from sans.models.BaseComponent import BaseComponent import numpy, math import copy from sans.models.pluginmodel import Model1DPlugin class MultiplicationModel(BaseComponent): """ Use for P(Q)*S(Q); function call must be in the order of P(Q) and then S(Q): The model parameters are combined from both models, P(Q) and S(Q), except 1) 'effect_radius' of S(Q) which will be calculated from P(Q) via calculate_ER(), and 2) 'scale' in P model which is synchronized w/ volfraction in S then P*S is multiplied by a new param, 'scale_factor'. The polydispersion is applicable only to P(Q), not to S(Q). Note: P(Q) refers to 'form factor' model while S(Q) does to 'structure factor'. """ def __init__(self, p_model, s_model ): BaseComponent.__init__(self) """ @param p_model: form factor, P(Q) @param s_model: structure factor, S(Q) """ ## Setting model name model description self.description="" self.name = p_model.name +" * "+ s_model.name self.description= self.name+"\n" self.fill_description(p_model, s_model) ## Define parameters self.params = {} ## Parameter details [units, min, max] self.details = {} # non-fittable parameters self.non_fittable = p_model.non_fittable ##models self.p_model= p_model self.s_model= s_model ## dispersion self._set_dispersion() ## Define parameters self._set_params() ## New parameter:Scaling factor self.params['scale_factor'] = 1 ## Parameter details [units, min, max] self._set_details() self.details['scale_factor'] = ['', None, None] #list of parameter that can be fitted self._set_fixed_params() ## parameters with orientation for item in self.p_model.orientation_params: self.orientation_params.append(item) for item in self.s_model.orientation_params: if not item in self.orientation_params: self.orientation_params.append(item) # get multiplicity if model provide it, else 1. try: multiplicity = p_model.multiplicity except: multiplicity = 1 ## functional multiplicity of the model self.multiplicity = multiplicity def _clone(self, obj): """ Internal utility function to copy the internal data members to a fresh copy. """ obj.params = copy.deepcopy(self.params) obj.description = copy.deepcopy(self.description) obj.details = copy.deepcopy(self.details) obj.dispersion = copy.deepcopy(self.dispersion) obj.p_model = self.p_model.clone() obj.s_model = self.s_model.clone() #obj = copy.deepcopy(self) return obj def _set_dispersion(self): """ combined the two models dispersions Polydispersion should not be applied to s_model """ ##set dispersion only from p_model for name , value in self.p_model.dispersion.iteritems(): self.dispersion[name]= value def getProfile(self): """ Get SLD profile of p_model if exists : return: (r, beta) where r is a list of radius of the transition points beta is a list of the corresponding SLD values : Note: This works only for func_shell# = 2 (exp function). """ try: x,y = self.p_model.getProfile() except: x = None y = None return x, y def _set_params(self): """ Concatenate the parameters of the two models to create this model parameters """ for name , value in self.p_model.params.iteritems(): if not name in self.params.keys() and name != 'scale': self.params[name]= value for name , value in self.s_model.params.iteritems(): #Remove the effect_radius from the (P*S) model parameters. if not name in self.params.keys() and name != 'effect_radius': self.params[name]= value # Set "scale and effec_radius to P and S model as initializing # since run P*S comes from P and S separately. self._set_scale_factor() self._set_effect_radius() def _set_details(self): """ Concatenate details of the two models to create this model details """ for name ,detail in self.p_model.details.iteritems(): if name != 'scale': self.details[name]= detail for name , detail in self.s_model.details.iteritems(): if not name in self.details.keys() or name != 'effect_radius': self.details[name]= detail def _set_scale_factor(self): """ Set scale=volfraction to P model """ value = self.params['volfraction'] if value != None: self.p_model.setParam( 'scale', value) def _set_effect_radius(self): """ Set effective radius to S(Q) model """ effective_radius = self.p_model.calculate_ER() #Reset the effective_radius of s_model just before the run if effective_radius != None and effective_radius != NotImplemented: self.s_model.setParam('effect_radius',effective_radius) def setParam(self, name, value): """ Set the value of a model parameter @param name: name of the parameter @param value: value of the parameter """ # set param to P*S model self._setParamHelper( name, value) ## setParam to p model # set 'scale' in P(Q) equal to volfraction if name == 'volfraction': self._set_scale_factor() elif name in self.p_model.getParamList(): self.p_model.setParam( name, value) ## setParam to s model # This is a little bit abundant: Todo: find better way self._set_effect_radius() if name in self.s_model.getParamList(): self.s_model.setParam( name, value) #self._setParamHelper( name, value) def _setParamHelper(self, name, value): """ Helper function to setparam """ # Look for dispersion parameters toks = name.split('.') if len(toks)==2: for item in self.dispersion.keys(): if item.lower()==toks[0].lower(): for par in self.dispersion[item]: if par.lower() == toks[1].lower(): self.dispersion[item][par] = value return else: # Look for standard parameter for item in self.params.keys(): if item.lower()==name.lower(): self.params[item] = value return raise ValueError, "Model does not contain parameter %s" % name def _set_fixed_params(self): """ fill the self.fixed list with the p_model fixed list """ for item in self.p_model.fixed: self.fixed.append(item) self.fixed.sort() def run(self, x = 0.0): """ Evaluate the model @param x: input q-value (float or [float, float] as [r, theta]) @return: (scattering function value) """ # set effective radius and scaling factor before run self._set_effect_radius() self._set_scale_factor() return self.params['scale_factor']*self.p_model.run(x)*self.s_model.run(x) def runXY(self, x = 0.0): """ Evaluate the model @param x: input q-value (float or [float, float] as [qx, qy]) @return: scattering function value """ # set effective radius and scaling factor before run self._set_effect_radius() self._set_scale_factor() return self.params['scale_factor']*self.p_model.runXY(x)* self.s_model.runXY(x) ## Now (May27,10) directly uses the model eval function ## instead of the for-loop in Base Component. def evalDistribution(self, x = []): """ Evaluate the model in cartesian coordinates @param x: input q[], or [qx[], qy[]] @return: scattering function P(q[]) """ # set effective radius and scaling factor before run self._set_effect_radius() self._set_scale_factor() return self.params['scale_factor']*self.p_model.evalDistribution(x)* self.s_model.evalDistribution(x) def set_dispersion(self, parameter, dispersion): """ Set the dispersion object for a model parameter @param parameter: name of the parameter [string] @dispersion: dispersion object of type DispersionModel """ value= None try: if parameter in self.p_model.dispersion.keys(): value= self.p_model.set_dispersion(parameter, dispersion) self._set_dispersion() return value except: raise def fill_description(self, p_model, s_model): """ Fill the description for P(Q)*S(Q) """ description = "" description += "Note:1) The effect_radius (effective radius) of %s \n"% (s_model.name) description +=" is automatically calculated from size parameters (radius...).\n" description += " 2) For non-spherical shape, this approximation is valid \n" description += " only for limited systems. Thus, use it at your own risk.\n" description +="See %s description and %s description \n"%( p_model.name, s_model.name ) description += " for details of individual models." self.description += description