def test_default_idw(self):
cml_list = load_processed_cml_list()
interpolator = pycml.spatial.interpolator.ComlinkGridInterpolator(
cml_list=cml_list,
variable='R',
resolution=0.05,
interpolator=pycml.spatial.interpolator.IdwKdtreeInterpolator())
ds = interpolator.loop_over_time()
assert ds.R.isel(time=0).values.shape == (16, 23)
np.testing.assert_array_almost_equal(
ds.R.isel(time=10).values[1:4, 1:4],
np.array([[0., 0., 0.], [0.02679148, 0., 0.],
[0.17902197, 0.10871572, 0.]]))
np.testing.assert_array_almost_equal(
ds.R.isel(time=20).values[1:4, 1:4],
np.array([[0.01852385, 0.08876891, 0.14781145],
[0.05900791, 0.15852366, 0.14716676],
[0.16692699, 0.16040398, 0.11649888]]))
zgrid_i = interpolator.interpolate_for_i(3)
np.testing.assert_array_almost_equal(
zgrid_i[1:4, 1:4],
np.array([[0.04484675, 0.12339762, 0.1185833],
[0.07857356, 0.26151277, 0.13641106],
[0.2014109, 0.19460449, 0.08598332]]))
def test_default_kriging(self):
cml_list = load_processed_cml_list()
# TODO: Use 'C' backend when available via pip installed pykirge
interpolator = pycml.spatial.interpolator.ComlinkGridInterpolator(
cml_list=cml_list,
variable="R",
resolution=0.05,
interpolator=(
pycml.spatial.interpolator.OrdinaryKrigingInterpolator(backend="loop")
),
)
ds = interpolator.loop_over_time()
assert ds.R.isel(time=0).values.shape == (16, 23)
# TODO: Check if theses results change when using the 'C' backend
#
# Note, that the reference values had to be updated after switching
# from pykrige version 1.3.1 to version 1.4.0. See this PR
# https://github.com/bsmurphy/PyKrige/issues/110
np.testing.assert_array_almost_equal(
ds.R.isel(time=24).values[1:4, 1:4],
np.array(
[
[2.312077, 3.591382, 5.370885],
[2.410231, 3.689951, 4.975776],
[2.661749, 3.460712, 4.415959],
]
),
)
def test_default_idw_aggregated_to_new_index(self):
t_str_list = [
"2017-06-27 21:50:00",
"2017-06-27 22:50:00",
"2017-06-27 23:50:00",
"2017-06-28 00:50:00",
"2017-06-28 01:50:00",
"2017-06-28 02:50:00",
"2017-06-28 03:50:00",
"2017-06-28 04:50:00",
"2017-06-28 05:50:00",
"2017-06-28 06:50:00",
"2017-06-28 07:50:00",
"2017-06-28 08:50:00",
"2017-06-28 09:50:00",
]
cml_list = load_processed_cml_list()
interpolator = pycml.spatial.interpolator.ComlinkGridInterpolator(
cml_list=cml_list,
variable="R",
resample_to_new_index=pd.to_datetime(t_str_list),
resolution=0.05,
interpolator=pycml.spatial.interpolator.IdwKdtreeInterpolator(),
)
ds = interpolator.loop_over_time()
assert ds.R.isel(time=0).values.shape == (16, 23)
assert ds.time[0].values == np.datetime64("2017-06-28T00:50:00")
assert ds.time[-1].values == np.datetime64("2017-06-28T09:50:00")
np.testing.assert_array_almost_equal(
ds.R.isel(time=9).values[1:4, 1:4],
np.array(
[
[0.01552989, 0.09481763, 0.15710547],
[0.035198, 0.06016459, 0.19047063],
[0.06757572, 0.09647522, 0.08332671],
]
),
)
np.testing.assert_array_almost_equal(
ds.R.isel(time=6).values[-3:, -3:],
np.array(
[
[2.64776423, 2.27544954, 1.96875187],
[3.26109505, 2.83678722, 2.45349594],
[3.77713447, 3.2373787, 2.96544491],
]
),
)
def test_short_new_hourly_index_left_edge(self):
cml_list = load_processed_cml_list()
cml = cml_list[0]
# Test with label 'right'
df_resample = pycml.util.temporal.aggregate_df_onto_DatetimeIndex(
df=cml.channel_1.data,
new_index=pd.to_datetime(t_str_list),
label='right',
method=np.mean)
pd.util.testing.assert_frame_equal(
pd.DataFrame(df_resample.txrx),
pd.DataFrame(
index=pd.DatetimeIndex([
'2017-06-28 00:50:00+00:00', '2017-06-28 01:50:00+00:00',
'2017-06-28 02:50:00+00:00', '2017-06-28 03:50:00+00:00',
'2017-06-28 04:50:00+00:00', '2017-06-28 05:50:00+00:00',
'2017-06-28 06:50:00+00:00', '2017-06-28 07:50:00+00:00',
'2017-06-28 08:50:00+00:00', '2017-06-28 09:50:00+00:00'],
dtype='datetime64[ns, UTC]', name=u'time', freq=None),
data={'txrx': [60.988, 60.755, 60.88813559, 60.87368421,
60.97241379, 70.48305085, 63.42807018, 61.2,
60.62166667, 60.92]})
)
# Test with label 'left'
df_resample = pycml.util.temporal.aggregate_df_onto_DatetimeIndex(
df=cml.channel_1.data,
new_index=pd.to_datetime(t_str_list),
label='left',
method=np.mean)
pd.util.testing.assert_frame_equal(
pd.DataFrame(df_resample.txrx),
pd.DataFrame(
index=pd.DatetimeIndex([
'2017-06-28 00:50:00+00:00', '2017-06-28 01:50:00+00:00',
'2017-06-28 02:50:00+00:00', '2017-06-28 03:50:00+00:00',
'2017-06-28 04:50:00+00:00', '2017-06-28 05:50:00+00:00',
'2017-06-28 06:50:00+00:00', '2017-06-28 07:50:00+00:00',
'2017-06-28 08:50:00+00:00'],
dtype='datetime64[ns, UTC]', name=u'time', freq=None),
data={'txrx': [60.755, 60.88813559, 60.87368421,
60.97241379, 70.48305085, 63.42807018, 61.2,
60.62166667, 60.92]})
)
def test_cml_id_order(self):
cml_list = load_processed_cml_list()
# Add two CMLs with new fake cml_id (which are very different from
# the example cml_ids to cause mixing order of their hashed values)
cml_temp = copy.deepcopy(cml_list[0])
cml_temp.metadata['cml_id'] = b'some_really_long_cml_id_122334567'
cml_list.insert(0, cml_temp)
cml_temp = copy.deepcopy(cml_list[0])
cml_temp.metadata[
'cml_id'] = b'another_really_long_cml_id_999222123123'
cml_list.append(cml_temp)
df = pycml.spatial.interpolator.get_dataframe_for_cml_variable(
cml_list=cml_list)
for df_column_name, cml in zip(df.columns.tolist(), cml_list):
assert df_column_name == cml.metadata['cml_id']
def test_default_kriging(self):
cml_list = load_processed_cml_list()
# TODO: Use 'C' backend when available via pip installed pykirge
interpolator = pycml.spatial.interpolator.ComlinkGridInterpolator(
cml_list=cml_list,
variable='R',
resolution=0.05,
interpolator=(
pycml.spatial.interpolator.OrdinaryKrigingInterpolator(
backend='loop')))
ds = interpolator.loop_over_time()
assert ds.R.isel(time=0).values.shape == (16, 23)
# TODO: Check if theses results change when using the 'C' backend
np.testing.assert_array_almost_equal(
ds.R.isel(time=24).values[1:4, 1:4],
np.array([[2.41876631, 3.55747763, 5.21045399],
[2.59474309, 3.74337047, 4.95187848],
[2.89174467, 3.57024647, 4.49247846]]))
