def test_calculate(self):
## daily data for three years is wanted for the test. subset a CMIP5
## decadal simulation to use for input into the computation.
rd = self.test_data.get_rd('cancm4_tas')
field = rd.get()
field = field.get_between('temporal',datetime.datetime(2001,1,1),datetime.datetime(2003,12,31,23,59))
## the calculation will be for months and years. set the temporal grouping.
temporal_group = field.temporal.get_grouping(['month','year'])
## create calculation object
percentile = 10
width = 5
operation = 'lt'
kwds = dict(percentile=percentile,width=width,operation=operation)
value = field.variables['tas'].value
dkp = DynamicDailyKernelPercentileThreshold(tgd=temporal_group,
parms=kwds,field=field,alias='tg10p')
dperc = dkp.get_daily_percentile(value,field.temporal.value_datetime,percentile,width)
select = np.logical_and(dperc['month'] == 6,dperc['day'] == 5)
to_test = dperc[select]['percentile'][0]
ref = self.get_percentile_reference()
self.assertNumpyAll(to_test,ref)
ret = dkp.execute()
self.assertEqual(ret['tg10p'].value.shape,(1,36,1,64,128))
self.assertAlmostEqual(ret['tg10p'].value.mean(),3.6267225477430554)
def test_operations_two_steps(self):
## get the request dataset to use as the basis for the percentiles
uri = self.test_data.get_uri('cancm4_tas')
variable = 'tas'
rd = RequestDataset(uri=uri,variable=variable)
## this is the underly OCGIS dataset object
nc_basis = rd.get()
## NOTE: if you want to subset the basis by time, this step is necessary
# nc_basis = nc_basis.get_between('temporal',datetime.datetime(2001,1,1),datetime.datetime(2003,12,31,23,59))
## these are the values to use when calculating the percentile basis. it
## may be good to wrap this in a function to have memory freed after the
## percentile structure array is computed.
all_values = nc_basis.variables[variable].value
## these are the datetime objects used for window creation
temporal = nc_basis.temporal.value_datetime
## additional parameters for calculating the basis
percentile = 10
width = 5
## get the structure array
from ocgis.calc.library.index.dynamic_kernel_percentile import DynamicDailyKernelPercentileThreshold
daily_percentile = DynamicDailyKernelPercentileThreshold.get_daily_percentile(all_values,temporal,percentile,width)
## perform the calculation using the precomputed basis. in this case,
## the basis and target datasets are the same, so the RequestDataset is
## reused.
calc_grouping = ['month','year']
kwds = {'percentile':percentile,'width':width,'operation':'lt','daily_percentile':daily_percentile}
calc = [{'func':'dynamic_kernel_percentile_threshold','name':'tg10p','kwds':kwds}]
ops = OcgOperations(dataset=rd,calc_grouping=calc_grouping,calc=calc,
output_format='nc')
ret = ops.execute()
## if we want to return the values as a three-dimenional numpy array the
## method below will do this. note the interface arrangement for the next
## release will alter this slightly.
ops = OcgOperations(dataset=rd,calc_grouping=calc_grouping,calc=calc,
output_format='numpy')
arrs = ops.execute()
## reference the returned numpy data. the first key is the geometry identifier.
## 1 in this case as this is the default for no selection geometry. the second
## key is the request dataset alias and the third is the calculation name.
## the variable name is appended to the end of the calculation to maintain
## a unique identifier.
tg10p = arrs[1]['tas'].variables['tg10p'].value
## if we want the date information for the temporal groups date attributes
date_parts = arrs[1]['tas'].temporal.date_parts
assert(date_parts.shape[0] == tg10p.shape[1])
## these are the representative datetime objects
rep_dt = arrs[1]['tas'].temporal.value_datetime
## and these are the lower and upper time bounds on the date groups
bin_bounds = arrs[1]['tas'].temporal.bounds_datetime
## confirm we have values for each month and year (12*10)
ret_ds = nc.Dataset(ret)
try:
self.assertEqual(ret_ds.variables['tg10p'].shape,(120,64,128))
finally:
ret_ds.close()
