class ExecutionEngine(object):
REDUCTION_FNS = {"all": xr.DataArray.all,
"any": xr.DataArray.any,
"argmax": xr.DataArray.argmax,
"argmin": xr.DataArray.argmin,
"max": xr.DataArray.max,
"mean": xr.DataArray.mean,
"median": xr.DataArray.median,
"min": xr.DataArray.min,
"prod": xr.DataArray.prod,
"sum": xr.DataArray.sum,
"std": xr.DataArray.std,
"var": xr.DataArray.var}
def __init__(self, api=None, index=None):
LOG.info('Initialise Execution Module.')
self.cache = {}
self.nd = NDexpr()
self.nd.set_ae(True)
self.api = api or API(index=index)
self.udfuncs = {}
def add_function(self, name, func):
self.udfuncs[name] = func
def execute_plan(self, plan):
for task in plan:
function = next(iter(task.values()))['orig_function']
op_type = next(iter(task.values()))['operation_type']
if op_type == OperationType.Get_Data:
self.execute_get_data(task)
elif op_type == OperationType.Expression:
self.execute_expression(task)
elif op_type == OperationType.Cloud_Mask:
self.execute_cloud_mask(task)
elif op_type == OperationType.Reduction and \
len([s for s in self.REDUCTION_FNS.keys() if s in function]) > 0:
self.execute_reduction(task)
elif op_type == OperationType.Bandmath:
self.execute_bandmath(task)
def execute_get_data(self, task):
value = next(iter(task.values()))
array_input_value = next(iter(value['array_input'][0].values()))
data_request_param = {}
data_request_param['dimensions'] = array_input_value['dimensions']
if 'storage_type' in array_input_value:
data_request_param['storage_type'] = array_input_value['storage_type']
else:
data_request_param['product'] = array_input_value['product']
data_request_param['platform'] = array_input_value['platform']
data_request_param['variables'] = ()
for array in value['array_input']:
data_request_param['variables'] += (next(iter(array.values()))['variable'],)
data_response = self.api.get_data(data_request_param)
no_data_value = value['array_output']['no_data_value']
if no_data_value is not None:
for k, v in data_response['arrays'].items():
data_response['arrays'][k] = data_response['arrays'][k].where(v != no_data_value)
key = next(iter(task.keys()))
self.cache[key] = {}
self.cache[key]['array_result'] = copy.deepcopy(data_response['arrays'])
self.cache[key]['array_indices'] = copy.deepcopy(data_response['indices'])
self.cache[key]['array_dimensions'] = copy.deepcopy(data_response['dimensions'])
self.cache[key]['array_output'] = copy.deepcopy(value['array_output'])
self.cache[key]['crs'] = copy.deepcopy(data_response['coordinate_reference_systems'])
del data_request_param
del data_response
return self.cache[key]
def execute_cloud_mask(self, task):
key = next(iter(task.keys()))
value = next(iter(task.values()))
data_key = value['array_input'][0]
mask_key = value['array_mask']
no_data_value = value['array_output']['no_data_value']
array_desc = self.cache[value['array_input'][0]]
data_array = next(iter(self.cache[data_key]['array_result'].values()))
mask_array = next(iter(self.cache[mask_key]['array_result'].values()))
pqa_mask = get_pqa_mask(mask_array.values)
masked_array = xr.DataArray.where(data_array, pqa_mask)
#masked_array = masked_array.fillna(no_data_value)
self.cache[key] = {}
self.cache[key]['array_result'] = {}
self.cache[key]['array_result'][key] = masked_array
self.cache[key]['array_indices'] = copy.deepcopy(array_desc['array_indices'])
self.cache[key]['array_dimensions'] = copy.deepcopy(array_desc['array_dimensions'])
self.cache[key]['array_output'] = copy.deepcopy(value['array_output'])
self.cache[key]['crs'] = copy.deepcopy(array_desc['crs'])
def execute_expression(self, task):
key = next(iter(task.keys()))
value = next(iter(task.values()))
no_data_value = value['array_output']['no_data_value']
arrays = {}
for task_name in value['array_input']:
arrays[task_name] = next(iter(self.cache[task_name]['array_result'].values()))
for i in arrays:
if arrays[i].dtype == type(no_data_value):
arrays[i] = arrays[i].where(arrays[i] != no_data_value)
array_result = {}
array_result['array_result'] = {}
array_result['array_result'][key] = self.nd.evaluate(value['function'],
local_dict=arrays,
user_functions=self.udfuncs)
#array_result['array_result'][key] = array_result['array_result'][key].fillna(no_data_value)
array_desc = self.cache[value['array_input'][0]]
array_result['array_indices'] = copy.deepcopy(array_desc['array_indices'])
array_result['array_dimensions'] = copy.deepcopy(array_desc['array_dimensions'])
array_result['array_output'] = copy.deepcopy(value['array_output'])
array_result['crs'] = copy.deepcopy(array_desc['crs'])
self.cache[key] = array_result
return self.cache[key]
def execute_bandmath(self, task):
key = next(iter(task.keys()))
value = next(iter(task.values()))
arrays = {}
for task_name in value['array_input']:
#for k, v in self.cache[task_name]['array_result'].items():
# arrays[k] = v.astype(float).values
arrays.update(self.cache[task_name]['array_result'])
array_result = {}
array_result['array_result'] = {}
array_result['array_result'][key] = xr.DataArray(ne.evaluate(value['function'], arrays))
#array_result['array_result'][key] = self.nd.evaluate(value['function'], arrays)
array_desc = self.cache[value['array_input'][0]]
array_result['array_indices'] = copy.deepcopy(array_desc['array_indices'])
array_result['array_dimensions'] = copy.deepcopy(array_desc['array_dimensions'])
array_result['array_output'] = copy.deepcopy(value['array_output'])
array_result['crs'] = copy.deepcopy(array_desc['crs'])
self.cache[key] = array_result
return self.cache[key]
def execute_reduction(self, task):
key = next(iter(task.keys()))
value = next(iter(task.values()))
function_name = value['orig_function'].replace(")", " ").replace("(", " ").split()[0]
func = self.REDUCTION_FNS[function_name]
data_key = value['array_input'][0]
data = self.cache[data_key]['array_dimensions']
no_data_value = value['array_output']['no_data_value']
array_data = next(iter(self.cache[data_key]['array_result'].values()))
array_desc = self.cache[value['array_input'][0]]
array_result = {}
array_result['array_result'] = {}
array_result['array_output'] = copy.deepcopy(value['array_output'])
dims = tuple((self.cache[data_key]['array_dimensions'].index(p) for p in value['dimension']))
args = {}
if function_name == 'argmax' or function_name == 'argmin':
if len(dims) != 1:
args['axis'] = dims[0]
else:
args['axis'] = dims
if sys.version_info >= (3, 0):
if 'skipna' in list(inspect.signature(self.REDUCTION_FNS[function_name]).parameters.keys()) and \
function_name != 'prod':
args['skipna'] = True
else:
if 'skipna' in inspect.getargspec(self.REDUCTION_FNS[function_name])[0] and \
function_name != 'prod':
args['skipna'] = True
array_result['array_result'][key] = func(xr.DataArray(array_data), **args)
array_result['array_indices'] = copy.deepcopy(array_desc['array_indices'])
array_result['array_dimensions'] = copy.deepcopy(array_result['array_output']['dimensions_order'])
array_result['crs'] = copy.deepcopy(array_desc['crs'])
self.cache[key] = array_result
return self.cache[key]
class ExecutionEngine(object):
REDUCTION_FNS = {"all": xr.DataArray.all,
"any": xr.DataArray.any,
"argmax": xr.DataArray.argmax,
"argmin": xr.DataArray.argmin,
"max": xr.DataArray.max,
"mean": xr.DataArray.mean,
"median": xr.DataArray.median,
"min": xr.DataArray.min,
"prod": xr.DataArray.prod,
"sum": xr.DataArray.sum,
"std": xr.DataArray.std,
"var": xr.DataArray.var}
def __init__(self, api=None, index=None):
LOG.info('Initialise Execution Module.')
self.cache = {}
self.nd = NDexpr()
self.nd.set_ae(True)
self.api = api or API(index=index)
self.udfuncs = {}
def add_function(self, name, func):
self.udfuncs[name] = func
def execute_plan(self, plan):
for task in plan:
function = next(iter(task.values()))['orig_function']
op_type = next(iter(task.values()))['operation_type']
if op_type == OperationType.Get_Data:
self.execute_get_data(task)
elif op_type == OperationType.Expression:
self.execute_expression(task)
elif op_type == OperationType.Cloud_Mask:
self.execute_cloud_mask(task)
elif op_type == OperationType.Reduction and \
len([s for s in self.REDUCTION_FNS.keys() if s in function]) > 0:
self.execute_reduction(task)
elif op_type == OperationType.Bandmath:
self.execute_bandmath(task)
def execute_get_data(self, task):
value = next(iter(task.values()))
array_input_value = next(iter(value['array_input'][0].values()))
data_request_param = {}
data_request_param['dimensions'] = array_input_value['dimensions']
if 'storage_type' in array_input_value:
data_request_param['storage_type'] = array_input_value['storage_type']
else:
data_request_param['product'] = array_input_value['product']
data_request_param['platform'] = array_input_value['platform']
data_request_param['variables'] = ()
for array in value['array_input']:
data_request_param['variables'] += (next(iter(array.values()))['variable'],)
data_response = self.api.get_data(data_request_param)
no_data_value = value['array_output']['no_data_value']
if no_data_value is not None:
for k, v in data_response['arrays'].items():
data_response['arrays'][k] = data_response['arrays'][k].where(v != no_data_value)
key = next(iter(task.keys()))
self.cache[key] = {}
self.cache[key]['array_result'] = copy.deepcopy(data_response['arrays'])
self.cache[key]['array_indices'] = copy.deepcopy(data_response['indices'])
self.cache[key]['array_dimensions'] = copy.deepcopy(data_response['dimensions'])
self.cache[key]['array_output'] = copy.deepcopy(value['array_output'])
self.cache[key]['crs'] = copy.deepcopy(data_response['coordinate_reference_systems'])
del data_request_param
del data_response
return self.cache[key]
def execute_cloud_mask(self, task):
key = next(iter(task.keys()))
value = next(iter(task.values()))
data_key = value['array_input'][0]
mask_key = value['array_mask']
no_data_value = value['array_output']['no_data_value']
array_desc = self.cache[value['array_input'][0]]
data_array = next(iter(self.cache[data_key]['array_result'].values()))
mask_array = next(iter(self.cache[mask_key]['array_result'].values()))
pqa_mask = get_pqa_mask(mask_array.values)
masked_array = xr.DataArray.where(data_array, pqa_mask)
#masked_array = masked_array.fillna(no_data_value)
self.cache[key] = {}
self.cache[key]['array_result'] = {}
self.cache[key]['array_result'][key] = masked_array
self.cache[key]['array_indices'] = copy.deepcopy(array_desc['array_indices'])
self.cache[key]['array_dimensions'] = copy.deepcopy(array_desc['array_dimensions'])
self.cache[key]['array_output'] = copy.deepcopy(value['array_output'])
self.cache[key]['crs'] = copy.deepcopy(array_desc['crs'])
def execute_expression(self, task):
key = next(iter(task.keys()))
value = next(iter(task.values()))
no_data_value = value['array_output']['no_data_value']
arrays = {}
for task_name in value['array_input']:
arrays[task_name] = next(iter(self.cache[task_name]['array_result'].values()))
for i in arrays.keys():
if arrays[i].dtype == type(no_data_value):
arrays[i] = arrays[i].where(arrays[i] != no_data_value)
array_result = {}
array_result['array_result'] = {}
array_result['array_result'][key] = self.nd.evaluate(value['function'],
local_dict=arrays,
user_functions=self.udfuncs)
#array_result['array_result'][key] = array_result['array_result'][key].fillna(no_data_value)
array_desc = self.cache[value['array_input'][0]]
array_result['array_indices'] = copy.deepcopy(array_desc['array_indices'])
array_result['array_dimensions'] = copy.deepcopy(array_desc['array_dimensions'])
array_result['array_output'] = copy.deepcopy(value['array_output'])
array_result['crs'] = copy.deepcopy(array_desc['crs'])
self.cache[key] = array_result
return self.cache[key]
def execute_bandmath(self, task):
key = next(iter(task.keys()))
value = next(iter(task.values()))
arrays = {}
for task_name in value['array_input']:
#for k, v in self.cache[task_name]['array_result'].items():
# arrays[k] = v.astype(float).values
arrays.update(self.cache[task_name]['array_result'])
array_result = {}
array_result['array_result'] = {}
array_result['array_result'][key] = xr.DataArray(ne.evaluate(value['function'], arrays))
#array_result['array_result'][key] = self.nd.evaluate(value['function'], arrays)
array_desc = self.cache[value['array_input'][0]]
array_result['array_indices'] = copy.deepcopy(array_desc['array_indices'])
array_result['array_dimensions'] = copy.deepcopy(array_desc['array_dimensions'])
array_result['array_output'] = copy.deepcopy(value['array_output'])
array_result['crs'] = copy.deepcopy(array_desc['crs'])
self.cache[key] = array_result
return self.cache[key]
def execute_reduction(self, task):
key = next(iter(task.keys()))
value = next(iter(task.values()))
function_name = value['orig_function'].replace(")", " ").replace("(", " ").split()[0]
func = self.REDUCTION_FNS[function_name]
data_key = value['array_input'][0]
data = self.cache[data_key]['array_dimensions']
no_data_value = value['array_output']['no_data_value']
array_data = next(iter(self.cache[data_key]['array_result'].values()))
array_desc = self.cache[value['array_input'][0]]
array_result = {}
array_result['array_result'] = {}
array_result['array_output'] = copy.deepcopy(value['array_output'])
dims = tuple((self.cache[data_key]['array_dimensions'].index(p) for p in value['dimension']))
args = {}
if function_name == 'argmax' or function_name == 'argmin':
if len(dims) != 1:
args['axis'] = dims[0]
else:
args['axis'] = dims
if sys.version_info >= (3, 0):
if 'skipna' in list(inspect.signature(self.REDUCTION_FNS[function_name]).parameters.keys()) and \
function_name != 'prod':
args['skipna'] = True
else:
if 'skipna' in inspect.getargspec(self.REDUCTION_FNS[function_name])[0] and \
function_name != 'prod':
args['skipna'] = True
array_result['array_result'][key] = func(xr.DataArray(array_data), **args)
array_result['array_indices'] = copy.deepcopy(array_desc['array_indices'])
array_result['array_dimensions'] = copy.deepcopy(array_result['array_output']['dimensions_order'])
array_result['crs'] = copy.deepcopy(array_desc['crs'])
self.cache[key] = array_result
return self.cache[key]
def test_2():
# perform language test
ne = NDexpr()
ne.set_ae(False)
x1 = xr.DataArray(np.random.randn(2, 3))
y1 = xr.DataArray(np.random.randn(2, 3))
z1 = xr.DataArray(
np.array([[[0, 1, 2], [3, 4, 5], [6, 7, 8]],
[[9, 10, 11], [12, 13, 14], [15, 16, 17]],
[[18, 19, 20], [21, 22, 23], [24, 25, 26]]]))
z2 = z1 * 2
z3 = np.arange(27)
mask1 = z1 > 4
assert ne.test("arccos(z1)", xr.ufuncs.arccos(z1))
assert ne.test("angle(z1)", xr.ufuncs.angle(z1))
assert ne.test("arccos(z1)", xr.ufuncs.arccos(z1))
assert ne.test("arccosh(z1)", xr.ufuncs.arccosh(z1))
assert ne.test("arcsin(z1)", xr.ufuncs.arcsin(z1))
assert ne.test("arcsinh(z1)", xr.ufuncs.arcsinh(z1))
assert ne.test("arctan(z1)", xr.ufuncs.arctan(z1))
assert ne.test("arctanh(z1)", xr.ufuncs.arctanh(z1))
assert ne.test("ceil(z1)", xr.ufuncs.ceil(z1))
assert ne.test("conj(z1)", xr.ufuncs.conj(z1))
assert ne.test("cos(z1)", xr.ufuncs.cos(z1))
assert ne.test("cosh(z1)", xr.ufuncs.cosh(z1))
assert ne.test("deg2rad(z1)", xr.ufuncs.deg2rad(z1))
assert ne.test("degrees(z1)", xr.ufuncs.degrees(z1))
assert ne.test("exp(z1)", xr.ufuncs.exp(z1))
assert ne.test("expm1(z1)", xr.ufuncs.expm1(z1))
assert ne.test("fabs(z1)", xr.ufuncs.fabs(z1))
assert ne.test("fix(z1)", xr.ufuncs.fix(z1))
assert ne.test("floor(z1)", xr.ufuncs.floor(z1))
assert ne.test("frexp(z3)", xr.ufuncs.frexp(z3))
assert ne.test("imag(z1)", xr.ufuncs.imag(z1))
assert ne.test("iscomplex(z1)", xr.ufuncs.iscomplex(z1))
assert ne.test("isfinite(z1)", xr.ufuncs.isfinite(z1))
assert ne.test("isinf(z1)", xr.ufuncs.isinf(z1))
assert ne.test("isnan(z1)", xr.ufuncs.isnan(z1))
assert ne.test("isreal(z1)", xr.ufuncs.isreal(z1))
assert ne.test("log(z1)", xr.ufuncs.log(z1))
assert ne.test("log10(z1)", xr.ufuncs.log10(z1))
assert ne.test("log1p(z1)", xr.ufuncs.log1p(z1))
assert ne.test("log2(z1)", xr.ufuncs.log2(z1))
assert ne.test("pow(z1, 2.0)", np.power(z1, 2.0))
assert ne.test("pow(z1, 2)", np.power(z1, 2))
assert ne.test("z1^2", np.power(z1, 2))
assert ne.test("z1**2", np.power(z1, 2))
assert ne.test("rad2deg(z1)", xr.ufuncs.rad2deg(z1))
assert ne.test("radians(z1)", xr.ufuncs.radians(z1))
assert ne.test("real(z1)", xr.ufuncs.real(z1))
assert ne.test("rint(z1)", xr.ufuncs.rint(z1))
assert ne.test("sign(z1)", xr.ufuncs.sign(z1))
assert ne.test("signbit(z1)", xr.ufuncs.signbit(z1))
assert ne.test("sin(z1)", xr.ufuncs.sin(z1))
assert ne.test("sinh(z1)", xr.ufuncs.sinh(z1))
assert ne.test("sqrt(z1)", xr.ufuncs.sqrt(z1))
assert ne.test("square(z1)", xr.ufuncs.square(z1))
assert ne.test("tan(z1)", xr.ufuncs.tan(z1))
assert ne.test("tanh(z1)", xr.ufuncs.tanh(z1))
assert ne.test("trunc(z1)", xr.ufuncs.trunc(z1))
assert ne.test("arctan2(z1, z2)", xr.ufuncs.arctan2(z1, z2))
assert ne.test("copysign(z1, z2)", xr.ufuncs.copysign(z1, z2))
assert ne.test("fmax(z1, z2)", xr.ufuncs.fmax(z1, z2))
assert ne.test("fmin(z1, z2)", xr.ufuncs.fmin(z1, z2))
assert ne.test("fmod(z1, z2)", xr.ufuncs.fmod(z1, z2))
assert ne.test("hypot(z1, z2)", xr.ufuncs.hypot(z1, z2))
assert ne.test("ldexp(z1, z2)", xr.DataArray(xr.ufuncs.ldexp(z1, z2)))
assert ne.test("logaddexp(z1, z2)", xr.ufuncs.logaddexp(z1, z2))
assert ne.test("logaddexp2(z1, z2)", xr.ufuncs.logaddexp2(z1, z2))
assert ne.test("logicaland(z1, z2)", xr.ufuncs.logical_and(z1, z2))
assert ne.test("logicalnot(z1, z2)", xr.ufuncs.logical_not(z1, z2))
assert ne.test("logicalor(z1, z2)", xr.ufuncs.logical_or(z1, z2))
assert ne.test("logicalxor(z1, z2)", xr.ufuncs.logical_xor(z1, z2))
assert ne.test("maximum(z1, z2)", xr.ufuncs.maximum(z1, z2))
assert ne.test("minimum(z1, z2)", xr.ufuncs.minimum(z1, z2))
assert ne.test("nextafter(z1, z2)", xr.ufuncs.nextafter(z1, z2))
assert ne.test("all(z1)", xr.DataArray.all(z1))
assert ne.test("all(z1, 0)", xr.DataArray.all(z1, axis=0))
assert ne.test("all(z1, 0, 1)", xr.DataArray.all(z1, axis=(0, 1)))
assert ne.test("all(z1, 0, 1, 2)", xr.DataArray.all(z1, axis=(0, 1, 2)))
assert ne.test("any(z1)", xr.DataArray.any(z1))
assert ne.test("any(z1, 0)", xr.DataArray.any(z1, axis=0))
assert ne.test("any(z1, 0, 1)", xr.DataArray.any(z1, axis=(0, 1)))
assert ne.test("any(z1, 0, 1, 2)", xr.DataArray.any(z1, axis=(0, 1, 2)))
assert ne.test("argmax(z1)", xr.DataArray.argmax(z1))
assert ne.test("argmax(z1, 0)", xr.DataArray.argmax(z1, axis=0))
assert ne.test("argmax(z1, 1)", xr.DataArray.argmax(z1, axis=1))
assert ne.test("argmax(z1, 2)", xr.DataArray.argmax(z1, axis=2))
assert ne.test("argmin(z1)", xr.DataArray.argmin(z1))
assert ne.test("argmin(z1, 0)", xr.DataArray.argmin(z1, axis=0))
assert ne.test("argmin(z1, 1)", xr.DataArray.argmin(z1, axis=1))
assert ne.test("argmin(z1, 2)", xr.DataArray.argmin(z1, axis=2))
assert ne.test("max(z1)", xr.DataArray.max(z1))
assert ne.test("max(z1, 0)", xr.DataArray.max(z1, axis=0))
assert ne.test("max(z1, 0, 1)", xr.DataArray.max(z1, axis=(0, 1)))
assert ne.test("max(z1, 0, 1, 2)", xr.DataArray.max(z1, axis=(0, 1, 2)))
assert ne.test("mean(z1)", xr.DataArray.mean(z1))
assert ne.test("mean(z1, 0)", xr.DataArray.mean(z1, axis=0))
assert ne.test("mean(z1, 0, 1)", xr.DataArray.mean(z1, axis=(0, 1)))
assert ne.test("mean(z1, 0, 1, 2)", xr.DataArray.mean(z1, axis=(0, 1, 2)))
assert ne.test("median(z1)", xr.DataArray.median(z1))
assert ne.test("median(z1, 0)", xr.DataArray.median(z1, axis=0))
assert ne.test("median(z1, 0, 1)", xr.DataArray.median(z1, axis=(0, 1)))
assert ne.test("median(z1, 0, 1, 2)",
xr.DataArray.median(z1, axis=(0, 1, 2)))
assert ne.test("min(z1)", xr.DataArray.min(z1))
assert ne.test("min(z1, 0)", xr.DataArray.min(z1, axis=0))
assert ne.test("min(z1, 0, 1)", xr.DataArray.min(z1, axis=(0, 1)))
assert ne.test("min(z1, 0, 1, 2)", xr.DataArray.min(z1, axis=(0, 1, 2)))
assert ne.test("prod(z1)", xr.DataArray.prod(z1))
assert ne.test("prod(z1, 0)", xr.DataArray.prod(z1, axis=0))
assert ne.test("prod(z1, 0, 1)", xr.DataArray.prod(z1, axis=(0, 1)))
assert ne.test("prod(z1, 0, 1, 2)", xr.DataArray.prod(z1, axis=(0, 1, 2)))
assert ne.test("sum(z1)", xr.DataArray.sum(z1))
assert ne.test("sum(z1, 0)", xr.DataArray.sum(z1, axis=0))
assert ne.test("sum(z1, 0, 1)", xr.DataArray.sum(z1, axis=(0, 1)))
assert ne.test("sum(z1, 0, 1, 2)", xr.DataArray.sum(z1, axis=(0, 1, 2)))
assert ne.test("std(z1)", xr.DataArray.std(z1))
assert ne.test("std(z1, 0)", xr.DataArray.std(z1, axis=0))
assert ne.test("std(z1, 0, 1)", xr.DataArray.std(z1, axis=(0, 1)))
assert ne.test("std(z1, 0, 1, 2)", xr.DataArray.std(z1, axis=(0, 1, 2)))
assert ne.test("var(z1)", xr.DataArray.var(z1))
assert ne.test("var(z1, 0)", xr.DataArray.var(z1, axis=0))
assert ne.test("var(z1, 0, 1)", xr.DataArray.var(z1, axis=(0, 1)))
assert ne.test("var(z1, 0, 1, 2)", xr.DataArray.var(z1, axis=(0, 1, 2)))
assert ne.test("percentile(z1, 50)", np.percentile(z1, 50))
assert ne.test("percentile(z1, 50)+percentile(z1, 50)",
np.percentile(z1, 50) + np.percentile(z1, 50))
assert ne.test("percentile(z1, (50))", np.percentile(z1, (50)))
assert ne.test("percentile(z1, (50, 60))", np.percentile(z1, (50, 60)))
assert ne.test("percentile(z1, (50, 60, 70))",
np.percentile(z1, (50, 60, 70)))
assert ne.test(
"percentile(z1, (50, 60, 70)) + percentile(z1, (50, 60, 70))",
np.percentile(z1, (50, 60, 70)) + np.percentile(z1, (50, 60, 70)))
assert ne.test("1 + var(z1, 0, 0+1, 2) + 1",
1 + xr.DataArray.var(z1, axis=(0, 0 + 1, 2)) + 1)
assert ne.test("1 + ((z1+z1)*0.0005)**2 + 1", 1 + np.power(
(z1 + z1) * 0.0005, 2) + 1)
assert ne.test("z1{mask1}", xr.DataArray.where(z1, mask1))
assert ne.test("z1{z1>2}", xr.DataArray.where(z1, z1 > 2))
assert ne.test("z1{z1>=2}", xr.DataArray.where(z1, z1 >= 2))
assert ne.test("z1{z1<2}", xr.DataArray.where(z1, z1 < 2))
assert ne.test("z1{z1<=2}", xr.DataArray.where(z1, z1 <= 2))
assert ne.test("z1{z1==2}", xr.DataArray.where(z1, z1 == 2))
assert ne.test("z1{z1!=2}", xr.DataArray.where(z1, z1 != 2))
assert ne.test("z1{z1<2 | z1>5}",
xr.DataArray.where(z1, (z1 < 2) | (z1 > 5)))
assert ne.test("z1{z1>2 & z1<5}",
xr.DataArray.where(z1, (z1 > 2) & (z1 < 5)))
ne.evaluate("m = z1+1")
assert ne.test("m", z1 + 1)
assert ne.test("z1{~mask1}", xr.DataArray.where(z1, ~mask1))
assert ne.test("(1<0?1+1;2+2)", 4)
assert ne.test("(0<1?1+1;2+2)", 2)
assert ne.test("z1+mask1", xr.DataArray.where(z1, mask1))
assert ne.is_number(1) is True
assert ne.test("-z1", -z1)
assert ne.test("z1{!(z1<2)}",
xr.DataArray.where(z1, xr.ufuncs.logical_not(z1 < 2)))
assert ne.test("z1>>1", np.right_shift(z1, 1))
assert ne.test("z1<<1", np.left_shift(z1, 1))
def test_2():
# perform language test
ne = NDexpr()
ne.set_ae(False)
x1 = xr.DataArray(np.random.randn(2, 3))
y1 = xr.DataArray(np.random.randn(2, 3))
z1 = xr.DataArray(np.array([[[0, 1, 2], [3, 4, 5], [6, 7, 8]],
[[9, 10, 11], [12, 13, 14], [15, 16, 17]],
[[18, 19, 20], [21, 22, 23], [24, 25, 26]]
]))
z2 = z1*2
z3 = np.arange(27)
mask1 = z1 > 4
assert ne.test("arccos(z1)", xr.ufuncs.arccos(z1))
assert ne.test("angle(z1)", xr.ufuncs.angle(z1))
assert ne.test("arccos(z1)", xr.ufuncs.arccos(z1))
assert ne.test("arccosh(z1)", xr.ufuncs.arccosh(z1))
assert ne.test("arcsin(z1)", xr.ufuncs.arcsin(z1))
assert ne.test("arcsinh(z1)", xr.ufuncs.arcsinh(z1))
assert ne.test("arctan(z1)", xr.ufuncs.arctan(z1))
assert ne.test("arctanh(z1)", xr.ufuncs.arctanh(z1))
assert ne.test("ceil(z1)", xr.ufuncs.ceil(z1))
assert ne.test("conj(z1)", xr.ufuncs.conj(z1))
assert ne.test("cos(z1)", xr.ufuncs.cos(z1))
assert ne.test("cosh(z1)", xr.ufuncs.cosh(z1))
assert ne.test("deg2rad(z1)", xr.ufuncs.deg2rad(z1))
assert ne.test("degrees(z1)", xr.ufuncs.degrees(z1))
assert ne.test("exp(z1)", xr.ufuncs.exp(z1))
assert ne.test("expm1(z1)", xr.ufuncs.expm1(z1))
assert ne.test("fabs(z1)", xr.ufuncs.fabs(z1))
assert ne.test("fix(z1)", xr.ufuncs.fix(z1))
assert ne.test("floor(z1)", xr.ufuncs.floor(z1))
assert ne.test("frexp(z3)", xr.DataArray(xr.ufuncs.frexp(z3)))
assert ne.test("imag(z1)", xr.ufuncs.imag(z1))
assert ne.test("iscomplex(z1)", xr.ufuncs.iscomplex(z1))
assert ne.test("isfinite(z1)", xr.ufuncs.isfinite(z1))
assert ne.test("isinf(z1)", xr.ufuncs.isinf(z1))
assert ne.test("isnan(z1)", xr.ufuncs.isnan(z1))
assert ne.test("isreal(z1)", xr.ufuncs.isreal(z1))
assert ne.test("log(z1)", xr.ufuncs.log(z1))
assert ne.test("log10(z1)", xr.ufuncs.log10(z1))
assert ne.test("log1p(z1)", xr.ufuncs.log1p(z1))
assert ne.test("log2(z1)", xr.ufuncs.log2(z1))
assert ne.test("rad2deg(z1)", xr.ufuncs.rad2deg(z1))
assert ne.test("radians(z1)", xr.ufuncs.radians(z1))
assert ne.test("real(z1)", xr.ufuncs.real(z1))
assert ne.test("rint(z1)", xr.ufuncs.rint(z1))
assert ne.test("sign(z1)", xr.ufuncs.sign(z1))
assert ne.test("signbit(z1)", xr.ufuncs.signbit(z1))
assert ne.test("sin(z1)", xr.ufuncs.sin(z1))
assert ne.test("sinh(z1)", xr.ufuncs.sinh(z1))
assert ne.test("sqrt(z1)", xr.ufuncs.sqrt(z1))
assert ne.test("square(z1)", xr.ufuncs.square(z1))
assert ne.test("tan(z1)", xr.ufuncs.tan(z1))
assert ne.test("tanh(z1)", xr.ufuncs.tanh(z1))
assert ne.test("trunc(z1)", xr.ufuncs.trunc(z1))
assert ne.test("arctan2(z1, z2)", xr.ufuncs.arctan2(z1, z2))
assert ne.test("copysign(z1, z2)", xr.ufuncs.copysign(z1, z2))
assert ne.test("fmax(z1, z2)", xr.ufuncs.fmax(z1, z2))
assert ne.test("fmin(z1, z2)", xr.ufuncs.fmin(z1, z2))
assert ne.test("fmod(z1, z2)", xr.ufuncs.fmod(z1, z2))
assert ne.test("hypot(z1, z2)", xr.ufuncs.hypot(z1, z2))
assert ne.test("ldexp(z1, z2)", xr.DataArray(xr.ufuncs.ldexp(z1, z2)))
assert ne.test("logaddexp(z1, z2)", xr.ufuncs.logaddexp(z1, z2))
assert ne.test("logaddexp2(z1, z2)", xr.ufuncs.logaddexp2(z1, z2))
assert ne.test("logicaland(z1, z2)", xr.ufuncs.logical_and(z1, z2))
assert ne.test("logicalnot(z1, z2)", xr.ufuncs.logical_not(z1, z2))
assert ne.test("logicalor(z1, z2)", xr.ufuncs.logical_or(z1, z2))
assert ne.test("logicalxor(z1, z2)", xr.ufuncs.logical_xor(z1, z2))
assert ne.test("maximum(z1, z2)", xr.ufuncs.maximum(z1, z2))
assert ne.test("minimum(z1, z2)", xr.ufuncs.minimum(z1, z2))
assert ne.test("nextafter(z1, z2)", xr.ufuncs.nextafter(z1, z2))
assert ne.test("all(z1)", xr.DataArray.all(z1))
assert ne.test("all(z1, 0)", xr.DataArray.all(z1, axis=0))
assert ne.test("all(z1, 0, 1)", xr.DataArray.all(z1, axis=(0, 1)))
assert ne.test("all(z1, 0, 1, 2)", xr.DataArray.all(z1, axis=(0, 1, 2)))
assert ne.test("any(z1)", xr.DataArray.any(z1))
assert ne.test("any(z1, 0)", xr.DataArray.any(z1, axis=0))
assert ne.test("any(z1, 0, 1)", xr.DataArray.any(z1, axis=(0, 1)))
assert ne.test("any(z1, 0, 1, 2)", xr.DataArray.any(z1, axis=(0, 1, 2)))
assert ne.test("argmax(z1)", xr.DataArray.argmax(z1))
assert ne.test("argmax(z1, 0)", xr.DataArray.argmax(z1, axis=0))
assert ne.test("argmax(z1, 1)", xr.DataArray.argmax(z1, axis=1))
assert ne.test("argmax(z1, 2)", xr.DataArray.argmax(z1, axis=2))
assert ne.test("argmin(z1)", xr.DataArray.argmin(z1))
assert ne.test("argmin(z1, 0)", xr.DataArray.argmin(z1, axis=0))
assert ne.test("argmin(z1, 1)", xr.DataArray.argmin(z1, axis=1))
assert ne.test("argmin(z1, 2)", xr.DataArray.argmin(z1, axis=2))
assert ne.test("max(z1)", xr.DataArray.max(z1))
assert ne.test("max(z1, 0)", xr.DataArray.max(z1, axis=0))
assert ne.test("max(z1, 0, 1)", xr.DataArray.max(z1, axis=(0, 1)))
assert ne.test("max(z1, 0, 1, 2)", xr.DataArray.max(z1, axis=(0, 1, 2)))
assert ne.test("mean(z1)", xr.DataArray.mean(z1))
assert ne.test("mean(z1, 0)", xr.DataArray.mean(z1, axis=0))
assert ne.test("mean(z1, 0, 1)", xr.DataArray.mean(z1, axis=(0, 1)))
assert ne.test("mean(z1, 0, 1, 2)", xr.DataArray.mean(z1, axis=(0, 1, 2)))
assert ne.test("median(z1)", xr.DataArray.median(z1))
assert ne.test("median(z1, 0)", xr.DataArray.median(z1, axis=0))
assert ne.test("median(z1, 0, 1)", xr.DataArray.median(z1, axis=(0, 1)))
assert ne.test("median(z1, 0, 1, 2)", xr.DataArray.median(z1, axis=(0, 1, 2)))
assert ne.test("min(z1)", xr.DataArray.min(z1))
assert ne.test("min(z1, 0)", xr.DataArray.min(z1, axis=0))
assert ne.test("min(z1, 0, 1)", xr.DataArray.min(z1, axis=(0, 1)))
assert ne.test("min(z1, 0, 1, 2)", xr.DataArray.min(z1, axis=(0, 1, 2)))
assert ne.test("prod(z1)", xr.DataArray.prod(z1))
assert ne.test("prod(z1, 0)", xr.DataArray.prod(z1, axis=0))
assert ne.test("prod(z1, 0, 1)", xr.DataArray.prod(z1, axis=(0, 1)))
assert ne.test("prod(z1, 0, 1, 2)", xr.DataArray.prod(z1, axis=(0, 1, 2)))
assert ne.test("sum(z1)", xr.DataArray.sum(z1))
assert ne.test("sum(z1, 0)", xr.DataArray.sum(z1, axis=0))
assert ne.test("sum(z1, 0, 1)", xr.DataArray.sum(z1, axis=(0, 1)))
assert ne.test("sum(z1, 0, 1, 2)", xr.DataArray.sum(z1, axis=(0, 1, 2)))
assert ne.test("std(z1)", xr.DataArray.std(z1))
assert ne.test("std(z1, 0)", xr.DataArray.std(z1, axis=0))
assert ne.test("std(z1, 0, 1)", xr.DataArray.std(z1, axis=(0, 1)))
assert ne.test("std(z1, 0, 1, 2)", xr.DataArray.std(z1, axis=(0, 1, 2)))
assert ne.test("var(z1)", xr.DataArray.var(z1))
assert ne.test("var(z1, 0)", xr.DataArray.var(z1, axis=0))
assert ne.test("var(z1, 0, 1)", xr.DataArray.var(z1, axis=(0, 1)))
assert ne.test("var(z1, 0, 1, 2)", xr.DataArray.var(z1, axis=(0, 1, 2)))
assert ne.test("percentile(z1, 50)", np.percentile(z1, 50))
assert ne.test("percentile(z1, 50)+percentile(z1, 50)", np.percentile(z1, 50) + np.percentile(z1, 50))
assert ne.test("percentile(z1, (50))", np.percentile(z1, (50)))
assert ne.test("percentile(z1, (50, 60))", np.percentile(z1, (50, 60)))
assert ne.test("percentile(z1, (50, 60, 70))", np.percentile(z1, (50, 60, 70)))
assert ne.test("percentile(z1, (50, 60, 70)) + percentile(z1, (50, 60, 70))",
np.percentile(z1, (50, 60, 70)) + np.percentile(z1, (50, 60, 70)))
assert ne.test("1 + var(z1, 0, 0+1, 2) + 1", 1+xr.DataArray.var(z1, axis=(0, 0+1, 2))+1)
assert ne.test("z1{mask1}", xr.DataArray.where(z1, mask1))
assert ne.test("z1{z1>2}", xr.DataArray.where(z1, z1 > 2))
assert ne.test("z1{z1>=2}", xr.DataArray.where(z1, z1 >= 2))
assert ne.test("z1{z1<2}", xr.DataArray.where(z1, z1 < 2))
assert ne.test("z1{z1<=2}", xr.DataArray.where(z1, z1 <= 2))
assert ne.test("z1{z1==2}", xr.DataArray.where(z1, z1 == 2))
assert ne.test("z1{z1!=2}", xr.DataArray.where(z1, z1 != 2))
assert ne.test("z1{z1<2 | z1>5}", xr.DataArray.where(z1, (z1 < 2) | (z1 > 5)))
assert ne.test("z1{z1>2 & z1<5}", xr.DataArray.where(z1, (z1 > 2) & (z1 < 5)))
ne.evaluate("m = z1+1")
assert ne.test("m", z1+1)
assert ne.test("z1{~mask1}", xr.DataArray.where(z1, ~mask1))
assert ne.test("(1<0?1+1;2+2)", 4)
assert ne.test("(0<1?1+1;2+2)", 2)
assert ne.test("z1+mask1", xr.DataArray.where(z1, mask1))
assert ne.is_number(1) is True
assert ne.test("-z1", -z1)
assert ne.test("z1{!(z1<2)}", xr.DataArray.where(z1, xr.ufuncs.logical_not(z1 < 2)))