escript/py_src/linearPDEs.py


########################################################
#
# Copyright (c) 2003-2008 by University of Queensland
# Earth Systems Science Computational Center (ESSCC)
# http://www.uq.edu.au/esscc
#
# Primary Business: Queensland, Australia
# Licensed under the Open Software License version 3.0
# http://www.opensource.org/licenses/osl-3.0.php
#
########################################################

__copyright__="""Copyright (c) 2003-2008 by University of Queensland
Earth Systems Science Computational Center (ESSCC)
http://www.uq.edu.au/esscc
Primary Business: Queensland, Australia"""
__license__="""Licensed under the Open Software License version 3.0
http://www.opensource.org/licenses/osl-3.0.php"""
__url__="https://launchpad.net/escript-finley"

"""
The module provides an interface to define and solve linear partial
differential equations (PDEs) and Transport problems within L{escript}.
L{linearPDEs} does not provide any solver capabilities in itself but hands the
PDE over to the PDE solver library defined through the L{Domain<escript.Domain>}
of the PDE. The general interface is provided through the L{LinearPDE} class.
L{TransportProblem} provides an interface to initial value problems dominated
by its advective terms.

@var __author__: name of author
@var __copyright__: copyrights
@var __license__: licence agreement
@var __url__: url entry point on documentation
@var __version__: version
@var __date__: date of the version
"""

import math
import escript
import util
import numpy

__author__="Lutz Gross, l.gross@uq.edu.au"


class SolverOptions(object):
    """
    this class defines the solver options for a linear or non-linear solver.
    
    The option also supports the handling of diagnostic informations. 
    
    Typical usage is 
    
    opts=SolverOptions()
    print opts
    opts.resetDiagnostics()
    u=solver(opts)
    print "number of iteration steps: =",opts.getDiagnostics("num_iter")
    

    @cvar DEFAULT: The default method used to solve the system of linear equations
    @cvar DIRECT: The direct solver based on LDU factorization
    @cvar CHOLEVSKY: The direct solver based on LDLt factorization (can only be applied for symmetric PDEs)
    @cvar PCG: The preconditioned conjugate gradient method (can only be applied for symmetric PDEs)
    @cvar CR: The conjugate residual method
    @cvar CGS: The conjugate gradient square method
    @cvar BICGSTAB: The stabilized Bi-Conjugate Gradient method
    @cvar TFQMR: Transport Free Quasi Minimal Residual method
    @cvar MINRES: Minimum residual method
    @cvar SSOR: The symmetric over-relaxation method
    @cvar ILU0: The incomplete LU factorization preconditioner with no fill-in
    @cvar ILUT: The incomplete LU factorization preconditioner with fill-in
    @cvar JACOBI: The Jacobi preconditioner
    @cvar GMRES: The Gram-Schmidt minimum residual method
    @cvar PRES20: Special GMRES with restart after 20 steps and truncation after 5 residuals
    @cvar LUMPING: Matrix lumping
    @cvar NO_REORDERING: No matrix reordering allowed
    @cvar MINIMUM_FILL_IN: Reorder matrix to reduce fill-in during factorization
    @cvar NESTED_DISSECTION: Reorder matrix to improve load balancing during factorization
    @cvar PASO: PASO solver package
    @cvar SCSL: SGI SCSL solver library
    @cvar MKL: Intel's MKL solver library
    @cvar UMFPACK: The UMFPACK library
    @cvar TRILINOS: The TRILINOS parallel solver class library from Sandia National Labs
    @cvar ITERATIVE: The default iterative solver
    @cvar AMG: Algebraic Multi Grid
    @cvar REC_ILU: recursive ILU0
    @cvar RILU: relaxed ILU0
    @cvar GAUSS_SEIDEL: Gauss-Seidel solver
    @cvar DEFAULT_REORDERING: the reordering method recommended by the solver
    @cvar SUPER_LU: the Super_LU solver package
    @cvar PASTIX: the Pastix direct solver_package
    @cvar YAIR_SHAPIRA_COARSENING: AMG coarsening method by Yair-Shapira
    @cvar RUGE_STUEBEN_COARSENING: AMG coarsening method by Ruge and Stueben
    @cvar AGGREGATION_COARSENING: AMG coarsening using (symmetric) aggregation
    @cvar MIN_COARSE_MATRIX_SIZE: minimum size of the coarsest level matrix to use direct solver.
    @cvar NO_PRECONDITIONER: no preconditioner is applied.
    """
    DEFAULT= 0
    DIRECT= 1
    CHOLEVSKY= 2
    PCG= 3
    CR= 4
    CGS= 5
    BICGSTAB= 6
    SSOR= 7
    ILU0= 8
    ILUT= 9
    JACOBI= 10
    GMRES= 11
    PRES20= 12
    LUMPING= 13
    NO_REORDERING= 17
    MINIMUM_FILL_IN= 18
    NESTED_DISSECTION= 19
    MKL= 15
    UMFPACK= 16
    ITERATIVE= 20
    PASO= 21
    AMG= 22
    REC_ILU = 23
    TRILINOS = 24
    NONLINEAR_GMRES = 25
    TFQMR = 26
    MINRES = 27
    GAUSS_SEIDEL=28
    RILU=29
    DEFAULT_REORDERING=30
    SUPER_LU=31
    PASTIX=32
    YAIR_SHAPIRA_COARSENING=33
    RUGE_STUEBEN_COARSENING=34
    AGGREGATION_COARSENING=35
    NO_PRECONDITIONER=36
    MIN_COARSE_MATRIX_SIZE=37
    
    def __init__(self):
        self.setLevelMax()
        self.setCoarseningThreshold()
        self.setNumSweeps()
        self.setNumPreSweeps()
        self.setNumPostSweeps()
        self.setTolerance()
        self.setAbsoluteTolerance()
        self.setInnerTolerance()
        self.setDropTolerance()
        self.setDropStorage()
        self.setIterMax()
        self.setInnerIterMax()
        self.setTruncation()
        self.setRestart()
        self.setSymmetry()
        self.setVerbosity()
        self.setInnerToleranceAdaption()
        self.setAcceptanceConvergenceFailure()
        self.setReordering()
        self.setPackage()
        self.setSolverMethod()
        self.setPreconditioner()
        self.setCoarsening()
        self.setMinCoarseMatrixSize()
        self.setRelaxationFactor()        
        self.resetDiagnostics(all=True)

    def __str__(self):
        return self.getSummary()
    def getSummary(self):
        """
        Returns a string reporting the current settings
        """
        out="Solver Package: %s"%(self.getName(self.getPackage()))
        out+="\nVerbosity = %s"%self.isVerbose()
        out+="\nAccept failed convergence = %s"%self.acceptConvergenceFailure()
        out+="\nRelative tolerance = %e"%self.getTolerance()
        out+="\nAbsolute tolerance = %e"%self.getAbsoluteTolerance()
        out+="\nSymmetric problem = %s"%self.isSymmetric()
        out+="\nMaximum number of iteration steps = %s"%self.getIterMax()
        out+="\nInner tolerance = %e"%self.getInnerTolerance()
        out+="\nAdapt innner tolerance = %s"%self.adaptInnerTolerance()
    
        if self.getPackage() == self.PASO:
            out+="\nSolver method = %s"%self.getName(self.getSolverMethod())
            if self.getSolverMethod() == self.GMRES:
                out+="\nTruncation  = %s"%self.getTruncation()
                out+="\nRestart  = %s"%self.getRestart()
            if self.getSolverMethod() == self.AMG:
                out+="\nNumber of pre / post sweeps = %s / %s, %s"%(self.getNumPreSweeps(), self.getNumPostSweeps(), self.getNumSweeps())
                out+="\nMaximum number of levels = %s"%self.LevelMax()
                out+="\nCoarsening threshold = %e"%self.getCoarseningThreshold()
                out+="\Coarsening method = %s"%self.getName(self.getCoarsening())
            out+="\nPreconditioner = %s"%self.getName(self.getPreconditioner())
            if self.getPreconditioner() == self.AMG:
                out+="\nMaximum number of levels = %s"%self.LevelMax()
                out+="\nCoarsening method = %s"%self.getName(self.getCoarsening())
                out+="\nCoarsening threshold = %e"%self.getMinCoarseMatrixSize()
                out+="\nMinimum size of the coarsest level matrix = %e"%self.getCoarseningThreshold()
                out+="\nNumber of pre / post sweeps = %s / %s, %s"%(self.getNumPreSweeps(), self.getNumPostSweeps(), self.getNumSweeps())
            if self.getPreconditioner() == self.GAUSS_SEIDEL:
                out+="\nNumber of sweeps = %s"%self.getNumSweeps()
            if self.getPreconditioner() == self.ILUT:
                out+="\nDrop tolerance = %e"%self.getDropTolerance()
                out+="\nStorage increase = %e"%self.getDropStorage()
            if self.getPreconditioner() == self.RILU:
                out+="\nRelaxation factor = %e"%self.getRelaxationFactor()
        return out
        
    def getName(self,key):
        """
        returns the name of a given key
        
        @param key: a valid key
        """
        if key == self.DEFAULT: return "DEFAULT"
        if key == self.DIRECT: return "DIRECT"
        if key == self.CHOLEVSKY: return "CHOLEVSKY"
        if key == self.PCG: return "PCG"
        if key == self.CR: return "CR"
        if key == self.CGS: return "CGS"
        if key == self.BICGSTAB: return "BICGSTAB"
        if key == self.SSOR: return "SSOR"
        if key == self.ILU0: return "ILU0:"
        if key == self.ILUT: return "ILUT"
        if key == self.JACOBI: return "JACOBI"
        if key == self.GMRES: return "GMRES"
        if key == self.PRES20: return "PRES20"
        if key == self.LUMPING: return "LUMPING"
        if key == self.NO_REORDERING: return "NO_REORDERING"
        if key == self.MINIMUM_FILL_IN: return "MINIMUM_FILL_IN"
        if key == self.NESTED_DISSECTION: return "NESTED_DISSECTION"
        if key == self.MKL: return "MKL"
        if key == self.UMFPACK: return "UMFPACK"
        if key == self.ITERATIVE: return "ITERATIVE"
        if key == self.PASO: return "PASO"
        if key == self.AMG: return "AMG"
        if key == self.REC_ILU: return "REC_ILU"
        if key == self.TRILINOS: return "TRILINOS"
        if key == self.NONLINEAR_GMRES: return "NONLINEAR_GMRES"
        if key == self.TFQMR: return "TFQMR"
        if key == self.MINRES: return "MINRES"
        if key == self.GAUSS_SEIDEL: return "GAUSS_SEIDEL"
        if key == self.RILU: return "RILU"
        if key == self.DEFAULT_REORDERING: return "DEFAULT_REORDERING"
        if key == self.SUPER_LU: return "SUPER_LU"
        if key == self.PASTIX: return "PASTIX"
        if key == self.YAIR_SHAPIRA_COARSENING: return "YAIR_SHAPIRA_COARSENING"
        if key == self.RUGE_STUEBEN_COARSENING: return "RUGE_STUEBEN_COARSENING"
        if key == self.AGGREGATION_COARSENING: return "AGGREGATION_COARSENING"
        if key == self.NO_PRECONDITIONER: return "NO_PRECONDITIONER"
        if key == self.MIN_COARSE_MATRIX_SIZE: return "MIN_COARSE_MATRIX_SIZE"
        
    def resetDiagnostics(self,all=False):
        """
        resets the diagnostics
        
        @param all: if C{all} is C{True} all diagnostics including accumulative counters are reset.
        @type all: C{bool}
        """
        self.__num_iter=None
        self.__num_level=None
        self.__num_inner_iter=None
        self.__time=None
        self.__set_up_time=None
        self.__residual_norm=None
        self.__converged=None
        if all: 
            self.__cum_num_inner_iter=0
            self.__cum_num_iter=0
            self.__cum_time=0
            self.__cum_set_up_time=0

    def _updateDiagnostics(self, name, value):
        """
        Updates diagnostic information 
        
        @param name: name of  diagnostic information
        @type name: C{str} in the list "num_iter", "num_level", "num_inner_iter", "time", "set_up_time", "residual_norm", "converged".
        @param vale: new value of the diagnostic information
        @note: this function is used by a solver to report diagnostics informations.
        """
        if name == "num_iter": 
            self.__num_iter=int(value)
            self.__cum_num_iter+=self.__num_iter
        if name == "num_level":
            self.__num_iter=int(value)
        if name == "num_inner_iter":
            self.__num_inner_iter=int(value)
            self.__cum_num_inner_iter+=self.__num_inner_iter
        if name == "time":
            self.__time=float(value)
            self.__cum_time+=self.__time
        if name == "set_up_time":
            self.__set_up_time=float(value)
            self.__cum_set_up_time+=self.__set_up_time
        if name == "residual_norm":
            self.__residual_norm=float(value)
        if name == "converged":
            self.__converged = (value == True)
    def getDiagnostics(self, name):
        """
        Returns the diagnostic information C{name} 
        
        @param name: name of diagnostic information where
        - "num_iter": the number of iteration steps
        - "cum_num_iter": the cumulative number of iteration steps
        - "num_level": the number of level in multi level solver
        - "num_inner_iter": the number of inner iteration steps
        - "cum_num_inner_iter": the cumulative number of inner iteration steps
        - "time": execution time 
        - "cum_time": cumulative execution time
        - "set_up_time": time to set up of the solver, typically this includes factorization and reordering
        - "cum_set_up_time": cumulative time to set up of the solver
        - "residual_norm": norm of the final residual
        - "converged": return self.__converged     
        @type name: C{str} in the list "num_iter", "num_level", "num_inner_iter", "time", "set_up_time", "residual_norm", "converged".
        @return: requested value. C{None} is returned if the value is undefined.
        @note: If the solver has thrown an exception diagnostic values have an undefined status.
        """
        if name == "num_iter": return self.__num_iter
        elif name == "cum_num_iter": return self.__cum_num_iter
        elif name == "num_level": return self.__num_level
        elif name == "num_inner_iter": return self.__num_inner_iter
        elif name == "cum_num_inner_iter": return self.__cum_num_inner_iter
        elif name == "time": return self.__time
        elif name == "cum_time": return self.__cum_time
        elif name == "set_up_time": return self.__set_up_time
        elif name == "cum_set_up_time": return self.__cum_set_up_time
        elif name == "residual_norm": return self.__residual_norm
        elif name == "converged": return self.__converged      
        else:
            raise ValueError,"unknown diagnostic item %s"%name
    def hasConverged(self):
        """
        Returns C{True} if the last solver call has been finalized successfully.
        @note: if an exception has been thrown by the solver the status of this flag is undefined.
        """
        return self.getDiagnostics("converged")
    def setCoarsening(self,method=0):
        """
        Sets the key of the coarsening method to be applied in AMG.

        @param method: selects the coarsening method .
        @type method: in {SolverOptions.DEFAULT}, L{SolverOptions.YAIR_SHAPIRA_COARSENING}, 
        L{SolverOptions.RUGE_STUEBEN_COARSENING}, L{SolverOptions.AGGREGATION_COARSENING}
        """
    if method==None: method=0
        if not method in [self.DEFAULT, self.YAIR_SHAPIRA_COARSENING, self.RUGE_STUEBEN_COARSENING, self.AGGREGATION_COARSENING]:
             raise ValueError,"unknown coarsening method %s"%method
        self.__coarsening=method
    
    def getCoarsening(self):
        """
        Returns the key of the coarsening algorithm to be applied AMG.

        @rtype: in the list L{SolverOptions.DEFAULT}, L{SolverOptions.YAIR_SHAPIRA_COARSENING}, 
        L{SolverOptions.RUGE_STUEBEN_COARSENING}, L{SolverOptions.AGGREGATION_COARSENING}
        """
        return self.__coarsening
      
    def setMinCoarseMatrixSize(self,size=500):
        """
        Sets the minumum size of the coarsest level matrix in AMG.

        @param size: minumum size of the coarsest level matrix .
        @type size: positive C{int} or C{None}
        """
        size=int(size)
        if size<0:
           raise ValueError,"minumum size of the coarsest level matrix must be non-negative."
    if size==None: size=500
        self.__MinCoarseMatrixSize=size
        
    def getMinCoarseMatrixSize(self):
        """
        Returns the minumum size of the coarsest level matrix in AMG.

        @rtype: C{int}
        """
        return self.__MinCoarseMatrixSize
      
    def setPreconditioner(self, preconditioner=10):
        """
        Sets the preconditioner to be used. 

        @param preconditioner: key of the preconditioner to be used.
        @type preconditioner: in L{SolverOptions.SSOR}, L{SolverOptions.ILU0}, L{SolverOptions.ILUT}, L{SolverOptions.JACOBI}, 
                                    L{SolverOptions.AMG}, L{SolverOptions.REC_ILU}, L{SolverOptions.GAUSS_SEIDEL}, L{SolverOptions.RILU},
                                    L{SolverOptions.NO_PRECONDITIONER}
        @note: Not all packages support all preconditioner. It can be assumed that a package makes a reasonable choice if it encounters
        an unknown preconditioner. 
        """
    if preconditioner==None: preconditioner=10
        if not preconditioner in [ SolverOptions.SSOR, SolverOptions.ILU0, SolverOptions.ILUT, SolverOptions.JACOBI, 
                                    SolverOptions.AMG, SolverOptions.REC_ILU, SolverOptions.GAUSS_SEIDEL, SolverOptions.RILU,
                                    SolverOptions.NO_PRECONDITIONER] :
             raise ValueError,"unknown preconditioner %s"%preconditioner
        self.__preconditioner=preconditioner    
    def getPreconditioner(self):
        """
        Returns key of the preconditioner to be used. 

        @rtype: in the list L{SolverOptions.SSOR}, L{SolverOptions.ILU0}, L{SolverOptions.ILUT}, L{SolverOptions.JACOBI}, 
                                    L{SolverOptions.AMG}, L{SolverOptions.REC_ILU}, L{SolverOptions.GAUSS_SEIDEL}, L{SolverOptions.RILU},
                                    L{SolverOptions.NO_PRECONDITIONER}
        """
        return self.__preconditioner
    def setSolverMethod(self, method=0):
        """
        Sets the solver method to be used. Use C{method}=C{SolverOptions.DIRECT} to indicate that a direct rather than an iterative
        solver should be used and Use C{method}=C{SolverOptions.ITERATIVE} to indicate that an iterative rather than a direct
        solver should be used. 

        @param method: key of the solver method to be used.
        @type method: in L{SolverOptions.DEFAULT}, L{SolverOptions.DIRECT}, L{SolverOptions.CHOLEVSKY}, L{SolverOptions.PCG}, 
                        L{SolverOptions.CR}, L{SolverOptions.CGS}, L{SolverOptions.BICGSTAB}, L{SolverOptions.SSOR}, 
                        L{SolverOptions.GMRES}, L{SolverOptions.PRES20}, L{SolverOptions.LUMPING}, L{SolverOptions.ITERATIVE}, 
                        L{SolverOptions.AMG}, L{SolverOptions.NONLINEAR_GMRES}, L{SolverOptions.TFQMR}, L{SolverOptions.MINRES}, 
                        L{SolverOptions.GAUSS_SEIDEL}
        @note: Not all packages support all solvers. It can be assumed that a package makes a reasonable choice if it encounters
        an unknown solver method. 
        """
    if method==None: method=0
        if not method in [ SolverOptions.DEFAULT, SolverOptions.DIRECT, SolverOptions.CHOLEVSKY, SolverOptions.PCG, 
                           SolverOptions.CR, SolverOptions.CGS, SolverOptions.BICGSTAB, SolverOptions.SSOR, 
                           SolverOptions.GMRES, SolverOptions.PRES20, SolverOptions.LUMPING, SolverOptions.ITERATIVE, SolverOptions.AMG, 
                           SolverOptions.NONLINEAR_GMRES, SolverOptions.TFQMR, SolverOptions.MINRES, SolverOptions.GAUSS_SEIDEL]:
             raise ValueError,"unknown solver method %s"%method
        self.__method=method
    def getSolverMethod(self):
        """
        Returns key of the solver method to be used. 

        @rtype: in the list L{SolverOptions.DEFAULT}, L{SolverOptions.DIRECT}, L{SolverOptions.CHOLEVSKY}, L{SolverOptions.PCG}, 
                        L{SolverOptions.CR}, L{SolverOptions.CGS}, L{SolverOptions.BICGSTAB}, L{SolverOptions.SSOR}, 
                        L{SolverOptions.GMRES}, L{SolverOptions.PRES20}, L{SolverOptions.LUMPING}, L{SolverOptions.ITERATIVE}, 
                        L{SolverOptions.AMG}, L{SolverOptions.NONLINEAR_GMRES}, L{SolverOptions.TFQMR}, L{SolverOptions.MINRES}, 
                        L{SolverOptions.GAUSS_SEIDEL}
        """
        return self.__method
        
    def setPackage(self, package=0):
        """
        Sets the solver package to be used as a solver.  

        @param package: key of the solver package to be used.
        @type package: in L{SolverOptions.DEFAULT}, L{SolverOptions.PASO}, L{SolverOptions.SUPER_LU}, L{SolverOptions.PASTIX}, L{SolverOptions.MKL}, L{SolverOptions.UMFPACK}, L{SolverOptions.TRILINOS}
        @note: Not all packages are support on all implementation. An exception may be thrown on some platforms if a particular is requested. 
        """
    if package==None: package=0
        if not package in [SolverOptions.DEFAULT, SolverOptions.PASO, SolverOptions.SUPER_LU, SolverOptions.PASTIX, SolverOptions.MKL, SolverOptions.UMFPACK, SolverOptions.TRILINOS]:
             raise ValueError,"unknown solver package %s"%package
        self.__package=package
    def getPackage(self):
        """
        Returns the solver package key

        @rtype: in the list L{SolverOptions.DEFAULT}, L{SolverOptions.PASO}, L{SolverOptions.SUPER_LU}, L{SolverOptions.PASTIX}, L{SolverOptions.MKL}, L{SolverOptions.UMFPACK}, L{SolverOptions.TRILINOS}
        """
        return self.__package
    def setReordering(self,ordering=30):
        """
        Sets the key of the reordering method to be applied if supported by the solver. Some direct solvers support reordering 
        to optimize compute time and storage use during elimination. 

        @param ordering: selects the reordering strategy.
        @type ordering: in L{SolverOptions.NO_REORDERING}, L{SolverOptions.NO_REORDERING}, 
        L{SolverOptions.NO_REORDERING}, L{SolverOptions.DEFAULT_REORDERING}
        """
        if not ordering in [self.NO_REORDERING, self.MINIMUM_FILL_IN, self.NESTED_DISSECTION, self.DEFAULT_REORDERING]:
             raise ValueError,"unknown reordering strategy %s"%ordering
        self.__reordering=ordering
    def getReordering(self):
        """
        Returns the key of the reordering method to be applied if supported by the solver.

        @rtype: in the list L{SolverOptions.NO_REORDERING}, L{SolverOptions.NO_REORDERING}, 
        L{SolverOptions.NO_REORDERING}, L{SolverOptions.DEFAULT_REORDERING}
        """
        return self.__reordering
    def setRestart(self,restart=None):
        """
        Sets the number of iterations steps after which GMRES is performing a restart.

        @param restart: number of iteration steps after which to perform a restart. If equal to C{None} no
                        restart is performed.
        @type restart: C{int} or C{None}
        """
        if restart == None:
            self.__restart=restart
        else:
            restart=int(restart)
            if restart<1:
                raise ValueError,"restart must be positive."
            self.__restart=restart
        
    def getRestart(self):
        """
        Returns the number of iterations steps after which GMRES is performing a restart.
        If C{None} is returned no restart is performed.

        @rtype: C{int} or C{None}
        """
        if self.__restart < 0:
            return None
        else:
            return self.__restart
    def _getRestartForC(self):
        r=self.getRestart()
        if r == None: 
            return -1
            else: 
            return r
    def setTruncation(self,truncation=20):
        """
        Sets the number of residuals in GMRES to be stored for orthogonalization.  The more residuals are stored
        the faster GMRES converged but

        @param truncation: truncation
        @type truncation: C{int}
        """
        truncation=int(truncation)
        if truncation<1:
           raise ValueError,"truncation must be positive."
        self.__truncation=truncation
    def getTruncation(self):
        """
        Returns the number of residuals in GMRES to be stored for orthogonalization

        @rtype: C{int}
        """
        return self.__truncation
    def setInnerIterMax(self,iter_max=10):
        """
        Sets the maximum number of iteration steps for the inner iteration.

        @param iter_max: maximum number of inner iterations
        @type iter_max: C{int}
        """
        iter_max=int(iter_max)
        if iter_max<1:
           raise ValueError,"maximum number of inner iteration must be positive."
        self.__inner_iter_max=iter_max
    def getInnerIterMax(self):
        """
        Returns maximum number of inner iteration steps

        @rtype: C{int}
        """
        return self.__inner_iter_max
    def setIterMax(self,iter_max=100000):
        """
        Sets the maximum number of iteration steps

        @param iter_max: maximum number of iteration steps
        @type iter_max: C{int}
        """
        iter_max=int(iter_max)
        if iter_max<1:
           raise ValueError,"maximum number of iteration steps must be positive."
        self.__iter_max=iter_max
    def getIterMax(self):
        """
        Returns maximum number of iteration steps

        @rtype: C{int}
        """
        return self.__iter_max
    def setLevelMax(self,level_max=10):
        """
        Sets the maximum number of coarsening levels to be used in an algebraic multi level solver or preconditioner

        @param level_max: maximum number of levels
        @type level_max: C{int}
        """
        level_max=int(level_max)
        if level_max<0:
           raise ValueError,"maximum number of coarsening levels must be non-negative."
        self.__level_max=level_max
    def getLevelMax(self):
        """
        Returns the maximum number of coarsening levels to be used in an algebraic multi level solver or preconditioner

        @rtype: C{int}
        """
        return self.__level_max
    def setCoarseningThreshold(self,theta=0.05):
        """
        Sets the threshold for coarsening in the algebraic multi level solver or preconditioner

        @param theta: threshold for coarsening
        @type theta: positive C{float}
        """
        theta=float(theta)
        if theta<0 or theta>1:
           raise ValueError,"threshold must be non-negative and less or equal 1."
        self.__coarsening_threshold=theta
    def getCoarseningThreshold(self):
        """
        Returns the threshold for coarsening in the algebraic multi level solver or preconditioner

        @rtype: C{float}
        """
        return self.__coarsening_threshold
    def setNumSweeps(self,sweeps=2):
        """
        Sets the number of sweeps in a Jacobi or Gauss-Seidel/SOR preconditioner.

        @param sweeps: number of sweeps
        @type theta: positive C{int}
        """
        sweeps=int(sweeps)
        if sweeps<1:
           raise ValueError,"number of sweeps must be positive."
        self.__sweeps=sweeps
    def getNumSweeps(self):
        """
        Returns the number of sweeps in a Jacobi or Gauss-Seidel/SOR preconditioner.

        @rtype: C{int}
        """
        return self.__sweeps
    def setNumPreSweeps(self,sweeps=2):
        """
        Sets the number of sweeps in the pre-smoothing step of a multi level solver or preconditioner

        @param sweeps: number of sweeps
        @type theta: positive C{int}
        """
        sweeps=int(sweeps)
        if sweeps<1:
           raise ValueError,"number of sweeps must be positive."
        self.__pre_sweeps=sweeps
    def getNumPreSweeps(self):
        """
        Returns he number of sweeps in the pre-smoothing step of a multi level solver or preconditioner

        @rtype: C{int}
        """
        return self.__pre_sweeps
    def setNumPostSweeps(self,sweeps=2):
        """
        Sets the number of sweeps in the post-smoothing step of a multi level solver or preconditioner

        @param sweeps: number of sweeps
        @type theta: positive C{int}
        """
        sweeps=int(sweeps)
        if sweeps<1:
           raise ValueError,"number of sweeps must be positive."
        self.__post_sweeps=sweeps
    def getNumPostSweeps(self):
        """
        Returns he number of sweeps in the post-smoothing step of a multi level solver or preconditioner

        @rtype: C{int}
        """
        return self.__post_sweeps

    def setTolerance(self,rtol=1.e-8):
        """
        Sets the relative tolerance for the solver

        @param rtol: relative tolerance
        @type rtol: non-negative C{float}
        """
        rtol=float(rtol)
        if rtol<0 or rtol>1:
           raise ValueError,"tolerance must be non-negative and less or equal 1."
        self.__tolerance=rtol
    def getTolerance(self):
        """
        Returns the relative tolerance for the solver

        @rtype: C{float}
        """
        return self.__tolerance
    def setAbsoluteTolerance(self,atol=0.):
        """
        Sets the absolute tolerance for the solver

        @param atol:  absolute tolerance
        @type atol: non-negative C{float}
        """
        atol=float(atol)
        if atol<0:
           raise ValueError,"tolerance must be non-negative."
        self.__absolute_tolerance=atol
    def getAbsoluteTolerance(self):
        """
        Returns the absolute tolerance for the solver

        @rtype: C{float}
        """
        return self.__absolute_tolerance

    def setInnerTolerance(self,rtol=0.9):
        """
         Sets the relative tolerance for an inner iteration scheme for instance
        on the coarsest level in a multi-level scheme.

        @param rtol: inner relative tolerance
        @type rtol: positive C{float}
        """
        rtol=float(rtol)
        if rtol<=0 or rtol>1:
            raise ValueError,"tolerance must be positive and less or equal 1."
        self.__inner_tolerance=rtol
    def getInnerTolerance(self):
        """
        Returns the relative tolerance for an inner iteration scheme

        @rtype: C{float}
        """
        return self.__inner_tolerance
    def setDropTolerance(self,drop_tol=0.01):
        """
        Sets the relative drop tolerance in ILUT

        @param drop_tol: drop tolerance
        @type drop_tol: positive C{float}
        """
        drop_tol=float(drop_tol)
        if drop_tol<=0 or drop_tol>1:
            raise ValueError,"drop tolerance must be positive and less or equal 1."
        self.__drop_tolerance=drop_tol
    def getDropTolerance(self):
        """
        Returns the relative drop tolerance in ILUT

        @rtype: C{float}
        """
        return self.__drop_tolerance
    def setDropStorage(self,storage=2.):
        """
        Sets the maximum allowed increase in storage for ILUT. C{storage}=2 would mean that
        a doubling of the storage needed for the coefficient matrix is allowed in the ILUT factorization.

        @param storage: allowed storage increase
        @type storage: C{float}
        """
        storage=float(storage)
        if storage<1:
            raise ValueError,"allowed storage increase must be greater or equal to 1."
        self.__drop_storage=storage
    def getDropStorage(self):
    
        """
        Returns the maximum allowed increase in storage for ILUT

        @rtype: C{float}
        """
        return self.__drop_storage
    def setRelaxationFactor(self,factor=0.3):
        """
        Sets the relaxation factor used to add dropped elements in RILU to the main diagonal.

        @param factor: relaxation factor
        @type factor: C{float}
        @note: RILU with a relaxation factor 0 is identical to ILU0
        """
        factor=float(factor)
        if factor<0: 
            raise ValueError,"relaxation factor must be non-negative."
        self.__relaxation=factor
    def getRelaxationFactor(self):
    
        """
        Returns the relaxation factor used to add dropped elements in RILU to the main diagonal.

        @rtype: C{float}
        """
        return self.__relaxation
    def isSymmetric(self):
        """
        Checks if symmetry of the  coefficient matrix is indicated.

        @return: True if a symmetric PDE is indicated, False otherwise
        @rtype: C{bool}
        """
        return self.__symmetric
    def setSymmetryOn(self):
        """
        Sets the symmetry flag to indicate that the coefficient matrix is symmetric.
        """
        self.__symmetric=True
    def setSymmetryOff(self):
        """
        Clears the symmetry flag for the coefficient matrix.
        """
        self.__symmetric=False
    def setSymmetry(self,flag=False):
        """
        Sets the symmetry flag for the coefficient matrix to C{flag}.

        @param flag: If True, the symmetry flag is set otherwise reset.
        @type flag: C{bool}
        """
        if flag:
            self.setSymmetryOn()
        else:
            self.setSymmetryOff()
    def isVerbose(self):
        """
        Returns C{True} if the solver is expected to be verbose.

        @return: True if verbosity of switched on.
        @rtype: C{bool}
        """
        return self.__verbose

    def setVerbosityOn(self):
        """
        Switches the verbosity of the solver on.
        """
        self.__verbose=True
    def setVerbosityOff(self):
        """
        Switches the verbosity of the solver off.
        """
        self.__verbose=False
    def setVerbosity(self,verbose=False):
        """
        Sets the verbosity flag for the solver to C{flag}.

        @param flag: If C{True}, the verbosity of the solver is switched on.
        @type flag: C{bool}
        """
        if verbose:
            self.setVerbosityOn()
        else:
            self.setVerbosityOff()
        
    def adaptInnerTolerance(self):
        """
        Returns C{True} if the tolerance of the inner solver is selected automatically. 
        Otherwise the inner tolerance set by L{setInnerTolerance} is used.

        @return: C{True} if inner tolerance adaption is chosen.
        @rtype: C{bool}
        """
        return self.__adapt_inner_tolerance

    def setInnerToleranceAdaptionOn(self):
        """
        Switches the automatic selection of inner tolerance on 
        """
        self.__adapt_inner_tolerance=True
    def setInnerToleranceAdaptionOff(self):
        """
        Switches the automatic selection of inner tolerance off.
        """
        self.__adapt_inner_tolerance=False
    def setInnerToleranceAdaption(self,adapt=True):
        """
        Sets a flag to indicate automatic selection of the inner tolerance. 

        @param adapt: If C{True}, the inner tolerance is selected automatically.
        @type adapt: C{bool}
        """
        if adapt:
            self.setInnerToleranceAdaptionOn()
        else:
            self.setInnerToleranceAdaptionOff()

    def acceptConvergenceFailure(self):
        """
        Returns C{True} if a failure to meet the stopping criteria within the
        given number of iteration steps is not raising in exception. This is useful 
        if a solver is used in a non-linear context where the non-linear solver can 
        continue even if the returned the solution does not necessarily meet the
        stopping criteria. One can use the L{hasConverged} method to check if the
        last call to the solver was successful.

        @return: C{True} if a failure to achieve convergence is accepted.
        @rtype: C{bool}
        """
        return self.__accept_convergence_failure

    def setAcceptanceConvergenceFailureOn(self):
        """
        Switches the acceptance of a failure of convergence on 
        """
        self.__accept_convergence_failure=True
    def setAcceptanceConvergenceFailureOff(self):
        """
        Switches the acceptance of a failure of convergence off.
        """
        self.__accept_convergence_failure=False
    def setAcceptanceConvergenceFailure(self,accept=False):
        """
        Sets a flag to indicate the acceptance of a failure of convergence. 

        @param accept: If C{True}, any failure to achieve convergence is accepted.
        @type accept: C{bool}
        """
        if accept:
            self.setAcceptanceConvergenceFailureOn()
        else:
            self.setAcceptanceConvergenceFailureOff()

class IllegalCoefficient(ValueError):
   """
   Exception that is raised if an illegal coefficient of the general or
   particular PDE is requested.
   """
   pass

class IllegalCoefficientValue(ValueError):
   """
   Exception that is raised if an incorrect value for a coefficient is used.
   """
   pass

class IllegalCoefficientFunctionSpace(ValueError):
   """
   Exception that is raised if an incorrect function space for a coefficient
   is used.
   """

class UndefinedPDEError(ValueError):
   """
   Exception that is raised if a PDE is not fully defined yet.
   """
   pass

class PDECoef(object):
    """
    A class for describing a PDE coefficient.

    @cvar INTERIOR: indicator that coefficient is defined on the interior of
                    the PDE domain
    @cvar BOUNDARY: indicator that coefficient is defined on the boundary of
                    the PDE domain
    @cvar CONTACT: indicator that coefficient is defined on the contact region
                   within the PDE domain
    @cvar INTERIOR_REDUCED: indicator that coefficient is defined on the
                            interior of the PDE domain using a reduced
                            integration order
    @cvar BOUNDARY_REDUCED: indicator that coefficient is defined on the
                            boundary of the PDE domain using a reduced
                            integration order
    @cvar CONTACT_REDUCED: indicator that coefficient is defined on the contact
                           region within the PDE domain using a reduced
                           integration order
    @cvar SOLUTION: indicator that coefficient is defined through a solution of
                    the PDE
    @cvar REDUCED: indicator that coefficient is defined through a reduced
                   solution of the PDE
    @cvar BY_EQUATION: indicator that the dimension of the coefficient shape is
                       defined by the number of PDE equations
    @cvar BY_SOLUTION: indicator that the dimension of the coefficient shape is
                       defined by the number of PDE solutions
    @cvar BY_DIM: indicator that the dimension of the coefficient shape is
                  defined by the spatial dimension
    @cvar OPERATOR: indicator that the the coefficient alters the operator of
                    the PDE
    @cvar RIGHTHANDSIDE: indicator that the the coefficient alters the right
                         hand side of the PDE
    @cvar BOTH: indicator that the the coefficient alters the operator as well
                as the right hand side of the PDE

    """
    INTERIOR=0
    BOUNDARY=1
    CONTACT=2
    SOLUTION=3
    REDUCED=4
    BY_EQUATION=5
    BY_SOLUTION=6
    BY_DIM=7
    OPERATOR=10
    RIGHTHANDSIDE=11
    BOTH=12
    INTERIOR_REDUCED=13
    BOUNDARY_REDUCED=14
    CONTACT_REDUCED=15

    def __init__(self, where, pattern, altering):
       """
       Initialises a PDE coefficient type.

       @param where: describes where the coefficient lives
       @type where: one of L{INTERIOR}, L{BOUNDARY}, L{CONTACT}, L{SOLUTION},
                    L{REDUCED}, L{INTERIOR_REDUCED}, L{BOUNDARY_REDUCED},
                    L{CONTACT_REDUCED}
       @param pattern: describes the shape of the coefficient and how the shape
                       is built for a given spatial dimension and numbers of
                       equations and solutions in then PDE. For instance,
                       (L{BY_EQUATION},L{BY_SOLUTION},L{BY_DIM}) describes a
                       rank 3 coefficient which is instantiated as shape (3,2,2)
                       in case of three equations and two solution components
                       on a 2-dimensional domain. In the case of single equation
                       and a single solution component the shape components
                       marked by L{BY_EQUATION} or L{BY_SOLUTION} are dropped.
                       In this case the example would be read as (2,).
       @type pattern: C{tuple} of L{BY_EQUATION}, L{BY_SOLUTION}, L{BY_DIM}
       @param altering: indicates what part of the PDE is altered if the
                        coefficient is altered
       @type altering: one of L{OPERATOR}, L{RIGHTHANDSIDE}, L{BOTH}
       """
       super(PDECoef, self).__init__()
       self.what=where
       self.pattern=pattern
       self.altering=altering
       self.resetValue()

    def resetValue(self):
       """
       Resets the coefficient value to the default.
       """
       self.value=escript.Data()

    def getFunctionSpace(self,domain,reducedEquationOrder=False,reducedSolutionOrder=False):
       """
       Returns the L{FunctionSpace<escript.FunctionSpace>} of the coefficient.

       @param domain: domain on which the PDE uses the coefficient
       @type domain: L{Domain<escript.Domain>}
       @param reducedEquationOrder: True to indicate that reduced order is used
                                    to represent the equation
       @type reducedEquationOrder: C{bool}
       @param reducedSolutionOrder: True to indicate that reduced order is used
                                    to represent the solution
       @type reducedSolutionOrder: C{bool}
       @return: L{FunctionSpace<escript.FunctionSpace>} of the coefficient
       @rtype: L{FunctionSpace<escript.FunctionSpace>}
       """
       if self.what==self.INTERIOR:
            return escript.Function(domain)
       elif self.what==self.INTERIOR_REDUCED:
            return escript.ReducedFunction(domain)
       elif self.what==self.BOUNDARY:
            return escript.FunctionOnBoundary(domain)
       elif self.what==self.BOUNDARY_REDUCED:
            return escript.ReducedFunctionOnBoundary(domain)
       elif self.what==self.CONTACT:
            return escript.FunctionOnContactZero(domain)
       elif self.what==self.CONTACT_REDUCED:
            return escript.ReducedFunctionOnContactZero(domain)
       elif self.what==self.SOLUTION:
            if reducedEquationOrder and reducedSolutionOrder:
                return escript.ReducedSolution(domain)
            else:
                return escript.Solution(domain)
       elif self.what==self.REDUCED:
            return escript.ReducedSolution(domain)

    def getValue(self):
       """
       Returns the value of the coefficient.

       @return: value of the coefficient
       @rtype: L{Data<escript.Data>}
       """
       return self.value

    def setValue(self,domain,numEquations=1,numSolutions=1,reducedEquationOrder=False,reducedSolutionOrder=False,newValue=None):
       """
       Sets the value of the coefficient to a new value.

       @param domain: domain on which the PDE uses the coefficient
       @type domain: L{Domain<escript.Domain>}
       @param numEquations: number of equations of the PDE
       @type numEquations: C{int}
       @param numSolutions: number of components of the PDE solution
       @type numSolutions: C{int}
       @param reducedEquationOrder: True to indicate that reduced order is used
                                    to represent the equation
       @type reducedEquationOrder: C{bool}
       @param reducedSolutionOrder: True to indicate that reduced order is used
                                    to represent the solution
       @type reducedSolutionOrder: C{bool}
       @param newValue: number of components of the PDE solution
       @type newValue: any object that can be converted into a
                       L{Data<escript.Data>} object with the appropriate shape
                       and L{FunctionSpace<escript.FunctionSpace>}
       @raise IllegalCoefficientValue: if the shape of the assigned value does
                                       not match the shape of the coefficient
       @raise IllegalCoefficientFunctionSpace: if unable to interpolate value
                                               to appropriate function space
       """
       if newValue==None:
           newValue=escript.Data()
       elif isinstance(newValue,escript.Data):
           if not newValue.isEmpty():
              if not newValue.getFunctionSpace() == self.getFunctionSpace(domain,reducedEquationOrder,reducedSolutionOrder):
                try:
                  newValue=escript.Data(newValue,self.getFunctionSpace(domain,reducedEquationOrder,reducedSolutionOrder))
                except:
                  raise IllegalCoefficientFunctionSpace,"Unable to interpolate coefficient to function space %s"%self.getFunctionSpace(domain)
       else:
           newValue=escript.Data(newValue,self.getFunctionSpace(domain,reducedEquationOrder,reducedSolutionOrder))
       if not newValue.isEmpty():
           if not self.getShape(domain,numEquations,numSolutions)==newValue.getShape():
               raise IllegalCoefficientValue,"Expected shape of coefficient is %s but actual shape is %s."%(self.getShape(domain,numEquations,numSolutions),newValue.getShape())
       self.value=newValue

    def isAlteringOperator(self):
        """
        Checks if the coefficient alters the operator of the PDE.

        @return: True if the operator of the PDE is changed when the
                 coefficient is changed
        @rtype: C{bool}
        """
        if self.altering==self.OPERATOR or self.altering==self.BOTH:
            return not None
        else:
            return None

    def isAlteringRightHandSide(self):
        """
        Checks if the coefficient alters the right hand side of the PDE.

        @rtype: C{bool}
        @return: True if the right hand side of the PDE is changed when the
                 coefficient is changed, C{None} otherwise.
        """
        if self.altering==self.RIGHTHANDSIDE or self.altering==self.BOTH:
            return not None
        else:
            return None

    def estimateNumEquationsAndNumSolutions(self,domain,shape=()):
       """
       Tries to estimate the number of equations and number of solutions if
       the coefficient has the given shape.

       @param domain: domain on which the PDE uses the coefficient
       @type domain: L{Domain<escript.Domain>}
       @param shape: suggested shape of the coefficient
       @type shape: C{tuple} of C{int} values
       @return: the number of equations and number of solutions of the PDE if
                the coefficient has given shape. If no appropriate numbers
                could be identified, C{None} is returned
       @rtype: C{tuple} of two C{int} values or C{None}
       """
       dim=domain.getDim()
       if len(shape)>0:
           num=max(shape)+1
       else:
           num=1
       search=[]
       if self.definesNumEquation() and self.definesNumSolutions():
          for u in range(num):
             for e in range(num):
                search.append((e,u))
          search.sort(self.__CompTuple2)
          for item in search:
             s=self.getShape(domain,item[0],item[1])
             if len(s)==0 and len(shape)==0:
                 return (1,1)
             else:
                 if s==shape: return item
       elif self.definesNumEquation():
          for e in range(num,0,-1):
             s=self.getShape(domain,e,0)
             if len(s)==0 and len(shape)==0:
                 return (1,None)
             else:
                 if s==shape: return (e,None)

       elif self.definesNumSolutions():
          for u in range(num,0,-1):
             s=self.getShape(domain,0,u)
             if len(s)==0 and len(shape)==0:
                 return (None,1)
             else:
                 if s==shape: return (None,u)
       return None

    def definesNumSolutions(self):
       """
       Checks if the coefficient allows to estimate the number of solution
       components.

       @return: True if the coefficient allows an estimate of the number of
                solution components, False otherwise
       @rtype: C{bool}
       """
       for i in self.pattern:
             if i==self.BY_SOLUTION: return True
       return False

    def definesNumEquation(self):
       """
       Checks if the coefficient allows to estimate the number of equations.

       @return: True if the coefficient allows an estimate of the number of
                equations, False otherwise
       @rtype: C{bool}
       """
       for i in self.pattern:
             if i==self.BY_EQUATION: return True
       return False

    def __CompTuple2(self,t1,t2):
      """
      Compares two tuples of possible number of equations and number of
      solutions.

      @param t1: the first tuple
      @param t2: the second tuple
      @return: 0, 1, or -1
      """

      dif=t1[0]+t1[1]-(t2[0]+t2[1])
      if dif<0: return 1
      elif dif>0: return -1
      else: return 0

    def getShape(self,domain,numEquations=1,numSolutions=1):
       """
       Builds the required shape of the coefficient.

       @param domain: domain on which the PDE uses the coefficient
       @type domain: L{Domain<escript.Domain>}
       @param numEquations: number of equations of the PDE
       @type numEquations: C{int}
       @param numSolutions: number of components of the PDE solution
       @type numSolutions: C{int}
       @return: shape of the coefficient
       @rtype: C{tuple} of C{int} values
       """
       dim=domain.getDim()
       s=()
       for i in self.pattern:
             if i==self.BY_EQUATION:
                if numEquations>1: s=s+(numEquations,)
             elif i==self.BY_SOLUTION:
                if numSolutions>1: s=s+(numSolutions,)
             else:
                s=s+(dim,)
       return s

#====================================================================================================================

class LinearProblem(object):
   """
   This is the base class to define a general linear PDE-type problem for
   for an unknown function M{u} on a given domain defined through a
   L{Domain<escript.Domain>} object. The problem can be given as a single
   equation or as a system of equations.

   The class assumes that some sort of assembling process is required to form
   a problem of the form

   M{L u=f}

   where M{L} is an operator and M{f} is the right hand side. This operator
   problem will be solved to get the unknown M{u}.

   """
   def __init__(self,domain,numEquations=None,numSolutions=None,debug=False):
     """
     Initializes a linear problem.

     @param domain: domain of the PDE
     @type domain: L{Domain<escript.Domain>}
     @param numEquations: number of equations. If C{None} the number of
                          equations is extracted from the coefficients.
     @param numSolutions: number of solution components. If C{None} the number
                          of solution components is extracted from the
                          coefficients.
     @param debug: if True debug information is printed

     """
     super(LinearProblem, self).__init__()

     self.__debug=debug
     self.__domain=domain
     self.__numEquations=numEquations
     self.__numSolutions=numSolutions
     self.__altered_coefficients=False
     self.__reduce_equation_order=False
     self.__reduce_solution_order=False
     self.__sym=False
     self.setSolverOptions()
     self.__is_RHS_valid=False
     self.__is_operator_valid=False
     self.__COEFFICIENTS={}
     self.__solution_rtol=1.e99
     self.__solution_atol=1.e99
     self.setSymmetryOff()
     # initialize things:
     self.resetAllCoefficients()
     self.initializeSystem()
   # ==========================================================================
   #    general stuff:
   # ==========================================================================
   def __str__(self):
     """
     Returns a string representation of the PDE.

     @return: a simple representation of the PDE
     @rtype: C{str}
     """
     return "<LinearProblem %d>"%id(self)
   # ==========================================================================
   #    debug :
   # ==========================================================================
   def setDebugOn(self):
     """
     Switches debug output on.
     """
     self.__debug=not None

   def setDebugOff(self):
     """
     Switches debug output off.
     """
     self.__debug=None

   def setDebug(self, flag):
     """
     Switches debug output on if C{flag} is True otherwise it is switched off.

     @param flag: desired debug status
     @type flag: C{bool}
     """
     if flag:
         self.setDebugOn()
     else:
         self.setDebugOff()

   def trace(self,text):
     """
     Prints the text message if debug mode is switched on.

     @param text: message to be printed
     @type text: C{string}
     """
     if self.__debug: print "%s: %s"%(str(self),text)

   # ==========================================================================
   # some service functions:
   # ==========================================================================
   def introduceCoefficients(self,**coeff):
       """
       Introduces new coefficients into the problem.

       Use::

       p.introduceCoefficients(A=PDECoef(...), B=PDECoef(...))

       to introduce the coefficients M{A} ans M{B}.
       """
       for name, type in coeff.items():
           if not isinstance(type,PDECoef):
              raise ValueError,"coefficient %s has no type."%name
           self.__COEFFICIENTS[name]=type
           self.__COEFFICIENTS[name].resetValue()
           self.trace("coefficient %s has been introduced."%name)

   def getDomain(self):
     """
     Returns the domain of the PDE.

     @return: the domain of the PDE
     @rtype: L{Domain<escript.Domain>}
     """
     return self.__domain
   def getDomainStatus(self):
     """
     Return the status indicator of the domain
     """
     return self.getDomain().getStatus()

   def getSystemStatus(self):
     """
     Return the domain status used to build the current system
     """
     return self.__system_status
   def setSystemStatus(self,status=None):
     """
     Sets the system status to C{status} if C{status} is not present the 
     current status of the domain is used.
     """
     if status == None:
         self.__system_status=self.getDomainStatus()
     else: 
         self.__system_status=status

   def getDim(self):
     """
     Returns the spatial dimension of the PDE.

     @return: the spatial dimension of the PDE domain
     @rtype: C{int}
     """
     return self.getDomain().getDim()

   def getNumEquations(self):
     """
     Returns the number of equations.

     @return: the number of equations
     @rtype: C{int}
     @raise UndefinedPDEError: if the number of equations is not specified yet
     """
     if self.__numEquations==None:
         if self.__numSolutions==None:
            raise UndefinedPDEError,"Number of equations is undefined. Please specify argument numEquations."
         else:
            self.__numEquations=self.__numSolutions
     return self.__numEquations

   def getNumSolutions(self):
     """
     Returns the number of unknowns.

     @return: the number of unknowns
     @rtype: C{int}
     @raise UndefinedPDEError: if the number of unknowns is not specified yet
     """
     if self.__numSolutions==None:
        if self.__numEquations==None:
            raise UndefinedPDEError,"Number of solution is undefined. Please specify argument numSolutions."
        else:
            self.__numSolutions=self.__numEquations
     return self.__numSolutions

   def reduceEquationOrder(self):
     """
     Returns the status of order reduction for the equation.

     @return: True if reduced interpolation order is used for the
              representation of the equation, False otherwise
     @rtype: L{bool}
     """
     return self.__reduce_equation_order

   def reduceSolutionOrder(self):
     """
     Returns the status of order reduction for the solution.

     @return: True if reduced interpolation order is used for the
              representation of the solution, False otherwise
     @rtype: L{bool}
     """
     return self.__reduce_solution_order

   def getFunctionSpaceForEquation(self):
     """
     Returns the L{FunctionSpace<escript.FunctionSpace>} used to discretize
     the equation.

     @return: representation space of equation
     @rtype: L{FunctionSpace<escript.FunctionSpace>}
     """
     if self.reduceEquationOrder():
         return escript.ReducedSolution(self.getDomain())
     else:
         return escript.Solution(self.getDomain())

   def getFunctionSpaceForSolution(self):
     """
     Returns the L{FunctionSpace<escript.FunctionSpace>} used to represent
     the solution.

     @return: representation space of solution
     @rtype: L{FunctionSpace<escript.FunctionSpace>}
     """
     if self.reduceSolutionOrder():
         return escript.ReducedSolution(self.getDomain())
     else:
         return escript.Solution(self.getDomain())

   # ==========================================================================
   #   solver settings:
   # ==========================================================================
   def setSolverOptions(self,options=None):
       """
       Sets the solver options.

       @param options: the new solver options. If equal C{None}, the solver options are set to the default.
       @type options: L{SolverOptions} or C{None}
       @note: The symmetry flag of options is overwritten by the symmetry flag of the L{LinearProblem}.
       """
       if options==None:
          self.__solver_options=SolverOptions()
       elif isinstance(options, SolverOptions):
          self.__solver_options=options
       else:
          raise ValueError,"options must be a SolverOptions object."
       self.__solver_options.setSymmetry(self.__sym)
     
   def getSolverOptions(self):
       """
       Returns the solver options
   
       @rtype: L{SolverOptions}
       """
       self.__solver_options.setSymmetry(self.__sym)
       return self.__solver_options
       
   def isUsingLumping(self):
      """
      Checks if matrix lumping is the current solver method.

      @return: True if the current solver method is lumping
      @rtype: C{bool}
      """
      return self.getSolverOptions().getSolverMethod()==self.getSolverOptions().LUMPING
   # ==========================================================================
   #    symmetry  flag:
   # ==========================================================================
   def isSymmetric(self):
      """
      Checks if symmetry is indicated.

      @return: True if a symmetric PDE is indicated, False otherwise
      @rtype: C{bool}
      @note: the method is equivalent to use getSolverOptions().isSymmetric()
      """
      self.getSolverOptions().isSymmetric()

   def setSymmetryOn(self):
      """
      Sets the symmetry flag. 
      @note: The method overwrites the symmetry flag set by the solver options
      """
      self.__sym=True
      self.getSolverOptions().setSymmetryOn()

   def setSymmetryOff(self):
      """
      Clears the symmetry flag.
      @note: The method overwrites the symmetry flag set by the solver options
      """
      self.__sym=False
      self.getSolverOptions().setSymmetryOff()

   def setSymmetry(self,flag=False):
      """
      Sets the symmetry flag to C{flag}.

      @param flag: If True, the symmetry flag is set otherwise reset.
      @type flag: C{bool}
      @note: The method overwrites the symmetry flag set by the solver options
      """
      self.getSolverOptions().setSymmetry(flag)
   # ==========================================================================
   # function space handling for the equation as well as the solution
   # ==========================================================================
   def setReducedOrderOn(self):
     """
     Switches reduced order on for solution and equation representation.

     @raise RuntimeError: if order reduction is altered after a coefficient has
                          been set
     """
     self.setReducedOrderForSolutionOn()
     self.setReducedOrderForEquationOn()

   def setReducedOrderOff(self):
     """
     Switches reduced order off for solution and equation representation

     @raise RuntimeError: if order reduction is altered after a coefficient has
                          been set
     """
     self.setReducedOrderForSolutionOff()
     self.setReducedOrderForEquationOff()

   def setReducedOrderTo(self,flag=False):
     """
     Sets order reduction state for both solution and equation representation
     according to flag.

     @param flag: if True, the order reduction is switched on for both solution
                  and equation representation, otherwise or if flag is not
                  present order reduction is switched off
     @type flag: C{bool}
     @raise RuntimeError: if order reduction is altered after a coefficient has
                          been set
     """
     self.setReducedOrderForSolutionTo(flag)
     self.setReducedOrderForEquationTo(flag)


   def setReducedOrderForSolutionOn(self):
     """
     Switches reduced order on for solution representation.

     @raise RuntimeError: if order reduction is altered after a coefficient has
                          been set
     """
     if not self.__reduce_solution_order:
         if self.__altered_coefficients:
              raise RuntimeError,"order cannot be altered after coefficients have been defined."
         self.trace("Reduced order is used for solution representation.")
         self.__reduce_solution_order=True
         self.initializeSystem()

   def setReducedOrderForSolutionOff(self):
     """
     Switches reduced order off for solution representation

     @raise RuntimeError: if order reduction is altered after a coefficient has
                          been set.
     """
     if self.__reduce_solution_order:
         if self.__altered_coefficients:
              raise RuntimeError,"order cannot be altered after coefficients have been defined."
         self.trace("Full order is used to interpolate solution.")
         self.__reduce_solution_order=False
         self.initializeSystem()

   def setReducedOrderForSolutionTo(self,flag=False):
     """
     Sets order reduction state for solution representation according to flag.

     @param flag: if flag is True, the order reduction is switched on for
                  solution representation, otherwise or if flag is not present
                  order reduction is switched off
     @type flag: C{bool}
     @raise RuntimeError: if order reduction is altered after a coefficient has
                          been set
     """
     if flag:
        self.setReducedOrderForSolutionOn()
     else:
        self.setReducedOrderForSolutionOff()

   def setReducedOrderForEquationOn(self):
     """
     Switches reduced order on for equation representation.

     @raise RuntimeError: if order reduction is altered after a coefficient has
                          been set
     """
     if not self.__reduce_equation_order:
         if self.__altered_coefficients:
              raise RuntimeError,"order cannot be altered after coefficients have been defined."
         self.trace("Reduced order is used for test functions.")
         self.__reduce_equation_order=True
         self.initializeSystem()

   def setReducedOrderForEquationOff(self):
     """
     Switches reduced order off for equation representation.

     @raise RuntimeError: if order reduction is altered after a coefficient has
                          been set
     """
     if self.__reduce_equation_order:
         if self.__altered_coefficients:
              raise RuntimeError,"order cannot be altered after coefficients have been defined."
         self.trace("Full order is used for test functions.")
         self.__reduce_equation_order=False
         self.initializeSystem()

   def setReducedOrderForEquationTo(self,flag=False):
     """
     Sets order reduction state for equation representation according to flag.

     @param flag: if flag is True, the order reduction is switched on for
                  equation representation, otherwise or if flag is not present
                  order reduction is switched off
     @type flag: C{bool}
     @raise RuntimeError: if order reduction is altered after a coefficient has
                          been set
     """
     if flag:
        self.setReducedOrderForEquationOn()
     else:
        self.setReducedOrderForEquationOff()
   def getOperatorType(self):
      """
      Returns the current system type.
      """
      return self.__operator_type

   def checkSymmetricTensor(self,name,verbose=True):
      """
      Tests a coefficient for symmetry.

      @param name: name of the coefficient
      @type name: C{str}
      @param verbose: if set to True or not present a report on coefficients
                      which break the symmetry is printed.
      @type verbose: C{bool}
      @return: True if coefficient C{name} is symmetric
      @rtype: C{bool}
      """
      SMALL_TOLERANCE=util.EPSILON*10.
      A=self.getCoefficient(name)
      verbose=verbose or self.__debug
      out=True
      if not A.isEmpty():
         tol=util.Lsup(A)*SMALL_TOLERANCE
         s=A.getShape()
         if A.getRank() == 4:
            if s[0]==s[2] and s[1] == s[3]:
               for i in range(s[0]):
                  for j in range(s[1]):
                     for k in range(s[2]):
                        for l in range(s[3]):
                            if util.Lsup(A[i,j,k,l]-A[k,l,i,j])>tol:
                               if verbose: print "non-symmetric problem as %s[%d,%d,%d,%d]!=%s[%d,%d,%d,%d]"%(name,i,j,k,l,name,k,l,i,j)
                               out=False
            else:
                 if verbose: print "non-symmetric problem because of inappropriate shape %s of coefficient %s."%(s,name)
                 out=False
         elif A.getRank() == 2:
            if s[0]==s[1]:
               for j in range(s[0]):
                  for l in range(s[1]):
                     if util.Lsup(A[j,l]-A[l,j])>tol:
                        if verbose: print "non-symmetric problem because %s[%d,%d]!=%s[%d,%d]"%(name,j,l,name,l,j)
                        out=False
            else:
                 if verbose: print "non-symmetric problem because of inappropriate shape %s of coefficient %s."%(s,name)
                 out=False
         elif A.getRank() == 0:
            pass
         else:
             raise ValueError,"Cannot check rank %s of %s."%(A.getRank(),name)
      return out

   def checkReciprocalSymmetry(self,name0,name1,verbose=True):
      """
      Tests two coefficients for reciprocal symmetry.

      @param name0: name of the first coefficient
      @type name0: C{str}
      @param name1: name of the second coefficient
      @type name1: C{str}
      @param verbose: if set to True or not present a report on coefficients
                      which break the symmetry is printed
      @type verbose: C{bool}
      @return: True if coefficients C{name0} and C{name1} are reciprocally
               symmetric.
      @rtype: C{bool}
      """
      SMALL_TOLERANCE=util.EPSILON*10.
      B=self.getCoefficient(name0)
      C=self.getCoefficient(name1)
      verbose=verbose or self.__debug
      out=True
      if B.isEmpty() and not C.isEmpty():
         if verbose: print "non-symmetric problem because %s is not present but %s is"%(name0,name1)
         out=False
      elif not B.isEmpty() and C.isEmpty():
         if verbose: print "non-symmetric problem because %s is not present but %s is"%(name0,name1)
         out=False
      elif not B.isEmpty() and not C.isEmpty():
         sB=B.getShape()
         sC=C.getShape()
         tol=(util.Lsup(B)+util.Lsup(C))*SMALL_TOLERANCE/2.
         if len(sB) != len(sC):
             if verbose: print "non-symmetric problem because ranks of %s (=%s) and %s (=%s) are different."%(name0,len(sB),name1,len(sC))
             out=False
         else:
             if len(sB)==0:
               if util.Lsup(B-C)>tol:
                  if verbose: print "non-symmetric problem because %s!=%s"%(name0,name1)
                  out=False
             elif len(sB)==1:
               if sB[0]==sC[0]:
                  for j in range(sB[0]):
                     if util.Lsup(B[j]-C[j])>tol:
                        if verbose: print "non-symmetric PDE because %s[%d]!=%s[%d]"%(name0,j,name1,j)
                        out=False
               else:
                 if verbose: print "non-symmetric problem because of inappropriate shapes %s and %s of coefficients %s and %s, respectively."%(sB,sC,name0,name1)
             elif len(sB)==3:
               if sB[0]==sC[1] and sB[1]==sC[2] and sB[2]==sC[0]:
                   for i in range(sB[0]):
                      for j in range(sB[1]):
                         for k in range(sB[2]):
                            if util.Lsup(B[i,j,k]-C[k,i,j])>tol:
                                 if verbose: print "non-symmetric problem because %s[%d,%d,%d]!=%s[%d,%d,%d]"%(name0,i,j,k,name1,k,i,j)
                                 out=False
               else:
                 if verbose: print "non-symmetric problem because of inappropriate shapes %s and %s of coefficients %s and %s, respectively."%(sB,sC,name0,name1)
             else:
                 raise ValueError,"Cannot check rank %s of %s and %s."%(len(sB),name0,name1)
      return out

   def getCoefficient(self,name):
     """
     Returns the value of the coefficient C{name}.

     @param name: name of the coefficient requested
     @type name: C{string}
     @return: the value of the coefficient
     @rtype: L{Data<escript.Data>}
     @raise IllegalCoefficient: if C{name} is not a coefficient of the PDE
     """
     if self.hasCoefficient(name):
         return self.__COEFFICIENTS[name].getValue()
     else:
        raise IllegalCoefficient,"illegal coefficient %s requested for general PDE."%name

   def hasCoefficient(self,name):
     """
     Returns True if C{name} is the name of a coefficient.

     @param name: name of the coefficient enquired
     @type name: C{string}
     @return: True if C{name} is the name of a coefficient of the general PDE,
              False otherwise
     @rtype: C{bool}
     """
     return self.__COEFFICIENTS.has_key(name)

   def createCoefficient(self, name):
     """
     Creates a L{Data<escript.Data>} object corresponding to coefficient
     C{name}.

     @return: the coefficient C{name} initialized to 0
     @rtype: L{Data<escript.Data>}
     @raise IllegalCoefficient: if C{name} is not a coefficient of the PDE
     """
     if self.hasCoefficient(name):
        return escript.Data(0.,self.getShapeOfCoefficient(name),self.getFunctionSpaceForCoefficient(name))
     else:
        raise IllegalCoefficient,"illegal coefficient %s requested for general PDE."%name

   def getFunctionSpaceForCoefficient(self,name):
     """
     Returns the L{FunctionSpace<escript.FunctionSpace>} to be used for
     coefficient C{name}.

     @param name: name of the coefficient enquired
     @type name: C{string}
     @return: the function space to be used for coefficient C{name}
     @rtype: L{FunctionSpace<escript.FunctionSpace>}
     @raise IllegalCoefficient: if C{name} is not a coefficient of the PDE
     """
     if self.hasCoefficient(name):
        return self.__COEFFICIENTS[name].getFunctionSpace(self.getDomain())
     else:
        raise ValueError,"unknown coefficient %s requested"%name

   def getShapeOfCoefficient(self,name):
     """
     Returns the shape of the coefficient C{name}.

     @param name: name of the coefficient enquired
     @type name: C{string}
     @return: the shape of the coefficient C{name}
     @rtype: C{tuple} of C{int}
     @raise IllegalCoefficient: if C{name} is not a coefficient of the PDE
     """
     if self.hasCoefficient(name):
        return self.__COEFFICIENTS[name].getShape(self.getDomain(),self.getNumEquations(),self.getNumSolutions())
     else:
        raise IllegalCoefficient,"illegal coefficient %s requested for general PDE."%name

   def resetAllCoefficients(self):
     """
     Resets all coefficients to their default values.
     """
     for i in self.__COEFFICIENTS.iterkeys():
         self.__COEFFICIENTS[i].resetValue()

   def alteredCoefficient(self,name):
     """
     Announces that coefficient C{name} has been changed.

     @param name: name of the coefficient affected
     @type name: C{string}
     @raise IllegalCoefficient: if C{name} is not a coefficient of the PDE
     @note: if C{name} is q or r, the method will not trigger a rebuild of the
            system as constraints are applied to the solved system.
     """
     if self.hasCoefficient(name):
        self.trace("Coefficient %s has been altered."%name)
        if not ((name=="q" or name=="r") and self.isUsingLumping()):
           if self.__COEFFICIENTS[name].isAlteringOperator(): self.invalidateOperator()
           if self.__COEFFICIENTS[name].isAlteringRightHandSide(): self.invalidateRightHandSide()
     else:
        raise IllegalCoefficient,"illegal coefficient %s requested for general PDE."%name

   def validSolution(self):
       """
       Marks the solution as valid.
       """
       self.__is_solution_valid=True

   def invalidateSolution(self):
       """
       Indicates the PDE has to be resolved if the solution is requested.
       """
       self.trace("System will be resolved.")
       self.__is_solution_valid=False

   def isSolutionValid(self):
       """
       Returns True if the solution is still valid.
       """
       if not self.getDomainStatus()==self.getSystemStatus(): self.invalidateSolution()
       if self.__solution_rtol>self.getSolverOptions().getTolerance() or \
          self.__solution_atol>self.getSolverOptions().getAbsoluteTolerance():
         self.invalidateSolution()  
       return self.__is_solution_valid

   def validOperator(self):
       """
       Marks the operator as valid.
       """
       self.__is_operator_valid=True

   def invalidateOperator(self):
       """
       Indicates the operator has to be rebuilt next time it is used.
       """
       self.trace("Operator will be rebuilt.")
       self.invalidateSolution()
       self.__is_operator_valid=False

   def isOperatorValid(self):
       """
       Returns True if the operator is still valid.
       """
       if not self.getDomainStatus()==self.getSystemStatus(): self.invalidateOperator()
       if not self.getRequiredOperatorType()==self.getOperatorType(): self.invalidateOperator()
       return self.__is_operator_valid

   def validRightHandSide(self):
       """
       Marks the right hand side as valid.
       """
       self.__is_RHS_valid=True

   def invalidateRightHandSide(self):
       """
       Indicates the right hand side has to be rebuilt next time it is used.
       """
       self.trace("Right hand side has to be rebuilt.")
       self.invalidateSolution()
       self.__is_RHS_valid=False

   def isRightHandSideValid(self):
       """
       Returns True if the operator is still valid.
       """
       if not self.getDomainStatus()==self.getSystemStatus(): self.invalidateRightHandSide()
       return self.__is_RHS_valid

   def invalidateSystem(self):
       """
       Announces that everything has to be rebuilt.
       """
       self.invalidateSolution()
       self.invalidateOperator()
       self.invalidateRightHandSide()

   def isSystemValid(self):
       """
       Returns True if the system (including solution) is still vaild.
       """
       return self.isSolutionValid() and self.isOperatorValid() and self.isRightHandSideValid()

   def initializeSystem(self):
       """
       Resets the system clearing the operator, right hand side and solution.
       """
       self.trace("New System has been created.")
       self.__operator_type=None
       self.setSystemStatus()
       self.__operator=escript.Operator()
       self.__righthandside=escript.Data()
       self.__solution=escript.Data()
       self.invalidateSystem()

   def getOperator(self):
     """
     Returns the operator of the linear problem.

     @return: the operator of the problem
     """
     return self.getSystem()[0]

   def getRightHandSide(self):
     """
     Returns the right hand side of the linear problem.

     @return: the right hand side of the problem
     @rtype: L{Data<escript.Data>}
     """
     return self.getSystem()[1]

   def createRightHandSide(self):
       """
       Returns an instance of a new right hand side.
       """
       self.trace("New right hand side is allocated.")
       if self.getNumEquations()>1:
           return escript.Data(0.,(self.getNumEquations(),),self.getFunctionSpaceForEquation(),True)
       else:
           return escript.Data(0.,(),self.getFunctionSpaceForEquation(),True)

   def createSolution(self):
       """
       Returns an instance of a new solution.
       """
       self.trace("New solution is allocated.")
       if self.getNumSolutions()>1:
           return escript.Data(0.,(self.getNumSolutions(),),self.getFunctionSpaceForSolution(),True)
       else:
           return escript.Data(0.,(),self.getFunctionSpaceForSolution(),True)

   def resetSolution(self):
       """
       Sets the solution to zero.
       """
       if self.__solution.isEmpty():
           self.__solution=self.createSolution()
       else:
           self.__solution.setToZero()
           self.trace("Solution is reset to zero.")

   def setSolution(self,u):
       """
       Sets the solution assuming that makes the system valid with the tolrance
       defined by the solver options
       """
       self.__solution_rtol=self.getSolverOptions().getTolerance()
       self.__solution_atol=self.getSolverOptions().getAbsoluteTolerance()
       self.__solution=u
       self.validSolution()

   def getCurrentSolution(self):
       """
       Returns the solution in its current state.
       """
       if self.__solution.isEmpty(): self.__solution=self.createSolution()
       return self.__solution

   def resetRightHandSide(self):
       """
       Sets the right hand side to zero.
       """
       if self.__righthandside.isEmpty():
           self.__righthandside=self.createRightHandSide()
       else:
           self.__righthandside.setToZero()
           self.trace("Right hand side is reset to zero.")

   def getCurrentRightHandSide(self):
       """
       Returns the right hand side in its current state.
       """
       if self.__righthandside.isEmpty(): self.__righthandside=self.createRightHandSide()
       return self.__righthandside

   def resetOperator(self):
       """
       Makes sure that the operator is instantiated and returns it initialized
       with zeros.
       """
       if self.getOperatorType() == None:
           if self.isUsingLumping():
               self.__operator=self.createSolution()
           else:
               self.__operator=self.createOperator()
       self.__operator_type=self.getRequiredOperatorType()
       else:
           if self.isUsingLumping():
               self.__operator.setToZero()
           else:
               self.__operator.resetValues()
           self.trace("Operator reset to zero")

   def getCurrentOperator(self):
       """
       Returns the operator in its current state.
       """
       return self.__operator

   def setValue(self,**coefficients):
      """
      Sets new values to coefficients.

      @raise IllegalCoefficient: if an unknown coefficient keyword is used
      """
      # check if the coefficients are  legal:
      for i in coefficients.iterkeys():
         if not self.hasCoefficient(i):
            raise IllegalCoefficient,"Attempt to set unknown coefficient %s"%i
      # if the number of unknowns or equations is still unknown we try to estimate them:
      if self.__numEquations==None or self.__numSolutions==None:
         for i,d in coefficients.iteritems():
            if hasattr(d,"shape"):
                s=d.shape
            elif hasattr(d,"getShape"):
                s=d.getShape()
            else:
                s=numpy.array(d).shape
            if s!=None:
                # get number of equations and number of unknowns:
                res=self.__COEFFICIENTS[i].estimateNumEquationsAndNumSolutions(self.getDomain(),s)
                if res==None:
                    raise IllegalCoefficientValue,"Illegal shape %s of coefficient %s"%(s,i)
                else:
                    if self.__numEquations==None: self.__numEquations=res[0]
                    if self.__numSolutions==None: self.__numSolutions=res[1]
      if self.__numEquations==None: raise UndefinedPDEError,"unidentified number of equations"
      if self.__numSolutions==None: raise UndefinedPDEError,"unidentified number of solutions"
      # now we check the shape of the coefficient if numEquations and numSolutions are set:
      for i,d in coefficients.iteritems():
        try:
           self.__COEFFICIENTS[i].setValue(self.getDomain(),
                     self.getNumEquations(),self.getNumSolutions(),
                     self.reduceEquationOrder(),self.reduceSolutionOrder(),d)
           self.alteredCoefficient(i)
        except IllegalCoefficientFunctionSpace,m:
            # if the function space is wrong then we try the reduced version:
            i_red=i+"_reduced"
            if (not i_red in coefficients.keys()) and i_red in self.__COEFFICIENTS.keys():
                try:
                    self.__COEFFICIENTS[i_red].setValue(self.getDomain(),
                                                      self.getNumEquations(),self.getNumSolutions(),
                                                      self.reduceEquationOrder(),self.reduceSolutionOrder(),d)
                    self.alteredCoefficient(i_red)
                except IllegalCoefficientValue,m:
                    raise IllegalCoefficientValue("Coefficient %s:%s"%(i,m))
                except IllegalCoefficientFunctionSpace,m:
                    raise IllegalCoefficientFunctionSpace("Coefficient %s:%s"%(i,m))
            else:
                raise IllegalCoefficientFunctionSpace("Coefficient %s:%s"%(i,m))
        except IllegalCoefficientValue,m:
           raise IllegalCoefficientValue("Coefficient %s:%s"%(i,m))
      self.__altered_coefficients=True

   # ==========================================================================
   # methods that are typically overwritten when implementing a particular
   # linear problem
   # ==========================================================================
   def getRequiredOperatorType(self):
      """
      Returns the system type which needs to be used by the current set up.

      @note: Typically this method is overwritten when implementing a
             particular linear problem.
      """
      return None

   def createOperator(self):
       """
       Returns an instance of a new operator.

       @note: This method is overwritten when implementing a particular
              linear problem.
       """
       return escript.Operator()

   def checkSymmetry(self,verbose=True):
      """
      Tests the PDE for symmetry.

      @param verbose: if set to True or not present a report on coefficients
                      which break the symmetry is printed
      @type verbose: C{bool}
      @return: True if the problem is symmetric
      @rtype: C{bool}
      @note: Typically this method is overwritten when implementing a
             particular linear problem.
      """
      out=True
      return out

   def getSolution(self,**options):
       """
       Returns the solution of the problem.

       @return: the solution
       @rtype: L{Data<escript.Data>}

       @note: This method is overwritten when implementing a particular
              linear problem.
       """
       return self.getCurrentSolution()

   def getSystem(self):
       """
       Returns the operator and right hand side of the PDE.

       @return: the discrete version of the PDE
       @rtype: C{tuple} of L{Operator,<escript.Operator>} and L{Data<escript.Data>}.

       @note: This method is overwritten when implementing a particular
              linear problem.
       """
       return (self.getCurrentOperator(), self.getCurrentRightHandSide())

class LinearPDE(LinearProblem):
   """
   This class is used to define a general linear, steady, second order PDE
   for an unknown function M{u} on a given domain defined through a
   L{Domain<escript.Domain>} object.

   For a single PDE having a solution with a single component the linear PDE
   is defined in the following form:

   M{-(grad(A[j,l]+A_reduced[j,l])*grad(u)[l]+(B[j]+B_reduced[j])u)[j]+(C[l]+C_reduced[l])*grad(u)[l]+(D+D_reduced)=-grad(X+X_reduced)[j,j]+(Y+Y_reduced)}

   where M{grad(F)} denotes the spatial derivative of M{F}. Einstein's
   summation convention, ie. summation over indexes appearing twice in a term
   of a sum performed, is used.
   The coefficients M{A}, M{B}, M{C}, M{D}, M{X} and M{Y} have to be specified
   through L{Data<escript.Data>} objects in L{Function<escript.Function>} and
   the coefficients M{A_reduced}, M{B_reduced}, M{C_reduced}, M{D_reduced},
   M{X_reduced} and M{Y_reduced} have to be specified through
   L{Data<escript.Data>} objects in L{ReducedFunction<escript.ReducedFunction>}.
   It is also allowed to use objects that can be converted into such
   L{Data<escript.Data>} objects. M{A} and M{A_reduced} are rank two, M{B},
   M{C}, M{X}, M{B_reduced}, M{C_reduced} and M{X_reduced} are rank one and
   M{D}, M{D_reduced}, M{Y} and M{Y_reduced} are scalar.

   The following natural boundary conditions are considered:

   M{n[j]*((A[i,j]+A_reduced[i,j])*grad(u)[l]+(B+B_reduced)[j]*u)+(d+d_reduced)*u=n[j]*(X[j]+X_reduced[j])+y}

   where M{n} is the outer normal field. Notice that the coefficients M{A},
   M{A_reduced}, M{B}, M{B_reduced}, M{X} and M{X_reduced} are defined in the
   PDE. The coefficients M{d} and M{y} are each a scalar in
   L{FunctionOnBoundary<escript.FunctionOnBoundary>} and the coefficients
   M{d_reduced} and M{y_reduced} are each a scalar in
   L{ReducedFunctionOnBoundary<escript.ReducedFunctionOnBoundary>}.

   Constraints for the solution prescribe the value of the solution at certain
   locations in the domain. They have the form

   M{u=r} where M{q>0}

   M{r} and M{q} are each scalar where M{q} is the characteristic function
   defining where the constraint is applied. The constraints override any
   other condition set by the PDE or the boundary condition.

   The PDE is symmetrical if

   M{A[i,j]=A[j,i]}  and M{B[j]=C[j]} and M{A_reduced[i,j]=A_reduced[j,i]}
   and M{B_reduced[j]=C_reduced[j]}

   For a system of PDEs and a solution with several components the PDE has the
   form

   M{-grad((A[i,j,k,l]+A_reduced[i,j,k,l])*grad(u[k])[l]+(B[i,j,k]+B_reduced[i,j,k])*u[k])[j]+(C[i,k,l]+C_reduced[i,k,l])*grad(u[k])[l]+(D[i,k]+D_reduced[i,k]*u[k] =-grad(X[i,j]+X_reduced[i,j])[j]+Y[i]+Y_reduced[i] }

   M{A} and M{A_reduced} are of rank four, M{B}, M{B_reduced}, M{C} and
   M{C_reduced} are each of rank three, M{D}, M{D_reduced}, M{X_reduced} and
   M{X} are each of rank two and M{Y} and M{Y_reduced} are of rank one.
   The natural boundary conditions take the form:

   M{n[j]*((A[i,j,k,l]+A_reduced[i,j,k,l])*grad(u[k])[l]+(B[i,j,k]+B_reduced[i,j,k])*u[k])+(d[i,k]+d_reduced[i,k])*u[k]=n[j]*(X[i,j]+X_reduced[i,j])+y[i]+y_reduced[i]}

   The coefficient M{d} is of rank two and M{y} is of rank one both in
   L{FunctionOnBoundary<escript.FunctionOnBoundary>}. The coefficients
   M{d_reduced} is of rank two and M{y_reduced} is of rank one both in
   L{ReducedFunctionOnBoundary<escript.ReducedFunctionOnBoundary>}.

   Constraints take the form

   M{u[i]=r[i]}  where  M{q[i]>0}

   M{r} and M{q} are each rank one. Notice that at some locations not
   necessarily all components must have a constraint.

   The system of PDEs is symmetrical if

      - M{A[i,j,k,l]=A[k,l,i,j]}
      - M{A_reduced[i,j,k,l]=A_reduced[k,l,i,j]}
      - M{B[i,j,k]=C[k,i,j]}
      - M{B_reduced[i,j,k]=C_reduced[k,i,j]}
      - M{D[i,k]=D[i,k]}
      - M{D_reduced[i,k]=D_reduced[i,k]}
      - M{d[i,k]=d[k,i]}
      - M{d_reduced[i,k]=d_reduced[k,i]}

   L{LinearPDE} also supports solution discontinuities over a contact region
   in the domain. To specify the conditions across the discontinuity we are
   using the generalised flux M{J} which, in the case of a system of PDEs
   and several components of the solution, is defined as

   M{J[i,j]=(A[i,j,k,l]+A_reduced[[i,j,k,l])*grad(u[k])[l]+(B[i,j,k]+B_reduced[i,j,k])*u[k]-X[i,j]-X_reduced[i,j]}

   For the case of single solution component and single PDE M{J} is defined as

   M{J[j]=(A[i,j]+A_reduced[i,j])*grad(u)[j]+(B[i]+B_reduced[i])*u-X[i]-X_reduced[i]}

   In the context of discontinuities M{n} denotes the normal on the
   discontinuity pointing from side 0 towards side 1 calculated from
   L{getNormal<escript.FunctionSpace.getNormal>} of L{FunctionOnContactZero<escript.FunctionOnContactZero>}.
   For a system of PDEs the contact condition takes the form

   M{n[j]*J0[i,j]=n[j]*J1[i,j]=(y_contact[i]+y_contact_reduced[i])- (d_contact[i,k]+d_contact_reduced[i,k])*jump(u)[k]}

   where M{J0} and M{J1} are the fluxes on side 0 and side 1 of the
   discontinuity, respectively. M{jump(u)}, which is the difference of the
   solution at side 1 and at side 0, denotes the jump of M{u} across
   discontinuity along the normal calculated by L{jump<util.jump>}.
   The coefficient M{d_contact} is of rank two and M{y_contact} is of rank one
   both in L{FunctionOnContactZero<escript.FunctionOnContactZero>} or
   L{FunctionOnContactOne<escript.FunctionOnContactOne>}.
   The coefficient M{d_contact_reduced} is of rank two and M{y_contact_reduced}
   is of rank one both in L{ReducedFunctionOnContactZero<escript.ReducedFunctionOnContactZero>}
   or L{ReducedFunctionOnContactOne<escript.ReducedFunctionOnContactOne>}.
   In case of a single PDE and a single component solution the contact
   condition takes the form

   M{n[j]*J0_{j}=n[j]*J1_{j}=(y_contact+y_contact_reduced)-(d_contact+y_contact_reduced)*jump(u)}

   In this case the coefficient M{d_contact} and M{y_contact} are each scalar
   both in L{FunctionOnContactZero<escript.FunctionOnContactZero>} or
   L{FunctionOnContactOne<escript.FunctionOnContactOne>} and the coefficient
   M{d_contact_reduced} and M{y_contact_reduced} are each scalar both in
   L{ReducedFunctionOnContactZero<escript.ReducedFunctionOnContactZero>} or
   L{ReducedFunctionOnContactOne<escript.ReducedFunctionOnContactOne>}.

   Typical usage::

       p = LinearPDE(dom)
       p.setValue(A=kronecker(dom), D=1, Y=0.5)
       u = p.getSolution()

   """

   def __init__(self,domain,numEquations=None,numSolutions=None,debug=False):
     """
     Initializes a new linear PDE.

     @param domain: domain of the PDE
     @type domain: L{Domain<escript.Domain>}
     @param numEquations: number of equations. If C{None} the number of
                          equations is extracted from the PDE coefficients.
     @param numSolutions: number of solution components. If C{None} the number
                          of solution components is extracted from the PDE
                          coefficients.
     @param debug: if True debug information is printed

     """
     super(LinearPDE, self).__init__(domain,numEquations,numSolutions,debug)
     #
     #   the coefficients of the PDE:
     #
     self.introduceCoefficients(
       A=PDECoef(PDECoef.INTERIOR,(PDECoef.BY_EQUATION,PDECoef.BY_DIM,PDECoef.BY_SOLUTION,PDECoef.BY_DIM),PDECoef.OPERATOR),
       B=PDECoef(PDECoef.INTERIOR,(PDECoef.BY_EQUATION,PDECoef.BY_DIM,PDECoef.BY_SOLUTION),PDECoef.OPERATOR),
       C=PDECoef(PDECoef.INTERIOR,(PDECoef.BY_EQUATION,PDECoef.BY_SOLUTION,PDECoef.BY_DIM),PDECoef.OPERATOR),
       D=PDECoef(PDECoef.INTERIOR,(PDECoef.BY_EQUATION,PDECoef.BY_SOLUTION),PDECoef.OPERATOR),
       X=PDECoef(PDECoef.INTERIOR,(PDECoef.BY_EQUATION,PDECoef.BY_DIM),PDECoef.RIGHTHANDSIDE),
       Y=PDECoef(PDECoef.INTERIOR,(PDECoef.BY_EQUATION,),PDECoef.RIGHTHANDSIDE),
       d=PDECoef(PDECoef.BOUNDARY,(PDECoef.BY_EQUATION,PDECoef.BY_SOLUTION),PDECoef.OPERATOR),
       y=PDECoef(PDECoef.BOUNDARY,(PDECoef.BY_EQUATION,),PDECoef.RIGHTHANDSIDE),
       d_contact=PDECoef(PDECoef.CONTACT,(PDECoef.BY_EQUATION,PDECoef.BY_SOLUTION),PDECoef.OPERATOR),
       y_contact=PDECoef(PDECoef.CONTACT,(PDECoef.BY_EQUATION,),PDECoef.RIGHTHANDSIDE),
       A_reduced=PDECoef(PDECoef.INTERIOR_REDUCED,(PDECoef.BY_EQUATION,PDECoef.BY_DIM,PDECoef.BY_SOLUTION,PDECoef.BY_DIM),PDECoef.OPERATOR),
       B_reduced=PDECoef(PDECoef.INTERIOR_REDUCED,(PDECoef.BY_EQUATION,PDECoef.BY_DIM,PDECoef.BY_SOLUTION),PDECoef.OPERATOR),
       C_reduced=PDECoef(PDECoef.INTERIOR_REDUCED,(PDECoef.BY_EQUATION,PDECoef.BY_SOLUTION,PDECoef.BY_DIM),PDECoef.OPERATOR),
       D_reduced=PDECoef(PDECoef.INTERIOR_REDUCED,(PDECoef.BY_EQUATION,PDECoef.BY_SOLUTION),PDECoef.OPERATOR),
       X_reduced=PDECoef(PDECoef.INTERIOR_REDUCED,(PDECoef.BY_EQUATION,PDECoef.BY_DIM),PDECoef.RIGHTHANDSIDE),
       Y_reduced=PDECoef(PDECoef.INTERIOR_REDUCED,(PDECoef.BY_EQUATION,),PDECoef.RIGHTHANDSIDE),
       d_reduced=PDECoef(PDECoef.BOUNDARY_REDUCED,(PDECoef.BY_EQUATION,PDECoef.BY_SOLUTION),PDECoef.OPERATOR),
       y_reduced=PDECoef(PDECoef.BOUNDARY_REDUCED,(PDECoef.BY_EQUATION,),PDECoef.RIGHTHANDSIDE),
       d_contact_reduced=PDECoef(PDECoef.CONTACT_REDUCED,(PDECoef.BY_EQUATION,PDECoef.BY_SOLUTION),PDECoef.OPERATOR),
       y_contact_reduced=PDECoef(PDECoef.CONTACT_REDUCED,(PDECoef.BY_EQUATION,),PDECoef.RIGHTHANDSIDE),
       r=PDECoef(PDECoef.SOLUTION,(PDECoef.BY_SOLUTION,),PDECoef.RIGHTHANDSIDE),
       q=PDECoef(PDECoef.SOLUTION,(PDECoef.BY_SOLUTION,),PDECoef.BOTH) )

   def __str__(self):
     """
     Returns the string representation of the PDE.

     @return: a simple representation of the PDE
     @rtype: C{str}
     """
     return "<LinearPDE %d>"%id(self)

   def getRequiredOperatorType(self):
      """
      Returns the system type which needs to be used by the current set up.
      """
      solver_options=self.getSolverOptions()
      return self.getDomain().getSystemMatrixTypeId(solver_options.getSolverMethod(), solver_options.getPreconditioner(),solver_options.getPackage(), solver_options.isSymmetric())

   def checkSymmetry(self,verbose=True):
      """
      Tests the PDE for symmetry.

      @param verbose: if set to True or not present a report on coefficients
                      which break the symmetry is printed.
      @type verbose: C{bool}
      @return: True if the PDE is symmetric
      @rtype: L{bool}
      @note: This is a very expensive operation. It should be used for
             degugging only! The symmetry flag is not altered.
      """
      out=True
      out=out and self.checkSymmetricTensor("A", verbose)
      out=out and self.checkSymmetricTensor("A_reduced", verbose)
      out=out and self.checkReciprocalSymmetry("B","C", verbose)
      out=out and self.checkReciprocalSymmetry("B_reduced","C_reduced", verbose)
      out=out and self.checkSymmetricTensor("D", verbose)
      out=out and self.checkSymmetricTensor("D_reduced", verbose)
      out=out and self.checkSymmetricTensor("d", verbose)
      out=out and self.checkSymmetricTensor("d_reduced", verbose)
      out=out and self.checkSymmetricTensor("d_contact", verbose)
      out=out and self.checkSymmetricTensor("d_contact_reduced", verbose)
      return out

   def createOperator(self):
       """
       Returns an instance of a new operator.
       """
       optype=self.getRequiredOperatorType()
       self.trace("New operator of type %s is allocated."%optype)
       return self.getDomain().newOperator( \
                           self.getNumEquations(), \
                           self.getFunctionSpaceForEquation(), \
                           self.getNumSolutions(), \
                           self.getFunctionSpaceForSolution(), \
                           optype)

   def getSolution(self):
       """
       Returns the solution of the PDE.

       @return: the solution
       @rtype: L{Data<escript.Data>}
       """
       option_class=self.getSolverOptions()
       if not self.isSolutionValid():
          mat,f=self.getSystem()
          if self.isUsingLumping():
             self.setSolution(f*1/mat)
          else:
             self.trace("PDE is resolved.")
             self.trace("solver options: %s"%str(option_class))
             self.setSolution(mat.solve(f,option_class))
       return self.getCurrentSolution()

   def getSystem(self):
       """
       Returns the operator and right hand side of the PDE.

       @return: the discrete version of the PDE
       @rtype: C{tuple} of L{Operator,<escript.Operator>} and
               L{Data<escript.Data>}
       """
       if not self.isOperatorValid() or not self.isRightHandSideValid():
          if self.isUsingLumping():
              if not self.isOperatorValid():
                 if not self.getFunctionSpaceForEquation()==self.getFunctionSpaceForSolution():
                      raise TypeError,"Lumped matrix requires same order for equations and unknowns"
                 if not self.getCoefficient("A").isEmpty():
                      raise ValueError,"coefficient A in lumped matrix may not be present."
                 if not self.getCoefficient("B").isEmpty():
                      raise ValueError,"coefficient B in lumped matrix may not be present."
                 if not self.getCoefficient("C").isEmpty():
                      raise ValueError,"coefficient C in lumped matrix may not be present."
                 if not self.getCoefficient("d_contact").isEmpty():
                      raise ValueError,"coefficient d_contact in lumped matrix may not be present."
                 if not self.getCoefficient("A_reduced").isEmpty():
                      raise ValueError,"coefficient A_reduced in lumped matrix may not be present."
                 if not self.getCoefficient("B_reduced").isEmpty():
                      raise ValueError,"coefficient B_reduced in lumped matrix may not be present."
                 if not self.getCoefficient("C_reduced").isEmpty():
                      raise ValueError,"coefficient C_reduced in lumped matrix may not be present."
                 if not self.getCoefficient("d_contact_reduced").isEmpty():