def _validate(self):
"""Validate my state.
This validates eq.
Raises SrFitError if validation fails.
"""
if self.eq is None:
raise SrFitError("eq is None")
from diffpy.srfit.equation.visitors import validate
try:
validate(self.eq)
except ValueError as e:
raise SrFitError(e)
# Try to get the value of eq.
try:
val = self.eq()
except TypeError:
raise SrFitError("eq cannot be evaluated")
finally:
if val is None:
raise SrFitError("eq evaluates to None")
return
def _validate(self):
"""Validate my state.
This validates that value and par are not None.
Raises SrFitError if validation fails.
"""
if self.value is None:
raise SrFitError("value is None")
if self.par is None:
raise SrFitError("par is None")
return
def _validate(self):
"""Validate my state.
This validates that the phase is not None.
This performs ProfileGenerator validations.
Raises SrFitError if validation fails.
"""
if self._calc is None:
raise SrFitError("_calc is None")
if self._phase is None:
raise SrFitError("_phase is None")
ProfileGenerator._validate(self)
return
def _validate(self):
"""Validate my state.
This validates eq.
Raises SrFitError if validation fails.
"""
if self.eq is None:
raise SrFitError("eq is None")
from diffpy.srfit.equation.visitors import validate
try:
validate(self.eq)
except ValueError, e:
raise SrFitError(e)
def _validate(self):
"""This evaluates the calculator.
Raises SrFitError if validation fails.
"""
from numpy import nan
p = self.penalty()
if p is None or p is nan:
raise SrFitError("Cannot evaluate penalty")
v = self._calc.value
if len(v) > 1 and not v.any():
emsg = ("Bond valence sums are all zero. Check atom symbols in "
"the structure or define custom bond-valence parameters.")
raise SrFitError(emsg)
return
def savetxt(self, fname, **kwargs):
"""Call `numpy.savetxt` with x, ycalc, y, dy.
Parameters
----------
fname : filename or file handle
This is passed to `numpy.savetxt`.
**kwargs
The keyword arguments that are passed to `numpy.savetxt`.
We preset file header "x ycalc y dy". Use ``header=''``
to save data without any header.
Raises
------
SrFitError
When `self.ycalc` has not been set.
See also
--------
numpy.savetxt
"""
x = self.x
ycalc = self.ycalc
if ycalc is None:
raise SrFitError("ycalc is None")
y = self.y
dy = self.dy
kwargs.setdefault('header', 'x ycalc y dy')
data = numpy.transpose([x, ycalc, y, dy])
numpy.savetxt(fname, data, **kwargs)
return
def _validate(self):
"""Validate my state.
This validates that x, y, dy, xobx, yobs and dyobs are not None.
This validates that x, y, and dy are the same length.
Raises SrFitError if validation fails.
"""
datanotset = any(v is None for v in
[self.x, self.y, self.dy, self.xobs, self.yobs, self.dyobs])
if datanotset:
raise SrFitError("Missing data")
if len(self.x) != len(self.y) or len(self.x) != len(self.dy):
raise SrFitError("Data are different lengths")
return
def _validate(self):
"""Validate my state.
This validates that value is not None.
Raises SrFitError if validation fails.
"""
if self.value is None:
raise SrFitError("value of '%s' is None" % self.name)
return
def setResidualEquation(self, eqstr):
"""Set the residual equation for the FitContribution.
eqstr -- A string representation of the residual. If eqstr is None
(default), then the previous residual equation will be
used, or the chi2 residual will be used if that does not
exist.
Two residuals are preset for convenience, "chiv" and "resv".
chiv is defined such that dot(chiv, chiv) = chi^2.
resv is defined such that dot(resv, resv) = Rw^2.
You can call on these in your residual equation. Note that the quantity
that will be optimized is the summed square of the residual equation.
Keep that in mind when defining a new residual or using the built-in
ones.
Raises SrFitError if the Profile is not yet defined.
Raises ValueError if eqstr depends on a Parameter that is not part of
the FitContribution.
"""
if self.profile is None:
raise SrFitError("Assign the Profile first")
if self._eq is None:
raise SrFitError("Assign the Equation first")
chivstr = "(eq - %s)/%s" % (self._yname, self._dyname)
resvstr = "(eq - %s)/sum(%s**2)**0.5" % (self._yname, self._yname)
# Get the equation string if it is not defined
if eqstr == "chiv":
eqstr = chivstr
elif eqstr == "resv":
eqstr = resvstr
reseq = equationFromString(eqstr, self._eqfactory)
self._eqfactory.wipeout(self._reseq)
self._reseq = reseq
return
def _validate(self):
"""Validate my state.
This performs profile validations.
This performs ParameterSet validations.
This does not validate the operation, since this could be costly. The
operation should be validated with a containing equation.
Raises SrFitError if validation fails.
"""
if self.profile is None:
raise SrFitError("profile is None")
self.profile._validate()
ParameterSet._validate(self)
return
def _validate(self):
"""Validate my state.
This performs profile validations.
This performs ProfileGenerator validations.
This validates _eq.
This validates _reseq and residual.
Raises SrFitError if validation fails.
"""
self.profile._validate()
ParameterSet._validate(self)
# Try to get the value of eq.
from diffpy.srfit.equation.visitors import validate
try:
validate(self._eq)
except ValueError, e:
raise SrFitError(e)
def _validate(self):
"""Validate my state.
This performs profile validations.
This performs ProfileGenerator validations.
This validates _eq.
This validates _reseq and residual.
Raises SrFitError if validation fails.
"""
self.profile._validate()
ParameterSet._validate(self)
# Try to get the value of eq.
from diffpy.srfit.equation.visitors import validate
try:
validate(self._eq)
except ValueError as e:
raise SrFitError(e)
if self._eq is None:
raise SrFitError("_eq is None")
val = None
try:
val = self._eq()
except TypeError as e:
raise SrFitError("_eq cannot be evaluated: %s"%e)
if val is None:
raise SrFitError("_eq evaluates to None")
# Try to get the value for residual
try:
validate(self._reseq)
except ValueError as e:
raise SrFitError(e)
try:
val = self.residual()
except TypeError:
raise SrFitError("residual cannot be evaluated")
if val is None:
raise SrFitError("residual evaluates to None")
return
class Constraint(Validatable):
"""Constraint class.
Constraints are designed to be stored in only one place. (The holder of the
constraint owns it).
Attributes
par -- A Parameter that is the subject of the constraint.
eq -- An equation whose evaluation is used to set the value of the
constraint.
"""
def __init__(self):
"""Initialization. """
self.par = None
self.eq = None
return
def constrain(self, par, eq):
"""Constrain a Parameter according to an Equation.
The parameter will be set constant once it is constrained. This will
keep it from being constrained multiple times.
Raises a ValueError if par is const.
"""
if par.const:
raise ValueError("The parameter '%s' is constant" % par)
if par.constrained:
raise ValueError("The parameter '%s' is already constrained" % par)
par.constrained = True
self.par = par
self.eq = eq
self.update()
return
def unconstrain(self):
"""Clear the constraint."""
self.par.constrained = False
self.par = None
self.eq = None
return
def update(self):
"""Update the parameter according to the equation."""
# This will be evaluated quickly thanks to the Equation class.
val = self.eq()
# This will only change the Parameter if val is different from the
# currently stored value.
self.par.setValue(val)
return
def _validate(self):
"""Validate my state.
This validates that par is not None.
This validates eq.
Raises SrFitError if validation fails.
"""
if self.par is None:
raise SrFitError("par is None")
if self.eq is None:
raise SrFitError("eq is None")
self.par._validate()
from diffpy.srfit.equation.visitors import validate
try:
validate(self.eq)
except ValueError, e:
raise SrFitError(e)
# Try to get the value of eq.
try:
val = self.eq()
self.par.setValue(val)
except TypeError, e:
raise SrFitError("eq cannot be evaluated")
class FitContribution(ParameterSet):
"""FitContribution class.
FitContributions organize an Equation that calculates the signal, and a
Profile that holds the signal. ProfileGenerators and Calculators can be
used as well. Contraints and Restraints can be created as part of a
FitContribution.
Attributes
name -- A name for this FitContribution.
profile -- A Profile that holds the measured (and calculated)
signal.
_calculators -- A managed dictionary of Calculators, indexed by name.
_constraints -- A set of constrained Parameters. Constraints can be
added using the 'constrain' methods.
_generators -- A managed dictionary of ProfileGenerators.
_parameters -- A managed OrderedDict of parameters.
_restraints -- A set of Restraints. Restraints can be added using the
'restrain' method.
_parsets -- A managed dictionary of ParameterSets.
_eqfactory -- A diffpy.srfit.equation.builder.EquationFactory
instance that is used to create constraints and
restraints from string
_eq -- The FitContribution equation that will be optimized.
_reseq -- The residual equation.
_xname -- Name of the x-variable
_yname -- Name of the y-variable
_dyname -- Name of the dy-variable
Properties
names -- Variable names (read only). See getNames.
values -- Variable values (read only). See getValues.
"""
def __init__(self, name):
"""Initialization."""
ParameterSet.__init__(self, name)
self._eq = None
self._reseq = None
self.profile = None
self._xname = None
self._yname = None
self._dyname = None
self._generators = {}
self._manage(self._generators)
return
def setProfile(self, profile, xname = None, yname = None, dyname = None):
"""Assign the Profile for this FitContribution.
profile -- A Profile that specifies the calculation points and that
will store the calculated signal.
xname -- The name of the independent variable from the Profile. If
this is None (default), then the name specified by the
Profile for this parameter will be used. This variable is
usable within string equations with the specified name.
yname -- The name of the observed Profile. If this is None
(default), then the name specified by the Profile for this
parameter will be used. This variable is usable within
string equations with the specified name.
dyname -- The name of the uncertainty in the observed Profile. If
this is None (default), then the name specified by the
Profile for this parameter will be used. This variable is
usable within string equations with the specified name.
"""
# Enforce type of the profile argument
if not isinstance(profile, Profile):
emsg = "Argument must be an instance of the Profile class."
raise TypeError(emsg)
# Set the Profile and add its parameters to this organizer.
self.profile = profile
if xname is None:
xname = self.profile.xpar.name
if yname is None:
yname = self.profile.ypar.name
if dyname is None:
dyname = self.profile.dypar.name
self._xname = xname
self._yname = yname
self._dyname = dyname
xpar = ParameterProxy(xname, self.profile.xpar)
ypar = ParameterProxy(yname, self.profile.ypar)
dypar = ParameterProxy(dyname, self.profile.dypar)
self.addParameter(xpar, check = False)
self.addParameter(ypar, check = False)
self.addParameter(dypar, check = False)
# If we have ProfileGenerators, set their Profiles.
for gen in self._generators.values():
gen.setProfile(profile)
# If we have _eq, but not _reseq, set the residual
if self._eq is not None and self._reseq is None:
self.setResidualEquation('chiv')
return
def addProfileGenerator(self, gen, name = None):
"""Add a ProfileGenerator to be used by this FitContribution.
The ProfileGenerator is given a name so that it can be used as part of
the profile equation (see setEquation). This can be different from the
name of the ProfileGenerator used for attribute access.
FitContributions should not share ProfileGenerator instances. Different
ProfileGenerators can share Parameters and ParameterSets, however.
Calling addProfileGenerator sets the profile equation to call the
calculator and if there is not a profile equation already.
gen -- A ProfileGenerator instance
name -- A name for the calculator. If name is None (default), then
the ProfileGenerator's name attribute will be used.
Raises ValueError if the ProfileGenerator has no name.
Raises ValueError if the ProfileGenerator has the same name as some
other managed object.
"""
if name is None:
name = gen.name
# Register the generator with the equation factory and add it as a
# managed object.
self._eqfactory.registerOperator(name, gen)
self._addObject(gen, self._generators, True)
# If we have a profile, set the profile of the generator.
if self.profile is not None:
gen.setProfile(self.profile)
# Make this our equation if we don't have one. This will set the
# residual equation if necessary.
if self._eq is None:
self.setEquation(name)
return
def setEquation(self, eqstr, ns = {}):
"""Set the profile equation for the FitContribution.
This sets the equation that will be used when generating the residual
for this FitContribution. The equation will be usable within
setResidualEquation as "eq", and it takes no arguments.
eqstr -- A string representation of the equation. Any Parameter
registered by addParameter or setProfile, or function
registered by setCalculator, registerFunction or
registerStringFunction can be can be used in the equation
by name. Other names will be turned into Parameters of this
FitContribution.
ns -- A dictionary of Parameters, indexed by name, that are used
in the eqstr, but not registered (default {}).
Raises ValueError if ns uses a name that is already used for a
variable.
"""
# Build the equation instance.
eq = equationFromString(eqstr, self._eqfactory,
buildargs=True, ns=ns)
eq.name = "eq"
# Register any new Parameters.
for par in self._eqfactory.newargs:
self._addParameter(par)
# Register eq as an operator
self._eqfactory.registerOperator("eq", eq)
self._eqfactory.wipeout(self._eq)
self._eq = eq
# Set the residual if we need to
if self.profile is not None and self._reseq is None:
self.setResidualEquation('chiv')
return
def getEquation(self):
"""Get math expression string for the active profile equation.
Return normalized math expression or an empty string if profile
equation has not been set yet.
"""
from diffpy.srfit.equation.visitors import getExpression
rv = ""
if self._eq is not None:
rv = getExpression(self._eq)
return rv
def setResidualEquation(self, eqstr):
"""Set the residual equation for the FitContribution.
eqstr -- A string representation of the residual. If eqstr is None
(default), then the previous residual equation will be
used, or the chi2 residual will be used if that does not
exist.
Two residuals are preset for convenience, "chiv" and "resv".
chiv is defined such that dot(chiv, chiv) = chi^2.
resv is defined such that dot(resv, resv) = Rw^2.
You can call on these in your residual equation. Note that the quantity
that will be optimized is the summed square of the residual equation.
Keep that in mind when defining a new residual or using the built-in
ones.
Raises SrFitError if the Profile is not yet defined.
Raises ValueError if eqstr depends on a Parameter that is not part of
the FitContribution.
"""
if self.profile is None:
raise SrFitError("Assign the Profile first")
if self._eq is None:
raise SrFitError("Assign the Equation first")
chivstr = "(eq - %s)/%s" % (self._yname, self._dyname)
resvstr = "(eq - %s)/sum(%s**2)**0.5" % (self._yname, self._yname)
# Get the equation string if it is not defined
if eqstr == "chiv":
eqstr = chivstr
elif eqstr == "resv":
eqstr = resvstr
reseq = equationFromString(eqstr, self._eqfactory)
self._eqfactory.wipeout(self._reseq)
self._reseq = reseq
return
def getResidualEquation(self):
"""Get math expression string for the active residual equation.
Return normalized math formula or an empty string if residual
equation has not been configured yet.
"""
from diffpy.srfit.equation.visitors import getExpression
rv = ""
if self._reseq is not None:
rv = getExpression(self._reseq, eqskip='eq$')
return rv
def residual(self):
"""Calculate the residual for this fitcontribution.
When this method is called, it is assumed that all parameters have been
assigned their most current values by the FitRecipe. This will be the
case when being called as part of a FitRecipe refinement.
The residual is by default an array chiv:
chiv = (eq() - self.profile.y) / self.profile.dy
The value that is optimized is dot(chiv, chiv).
The residual equation can be changed with the setResidualEquation
method.
"""
# Assign the calculated profile
self.profile.ycalc = self._eq()
# Note that equations only recompute when their inputs are modified, so
# the following will not recompute the equation.
return self._reseq()
def evaluate(self):
"""Evaluate the contribution equation and update profile.ycalc.
"""
yc = self._eq()
if self.profile is not None:
self.profile.ycalc = yc
return yc
def _validate(self):
"""Validate my state.
This performs profile validations.
This performs ProfileGenerator validations.
This validates _eq.
This validates _reseq and residual.
Raises SrFitError if validation fails.
"""
self.profile._validate()
ParameterSet._validate(self)
# Try to get the value of eq.
from diffpy.srfit.equation.visitors import validate
try:
validate(self._eq)
except ValueError, e:
raise SrFitError(e)
if self._eq is None:
raise SrFitError("_eq is None")
val = None
try:
val = self._eq()
except TypeError, e:
raise SrFitError("_eq cannot be evaluated: %s"%e)
def _validate(self):
"""Validate my state.
This validates eq.
Raises SrFitError if validation fails.
"""
if self.eq is None:
raise SrFitError("eq is None")
from diffpy.srfit.equation.visitors import validate
try:
validate(self.eq)
except ValueError, e:
raise SrFitError(e)
# Try to get the value of eq.
try:
val = self.eq()
except TypeError, e:
raise SrFitError("eq cannot be evaluated")
finally:
if val is None:
raise SrFitError("eq evaluates to None")
return
# End class Restraint
# End of file
from diffpy.srfit.equation.visitors import validate
try:
validate(self._eq)
except ValueError, e:
raise SrFitError(e)
if self._eq is None:
raise SrFitError("_eq is None")
val = None
try:
val = self._eq()
except TypeError, e:
raise SrFitError("_eq cannot be evaluated: %s"%e)
if val is None:
raise SrFitError("_eq evaluates to None")
# Try to get the value for residual
try:
validate(self._reseq)
except ValueError, e:
raise SrFitError(e)
try:
val = self.residual()
except TypeError, e:
raise SrFitError("residual cannot be evaluated")
if val is None:
raise SrFitError("residual evaluates to None")
return
# End of file
class Restraint(Validatable):
"""Restraint class.
Attributes
eq -- An equation whose evaluation is compared against the restraint
bounds.
lb -- The lower bound on the restraint evaluation (default -inf).
ub -- The lower bound on the restraint evaluation (default inf).
sig -- The uncertainty on the bounds (default 1).
scaled -- A flag indicating if the restraint is scaled (multiplied) by
the unrestrained point-average chi^2 (chi^2/numpoints)
(default False).
The penalty is calculated as
(max(0, lb - val, val - ub)/sig)**2
and val is the value of the calculated equation. This is multipled by the
average chi^2 if scaled is True.
"""
def __init__(self, eq, lb = -inf, ub = inf, sig = 1, scaled = False):
"""Restrain an equation to specified bounds.
eq -- An equation whose evaluation is compared against the
restraint bounds.
lb -- The lower bound on the restraint evaluation (float, default
-inf).
ub -- The lower bound on the restraint evaluation (float, default
inf).
sig -- The uncertainty on the bounds (default 1).
scaled -- A flag indicating if the restraint is scaled (multiplied)
by the unrestrained point-average chi^2 (chi^2/numpoints)
(bool, default False).
"""
self.eq = eq
self.lb = float(lb)
self.ub = float(ub)
self.sig = float(sig)
self.scaled = bool(scaled)
return
def penalty(self, w = 1.0):
"""Calculate the penalty of the restraint.
w -- The point-average chi^2 which is optionally used to scale the
penalty (default 1.0).
Returns the penalty as a float
"""
val = self.eq()
penalty = (max(0, self.lb - val, val - self.ub) / self.sig)**2
if self.scaled:
penalty *= w
return penalty
def _validate(self):
"""Validate my state.
This validates eq.
Raises SrFitError if validation fails.
"""
if self.eq is None:
raise SrFitError("eq is None")
from diffpy.srfit.equation.visitors import validate
try:
validate(self.eq)
except ValueError, e:
raise SrFitError(e)
# Try to get the value of eq.
try:
val = self.eq()
except TypeError, e:
raise SrFitError("eq cannot be evaluated")
