def test_multi_filter(dly):
"""
Since test.dly is artificial, has 3 distinct years where each year has 3 observations in 3 distinct months, expectation here is to get a single value. A normal .dly would yield 12 values, from the second of each month.
"""
dly.parse()
filters = [
Filter(column='year', value=1881, operator='lte'),
Filter(column='day', value=2, operator='eq'),
Filter(column='obs', value='TMAX', operator='eq'),
]
data = dly.get_data(filters)
values = dly.get_data_as_dict(data)['value']
assert 1 == len(values)
assert 317 == values[0]
def test_filter_lt(dly):
dly.parse()
filters = [
Filter(column='year', value=1881, operator='lt')
]
data = dly.get_data(filters)
years = dly.get_data_as_dict(data)['year']
assert 0 == len(years)
def test_filter_lte(dly):
dly.parse()
filters = [
Filter(column='year', value=1881, operator='lte')
]
data = dly.get_data(filters)
years = dly.get_data_as_dict(data)['year']
assert 92 == len(years)
assert 1881 in years
assert 1900 not in years
assert 1950 not in years
def test_filter_eq(dly):
dly.parse()
filters = [
Filter(column='year', value=1900)
]
data = dly.get_data(filters)
years = dly.get_data_as_dict(data)['year']
assert 90 == len(years)
assert 1881 not in years
assert 1900 in years
assert 1950 not in years
def test_interpolation(dly):
"""
GHCN uses -9999 for data points that don't exist, or weren't collected form some reason. Dly.parse() automatically replaces -9999 with np.nan (not a number), although this is controllable with the kwarg <code>replace_missing_data</code>.
Call Dly.interpolate(...) to replace nan values with interpolated. Test that this occurs.
"""
dly.parse()
filters = [
Filter(column='year', value=1950),
Filter(column='month', value=4),
Filter(column='obs', value='TMAX', operator='eq'),
]
data = dly.get_data(filters)
assert np.isnan(data['value'][2])
assert 56 == data['value'][0]
data = dly.interpolate(data, 'day')
assert not np.isnan(data['value'][2])
# floating point precision problem
assert 5.6 == pytest.approx(data['value'][0])
import matplotlib.pyplot as pp
from ghcn_py.dly import Dly, Filter
test_dir = os.path.abspath(os.path.join('..', os.path.dirname(__file__)))
if __name__ == "__main__":
"""
Generate 3 plots for demonstration.
This shows 1) you must call interpolate(...) in order to handle absent data values (-9999 in the .dly) and 2) the default behavior of interpolate is to divide TMIN and TMAX values by 10, since the .dly expresses temp values as tenths of a degree.
"""
file = os.path.join(test_dir, 'data', 'test.dly')
dly = Dly(file)
dly.parse()
filters = [
Filter(column='year', value=1950),
Filter(column='month', value=4),
Filter(column='obs', value='TMAX', operator='eq'),
]
data = dly.get_data(filters)
fig, axs = pp.subplots(3, 1, constrained_layout=True)
axs[0].plot(data['day'], data['value'])
axs[0].set_title('No interpolation (NaN still present)')
axs[0].set_xlabel('day')
axs[0].set_ylabel('TMAX')
data = dly.interpolate(data, 'day', adjust_temp_values=False)
axs[1].plot(data['day'], data['value'])
axs[1].set_title('Interpolated')