask to break, because
# the modification time was used to determine the dask token. This
# tests ensure we can still chunk such files when reading with
# rasterio.
with create_tmp_geotiff(
8, 10, 3, transform_args=[1, 2, 0.5, 2.0], crs="+proj=latlong"
) as (tmp_file, expected):
with mock.patch("os.path.getmtime", side_effect=OSError):
with xr.open_rasterio(tmp_file, chunks=(1, 2, 2)) as actual:
assert isinstance(actual.data, da.Array)
assert_allclose(actual, expected)
def test_transpose():
# test on DataArray
X_da = xr.DataArray(
np.random.random((100, 10)),
coords={'sample': range(100), 'feature': range(10)},
dims=('sample', 'feature')
)
Xt_da, estimator = transpose(
X_da, order=['feature', 'sample'], return_estimator=True)
xrt.assert_allclose(Xt_da, X_da.transpose())
Xt_da = estimator.inverse_transform(Xt_da)
xrt.assert_allclose(Xt_da, X_da)
# test on Dataset with subset of dimensions
X_ds = xr.Dataset(
{'var_1': (['sample', 'feat_1', 'feat_2'],
np.random.random((100, 10, 5))),
'var_2': (['feat_2', 'sample'],
np.random.random((5, 100)))},
coords={'sample': range(100), 'feat_1': range(10), 'feat_2': range(5)}
)
Xt_ds, estimator = transpose(
X_ds, order=['sample', 'feat_2'], return_estimator=True)
xrt.assert_allclose(Xt_ds, X_ds.transpose('sample', 'feat_1', 'feat_2'))
Xt_ds = estimator.inverse_transform(Xt_ds)
xrt.assert_allclose(Xt_ds, X_ds)
def test_split():
# test on DataArray with number of samples multiple of new length
X_da = xr.DataArray(
np.random.random((100, 10)),
coords={
"sample": range(100),
"feature": range(10),
"coord_1": (["sample", "feature"], np.tile("Test", (100, 10))),
},
dims=("sample", "feature"),
)
estimator = Splitter(
new_dim="split_sample",
new_len=5,
reduce_index="subsample",
axis=1,
keep_coords_as="sample_coord",
)
Xt_da = estimator.fit_transform(X_da)
assert Xt_da.shape == (20, 5, 10)
npt.assert_allclose(Xt_da[0, :, 0], X_da[:5, 0])
Xit_da = estimator.inverse_transform(Xt_da)
xrt.assert_allclose(X_da, Xit_da)
# test on Dataset with number of samples NOT multiple of new length
X_ds = xr.Dataset(
{"var_1": (["sample", "feature"], np.random.random((100, 10)))},
coords={
"sample": range(100),
"feature": range(10)
},
)
Xt_ds = split(
X_ds,
new_dim="split_sample",
new_len=7,
reduce_index="head",
axis=1,
new_index_func=None,
)
assert Xt_ds["var_1"].shape == (14, 7, 10)
npt.assert_allclose(Xt_ds.var_1[0, :, 0], X_ds.var_1[:7, 0])
def test_reduce():
X_da = xr.DataArray(
np.random.random((100, 10)),
coords={
"sample": range(100),
"feature": range(10)
},
dims=("sample", "feature"),
)
Xt_da = reduce(X_da)
xrt.assert_allclose(Xt_da, X_da.reduce(np.linalg.norm, dim="feature"))
def test_caching(self):
with create_tmp_geotiff(
8, 10, 3, transform_args=[1, 2, 0.5, 2.0], crs="+proj=latlong"
) as (tmp_file, expected):
# Cache is the default
with xr.open_rasterio(tmp_file) as actual:
# This should cache everything
assert_allclose(actual, expected)
# once cached, non-windowed indexing should become possible
ac = actual.isel(x=[2, 4])
ex = expected.isel(x=[2, 4])
assert_allclose(ac, ex)
def test_min_count(dim_num, dtype, dask, func, aggdim, contains_nan, skipna):
if dask and not has_dask:
pytest.skip("requires dask")
da = construct_dataarray(dim_num, dtype, contains_nan=contains_nan, dask=dask)
min_count = 3
# If using Dask, the function call should be lazy.
with raise_if_dask_computes():
actual = getattr(da, func)(dim=aggdim, skipna=skipna, min_count=min_count)
expected = series_reduce(da, func, skipna=skipna, dim=aggdim, min_count=min_count)
assert_allclose(actual, expected)
assert_dask_array(actual, dask)
def test_notransform(self):
# regression test for https://github.com/pydata/xarray/issues/1686
# Create a geotiff file
with warnings.catch_warnings():
# rasterio throws a NotGeoreferencedWarning here, which is
# expected since we test rasterio's defaults in this case.
warnings.filterwarnings(
"ignore",
category=UserWarning,
message="Dataset has no geotransform set",
)
with create_tmp_file(suffix=".tif") as tmp_file:
# data
nx, ny, nz = 4, 3, 3
data = np.arange(nx * ny * nz, dtype=rasterio.float32).reshape(
nz, ny, nx
)
with rasterio.open(
tmp_file,
"w",
driver="GTiff",
height=ny,
width=nx,
count=nz,
dtype=rasterio.float32,
) as s:
s.descriptions = ("nx", "ny", "nz")
s.units = ("cm", "m", "km")
s.write(data)
# Tests
expected = DataArray(
data,
dims=("band", "y", "x"),
coords={
"band": [1, 2, 3],
"y": [0.5, 1.5, 2.5],
"x": [0.5, 1.5, 2.5, 3.5],
},
)
with xr.open_rasterio(tmp_file) as rioda:
assert_allclose(rioda, expected)
assert rioda.attrs["scales"] == (1.0, 1.0, 1.0)
assert rioda.attrs["offsets"] == (0.0, 0.0, 0.0)
assert rioda.attrs["descriptions"] == ("nx", "ny", "nz")
assert rioda.attrs["units"] == ("cm", "m", "km")
assert isinstance(rioda.attrs["res"], tuple)
assert isinstance(rioda.attrs["is_tiled"], np.uint8)
assert isinstance(rioda.attrs["transform"], tuple)
assert len(rioda.attrs["transform"]) == 6
def test_wrapped_transformer():
from sklearn.preprocessing import StandardScaler
X = xr.DataArray(
np.random.random((100, 10)),
coords={'sample': range(100), 'feature': range(10)},
dims=['sample', 'feature']
)
estimator = wrap(StandardScaler()).partial_fit(X)
estimator.partial_fit(X)
assert_allclose(X, estimator.inverse_transform(estimator.transform(X)))
def test_ENVI_tags():
# Create an ENVI file with some tags in the ENVI namespace
# this test uses a custom driver, so we can't use create_tmp_geotiff
with create_tmp_file(suffix=".dat") as tmp_file:
# data
nx, ny, nz = 4, 3, 3
data = np.arange(nx * ny * nz, dtype=rasterio.float32).reshape(nz, ny, nx)
transform = from_origin(5000, 80000, 1000, 2000.0)
with rasterio.open(
tmp_file,
"w",
driver="ENVI",
height=ny,
width=nx,
count=nz,
crs="EPSG:32618",
transform=transform,
dtype=rasterio.float32,
) as s:
s.update_tags(
ns="ENVI",
description="{Tagged file}",
wavelength="{123.000000, 234.234000, 345.345678}",
fwhm="{1.000000, 0.234000, 0.000345}",
)
s.write(data)
dx, dy = s.res[0], -s.res[1]
crs_wkt = s.crs.to_wkt()
# Tests
coords = {
"band": [1, 2, 3],
"y": -np.arange(ny) * 2000 + 80000 + dy / 2,
"x": np.arange(nx) * 1000 + 5000 + dx / 2,
"wavelength": ("band", np.array([123, 234.234, 345.345678])),
"fwhm": ("band", np.array([1, 0.234, 0.000345])),
}
expected = DataArray(data, dims=("band", "y", "x"), coords=coords)
expected.coords[DEFAULT_GRID_MAP] = xr.Variable((), 0)
expected.coords[DEFAULT_GRID_MAP].attrs["crs_wkt"] = crs_wkt
with rioxarray.open_rasterio(tmp_file) as rioda:
assert_allclose(rioda, expected)
assert rioda.rio.crs == crs_wkt
assert isinstance(rioda.rio._cached_transform(), Affine)
# from ENVI tags
assert isinstance(rioda.attrs["description"], str)
assert isinstance(rioda.attrs["map_info"], str)
assert isinstance(rioda.attrs["samples"], str)
def test_chunks_with_mask_and_scale():
with create_tmp_geotiff(
10, 10, 4, transform_args=[1, 2, 0.5, 2.0], crs="EPSG:4326"
) as (tmp_file, expected):
# Chunk at open time
with rioxarray.open_rasterio(
tmp_file, mask_and_scale=True, chunks=(1, 2, 2)
) as actual:
assert isinstance(actual.data, dask.array.Array)
assert "open_rasterio" in actual.data.name
# do some arithmetic
ac = actual.mean().compute()
ex = expected.mean()
assert_allclose(ac, ex)
def test_min_count(dim_num, dtype, dask, func, aggdim):
if dask and not has_dask:
pytest.skip("requires dask")
da = construct_dataarray(dim_num, dtype, contains_nan=True, dask=dask)
min_count = 3
actual = getattr(da, func)(dim=aggdim, skipna=True, min_count=min_count)
expected = series_reduce(da,
func,
skipna=True,
dim=aggdim,
min_count=min_count)
assert_allclose(actual, expected)
assert_dask_array(actual, dask)
def test_min_count(dim_num, dtype, dask, func, aggdim):
if dask and not has_dask:
pytest.skip("requires dask")
da = construct_dataarray(dim_num, dtype, contains_nan=True, dask=dask)
min_count = 3
actual = getattr(da, func)(dim=aggdim, skipna=True, min_count=min_count)
if LooseVersion(pd.__version__) >= LooseVersion("0.22.0"):
# min_count is only implenented in pandas > 0.22
expected = series_reduce(da, func, skipna=True, dim=aggdim, min_count=min_count)
assert_allclose(actual, expected)
assert_dask_array(actual, dask)
def test_multi_loc(self):
data_type = '3-hour'
years = [1995]
lat_lon = ((44.624076,-124.280097),(43.489171,-125.152137))
parameters = 'mean_absolute_period'
wave_multiloc, meta= wave.io.hindcast.request_wpto_point_data(data_type,
parameters,lat_lon,years)
dir_multiyear, meta_dir = wave.io.hindcast.request_wpto_directional_spectrum(lat_lon,year='1995')
dir_multiyear = dir_multiyear.sel(time_index=slice(dir_multiyear.time_index[0],dir_multiyear.time_index[99]))
dir_multiyear = dir_multiyear.rename_vars({87:'87',58:'58'})
assert_frame_equal(self.ml,wave_multiloc)
assert_frame_equal(self.ml_meta,meta)
xrt.assert_allclose(self.my_dir,dir_multiyear)
assert_frame_equal(self.my_dir_meta,meta_dir)
def test_min_count_nd(dtype, dask, func):
if dask and not has_dask:
pytest.skip("requires dask")
min_count = 3
dim_num = 3
da = construct_dataarray(dim_num, dtype, contains_nan=True, dask=dask)
actual = getattr(da, func)(dim=["x", "y", "z"],
skipna=True,
min_count=min_count)
# Supplying all dims is equivalent to supplying `...` or `None`
expected = getattr(da, func)(dim=..., skipna=True, min_count=min_count)
assert_allclose(actual, expected)
assert_dask_array(actual, dask)
def test_compute_perfect_model_different_dims_quite_close(
perfectModelEnsemble_initialized_control, ):
"""Tests nearly equal dim=['init','member'] and dim='member'."""
stack_dims_true = perfectModelEnsemble_initialized_control.verify(
comparison="m2c",
metric="rmse",
dim=["init", "member"],
)["tos"]
stack_dims_false = perfectModelEnsemble_initialized_control.verify(
comparison="m2c",
metric="rmse",
dim="member",
).mean(["init"])["tos"]
# no more than 10% difference
assert_allclose(stack_dims_true, stack_dims_false, rtol=0.1, atol=0.03)
def test_calculate_rel_vorticity():
datadict = datasets()
coords = datadict["coords"]
ds_b = datadict["B"]
grid_b = Grid(ds_b, coords=coords)
ds_c = datadict["C"]
grid_c = Grid(ds_c, coords=coords)
test_b = (grid_b.diff(grid_b.interp(ds_b.v * ds_b.dy_ne, "Y"), "X") -
grid_b.diff(grid_b.interp(ds_b.u * ds_b.dx_ne, "X"),
"Y")) / ds_b.area_t
zeta_b = calculate_rel_vorticity(
grid_b,
ds_b.u,
ds_b.v,
ds_b.dx_ne,
ds_b.dy_ne,
ds_b.area_t,
gridtype=None,
)
test_c = (grid_c.diff(ds_c.v * ds_c.dy_n, "X") -
grid_c.diff(ds_c.u * ds_c.dx_e, "Y")) / ds_c.area_ne
zeta_c = calculate_rel_vorticity(
grid_c,
ds_c.u,
ds_c.v,
ds_c.dx_e,
ds_c.dy_n,
ds_c.area_ne,
gridtype=None,
)
assert_allclose(test_b, zeta_b)
assert_allclose(test_c, zeta_c)
with pytest.raises(RuntimeError):
zeta_c = calculate_rel_vorticity(
grid_b,
ds_c.u,
ds_c.v,
ds_c.dx_n, # wrong coordinate
ds_c.dy_n,
ds_c.area_ne,
gridtype=None,
)
def test_custom_metric_passed_to_bootstrap(
perfectModelEnsemble_initialized_control):
"""Test custom metric in PerfectModelEnsemble.bootstrap."""
comparison = "e2c"
np.random.seed(42)
kwargs = dict(comparison=comparison,
iterations=ITERATIONS,
dim="init",
resample_dim="init")
actual = perfectModelEnsemble_initialized_control.bootstrap(metric=my_mse,
**kwargs)
expected = perfectModelEnsemble_initialized_control.bootstrap(metric="mse",
**kwargs)
assert_allclose(actual, expected, rtol=0.1, atol=1)
def test_min_count(dim_num, dtype, dask, func, aggdim):
if dask and not has_dask:
pytest.skip('requires dask')
da = construct_dataarray(dim_num, dtype, contains_nan=True, dask=dask)
min_count = 3
actual = getattr(da, func)(dim=aggdim, skipna=True, min_count=min_count)
if LooseVersion(pd.__version__) >= LooseVersion('0.22.0'):
# min_count is only implenented in pandas > 0.22
expected = series_reduce(da, func, skipna=True, dim=aggdim,
min_count=min_count)
assert_allclose(actual, expected)
assert_dask_array(actual, dask)
def test_stats_functions_dask_chunks(PM_da_control_3d, func, step):
"""Check whether selected stats functions be chunked and computed along other
dim."""
dim = "time"
for chunk_dim in PM_da_control_3d.isel({dim: 0}).dims:
control_chunked = PM_da_control_3d.chunk({chunk_dim: step})
res_chunked = func(control_chunked, dim=dim)
res = func(PM_da_control_3d, dim=dim)
# check for chunks
assert dask.is_dask_collection(res_chunked)
assert res_chunked.chunks is not None
# check for no chunks
assert not dask.is_dask_collection(res)
assert res.chunks is None
# check for identical result
assert_allclose(res, res_chunked.compute())
def test_add_bounds(obj, dims):
expected = {}
expected["lat"] = xr.concat(
[
obj.lat.copy(data=np.arange(76.25, 16.0, -2.5)),
obj.lat.copy(data=np.arange(73.75, 13.6, -2.5)),
],
dim="bounds",
)
expected["lon"] = xr.concat(
[
obj.lon.copy(data=np.arange(198.75, 325 - 1.25, 2.5)),
obj.lon.copy(data=np.arange(201.25, 325 + 1.25, 2.5)),
],
dim="bounds",
)
t0 = pd.Timestamp("2013-01-01")
t1 = pd.Timestamp("2013-01-01 18:00")
dt = "6h"
dtb2 = pd.Timedelta("3h")
expected["time"] = xr.concat(
[
obj.time.copy(
data=pd.date_range(start=t0 - dtb2, end=t1 - dtb2, freq=dt)),
obj.time.copy(
data=pd.date_range(start=t0 + dtb2, end=t1 + dtb2, freq=dt)),
],
dim="bounds",
)
expected["lat"].attrs.clear()
expected["lon"].attrs.clear()
expected["time"].attrs.clear()
added = obj.cf.add_bounds(dims)
if isinstance(dims, str):
dims = (dims, )
for dim in dims:
name = f"{dim}_bounds"
assert name in added.coords
assert added[dim].attrs["bounds"] == name
assert_allclose(added[name].reset_coords(drop=True), expected[dim])
# Test multiple dimensions
assert not {"x1_bounds", "x2_bounds"} <= set(multiple.variables)
assert {"x1_bounds", "x2_bounds"} <= set(
multiple.cf.add_bounds("X").variables)
def test_utm():
with create_tmp_geotiff() as (tmp_file, expected):
with rioxarray.open_rasterio(tmp_file) as rioda:
assert_allclose(rioda, expected)
assert rioda.attrs["scale_factor"] == 1.0
assert rioda.attrs["add_offset"] == 0.0
assert rioda.attrs["long_name"] == ("d1", "d2", "d3")
assert rioda.attrs["units"] == ("u1", "u2", "u3")
assert rioda.rio.crs == expected.rio.crs
assert_array_equal(rioda.rio.resolution(), expected.rio.resolution())
assert isinstance(rioda.rio._cached_transform(), Affine)
assert rioda.rio.nodata is None
# Check no parse coords
with rioxarray.open_rasterio(tmp_file, parse_coordinates=False) as rioda:
assert "x" not in rioda.coords
assert "y" not in rioda.coords
def test_chunks(self):
with create_tmp_geotiff(
8, 10, 3, transform_args=[1, 2, 0.5, 2.0], crs="+proj=latlong"
) as (tmp_file, expected):
# Chunk at open time
with xr.open_rasterio(tmp_file, chunks=(1, 2, 2)) as actual:
assert isinstance(actual.data, da.Array)
assert "open_rasterio" in actual.data.name
# do some arithmetic
ac = actual.mean()
ex = expected.mean()
assert_allclose(ac, ex)
ac = actual.sel(band=1).mean(dim="x")
ex = expected.sel(band=1).mean(dim="x")
assert_allclose(ac, ex)
def test_true_phase():
"""Test if true phase"""
f0 = 2.0
T = 4.0
dx = 0.02
x = np.arange(-8 * T, 5 * T + dx, dx) # uncentered and odd number of points
y = np.cos(2 * np.pi * f0 * x)
y[np.abs(x) >= (T / 2.0)] = 0.0
s = xr.DataArray(y, dims=("x",), coords={"x": x})
lag = x[len(x) // 2]
f = np.fft.fftfreq(len(x), dx)
expected = np.fft.fft(np.fft.ifftshift(y)) * np.exp(-1j * 2.0 * np.pi * f * lag)
expected = xr.DataArray(expected, dims="freq_x", coords={"freq_x": f})
output = xrft.dft(
s, dim="x", true_phase=True, true_amplitude=False, shift=False, prefix="freq_"
)
xrt.assert_allclose(expected, output)
def test_stats_functions_dask_single_chunk(PM_da_control_3d, func):
"""Test stats functions when single chunk not along dim."""
step = -1 # single chunk
for chunk_dim in PM_da_control_3d.dims:
control_chunked = PM_da_control_3d.chunk({chunk_dim: step})
for dim in PM_da_control_3d.dims:
if dim != chunk_dim:
res_chunked = func(control_chunked, dim=dim)
res = func(PM_da_control_3d, dim=dim)
# check for chunks
assert dask.is_dask_collection(res_chunked)
assert res_chunked.chunks is not None
# check for no chunks
assert not dask.is_dask_collection(res)
assert res.chunks is None
# check for identical result
assert_allclose(res, res_chunked.compute())
def test_split():
# test on DataArray with number of samples multiple of new length
X_da = xr.DataArray(np.random.random((100, 10)),
coords={
'sample':
range(100),
'feature':
range(10),
'coord_1':
(['sample', 'feature'], np.tile('Test', (100, 10)))
},
dims=('sample', 'feature'))
estimator = Splitter(new_dim='split_sample',
new_len=5,
reduce_index='subsample',
axis=1,
keep_coords_as='sample_coord')
Xt_da = estimator.fit_transform(X_da)
assert Xt_da.shape == (20, 5, 10)
npt.assert_allclose(Xt_da[0, :, 0], X_da[:5, 0])
Xit_da = estimator.inverse_transform(Xt_da)
xrt.assert_allclose(X_da, Xit_da)
# test on Dataset with number of samples NOT multiple of new length
X_ds = xr.Dataset(
{'var_1': (['sample', 'feature'], np.random.random((100, 10)))},
coords={
'sample': range(100),
'feature': range(10)
})
Xt_ds = split(X_ds,
new_dim='split_sample',
new_len=7,
reduce_index='head',
axis=1,
new_index_func=None)
assert Xt_ds['var_1'].shape == (14, 7, 10)
npt.assert_allclose(Xt_ds.var_1[0, :, 0], X_ds.var_1[:7, 0])
def test_compute_perfect_model_stack_dims_True_and_False_quite_close(
pm_da_ds1d, pm_da_control1d):
"""Test whether dim=['init','member'] for stack_dims=False and
dim='member' for stack_dims=True give similar results."""
stack_dims_true = compute_perfect_model(
pm_da_ds1d,
pm_da_control1d,
comparison='m2c',
metric='rmse',
dim=['init', 'member'],
)
stack_dims_false = compute_perfect_model(pm_da_ds1d,
pm_da_control1d,
comparison='m2c',
metric='rmse',
dim='member').mean(['init'])
# no more than 10% difference
assert_allclose(stack_dims_true, stack_dims_false, rtol=0.1, atol=0.03)
def test_detrend_2D(array_dims, array_shape, chunks, detrend_type,
trend_amplitude):
da_original = noise(array_dims, array_shape)
da_trend = (da_original + trend_amplitude["x"] * da_original["x"] +
trend_amplitude["y"] * da_original["y"])
if chunks:
da_trend = da_trend.chunk(chunks)
detrend_dim = ["y", "x"]
detrended = detrend(da_trend, detrend_dim, detrend_type=detrend_type)
assert detrended.chunks == da_trend.chunks
if detrend_type is None:
xrt.assert_equal(detrended, da_trend)
elif detrend_type == "constant":
xrt.assert_allclose(detrended,
da_trend - da_trend.mean(dim=detrend_dim))
elif detrend_type == "linear":
xrt.assert_allclose(detrended, da_original)
def test_w_mean(axis, metric_list, gridtype):
fail_metric_list = ["fail"]
ds = datasets()[gridtype]
grid = Grid(ds)
for var in ds.data_vars:
metric = _find_metric(ds[var], metric_list)
dim = _find_dim(grid, ds[var], axis)
a = w_mean(grid, ds[var], axis, metric_list, verbose=True)
if dim is None: # no dimension found, return the input arrays
b = ds[var]
else:
b = weighted_mean(ds[var], ds[metric], dim=dim)
assert_allclose(a, b)
# original array should be returned if a non matching metric list
# is supplied
a_fail = w_mean(grid, ds[var], axis, fail_metric_list)
assert_allclose(a_fail, ds[var])
def test_custom_metric_passed_to_compute(PM_da_initialized_1d,
PM_da_control_1d, comparison):
"""Test custom metric in compute_perfect_model."""
actual = compute_perfect_model(
PM_da_initialized_1d,
PM_da_control_1d,
comparison=comparison,
metric=my_mse,
)
expected = compute_perfect_model(
PM_da_initialized_1d,
PM_da_control_1d,
comparison=comparison,
metric='mse',
)
assert_allclose(actual, expected)
def test_stats_functions_dask_many_chunks(PM_da_control_3d, func):
"""Check whether selected stats functions be chunked in multiple chunks and
computed along other dim."""
step = 1
for chunk_dim in PM_da_control_3d.dims:
control_chunked = PM_da_control_3d.chunk({chunk_dim: step})
for dim in PM_da_control_3d.dims:
if dim != chunk_dim and dim in control_chunked.dims:
res_chunked = func(control_chunked, dim=dim)
res = func(PM_da_control_3d, dim=dim)
# check for chunks
assert dask.is_dask_collection(res_chunked)
assert res_chunked.chunks is not None
# check for no chunks
assert not dask.is_dask_collection(res)
assert res.chunks is None
# check for identical result
assert_allclose(res, res_chunked.compute())
def assertDatasetAllClose(self, d1, d2, rtol=1e-05, atol=1e-08,
decode_bytes=True):
__tracebackhide__ = True # noqa: F841
assert_allclose(d1, d2, rtol=rtol, atol=atol,
decode_bytes=decode_bytes)
def assertDataArrayAllClose(self, ar1, ar2, rtol=1e-05, atol=1e-08):
assert_allclose(ar1, ar2, rtol=rtol, atol=atol)
def assertDatasetAllClose(self, d1, d2, rtol=1e-05, atol=1e-08):
assert_allclose(d1, d2, rtol=rtol, atol=atol)
def assertVariableAllClose(self, v1, v2, rtol=1e-05, atol=1e-08):
assert_allclose(v1, v2, rtol=rtol, atol=atol)
def test_min_count_dataset(func):
da = construct_dataarray(2, dtype=float, contains_nan=True, dask=False)
ds = Dataset({'var1': da}, coords={'scalar': 0})
actual = getattr(ds, func)(dim='x', skipna=True, min_count=3)['var1']
expected = getattr(ds['var1'], func)(dim='x', skipna=True, min_count=3)
assert_allclose(actual, expected)
def test_reduce(dim_num, dtype, dask, func, skipna, aggdim):
if aggdim == 'y' and dim_num < 2:
pytest.skip('dim not in this test')
if dtype == np.bool_ and func == 'mean':
pytest.skip('numpy does not support this')
if dask and not has_dask:
pytest.skip('requires dask')
rtol = 1e-04 if dtype == np.float32 else 1e-05
da = construct_dataarray(dim_num, dtype, contains_nan=True, dask=dask)
axis = None if aggdim is None else da.get_axis_num(aggdim)
# TODO: remove these after resolving
# https://github.com/dask/dask/issues/3245
with warnings.catch_warnings():
warnings.filterwarnings('ignore', 'All-NaN slice')
warnings.filterwarnings('ignore', 'invalid value encountered in')
if (LooseVersion(np.__version__) >= LooseVersion('1.13.0') and
da.dtype.kind == 'O' and skipna):
# Numpy < 1.13 does not handle object-type array.
try:
if skipna:
expected = getattr(np, 'nan{}'.format(func))(da.values,
axis=axis)
else:
expected = getattr(np, func)(da.values, axis=axis)
actual = getattr(da, func)(skipna=skipna, dim=aggdim)
assert np.allclose(actual.values, np.array(expected),
rtol=1.0e-4, equal_nan=True)
except (TypeError, AttributeError, ZeroDivisionError):
# TODO currently, numpy does not support some methods such as
# nanmean for object dtype
pass
# make sure the compatiblility with pandas' results.
actual = getattr(da, func)(skipna=skipna, dim=aggdim)
if func == 'var':
expected = series_reduce(da, func, skipna=skipna, dim=aggdim,
ddof=0)
assert_allclose(actual, expected, rtol=rtol)
# also check ddof!=0 case
actual = getattr(da, func)(skipna=skipna, dim=aggdim, ddof=5)
expected = series_reduce(da, func, skipna=skipna, dim=aggdim,
ddof=5)
assert_allclose(actual, expected, rtol=rtol)
else:
expected = series_reduce(da, func, skipna=skipna, dim=aggdim)
assert_allclose(actual, expected, rtol=rtol)
# make sure the dtype argument
if func not in ['max', 'min']:
actual = getattr(da, func)(skipna=skipna, dim=aggdim, dtype=float)
assert actual.dtype == float
# without nan
da = construct_dataarray(dim_num, dtype, contains_nan=False, dask=dask)
actual = getattr(da, func)(skipna=skipna)
expected = getattr(np, 'nan{}'.format(func))(da.values)
if actual.dtype == object:
assert actual.values == np.array(expected)
else:
assert np.allclose(actual.values, np.array(expected), rtol=rtol)
def assertVariableAllClose(self, v1, v2, rtol=1e-05, atol=1e-08):
__tracebackhide__ = True # noqa: F841
assert_allclose(v1, v2, rtol=rtol, atol=atol)
def test_generate_axis():
a = generate_axis(ds_original, 'X', 'lon', 'lon',
pos_from='center',
pos_to='right',
pad=None,
boundary_discontinuity=360)
b = generate_axis(ds_original, 'Y', 'lat', 'lat',
pos_from='center',
pos_to='left',
pad=None,
boundary_discontinuity=180)
c = generate_axis(ds_original, 'Z', 'z', 'z',
pos_from='center',
pos_to='left',
pad='auto')
d = generate_axis(ds_original_1D, 'Z', 'z', 'z',
pos_from='left',
pos_to='center',
pad=1.0+dz,
new_name='test')
e = generate_axis(ds_original_left, 'Z', 'z', 'z',
pos_from='left',
pos_to='center',
pad='auto')
assert_allclose(a['lon_right'], ds_out_right['lon_right'])
assert_allclose(b['lat_left'], ds_out_left['lat_left'])
assert_allclose(c['z_left'], ds_out_left['z_left'])
assert_allclose(d['test'], ds_original_1D_padded['test'])
assert_allclose(e['z_center'], ds_out_center['z_center'])
# Mulitdim cases
aa = generate_axis(a, 'X', 'llon', 'lon',
pos_from='center',
pos_to='right',
pad=None,
boundary_discontinuity=360,
attrs_from_scratch=False)
bb = generate_axis(b, 'Y', 'llat', 'lat',
pos_from='center',
pos_to='left',
pad=None,
boundary_discontinuity=180,
attrs_from_scratch=False)
ee = generate_axis(e, 'Z', 'zz', 'z',
pos_from='left',
pos_to='center',
pad='auto',
attrs_from_scratch=False)
assert_allclose(aa['llon_right'], ds_out_right['llon_right'])
assert_allclose(bb['llat_left'], ds_out_left['llat_left'])
assert_allclose(ee['zz_center'], ds_out_center['zz_center'])
with pytest.raises(RuntimeError):
# Check if generate axis fails when a DataArray is passed instead of
# Dataset
generate_axis(c['somedata'], 'Z', 'zz', 'z',
pos_from='left',
pos_to='center',
pad='auto',
attrs_from_scratch=False)
with pytest.raises(RuntimeError):
generate_axis(c, 'Z', 'zz', 'z', pad='auto',
boundary_discontinuity=360)
with pytest.raises(RuntimeError):
generate_axis(c, 'Z', 'zz', 'z', pad=None,
boundary_discontinuity=None)
def assertDataArrayAllClose(self, ar1, ar2, rtol=1e-05, atol=1e-08,
decode_bytes=True):
__tracebackhide__ = True # noqa: F841
assert_allclose(ar1, ar2, rtol=rtol, atol=atol,
decode_bytes=decode_bytes)
