def data_focal_stats():
data = np.arange(16).reshape(4, 4)
cellsize = (1, 1)
kernel = circle_kernel(*cellsize, 1.5)
expected_result = np.asarray([
# mean
[[1.66666667, 2., 3., 4.], [4.25, 5., 6., 6.75],
[8.25, 9., 10., 10.75], [11., 12., 13., 13.33333333]],
# max
[[4., 5., 6., 7.], [8., 9., 10., 11.], [12., 13., 14., 15.],
[13., 14., 15., 15.]],
# min
[[0., 0., 1., 2.], [0., 1., 2., 3.], [4., 5., 6., 7.],
[8., 9., 10., 11.]],
# range
[[4., 5., 5., 5.], [8., 8., 8., 8.], [8., 8., 8., 8.],
[5., 5., 5., 4.]],
# std
[[1.69967317, 1.87082869, 1.87082869, 2.1602469],
[2.86138079, 2.60768096, 2.60768096, 2.86138079],
[2.86138079, 2.60768096, 2.60768096, 2.86138079],
[2.1602469, 1.87082869, 1.87082869, 1.69967317]],
# var
[[2.88888889, 3.5, 3.5, 4.66666667], [8.1875, 6.8, 6.8, 8.1875],
[8.1875, 6.8, 6.8, 8.1875], [4.66666667, 3.5, 3.5, 2.88888889]],
# sum
[[5., 8., 12., 12.], [17., 25., 30., 27.], [33., 45., 50., 43.],
[33., 48., 52., 40.]]
])
return data, kernel, expected_result
def test_kernel(kernel_circle_1_1_1, kernel_annulus_2_2_2_1):
kernel_circle = circle_kernel(1, 1, 1)
assert isinstance(kernel_circle, np.ndarray)
np.testing.assert_allclose(kernel_circle,
kernel_circle_1_1_1,
equal_nan=True)
kernel_annulus = annulus_kernel(2, 2, 2, 1)
assert isinstance(kernel_annulus, np.ndarray)
np.testing.assert_allclose(kernel_annulus,
kernel_annulus_2_2_2_1,
equal_nan=True)
def test_hotspot_gpu_equals_cpu():
n, m = 10, 10
data = np.zeros((n, m), dtype=float)
nan_cells = [(i, i) for i in range(m)]
for cell in nan_cells:
data[cell[0], cell[1]] = np.nan
# add some extreme values
hot_region = [(1, 1), (1, 2), (1, 3),
(2, 1), (2, 2), (2, 3),
(3, 1), (3, 2), (3, 3)]
cold_region = [(7, 7), (7, 8), (7, 9),
(8, 7), (8, 8), (8, 9),
(9, 7), (9, 8), (9, 9)]
for p in hot_region:
data[p[0], p[1]] = 10000
for p in cold_region:
data[p[0], p[1]] = -10000
numpy_agg = xr.DataArray(data, dims=['y', 'x'])
numpy_agg['x'] = np.linspace(0, n, n)
numpy_agg['y'] = np.linspace(0, m, m)
cellsize_x, cellsize_y = calc_cellsize(numpy_agg)
kernel = circle_kernel(cellsize_x, cellsize_y, 2.0)
# numpy case
numpy_hotspots = hotspots(numpy_agg, kernel)
# cupy case
import cupy
cupy_agg = xr.DataArray(cupy.asarray(data))
cupy_hotspots = hotspots(cupy_agg, kernel)
assert isinstance(cupy_hotspots.data, cupy.ndarray)
assert np.isclose(
numpy_hotspots, cupy_hotspots.data.get(), equal_nan=True
).all()
# dask + cupy case not implemented
dask_cupy_agg = xr.DataArray(
da.from_array(cupy.asarray(data), chunks=(3, 3))
)
with pytest.raises(NotImplementedError) as e_info:
hotspots(dask_cupy_agg, kernel)
assert e_info
def test_focal_stats_cpu():
data = np.arange(16).reshape(4, 4)
numpy_agg = xr.DataArray(data)
dask_numpy_agg = xr.DataArray(da.from_array(data, chunks=(3, 3)))
cellsize = (1, 1)
kernel = circle_kernel(*cellsize, 1.5)
expected_results = np.asarray([
# mean
[[1.66666667, 2., 3., 4.], [4.25, 5., 6., 6.75],
[8.25, 9., 10., 10.75], [11., 12., 13., 13.33333333]],
# max
[[4., 5., 6., 7.], [8., 9., 10., 11.], [12., 13., 14., 15.],
[13., 14., 15., 15.]],
# min
[[0., 0., 1., 2.], [0., 1., 2., 3.], [4., 5., 6., 7.],
[8., 9., 10., 11.]],
# range
[[4., 5., 5., 5.], [8., 8., 8., 8.], [8., 8., 8., 8.],
[5., 5., 5., 4.]],
# std
[[1.69967317, 1.87082869, 1.87082869, 2.1602469],
[2.86138079, 2.60768096, 2.60768096, 2.86138079],
[2.86138079, 2.60768096, 2.60768096, 2.86138079],
[2.1602469, 1.87082869, 1.87082869, 1.69967317]],
# var
[[2.88888889, 3.5, 3.5, 4.66666667], [8.1875, 6.8, 6.8, 8.1875],
[8.1875, 6.8, 6.8, 8.1875], [4.66666667, 3.5, 3.5, 2.88888889]],
# sum
[[5., 8., 12., 12.], [17., 25., 30., 27.], [33., 45., 50., 43.],
[33., 48., 52., 40.]]
])
numpy_focalstats = focal_stats(numpy_agg, kernel)
general_output_checks(numpy_agg,
numpy_focalstats,
verify_attrs=False,
expected_results=expected_results)
assert numpy_focalstats.ndim == 3
dask_numpy_focalstats = focal_stats(dask_numpy_agg, kernel)
general_output_checks(dask_numpy_agg,
dask_numpy_focalstats,
verify_attrs=False,
expected_results=expected_results)
def test_kernel():
data = convolve_2d_data
m, n = data.shape
agg = xr.DataArray(data, dims=['y', 'x'])
agg['x'] = np.linspace(0, n, n)
agg['y'] = np.linspace(0, m, m)
cellsize_x, cellsize_y = calc_cellsize(agg)
kernel1 = circle_kernel(cellsize_x, cellsize_y, 2)
expected_kernel1 = np.array([[0, 1, 0], [1, 1, 1], [0, 1, 0]])
assert isinstance(kernel1, np.ndarray)
np.testing.assert_allclose(kernel1, expected_kernel1, equal_nan=True)
kernel2 = annulus_kernel(cellsize_x, cellsize_y, 2, 0.5)
expected_kernel2 = np.array([[0, 1, 0], [1, 0, 1], [0, 1, 0]])
assert isinstance(kernel2, np.ndarray)
np.testing.assert_allclose(kernel2, expected_kernel2, equal_nan=True)
def test_hotspot_gpu_equals_cpu():
n, m = 10, 10
data = np.asarray([[np.nan, 0., 0., 0., 0., 0., 0., 0., 0., 0.],
[0., 10000., 10000., 10000., 0., 0., 0., 0., 0., 0.],
[0., 10000., 10000., 10000., 0., 0., 0., 0., 0., 0.],
[0., 10000., 10000., 10000., 0., 0., 0., 0., 0., 0.],
[0., 0., 0., 0., np.nan, 0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0., np.nan, 0., 0., 0., 0.],
[0., 0., 0., 0., 0., 0., np.nan, 0., 0., 0.],
[0., 0., 0., 0., 0., 0., 0., -10000., -10000., -10000.],
[0., 0., 0., 0., 0., 0., 0., -10000., -10000., -10000.],
[0., 0., 0., 0., 0., 0., 0., -10000., -10000.,
-10000.]])
numpy_agg = xr.DataArray(data, dims=['y', 'x'])
numpy_agg['x'] = np.linspace(0, n, n)
numpy_agg['y'] = np.linspace(0, m, m)
cellsize_x, cellsize_y = calc_cellsize(numpy_agg)
kernel = circle_kernel(cellsize_x, cellsize_y, 2.0)
# numpy case
numpy_hotspots = hotspots(numpy_agg, kernel)
# cupy case
import cupy
cupy_agg = xr.DataArray(cupy.asarray(data))
cupy_hotspots = hotspots(cupy_agg, kernel)
np.testing.assert_allclose(numpy_hotspots.data,
cupy_hotspots.data.get(),
equal_nan=True)
# dask + cupy case not implemented
dask_cupy_agg = xr.DataArray(
da.from_array(cupy.asarray(data), chunks=(3, 3)))
with pytest.raises(NotImplementedError) as e_info:
hotspots(dask_cupy_agg, kernel)
assert e_info
def test_hotspots():
n, m = 10, 10
data = np.asarray([[np.nan, 0., 0., 0., 0., 0., 0., 0., 0., 0.],
[0., 10000., 10000., 10000., 0., 0., 0., 0., 0., 0.],
[0., 10000., 10000., 10000., 0., 0., 0., 0., 0., 0.],
[0., 10000., 10000., 10000., 0., 0., 0., 0., 0., 0.],
[0., 0., 0., 0., np.nan, 0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0., np.nan, 0., 0., 0., 0.],
[0., 0., 0., 0., 0., 0., np.nan, 0., 0., 0.],
[0., 0., 0., 0., 0., 0., 0., -10000., -10000., -10000.],
[0., 0., 0., 0., 0., 0., 0., -10000., -10000., -10000.],
[0., 0., 0., 0., 0., 0., 0., -10000., -10000.,
-10000.]])
numpy_agg = xr.DataArray(data, dims=['y', 'x'])
numpy_agg['x'] = np.linspace(0, n, n)
numpy_agg['y'] = np.linspace(0, m, m)
cellsize_x, cellsize_y = calc_cellsize(numpy_agg)
kernel = circle_kernel(cellsize_x, cellsize_y, 2.0)
expected_results = np.array(
[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 90, 0, 0, 0, 0, 0, 0, 0],
[0, 90, 95, 90, 0, 0, 0, 0, 0, 0], [0, 0, 90, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, -90, 0],
[0, 0, 0, 0, 0, 0, 0, -90, -95, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0]],
dtype=np.int8)
# numpy case
numpy_hotspots = hotspots(numpy_agg, kernel)
general_output_checks(numpy_agg, numpy_hotspots, expected_results)
# dask + numpy
dask_numpy_agg = xr.DataArray(da.from_array(data, chunks=(3, 3)))
dask_numpy_hotspots = hotspots(dask_numpy_agg, kernel)
general_output_checks(dask_numpy_agg, dask_numpy_hotspots,
expected_results)
def test_hotspot():
n, m = 10, 10
data = np.zeros((n, m), dtype=float)
all_idx = zip(*np.where(data == 0))
nan_cells = [(i, i) for i in range(m)]
for cell in nan_cells:
data[cell[0], cell[1]] = np.nan
# add some extreme values
hot_region = [(1, 1), (1, 2), (1, 3),
(2, 1), (2, 2), (2, 3),
(3, 1), (3, 2), (3, 3)]
cold_region = [(7, 7), (7, 8), (7, 9),
(8, 7), (8, 8), (8, 9),
(9, 7), (9, 8), (9, 9)]
for p in hot_region:
data[p[0], p[1]] = 10000
for p in cold_region:
data[p[0], p[1]] = -10000
numpy_agg = xr.DataArray(data, dims=['y', 'x'])
numpy_agg['x'] = np.linspace(0, n, n)
numpy_agg['y'] = np.linspace(0, m, m)
cellsize_x, cellsize_y = calc_cellsize(numpy_agg)
kernel = circle_kernel(cellsize_x, cellsize_y, 2.0)
no_significant_region = [id for id in all_idx if id not in hot_region and
id not in cold_region]
# numpy case
numpy_hotspots = hotspots(numpy_agg, kernel)
# dask + numpy
dask_numpy_agg = xr.DataArray(da.from_array(data, chunks=(3, 3)))
dask_numpy_hotspots = hotspots(dask_numpy_agg, kernel)
assert isinstance(dask_numpy_hotspots.data, da.Array)
# both output same results
assert np.isclose(numpy_hotspots.data, dask_numpy_hotspots.data.compute(),
equal_nan=True).all()
# check output's properties
# output must be an xarray DataArray
assert isinstance(numpy_hotspots, xr.DataArray)
assert isinstance(numpy_hotspots.values, np.ndarray)
assert issubclass(numpy_hotspots.values.dtype.type, np.int8)
# shape, dims, coords, attr preserved
assert numpy_agg.shape == numpy_hotspots.shape
assert numpy_agg.dims == numpy_hotspots.dims
assert numpy_agg.attrs == numpy_hotspots.attrs
for coord in numpy_agg.coords:
assert np.all(numpy_agg[coord] == numpy_hotspots[coord])
# no nan in output
assert not np.isnan(np.min(numpy_hotspots))
# output of extreme regions are non-zeros
# hot spots
hot_spot = np.asarray([numpy_hotspots[p] for p in hot_region])
assert np.all(hot_spot >= 0)
assert np.sum(hot_spot) > 0
# cold spots
cold_spot = np.asarray([numpy_hotspots[p] for p in cold_region])
assert np.all(cold_spot <= 0)
assert np.sum(cold_spot) < 0
# output of no significant regions are 0s
no_sign = np.asarray([numpy_hotspots[p] for p in no_significant_region])
assert np.all(no_sign == 0)