def __init__(self, value, gridData, gridName='grid'):
# if grid data is a dictionary, first convert to datacontainer
if isinstance(gridData, dict):
from src.DataContainer import DataContainer
dc = DataContainer()
dc.addData(gridName, gridData)
gridData = dc
del dc
# check if a grid is provided in DC data
dimensions = gridData.v(gridName, 'dimensions')
dimNames = [dimensions[i] for i in range(0, len(value.shape))]
NumericalFunctionBase.__init__(self, dimNames)
# add grid (with name 'grid') and value
self.addGrid(gridData)
self.addValue(value)
return
class NumericalFunctionBase(FunctionBase):
#Variables
#Methods
def __init__(self, dimNames):
FunctionBase.__init__(self, dimNames)
self.dataContainer = DataContainer()
self.valueSize = 0
return
def function(self, **kwargs):
if len([
i for i in kwargs.keys()
if i in self.dataContainer.v('grid', 'dimensions')
]) == 0:
return self.__setReturnReference(kwargs.get('operation'))
# evaluate function
try:
returnval = self.__evaluateFunction(**kwargs)
except FunctionEvaluationError:
returnval = self.__setReturnReference(kwargs.get('operation'))
return returnval
def __evaluateFunction(self, **kwargs):
"""Overrides the function method of FunctionBase, but is very similar.
The difference is only that FunctionBase transfers kwargs to args before calling the actual functions
Here we keep the kwargs as the actual functions also use this.
"""
requestSize = sum(
[dim in kwargs for dim in self.dimNames]
) # count the number of dimensions in kwargs (this makes sure that other parameters or irrelevant dimensions are ignored)
operation = kwargs.get('operation')
try:
kwargs.pop('operation')
except:
pass
# direct to actual function
if operation is None:
returnval = self.value(**kwargs)
elif operation == 'd':
returnval = self.derivative(**kwargs)
elif operation == 'n':
returnval = -self.value(**kwargs)
elif operation == 'dn':
returnval = -self.derivative(**kwargs)
else:
raise FunctionEvaluationError
return returnval
def __setReturnReference(self, operation=None):
'''No difference with FunctionBase, but required here to refer to its own functions
'''
if not operation:
returnval = self.function
elif operation == 'n':
returnval = self.negfunction
elif operation == 'd':
returnval = self.derfunction
elif operation == 'dn':
returnval = self.dnfunction
else:
raise KnownError(
'Function called with unknown operation. (This error indicates an incorrectly defined function)'
)
return returnval
def negfunction(self, **kwargs):
"""No difference with FunctionBase, but required here to refer to its own functions
"""
# reset operations
if kwargs.get(
'operation'
) == 'n': # if the negative of a negfunction is called, return to function.
kwargs.pop('operation')
elif kwargs.get('operation') == 'd':
kwargs['operation'] = 'dn'
else:
kwargs['operation'] = 'n'
# evaluate
returnval = self.function(**kwargs)
return returnval
def addGrid(self, gridData, gridName='grid'):
# set own DataContainer containing grid data
data = gridData.slice(gridName, excludeKey=True)
self.dataContainer.addData(
'grid', data.data
) # improper use of the DataContainer by accessing its data directly
return
def addValue(self, value):
"""Add a variable 'value' to the numerical function
Parameters:
value (ndarray) - value to be put in the internal DataContainer
"""
self.dataContainer.addData('value', value)
self.valueSize = len(value.shape)
return
def addDerivative(self, derivative, dim):
self.dataContainer.merge({'derivative': {dim: derivative}})
return
### Depreciated v2.2 [dep01] ###
def addSecondDerivative(self, derivative, dim):
""" Depreciated v2.2
"""
self.dataContainer.merge({'secondDerivative': {dim: derivative}})
return
### End ###
def value(self, **kwargs):
"""Return the value of the variable in this numerical function.
Parameters:
kwargs (dict) - coordinates
Returns:
Array value using DataContainer interpolation.
"""
return self.dataContainer.v('value', **kwargs)
def derivative(self, **kwargs):
"""Similar to .value(). Returns the derivative uploaded to the numerical function or makes a call
to a numerical derivation method if no derivative is uploaded.
"""
# kwargs.pop('operation') # obsolete
dim = kwargs.get('dim')
v = self.dataContainer.v(
'derivative', dim,
**kwargs) # try if analytical derivative is available
if v is None:
v = self.dataContainer.d(
'value', **kwargs) # else take numerical derivative
return v
### Depreciated v2.2 [dep01] ###
def secondDerivative(self, **kwargs):
"""See .derivative(). This method does the same for the second derivative
Depreciated 2.2 [dep01]
"""
# kwargs.pop('operation') # obsolete
dim = kwargs.get('dim')
v = self.dataContainer.v('secondDerivative', dim, **kwargs)
if v is None:
v = self.dataContainer.dd('value', **kwargs)
return v
def run(self):
"""invoke the saveData() method to save by using Pickle.
Returns:
Empty dictionary.
"""
self.logger.info('Saving output')
################################################################################################################
# Make all variables from config, input and modules available (note, config not available in output module, see Program.py)
################################################################################################################
# read input file
reader = Reader()
reader.open(self.input.v('inputFile'))
data = reader.read('module')
reader.close()
# merge the datacontainers of all modules & make the module tags into a list of modules
inputvars = DataContainer()
module = []
for d in data:
module.append(d.v('module'))
inputvars.merge(d)
inputvars.addData('module', module)
# merge input vars with self.input in hierarchy config, input, input for this module, vars calculated in other modules (low - high)
# + make a list of all keys of input and config vars; these are saved always and later appended by selected module calc. vars.
data = self.__loadConfig()
data.merge(inputvars)
outputKeys = self.__checkInputOverrides(
data
) # checks if input is overwritten and provides the keys of not or correctly overwritten input vars
data.merge(self.input)
del inputvars, reader
# now all variables from config, input and modules are in 'data'
################################################################################################################
# Isolate part of DC to write; put this in saveData
################################################################################################################
saveData = DataContainer()
# vars to save
outputVariables = toList(self.input.v('requirements'))
outputKeys = list(
set(outputKeys + self.__getSubmoduleRequirements(outputVariables))
) # convert the requested output variables to key tuples including submodule requirements
for key in outputKeys:
if len(key) > 1:
saveData.merge({key[0]: data.slice(*key).data})
else:
saveData.merge(data.slice(*key))
# add grid and outputgrid if available; needed for interpolating data to outputgrid later
saveData.merge(self.input.slice('grid'))
saveData.merge(self.input.slice(self.outputgridName))
# add reference level to outputgrid
if self.input.v('R') is not None:
self.input.merge({
self.outputgridName: {
'low': {
'z':
self.input.v('R',
x=self.input.v(self.outputgridName,
'axis', 'x'))
}
}
}) # add reference level to outputgrid
# make a deepcopy of the data to be saved
# NB. very memory inefficient, but needed not to overwrite existing data
saveData = deepcopy(saveData)
################################################################################################################
# Convert data using output grid (if this is provided)
################################################################################################################
grid = saveData.slice('grid')
outputgrid = saveData.slice(self.outputgridName)
saveAnalytical = toList(self.input.v('saveAnalytical')) or []
dontConvert = toList(self.input.v('dontConvert')) or []
if 'all' in saveAnalytical:
saveAnalytical = outputVariables
if 'all' in dontConvert:
dontConvert = outputVariables
self._convertData(saveData, grid, outputgrid, saveAnalytical,
dontConvert)
# rename the outputgrid to grid and replace the original grid in saveData
saveData.addData('grid', saveData.data[self.outputgridName])
saveData.data.pop(self.outputgridName)
################################################################################################################
# Make the output directory if it doesnt exist
################################################################################################################
cwdpath = cfm.CWD # path to working directory
self.path = os.path.join(cwdpath, self.input.v('path'))
if not self.path[-1] == '/':
self.path += '/'
if not os.path.exists(self.path):
os.makedirs(self.path)
################################################################################################################
# set file name and save data
################################################################################################################
filename = self.__makeFileName()
# write
filepath = (self.path + filename + self.ext)
try:
with open(filepath, 'wb') as fp:
pickle.dump(saveData.data,
fp,
protocol=pickle.HIGHEST_PROTOCOL)
except:
raise
################################################################################################################
# return
################################################################################################################
d = {}
d['outputDirectory'] = self.path
return d