source: sasview/src/sas/sascalc/realspace/VolumeCanvas.py @ 14ecc24e

#!/usr/bin/env python
""" Volume Canvas

    Simulation canvas for real-space simulation of SAS scattering intensity.
    The user can create an arrangement of basic shapes and estimate I(q) and
    I(q_x, q_y). Error estimates on the simulation are also available. 
    
    Example:
    
    import sas.sascalc.realspace.VolumeCanvas as VolumeCanvas
    canvas = VolumeCanvas.VolumeCanvas()
    canvas.setParam('lores_density', 0.01)
    
    sphere = SphereDescriptor()
    handle = canvas.addObject(sphere)

    output, error = canvas.getIqError(q=0.1)
    output, error = canvas.getIq2DError(0.1, 0.1)
    
    or alternatively:
    iq = canvas.run(0.1)
    i2_2D = canvas.run([0.1, 1.57])
    
"""

from sas.models.BaseComponent import BaseComponent
from sas.sascalc.simulation.pointsmodelpy import pointsmodelpy
from sas.sascalc.simulation.geoshapespy import geoshapespy


import os.path, math

class ShapeDescriptor:
    """
        Class to hold the information about a shape
        The descriptor holds a dictionary of parameters.
        
        Note: if shape parameters are accessed directly
        from outside VolumeCanvas. The getPr method
        should be called before evaluating I(q).
                
    """
    def __init__(self):
        """
            Initialization
        """
        ## Real space object
        self.shapeObject = None
        ## Parameters of the object
        self.params = {}
        self.params["center"] = [0, 0, 0]
        # Orientation are angular offsets in degrees with respect to X, Y, Z
        self.params["orientation"] = [0, 0, 0]
        # Default to lores shape
        self.params['is_lores'] = True
        self.params['order'] = 0
            
    def create(self):
        """
            Create an instance of the shape
        """
        # Set center
        x0 = self.params["center"][0]
        y0 = self.params["center"][1]
        z0 = self.params["center"][2]
        geoshapespy.set_center(self.shapeObject, x0, y0, z0)
        
        # Set orientation
        x0 = self.params["orientation"][0]
        y0 = self.params["orientation"][1]
        z0 = self.params["orientation"][2]
        geoshapespy.set_orientation(self.shapeObject, x0, y0, z0)
               
class SphereDescriptor(ShapeDescriptor):
    """
        Descriptor for a sphere
        
        The parameters are:
            - radius [Angstroem] [default = 20 A]
            - Contrast [A-2] [default = 1 A-2]
            
    """
    def __init__(self):
        """
            Initialization
        """ 
        ShapeDescriptor.__init__(self)
        # Default parameters
        self.params["type"]   = "sphere"
        # Radius of the sphere
        self.params["radius"] = 20.0
        # Constrast parameter
        self.params["contrast"] = 1.0

    def create(self):
        """
            Create an instance of the shape
            @return: instance of the shape
        """
        self.shapeObject = geoshapespy.new_sphere(\
            self.params["radius"])
        
        ShapeDescriptor.create(self)   
        return self.shapeObject
    
class CylinderDescriptor(ShapeDescriptor):
    """
        Descriptor for a cylinder
        Orientation: Default cylinder is along Y
        
        Parameters:
            - Length [default = 40 A]
            - Radius [default = 10 A]
            - Contrast [default = 1 A-2]
    """
    def __init__(self):
        """
            Initialization
        """ 
        ShapeDescriptor.__init__(self)
        # Default parameters
        self.params["type"]   = "cylinder"
        # Length of the cylinder
        self.params["length"] = 40.0
        # Radius of the cylinder
        self.params["radius"] = 10.0
        # Constrast parameter
        self.params["contrast"] = 1.0
        
    def create(self):
        """
            Create an instance of the shape
            @return: instance of the shape
        """
        self.shapeObject = geoshapespy.new_cylinder(\
            self.params["radius"], self.params["length"])

        ShapeDescriptor.create(self)
        return self.shapeObject
        

class EllipsoidDescriptor(ShapeDescriptor):
    """
        Descriptor for an ellipsoid
        
        Parameters:
            - Radius_x along the x-axis [default = 30 A]
            - Radius_y along the y-axis [default = 20 A]
            - Radius_z along the z-axis [default = 10 A]
            - contrast [default = 1 A-2]
    """
    def __init__(self):
        """
            Initialization
        """ 
        ShapeDescriptor.__init__(self)
        # Default parameters
        self.params["type"]   = "ellipsoid"
        self.params["radius_x"] = 30.0
        self.params["radius_y"] = 20.0
        self.params["radius_z"] = 10.0
        self.params["contrast"] = 1.0
        
    def create(self):
        """
            Create an instance of the shape
            @return: instance of the shape
        """
        self.shapeObject = geoshapespy.new_ellipsoid(\
            self.params["radius_x"], self.params["radius_y"], 
            self.params["radius_z"])
        
        ShapeDescriptor.create(self)   
        return self.shapeObject
        
class HelixDescriptor(ShapeDescriptor):
    """
        Descriptor for an helix
        
        Parameters:
            -radius_helix: the radius of the helix [default = 10 A]
            -radius_tube: radius of the "tube" that forms the helix [default = 3 A]
            -pitch: distance between two consecutive turns of the helix [default = 34 A]
            -turns: number of turns of the helix [default = 3]
            -contrast: contrast parameter [default = 1 A-2]
    """
    def __init__(self):
        """
            Initialization
        """ 
        ShapeDescriptor.__init__(self)
        # Default parameters
        self.params["type"]   = "singlehelix"
        self.params["radius_helix"] = 10.0
        self.params["radius_tube"] = 3.0
        self.params["pitch"] = 34.0
        self.params["turns"] = 3.0
        self.params["contrast"] = 1.0

    def create(self):
        """
            Create an instance of the shape
            @return: instance of the shape
        """
        self.shapeObject = geoshapespy.new_singlehelix(\
            self.params["radius_helix"], self.params["radius_tube"], 
            self.params["pitch"], self.params["turns"])
        
        ShapeDescriptor.create(self)   
        return self.shapeObject
        
class PDBDescriptor(ShapeDescriptor):
    """
        Descriptor for a PDB set of points
        
        Parameter:
            - file = name of the PDB file
    """
    def __init__(self, filename):
        """
            Initialization
            @param filename: name of the PDB file to load
        """ 
        ShapeDescriptor.__init__(self)
        # Default parameters
        self.params["type"]   = "pdb"
        self.params["file"] = filename
        self.params['is_lores'] = False

    def create(self):
        """
            Create an instance of the shape
            @return: instance of the shape
        """
        self.shapeObject = pointsmodelpy.new_pdbmodel()
        pointsmodelpy.pdbmodel_add(self.shapeObject, self.params['file'])       
        
        #ShapeDescriptor.create(self)   
        return self.shapeObject
        
# Define a dictionary for the shape until we find
# a better way to create them
shape_dict = {'sphere':SphereDescriptor,
              'cylinder':CylinderDescriptor,
              'ellipsoid':EllipsoidDescriptor,
              'singlehelix':HelixDescriptor}
        
class VolumeCanvas(BaseComponent):
    """
        Class representing an empty space volume to add 
        geometrical object to.
        
        For 1D I(q) simulation, getPr() is called internally for the
        first call to getIq().
        
    """
    
    def __init__(self):
        """
            Initialization
        """
        BaseComponent.__init__(self)
        
        ## Maximum value of q reachable
        self.params['q_max'] = 0.1
        self.params['lores_density'] = 0.1
        self.params['scale'] = 1.0
        self.params['background'] = 0.0
        
        self.lores_model = pointsmodelpy.new_loresmodel(self.params['lores_density'])
        self.complex_model = pointsmodelpy.new_complexmodel()
        self.shapes = {}
        self.shapecount = 0        
        self.points = None
        self.npts = 0
        self.hasPr = False        
        
    def _model_changed(self):
        """
            Reset internal data members to reflect the fact that the 
            real-space model has changed
        """
        self.hasPr  = False
        self.points = None
        
    def addObject(self, shapeDesc, id = None):
        """
            Adds a real-space object to the canvas.
        
            @param shapeDesc: object to add to the canvas [ShapeDescriptor]
            @param id: string handle for the object [string] [optional]
            @return: string handle for the object
        """
        # If the handle is not provided, create one
        if id is None:
            id = shapeDesc.params["type"]+str(self.shapecount)
         
        # Self the order number
        shapeDesc.params['order'] = self.shapecount
        # Store the shape in a dictionary entry associated
        # with the handle
        self.shapes[id] = shapeDesc
        self.shapecount += 1

        # model changed, need to recalculate P(r)
        self._model_changed()

        return id
            
    
    def add(self, shape, id = None):
        """
            The intend of this method is to eventually be able to use it
            as a factory for the canvas and unify the simulation with the
            analytical solutions. For instance, if one adds a cylinder and
            it is the only shape on the canvas, the analytical solution
            could be called. If multiple shapes are involved, then 
            simulation has to be performed.
            
            This function is deprecated, use addObject().
        
            @param shape: name of the object to add to the canvas [string]
            @param id: string handle for the object [string] [optional]
            @return: string handle for the object
        """
        # If the handle is not provided, create one
        if id is None:
            id = "shape"+str(self.shapecount)
 
        # shapeDesc = ShapeDescriptor(shape.lower())
        if shape.lower() in shape_dict:
            shapeDesc = shape_dict[shape.lower()]()
        elif os.path.isfile(shape):
            # A valid filename was supplier, create a PDB object
            shapeDesc = PDBDescriptor(shape)
        else:
            raise ValueError("VolumeCanvas.add: Unknown shape %s" % shape)
        
        return self.addObject(shapeDesc, id)

    def delete(self, id):
        """
            Delete a shape. The ID for the shape is required.
            @param id: string handle for the object [string] [optional]
        """

        if id in self.shapes:
            del self.shapes[id]
        else:
            raise KeyError("VolumeCanvas.delete: could not find shape ID")

        # model changed, need to recalculate P(r)
        self._model_changed()


    def setParam(self, name, value):    
        """
            Function to set the value of a parameter. 
            Both VolumeCanvas parameters and shape parameters
            are accessible. 
            
            Note: if shape parameters are accessed directly
            from outside VolumeCanvas. The getPr method
            should be called before evaluating I(q).
        
            TODO: implemented a check method to protect
            against that.
        
            @param name: name of the parameter to change
            @param value: value to give the parameter
        """
        
        # Lowercase for case insensitivity
        name = name.lower()
        
        # Look for shape access
        toks = name.split('.')
        
        # If a shape identifier was given, look the shape up
        # in the dictionary
        if len(toks):
            if toks[0] in self.shapes:
                # The shape was found, now look for the parameter
                if toks[1] in self.shapes[toks[0]].params:
                    # The parameter was found, now change it
                    self.shapes[toks[0]].params[toks[1]] = value
                    self._model_changed()
                else:
                    raise ValueError("Could not find parameter %s" % name)
            else:
                raise ValueError("Could not find shape %s" % toks[0])
        
        else:
            # If we are not accessing the parameters of a 
            # shape, see if the parameter is part of this object
            BaseComponent.setParam(self, name, value)
            self._model_changed()

    def getParam(self, name):    
        """
            @param name: name of the parameter to change
        """
        #TODO: clean this up
        
        # Lowercase for case insensitivity
        name = name.lower()
        
        # Look for sub-model access
        toks = name.split('.')
        if len(toks) == 1:
            try:
                value = self.params[toks[0]]
            except KeyError:
                raise ValueError("VolumeCanvas.getParam: Could not find"
                                 " %s" % name)
            if isinstance(value, ShapeDescriptor):
                raise ValueError("VolumeCanvas.getParam: Cannot get parameter"
                                 " value.")
            else:
                return value

        elif len(toks) == 2:
            try:
                shapeinstance = self.shapes[toks[0]]
            except KeyError:
                raise ValueError("VolumeCanvas.getParam: Could not find "
                                 "%s" % name)

            if not toks[1] in shapeinstance.params:
                raise ValueError("VolumeCanvas.getParam: Could not find "
                                 "%s" % name)

            return shapeinstance.params[toks[1]]

        else:
            raise ValueError("VolumeCanvas.getParam: Could not find %s" % name)
            
    def getParamList(self, shapeid=None):
        """
               return a full list of all available parameters from 
           self.params.keys(). If a key in self.params is a instance
           of ShapeDescriptor, extend the return list to: 
           [param1,param2,shapeid.param1,shapeid.param2.......]

           If shapeid is provided, return the list of parameters that
           belongs to that shape id only : [shapeid.param1, shapeid.param2...]
        """

        param_list = []
        if shapeid is None:
            for key1 in self.params:
                #value1 = self.params[key1]
                param_list.append(key1)
            for key2 in self.shapes:
                value2 = self.shapes[key2]
                header = key2 + '.'
                for key3 in value2.params:
                    fullname = header + key3                 
                    param_list.append(fullname)
     
        else:
            if not shapeid in self.shapes:
                raise ValueError("VolumeCanvas: getParamList: Could not find "
                                 "%s" % shapeid)

            header = shapeid + '.'
            param_list = [header + param for param in self.shapes[shapeid].params]
        return param_list

    def getShapeList(self):
        """
            Return a list of the shapes 
        """
        return self.shapes.keys()

    def _addSingleShape(self, shapeDesc):
        """
            create shapeobject based on shapeDesc
            @param shapeDesc: shape description
        """
        # Create the object model
        shapeDesc.create()
                    
        if shapeDesc.params['is_lores']:
            # Add the shape to the lores_model
            pointsmodelpy.lores_add(self.lores_model, 
                shapeDesc.shapeObject, shapeDesc.params['contrast'])  

    def _createVolumeFromList(self):
        """
            Create a new lores model with all the shapes in our internal list
            Whenever we change a parameter of a shape, we have to re-create
            the whole thing.
            
            Items with higher 'order' number take precedence for regions
            of space that are shared with other objects. Points in the 
            overlapping region belonging to objects with lower 'order'
            will be ignored.
            
            Items are added in decreasing 'order' number.
            The item with the highest 'order' will be added *first*.
            [That conventions is prescribed by the realSpaceModeling module]
        """
        
        # Create empty model
        self.lores_model = \
            pointsmodelpy.new_loresmodel(self.params['lores_density'])

        # Create empty complex model
        self.complex_model = pointsmodelpy.new_complexmodel()
       
        # Order the object first
        obj_list = []
    
        for shape in self.shapes:
            order = self.shapes[shape].params['order']
            # find where to place it in the list
            stored = False
            
            for i in range(len(obj_list)):
                if obj_list[i][0] > order:
                    obj_list.insert(i, [order, shape])
                    stored = True
                    break
            
            if not stored:
                obj_list.append([order, shape])
                
        # Add each shape
        len_list = len(obj_list)
        for i in range(len_list-1, -1, -1):
            shapedesc = self.shapes[obj_list[i][1]]
            self._addSingleShape(shapedesc)

        return 0     
    
    def getPr(self):
        """
            Calculate P(r) from the objects on the canvas.
            This method should always be called after the shapes
            on the VolumeCanvas have changed.
            
            @return: calculation output flag 
        """
        # To find a complete example of the correct call order:
        # In LORES2, in actionclass.py, method CalculateAction._get_iq()
        
        # If there are not shapes, do nothing
        if len(self.shapes) == 0:
            self._model_changed()
            return 0
        
        # generate space filling points from shape list
        self._createVolumeFromList()

        self.points = pointsmodelpy.new_point3dvec()

        pointsmodelpy.complexmodel_add(self.complex_model, 
                                        self.lores_model, "LORES")
        for shape in self.shapes:
            if self.shapes[shape].params['is_lores'] == False:
                pointsmodelpy.complexmodel_add(self.complex_model, 
                    self.shapes[shape].shapeObject, "PDB")
        
        #pointsmodelpy.get_lorespoints(self.lores_model, self.points)
        self.npts = pointsmodelpy.get_complexpoints(self.complex_model, self.points)
        
        # expecting the rmax is a positive float or 0. The maximum distance.
        #rmax = pointsmodelpy.get_lores_pr(self.lores_model, self.points)   
         
        rmax = pointsmodelpy.get_complex_pr(self.complex_model, self.points) 
        self.hasPr = True   

        return rmax
        
    def run(self, q = 0):
        """
            Returns the value of I(q) for a given q-value
            @param q: q-value ([float] or [list]) ([A-1] or [[A-1], [rad]])
            @return: I(q) [float] [cm-1]
        """
        # Check for 1D q length
        if q.__class__.__name__ == 'int' \
            or q.__class__.__name__ == 'float':
            return self.getIq(q)
        # Check for 2D q-value
        elif q.__class__.__name__ == 'list':
            # Compute (Qx, Qy) from (Q, phi)
            # Phi is in radian and Q-values are in A-1
            qx = q[0]*math.cos(q[1])
            qy = q[0]*math.sin(q[1])
            return self.getIq2D(qx, qy)
        # Through an exception if it's not a
        # type we recognize
        else:
            raise ValueError("run(q): bad type for q")
    
    def runXY(self, q = 0):
        """
            Standard run command for the canvas.
            Redirects to the correct method 
            according to the input type.
            @param q: q-value [float] or [list] [A-1]
            @return: I(q) [float] [cm-1]
        """
        # Check for 1D q length
        if q.__class__.__name__ == 'int' \
            or q.__class__.__name__ == 'float':
            return self.getIq(q)
        # Check for 2D q-value
        elif q.__class__.__name__ == 'list':
            return self.getIq2D(q[0], q[1])
        # Through an exception if it's not a
        # type we recognize
        else:
            raise ValueError("runXY(q): bad type for q")
    
    def _create_modelObject(self):
        """
            Create the simulation model obejct from the list
            of shapes.
            
            This method needs to be called each time a parameter
            changes because of the way the underlying library
            was (badly) written. It is impossible to change a 
            parameter, or remove a shape without having to 
            refill the space points.
            
            TODO: improve that.
        """
        # To find a complete example of the correct call order:
        # In LORES2, in actionclass.py, method CalculateAction._get_iq()
        
        # If there are not shapes, do nothing
        if len(self.shapes) == 0:
            self._model_changed()
            return 0
        
        # generate space filling points from shape list
        self._createVolumeFromList()

        self.points = pointsmodelpy.new_point3dvec()

        pointsmodelpy.complexmodel_add(self.complex_model, 
                                        self.lores_model, "LORES")
        for shape in self.shapes:
            if self.shapes[shape].params['is_lores'] == False:
                pointsmodelpy.complexmodel_add(self.complex_model, 
                    self.shapes[shape].shapeObject, "PDB")
        
        #pointsmodelpy.get_lorespoints(self.lores_model, self.points)
        self.npts = pointsmodelpy.get_complexpoints(self.complex_model, self.points)
        
        
    def getIq2D(self, qx, qy):
        """
            Returns simulate I(q) for given q_x and q_y values.
            @param qx: q_x [A-1]
            @param qy: q_y [A-1]
            @return: I(q) [cm-1]
        """
        
        # If this is the first simulation call, we need to generate the
        # space points
        if self.points is None:
            self._create_modelObject()
            
            # Protect against empty model
            if self.points is None:
                return 0
               
        # Evalute I(q) 
        norm =  1.0e8/self.params['lores_density']*self.params['scale']
        return norm*pointsmodelpy.get_complex_iq_2D(self.complex_model, self.points, qx, qy)\
            + self.params['background']
                
    def write_pr(self, filename):
        """
            Write P(r) to an output file
            @param filename: file name for P(r) output
        """   
        if self.hasPr == False:
            self.getPr()
     
        pointsmodelpy.outputPR(self.complex_model, filename)
     
    def getPrData(self):
        """
            Write P(r) to an output file
            @param filename: file name for P(r) output
        """   
        if self.hasPr == False:
            self.getPr()
     
        return pointsmodelpy.get_pr(self.complex_model)
     
    def getIq(self, q):
        """
            Returns the value of I(q) for a given q-value
            
            This method should remain internal to the class
            and the run() method should be used instead.
            
            @param q: q-value [float]
            @return: I(q) [float]
        """
        
        if self.hasPr == False:
            self.getPr()

        # By dividing by the density instead of the actuall V/N, 
        # we have an uncertainty of +-1 on N because the number 
        # of points chosen for the simulation is int(density*volume).
        # Propagation of error gives:
        #   delta(1/density^2) = 2*(1/density^2)/N
        # where N is stored in self.npts

        norm =  1.0e8/self.params['lores_density']*self.params['scale']
        #return norm*pointsmodelpy.get_lores_i(self.lores_model, q)
        return norm*pointsmodelpy.get_complex_i(self.complex_model, q)\
            + self.params['background']
    
    def getError(self, q):
        """
            Returns the error of I(q) for a given q-value
            @param q: q-value [float]
            @return: I(q) [float]
        """
        
        if self.hasPr == False:
            self.getPr()

        # By dividing by the density instead of the actual V/N, 
        # we have an uncertainty of +-1 on N because the number 
        # of points chosen for the simulation is int(density*volume).
        # Propagation of error gives:
        #   delta(1/density^2) = 2*(1/density^2)/N
        # where N is stored in self.npts

        norm =  1.0e8/self.params['lores_density']*self.params['scale']
        #return norm*pointsmodelpy.get_lores_i(self.lores_model, q)
        return norm*pointsmodelpy.get_complex_i_error(self.complex_model, q)\
            + self.params['background']
    
    def getIqError(self, q):
        """
            Return the simulated value along with its estimated
            error for a given q-value
            
            Propagation of errors is used to evaluate the
            uncertainty.
            
            @param q: q-value [float]
            @return: mean, error [float, float]
        """
        val = self.getIq(q)
        # Simulation error (statistical)
        err = self.getError(q)
        # Error on V/N
        simerr = 2*val/self.npts
        return val, err+simerr

    def getIq2DError(self, qx, qy):
        """
            Return the simulated value along with its estimated
            error for a given q-value
            
            Propagation of errors is used to evaluate the
            uncertainty.
            
            @param qx: qx-value [float]
            @param qy: qy-value [float]
            @return: mean, error [float, float]
        """
        self._create_modelObject()
                
        norm =  1.0e8/self.params['lores_density']*self.params['scale']
        val = norm*pointsmodelpy.get_complex_iq_2D(self.complex_model, self.points, qx, qy)\
            + self.params['background']
        
        # Simulation error (statistical)
        norm =  1.0e8/self.params['lores_density']*self.params['scale'] \
                * math.pow(self.npts/self.params['lores_density'], 1.0/3.0)/self.npts
        err = norm*pointsmodelpy.get_complex_iq_2D_err(self.complex_model, self.points, qx, qy)
        # Error on V/N
        simerr = 2*val/self.npts
        
        # The error used for the position is over-simplified.
        # The actual error was empirically found to be about
        # an order of magnitude larger.
        return val, 10.0*err+simerr