def test_average_08(self):
from_axis = DailyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_n_interval(14) \
.build()
to_axis = WeeklyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_n_interval(2) \
.build()
tc = AxisRemapper(from_axis=from_axis, to_axis=to_axis)
from_data = np.moveaxis(
np.asarray(list(range(1, 15)) * 12).reshape((3, 4, 14)), 2, 1)
to_data = tc.average(from_data, dimension=1)
self.assertEqual((3, 2, 4), to_data.shape)
self.assertListEqual([4] * 12,
np.round(to_data[:, 0, :].reshape(
(-1, 1))).astype(np.int).flatten().tolist())
self.assertListEqual([11] * 12,
np.round(to_data[:, 1, :].reshape(
(-1, 1))).astype(np.int).flatten().tolist())
def test_dask_max(self):
from_axis = DailyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=14).build()
to_axis = WeeklyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=3).build()
tc = AxisRemapper(from_axis=from_axis,
to_axis=to_axis,
assure_no_bound_mismatch=False)
from_data = da.arange(14, dtype='float64')
to_data = tc.max(from_data).compute()
np.testing.assert_almost_equal(
to_data,
np.array([6.0, 13.0, np.nan]).reshape(3, 1))
def test_max_01(self):
from_axis = DailyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=14).build()
to_axis = WeeklyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=2).build()
tc = AxisRemapper(from_axis=from_axis, to_axis=to_axis)
from_data = list(range(1, 15))
to_data = tc.max(from_data)
self.assertAlmostEqual(7.0, to_data[0, 0], 0)
self.assertAlmostEqual(14.0, to_data[1, 0], 0)
def test_apply_function_01(self):
daily_axis = DailyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=14).build()
weekly_axis = WeeklyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=2).build()
ac = AxisRemapper(from_axis=daily_axis, to_axis=weekly_axis)
daily_data = np.random.random((3, 4, 5, 14))
def user_defined_function(data):
return np.nansum(data, axis=0) * 42
weekly_user_defined = ac.apply_function(daily_data,
user_defined_function,
dimension=3)
def test_average_11(self):
from_axis = DailyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=14).build()
to_axis = RollingWindowTimeAxisBuilder(start_date=date(2019, 1, 1),
end_date=date(2019, 1, 15),
window_size=7).build()
tc = AxisRemapper(from_axis=from_axis, to_axis=to_axis)
from_data = list(range(1, 15))
to_data = tc.average(from_data)
self.assertEqual(8, to_data.size)
for idx in range(8):
self.assertAlmostEqual(idx + 4, to_data[idx, 0], 0)
def test_dask_array_01(self):
daily_axis = DailyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=14).build()
weekly_axis = WeeklyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=2).build()
tc = AxisRemapper(from_axis=daily_axis, to_axis=weekly_axis)
from_data = da.arange(14)
to_data = tc.average(from_data)
self.assertTrue(isinstance(to_data, da.Array))
to_data = to_data.compute()
self.assertAlmostEqual(3.0, to_data[0, 0], 0)
self.assertAlmostEqual(10.0, to_data[1, 0], 0)
def test_max_05(self):
from_axis = DailyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=14).build()
to_axis = WeeklyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=3).build()
tc = AxisRemapper(from_axis=from_axis,
to_axis=to_axis,
assure_no_bound_mismatch=False)
from_data = list(range(1, 15))
to_data = tc.max(from_data)
self.assertAlmostEqual(7.0, to_data[0, 0], 0)
self.assertAlmostEqual(14.0, to_data[1, 0], 0)
self.assertTrue(np.isnan(to_data[2, 0]))
def test_average_01(self):
from_axis = DailyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_end_date(date(2019, 1, 8)) \
.build()
to_axis = WeeklyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_end_date(date(2019, 1, 8)) \
.build()
tc = AxisRemapper(from_axis=from_axis, to_axis=to_axis)
from_data = [1, 2, 3, 4, 5, 6, 7]
to_data = tc.average(from_data)
self.assertAlmostEqual(4.0, to_data[0, 0], 0)
def test_average_05(self):
from_axis = WeeklyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_n_interval(2) \
.build()
to_axis = DailyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_n_interval(14) \
.build()
tc = AxisRemapper(from_axis=from_axis, to_axis=to_axis)
from_data = [4.0, 11]
to_data = tc.average(from_data)
expected = [4] * 7 + [11] * 7
self.assertListEqual(expected, to_data.flatten().tolist())
def test_average_04(self):
from_axis = DailyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_n_interval(14) \
.build()
to_axis = WeeklyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_n_interval(2) \
.build()
tc = AxisRemapper(from_axis=from_axis, to_axis=to_axis)
from_data = list(range(1, 15))
to_data = tc.average(from_data)
self.assertAlmostEqual(4.0, to_data[0, 0], 0)
self.assertAlmostEqual(11.0, to_data[1, 0], 0)
def test_dask_array_05(self):
daily_axis = DailyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=14).build()
weekly_axis = WeeklyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=2).build()
tc = AxisRemapper(from_axis=daily_axis, to_axis=weekly_axis)
from_data = da.from_array(np.moveaxis(
np.asarray(np.arange(14).tolist() * 12).reshape(3, 4, 14), 2, 0),
chunks=(2, 1, 1))
to_data = tc.average(from_data)
self.assertTrue(isinstance(to_data, da.Array))
to_data = to_data.compute()
np.testing.assert_almost_equal(
to_data,
np.asarray([3] * 12 + [10] * 12).reshape(2, 3, 4))
def test_min_02(self):
from_axis = DailyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=14).build()
to_axis = WeeklyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=2).build()
tc = AxisRemapper(from_axis=from_axis, to_axis=to_axis)
from_data = np.moveaxis(
np.asarray(list(range(1, 15)) * 12).reshape((3, 4, 14)), 2,
0).tolist()
to_data = tc.min(from_data)
self.assertTrue(
np.all(np.ones((3, 4), dtype="int") - to_data[0].round() == 0.0))
self.assertTrue(
np.all(
np.ones((3, 4), dtype="int") * 8 - to_data[1].round() == 0.0))
def test_average_02(self):
from_axis = WeeklyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_end_date(date(2019, 1, 8)) \
.build()
to_axis = DailyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_end_date(date(2019, 1, 8)) \
.build()
tc = AxisRemapper(from_axis=from_axis, to_axis=to_axis)
from_data = [4.0]
to_data = tc.average(from_data)
self.assertEqual((tc.to_nelem, 1), to_data.shape)
for e in to_data:
self.assertAlmostEqual(4.0, e[0], 0)
def test_max_04(self):
from_axis = DailyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=14).build()
to_axis = RollingWindowTimeAxisBuilder(start_date=date(2019, 1, 1),
end_date=date(2019, 1, 15),
window_size=7).build()
tc = AxisRemapper(from_axis=from_axis, to_axis=to_axis)
from_data = np.moveaxis(
np.asarray(list(range(1, 15)) * 12).reshape((3, 4, 14)), 2,
1).tolist()
to_data = tc.max(from_data, dimension=1)
for i in range(tc.to_nelem):
self.assertTrue(
np.all(
np.ones((3, 4), dtype="int") * (7 + i) -
to_data[:, i, :].round() == 0.0))
def test_creation_01(self):
from_axis = DailyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_end_date(date(2019, 1, 8)) \
.build()
to_axis = WeeklyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_end_date(date(2019, 1, 8)) \
.build()
tc = AxisRemapper(from_axis=from_axis, to_axis=to_axis)
def test_speed_01(self):
from_axis = DailyTimeAxisBuilder(start_date=date(2000, 1, 1),
end_date=date(2020, 1, 1)).build()
to_axis = MonthlyTimeAxisBuilder(start_year=2000,
end_year=2019).build()
tc = AxisRemapper(from_axis=from_axis, to_axis=to_axis)
from_data = np.random.random((from_axis.nelem, 100, 100))
import time
start = time.time()
for i in range(1):
tc.min(from_data)
end = time.time()
print(f"Took: %f [s]", end - start)
import time
start = time.time()
for i in range(1):
tc.min(from_data)
end = time.time()
print(f"Took: %f [s]", end - start)
def test_max_03(self):
from_axis = DailyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=14).build()
to_axis = WeeklyTimeAxisBuilder(start_date=date(2019, 1, 1),
n_interval=2).build()
tc = AxisRemapper(from_axis=from_axis, to_axis=to_axis)
from_data = np.moveaxis(
np.asarray(list(range(1, 15)) * 12).reshape((3, 4, 14)), 2,
1).tolist()
to_data = tc.max(from_data, dimension=1)
self.assertTrue(
np.all(
np.ones((3, 4), dtype="int") * 7 -
to_data[:, 0, :].round() == 0.0))
self.assertTrue(
np.all(
np.ones((3, 4), dtype="int") * 14 -
to_data[:, 1, :].round() == 0.0))
def test_get_coverage_01(self):
from_axis = Axis(lower_bound=np.arange(0, 7) * 24,
upper_bound=np.arange(1, 8) * 24,
fraction=0.5)
to_axis = Axis(lower_bound=[0], upper_bound=[168], data_ticks=[84])
row_idx, col_idx, weights = AxisRemapper._get_coverage(
from_axis.lower_bound, from_axis.upper_bound, to_axis.lower_bound,
to_axis.upper_bound)
self.assertListEqual([0] * 7, row_idx)
self.assertListEqual(list(range(7)), col_idx)
self.assertListEqual([1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0], weights)
def test_get_coverage_03(self):
from_axis = DailyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_end_date(date(2019, 1, 8)) \
.build()
to_axis = WeeklyTimeAxisBuilder()\
.set_start_date(date(2019, 1, 1)) \
.set_end_date(date(2019, 1, 8)) \
.build()
row_idx, col_idx, weights = AxisRemapper._get_coverage(
from_axis.lower_bound, from_axis.upper_bound, to_axis.lower_bound,
to_axis.upper_bound)
self.assertListEqual([0] * 7, row_idx)
self.assertListEqual(list(range(7)), col_idx)
self.assertListEqual([1.0] * 7, weights)