/// <summary>
/// Parses a z3 solution into a configuration.
/// This method also supports numeric variables.
/// </summary>
/// <param name="variables">List of all variables in the z3 context.</param>
/// <param name="model">Solution of the context.</param>
/// <param name="termToOption">Map from variables to binary options.</param>
/// <param name="optionsToConsider">The options that are considered for the solution.</param>
/// <returns>Configuration parsed from the solution.</returns>
public static Tuple <List <BinaryOption>, Dictionary <NumericOption, double> > RetrieveConfiguration(
List <Expr> variables, Model model, Dictionary <Expr, ConfigurationOption> termToOption,
List <ConfigurationOption> optionsToConsider = null)
{
List <BinaryOption> binOpts = new List <BinaryOption>();
Dictionary <NumericOption, double> config = new Dictionary <NumericOption, double>();
foreach (Expr variable in variables)
{
if (optionsToConsider != null && !optionsToConsider.Contains(termToOption[variable]))
{
continue;
}
Expr allocation = model.Eval(variable, completion: true);
if (allocation.GetType() == typeof(BoolExpr))
{
BoolExpr boolExpr = (BoolExpr)allocation;
if (boolExpr.IsTrue)
{
binOpts.Add((BinaryOption)termToOption[variable]);
}
}
else
{
// In this case, we have a numeric variable
FPNum fpNum = (FPNum)allocation;
config.Add((NumericOption)termToOption[variable], Z3Solver.lookUpNumericValue(fpNum.ToString()));
}
}
return(new Tuple <List <BinaryOption>, Dictionary <NumericOption, double> >(binOpts, config));
}
/// <summary>
/// The method aims at finding a configuration which is similar to the given configuration, but does not contain the optionToBeRemoved. If further options need to be removed from the given configuration, they are outputed in removedElements.
/// </summary>
/// <param name="optionToBeRemoved">The binary configuration option that must not be part of the new configuration.</param>
/// <param name="originalConfig">The configuration for which we want to find a similar one.</param>
/// <param name="removedElements">If further options need to be removed from the given configuration to build a valid configuration, they are outputed in this list.</param>
/// <param name="vm">The variability model containing all options and their constraints.</param>
/// <returns>A configuration that is valid, similar to the original configuration and does not contain the optionToBeRemoved.</returns>
public List <BinaryOption> GenerateConfigWithoutOption(BinaryOption optionToBeRemoved, List <BinaryOption> originalConfig, out List <BinaryOption> removedElements, VariabilityModel vm)
{
removedElements = new List <BinaryOption>();
var originalConfigWithoutRemoved = originalConfig.Where(x => !x.Equals(optionToBeRemoved));
List <BoolExpr> variables;
Dictionary <BoolExpr, BinaryOption> termToOption;
Dictionary <BinaryOption, BoolExpr> optionToTerm;
Tuple <Context, BoolExpr> z3Tuple = Z3Solver.GetInitializedBooleanSolverSystem(out variables, out optionToTerm, out termToOption, vm, this.henard);
Context z3Context = z3Tuple.Item1;
BoolExpr z3Constraints = z3Tuple.Item2;
List <BoolExpr> constraints = new List <BoolExpr>();
constraints.Add(z3Constraints);
constraints.Add(z3Context.MkNot(optionToTerm[optionToBeRemoved]));
ArithExpr[] minGoals = new ArithExpr[variables.Count];
for (int r = 0; r < variables.Count; r++)
{
BinaryOption currOption = termToOption[variables[r]];
ArithExpr numericVariable = z3Context.MkIntConst(currOption.Name);
int weight = -1000;
if (!originalConfigWithoutRemoved.Contains(currOption))
{
weight = 1000;
}
else if (currOption.Equals(optionToBeRemoved))
{
weight = 100000;
}
constraints.Add(z3Context.MkEq(numericVariable, z3Context.MkITE(variables[r], z3Context.MkInt(weight), z3Context.MkInt(0))));
minGoals[r] = numericVariable;
}
Optimize optimizer = z3Context.MkOptimize();
optimizer.Assert(constraints.ToArray());
optimizer.MkMinimize(z3Context.MkAdd(minGoals));
if (optimizer.Check() != Status.SATISFIABLE)
{
return(null);
}
else
{
Model model = optimizer.Model;
List <BinaryOption> similarConfig = RetrieveConfiguration(variables, model, termToOption);
removedElements = originalConfigWithoutRemoved.Except(similarConfig).ToList();
return(similarConfig);
}
}
/// <summary>
/// Based on a given (partial) configuration and a variability, we aim at finding all optimally maximal or minimal (in terms of selected binary options) configurations.
/// </summary>
/// <param name="config">The (partial) configuration which needs to be expanded to be valid.</param>
/// <param name="vm">Variability model containing all options and their constraints.</param>
/// <param name="minimize">If true, we search for the smallest (in terms of selected options) valid configuration. If false, we search for the largest one.</param>
/// <param name="unwantedOptions">Binary options that we do not want to become part of the configuration. Might be part if there is no other valid configuration without them</param>
/// <returns>A list of configurations that satisfies the VM and the goal (or null if there is none).</returns>
public List <List <BinaryOption> > MaximizeConfig(List <BinaryOption> config, VariabilityModel vm, bool minimize, List <BinaryOption> unwantedOptions)
{
List <BoolExpr> variables;
Dictionary <BoolExpr, BinaryOption> termToOption;
Dictionary <BinaryOption, BoolExpr> optionToTerm;
Tuple <Context, BoolExpr> z3Tuple = Z3Solver.GetInitializedBooleanSolverSystem(out variables, out optionToTerm, out termToOption, vm, this.henard);
Context z3Context = z3Tuple.Item1;
BoolExpr z3Constraints = z3Tuple.Item2;
List <BoolExpr> constraints = new List <BoolExpr>();
constraints.Add(z3Constraints);
List <BoolExpr> requireConfigs = new List <BoolExpr>();
if (config != null)
{
foreach (BinaryOption option in config)
{
requireConfigs.Add(optionToTerm[option]);
}
constraints.Add(z3Context.MkAnd(requireConfigs.ToArray()));
}
ArithExpr[] optimizationGoals = new ArithExpr[variables.Count];
for (int r = 0; r < variables.Count; r++)
{
BinaryOption currOption = termToOption[variables[r]];
ArithExpr numericVariable = z3Context.MkIntConst(currOption.Name);
constraints.Add(z3Context.MkEq(numericVariable, z3Context.MkITE(variables[r], z3Context.MkInt(1), z3Context.MkInt(0))));
optimizationGoals[r] = numericVariable;
}
Optimize optimizer = z3Context.MkOptimize();
optimizer.Assert(constraints.ToArray());
optimizer.MkMaximize(z3Context.MkAdd(optimizationGoals));
if (optimizer.Check() != Status.SATISFIABLE)
{
return(null);
}
List <BinaryOption> solution = RetrieveConfiguration(variables, optimizer.Model, termToOption);
return(new List <List <BinaryOption> >
{
solution
});
}
/// <summary>
/// Generates all valid combinations of all configuration options in the given model.
/// </summary>
/// <param name="vm">the variability model containing the binary options and their constraints</param>
/// <param name="optionsToConsider">the options that should be considered. All other options are ignored</param>
/// <returns>Returns a list of <see cref="Configuration"/></returns>
public List <Configuration> GenerateAllVariants(VariabilityModel vm, List <ConfigurationOption> optionsToConsider)
{
List <Configuration> allConfigurations = new List <Configuration>();
List <BoolExpr> variables;
Dictionary <BoolExpr, BinaryOption> termToOption;
Dictionary <BinaryOption, BoolExpr> optionToTerm;
Tuple <Context, BoolExpr> z3Tuple = Z3Solver.GetInitializedBooleanSolverSystem(out variables, out optionToTerm, out termToOption, vm, this.henard);
Context z3Context = z3Tuple.Item1;
BoolExpr z3Constraints = z3Tuple.Item2;
Microsoft.Z3.Solver solver = z3Context.MkSolver();
// TODO: The following line works for z3Solver version >= 4.6.0
//solver.Set (RANDOM_SEED, z3RandomSeed);
Params solverParameter = z3Context.MkParams();
solverParameter.Add(RANDOM_SEED, z3RandomSeed);
solver.Parameters = solverParameter;
solver.Assert(z3Constraints);
while (solver.Check() == Status.SATISFIABLE)
{
Model model = solver.Model;
List <BinaryOption> binOpts = RetrieveConfiguration(variables, model, termToOption, optionsToConsider);
Configuration c = new Configuration(binOpts);
// Check if the non-boolean constraints are satisfied
if (vm.configurationIsValid(c) && !VariantGenerator.IsInConfigurationFile(c, allConfigurations) && VariantGenerator.FulfillsMixedConstraints(c, vm))
{
allConfigurations.Add(c);
}
solver.Push();
solver.Assert(Z3Solver.NegateExpr(z3Context, Z3Solver.ConvertConfiguration(z3Context, binOpts, optionToTerm, vm)));
}
solver.Push();
solver.Pop(Convert.ToUInt32(allConfigurations.Count() + 1));
return(allConfigurations);
}
private void InitializeZ3Cache(VariabilityModel vm, int numberSelectedFeatures)
{
Dictionary <BoolExpr, BinaryOption> termToOption;
Microsoft.Z3.Solver solver;
Tuple <Context, BoolExpr> z3Tuple;
List <BoolExpr> variables;
Dictionary <BinaryOption, BoolExpr> optionToTerm;
Context z3Context;
z3Tuple = Z3Solver.GetInitializedBooleanSolverSystem(out variables, out optionToTerm, out termToOption, vm,
this.henard);
z3Context = z3Tuple.Item1;
BoolExpr z3Constraints = z3Tuple.Item2;
solver = z3Context.MkSolver();
// TODO: The following line works for z3Solver version >= 4.6.0
//solver.Set (RANDOM_SEED, z3RandomSeed);
Params solverParameter = z3Context.MkParams();
solverParameter.Add(RANDOM_SEED, z3RandomSeed);
solver.Parameters = solverParameter;
solver.Assert(z3Constraints);
// The goal of this method is, to have an exact number of features selected
// Therefore, initialize an integer array with the value '1' for the pseudo-boolean equal function
int[] neutralWeights = new int[variables.Count];
for (int i = 0; i < variables.Count; i++)
{
neutralWeights[i] = 1;
}
solver.Assert(z3Context.MkPBEq(neutralWeights, variables.ToArray(), numberSelectedFeatures));
// Create a backtracking point before adding the optimization goal
solver.Push();
this._z3Cache[numberSelectedFeatures] = new Z3Cache(z3Context, solver, variables, optionToTerm, termToOption);
}
public bool checkConfigurationSAT(List <BinaryOption> config, VariabilityModel vm, bool partialConfiguration)
{
List <BoolExpr> variables;
Dictionary <BoolExpr, BinaryOption> termToOption;
Dictionary <BinaryOption, BoolExpr> optionToTerm;
Tuple <Context, BoolExpr> z3Tuple = Z3Solver.GetInitializedBooleanSolverSystem(out variables, out optionToTerm, out termToOption, vm, false);
Context z3Context = z3Tuple.Item1;
BoolExpr z3Constraints = z3Tuple.Item2;
List <BoolExpr> constraints = new List <BoolExpr>();
Microsoft.Z3.Solver solver = z3Context.MkSolver();
solver.Assert(z3Constraints);
solver.Assert(Z3Solver.ConvertConfiguration(z3Context, config, optionToTerm, vm, partialConfiguration));
if (solver.Check() == Status.SATISFIABLE)
{
return(true);
}
return(false);
}
/// <summary>
/// Creates a sample of configurations, by iteratively adding a configuration that has the maximal manhattan distance
/// to the configurations that were previously selected.
/// </summary>
/// <param name="vm">The domain for sampling.</param>
/// <param name="minimalConfiguration">A minimal configuration that will be used as starting point.</param>
/// <param name="numberToSample">The number of configurations that should be sampled.</param>
/// <param name="optionWeight">Weight assigned to optional binary options.</param>
/// <returns>A list of distance maximized configurations.</returns>
public List <List <BinaryOption> > DistanceMaximization(VariabilityModel vm, List <BinaryOption> minimalConfiguration, int numberToSample, int optionWeight)
{
List <BoolExpr> variables;
Dictionary <BoolExpr, BinaryOption> termToOption;
Dictionary <BinaryOption, BoolExpr> optionToTerm;
Tuple <Context, BoolExpr> z3Tuple = Z3Solver.GetInitializedBooleanSolverSystem(out variables, out optionToTerm, out termToOption, vm, this.henard);
Context z3Context = z3Tuple.Item1;
BoolExpr z3Constraints = z3Tuple.Item2;
List <BoolExpr> constraints = new List <BoolExpr>();
constraints.Add(z3Constraints);
List <List <BinaryOption> > sample = new List <List <BinaryOption> >();
sample.Add(minimalConfiguration);
Dictionary <BinaryOption, ArithExpr> optionToInt = generateIntConstants(z3Context, constraints, variables, termToOption);
while (numberToSample > sample.Count)
{
List <ArithExpr> goals = generateDistMaximizationGoals(sample, optionToInt, z3Context, vm, optionWeight);
Optimize optimizer = z3Context.MkOptimize();
optimizer.Assert(constraints);
optimizer.MkMaximize(z3Context.MkAdd(goals));
if (optimizer.Check() != Status.SATISFIABLE)
{
GlobalState.logInfo.logLine("No more solutions available.");
return(sample);
}
else
{
Model model = optimizer.Model;
List <BinaryOption> maxDist = RetrieveConfiguration(variables, model, termToOption);
sample.Add(maxDist);
constraints.Add(z3Context.MkNot(Z3Solver.ConvertConfiguration(z3Context, maxDist, optionToTerm, vm)));
}
}
return(sample);
}
/// <summary>
/// Generates all valid combinations of all configuration options in the given model.
/// </summary>
/// <param name="vm">the variability model containing the binary options and their constraints</param>
/// <param name="optionsToConsider">the options that should be considered. All other options are ignored</param>
/// <returns>Returns a list of <see cref="Configuration"/></returns>
public List <Configuration> GenerateAllVariants(VariabilityModel vm, List <ConfigurationOption> optionsToConsider)
{
List <Configuration> allConfigurations = new List <Configuration>();
List <Expr> variables;
Dictionary <Expr, ConfigurationOption> termToOption;
Dictionary <ConfigurationOption, Expr> optionToTerm;
Tuple <Context, BoolExpr> z3Tuple = Z3Solver.GetInitializedSolverSystem(out variables, out optionToTerm, out termToOption, vm);
Context z3Context = z3Tuple.Item1;
BoolExpr z3Constraints = z3Tuple.Item2;
Microsoft.Z3.Solver solver = z3Context.MkSolver();
solver.Set(RANDOM_SEED, z3RandomSeed);
solver.Assert(z3Constraints);
while (solver.Check() == Status.SATISFIABLE)
{
Model model = solver.Model;
Tuple <List <BinaryOption>, Dictionary <NumericOption, double> > confOpts = RetrieveConfiguration(variables, model, termToOption, optionsToConsider);
Configuration c = new Configuration(confOpts.Item1, confOpts.Item2);
// Check if the non-boolean constraints are satisfied
bool configIsValid = vm.configurationIsValid(c);
bool isInConfigurationFile = !VariantGenerator.IsInConfigurationFile(c, allConfigurations);
bool fulfillsMixedConstraintrs = VariantGenerator.FulfillsMixedConstraints(c, vm);
if (configIsValid && isInConfigurationFile && fulfillsMixedConstraintrs)
{
allConfigurations.Add(c);
}
solver.Push();
solver.Assert(Z3Solver.NegateExpr(z3Context, Z3Solver.ConvertConfiguration(z3Context, confOpts.Item1, optionToTerm, vm, numericValues: confOpts.Item2)));
}
solver.Push();
solver.Pop(Convert.ToUInt32(allConfigurations.Count() + 1));
return(allConfigurations);
}
public bool checkConfigurationSAT(Configuration c, VariabilityModel vm, bool partialConfiguration = false)
{
List <Expr> variables;
Dictionary <Expr, ConfigurationOption> termToOption;
Dictionary <ConfigurationOption, Expr> optionToTerm;
Tuple <Context, BoolExpr> z3Tuple = Z3Solver.GetInitializedSolverSystem(out variables, out optionToTerm, out termToOption, vm);
Context z3Context = z3Tuple.Item1;
BoolExpr z3Constraints = z3Tuple.Item2;
List <Expr> constraints = new List <Expr>();
Microsoft.Z3.Solver solver = z3Context.MkSolver();
solver.Assert(z3Constraints);
solver.Assert(Z3Solver.ConvertConfiguration(z3Context, c.getBinaryOptions(BinaryOption.BinaryValue.Selected), optionToTerm, vm, partialConfiguration, c.NumericOptions));
if (solver.Check() == Status.SATISFIABLE)
{
return(true);
}
return(false);
}
/// <summary>
/// Adds the given configuration as a constraint to the solver.
/// </summary>
/// <param name="vm">The variability model</param>
/// <param name="configToExclude">The configuration to exclude</param>
/// <param name="numberSelectedFeatures">The number of features to be selected</param>
public void AddConstraint(VariabilityModel vm, Configuration configToExclude, int numberSelectedFeatures)
{
if (!constraintsCache.ContainsKey(numberSelectedFeatures))
{
constraintsCache[numberSelectedFeatures] = new HashSet <Configuration>();
}
if (constraintsCache[numberSelectedFeatures].Contains(configToExclude))
{
return;
}
constraintsCache[numberSelectedFeatures].Add(configToExclude);
if (_z3Cache == null)
{
_z3Cache = new Dictionary <int, Z3Cache>();
}
Dictionary <BinaryOption, BoolExpr> optionToTerm;
Context z3Context;
Microsoft.Z3.Solver solver;
if (!this._z3Cache.Keys.Contains(numberSelectedFeatures))
{
InitializeZ3Cache(vm, numberSelectedFeatures);
}
Z3Cache cache = this._z3Cache[numberSelectedFeatures];
z3Context = cache.GetContext();
solver = cache.GetSolver();
optionToTerm = cache.GetOptionToTermMapping();
// Add the previous configurations as constraints
solver.Assert(Z3Solver.NegateExpr(z3Context,
Z3Solver.ConvertConfiguration(z3Context,
configToExclude.getBinaryOptions(BinaryOption.BinaryValue.Selected), optionToTerm, vm)));
// Create a new backtracking point for the next run
solver.Push();
}
public List <BinaryOption> GenerateConfigurationFromBucket(VariabilityModel vm, int numberSelectedFeatures, Dictionary <List <BinaryOption>, int> featureWeight, Configuration lastSampledConfiguration)
{
if (_z3Cache == null)
{
_z3Cache = new Dictionary <int, Z3Cache>();
}
List <KeyValuePair <List <BinaryOption>, int> > featureRanking;
if (featureWeight != null)
{
featureRanking = featureWeight.ToList();
featureRanking.Sort((first, second) => first.Value.CompareTo(second.Value));
}
else
{
featureRanking = new List <KeyValuePair <List <BinaryOption>, int> >();
}
List <BoolExpr> variables = null;
Dictionary <BoolExpr, BinaryOption> termToOption = null;
Dictionary <BinaryOption, BoolExpr> optionToTerm = null;
Tuple <Context, BoolExpr> z3Tuple;
Context z3Context;
Microsoft.Z3.Solver solver;
// Reuse the solver if it is already in the cache
if (this._z3Cache.Keys.Contains(numberSelectedFeatures))
{
Z3Cache cache = this._z3Cache[numberSelectedFeatures];
z3Context = cache.GetContext();
solver = cache.GetSolver();
variables = cache.GetVariables();
termToOption = cache.GetTermToOptionMapping();
optionToTerm = cache.GetOptionToTermMapping();
if (lastSampledConfiguration != null)
{
// Add the previous configurations as constraints
solver.Assert(Z3Solver.NegateExpr(z3Context, Z3Solver.ConvertConfiguration(z3Context, lastSampledConfiguration.getBinaryOptions(BinaryOption.BinaryValue.Selected), optionToTerm, vm)));
// Create a new backtracking point for the next run
solver.Push();
}
}
else
{
z3Tuple = Z3Solver.GetInitializedBooleanSolverSystem(out variables, out optionToTerm, out termToOption, vm, this.henard);
z3Context = z3Tuple.Item1;
BoolExpr z3Constraints = z3Tuple.Item2;
solver = z3Context.MkSolver();
// TODO: The following line works for z3Solver version >= 4.6.0
//solver.Set (RANDOM_SEED, z3RandomSeed);
Params solverParameter = z3Context.MkParams();
solverParameter.Add(RANDOM_SEED, z3RandomSeed);
solver.Parameters = solverParameter;
solver.Assert(z3Constraints);
if (lastSampledConfiguration != null)
{
// Add the previous configurations as constraints
solver.Assert(Z3Solver.NegateExpr(z3Context, Z3Solver.ConvertConfiguration(z3Context, lastSampledConfiguration.getBinaryOptions(BinaryOption.BinaryValue.Selected), optionToTerm, vm)));
}
// The goal of this method is, to have an exact number of features selected
// Therefore, initialize an integer array with the value '1' for the pseudo-boolean equal function
int[] neutralWeights = new int[variables.Count];
for (int i = 0; i < variables.Count; i++)
{
neutralWeights[i] = 1;
}
solver.Assert(z3Context.MkPBEq(neutralWeights, variables.ToArray(), numberSelectedFeatures));
// Create a backtracking point before adding the optimization goal
solver.Push();
this._z3Cache[numberSelectedFeatures] = new Z3Cache(z3Context, solver, variables, optionToTerm, termToOption);
}
// Check if there is still a solution available by finding the first satisfiable configuration
if (solver.Check() == Status.SATISFIABLE)
{
Model model = solver.Model;
List <BinaryOption> possibleSolution = RetrieveConfiguration(variables, model, termToOption);
// Disable finding a configuration where the least frequent feature/feature combinations are selected
// if no featureWeight is given.
List <BinaryOption> approximateOptimal = null;
if (featureRanking.Count != 0)
{
approximateOptimal = WeightMinimizer
.getSmallWeightConfig(featureRanking, this._z3Cache[numberSelectedFeatures], vm);
}
if (approximateOptimal == null)
{
return(possibleSolution);
}
else
{
return(approximateOptimal);
}
}
else
{
return(null);
}
}
/// <summary>
/// This method searches for a corresponding methods in the dynamically loaded assemblies and calls it if found. It prefers due to performance reasons the Microsoft Solver Foundation implementation.
/// </summary>
/// <param name="config">The (partial) configuration which needs to be expaned to be valid.</param>
/// <param name="vm">Variability model containing all options and their constraints.</param>
/// <param name="minimize">If true, we search for the smallest (in terms of selected options) valid configuration. If false, we search for the largest one.</param>
/// <param name="unWantedOptions">Binary options that we do not want to become part of the configuration. Might be part if there is no other valid configuration without them.</param>
/// <returns>The valid configuration (or null if there is none) that satisfies the VM and the goal.</returns>
public List <BinaryOption> MinimizeConfig(List <BinaryOption> config, VariabilityModel vm, bool minimize, List <BinaryOption> unWantedOptions)
{
List <BoolExpr> variables;
Dictionary <BoolExpr, BinaryOption> termToOption;
Dictionary <BinaryOption, BoolExpr> optionToTerm;
Tuple <Context, BoolExpr> z3Tuple = Z3Solver.GetInitializedBooleanSolverSystem(out variables, out optionToTerm, out termToOption, vm, this.henard);
Context z3Context = z3Tuple.Item1;
BoolExpr z3Constraints = z3Tuple.Item2;
List <BoolExpr> constraints = new List <BoolExpr>();
constraints.Add(z3Constraints);
//Feature Selection
foreach (BinaryOption binOpt in config)
{
BoolExpr term = optionToTerm[binOpt];
constraints.Add(term);
}
Model model = null;
if (minimize == true)
{
//Defining Goals
ArithExpr[] optimizationGoals = new ArithExpr[variables.Count];
for (int r = 0; r < variables.Count; r++)
{
BinaryOption currOption = termToOption[variables[r]];
ArithExpr numericVariable = z3Context.MkIntConst(currOption.Name);
int weight = 1;
if (unWantedOptions != null && (unWantedOptions.Contains(termToOption[variables[r]]) && !config.Contains(termToOption[variables[r]])))
{
weight = 1000;
}
constraints.Add(z3Context.MkEq(numericVariable, z3Context.MkITE(variables[r], z3Context.MkInt(weight), z3Context.MkInt(0))));
optimizationGoals[r] = numericVariable;
}
// For minimization, we need the class 'Optimize'
Optimize optimizer = z3Context.MkOptimize();
optimizer.Assert(constraints.ToArray());
optimizer.MkMinimize(z3Context.MkAdd(optimizationGoals));
if (optimizer.Check() != Status.SATISFIABLE)
{
return(new List <BinaryOption>());
}
else
{
model = optimizer.Model;
}
}
else
{
// Return the first configuration returned by the solver
Microsoft.Z3.Solver solver = z3Context.MkSolver();
// TODO: The following line works for z3Solver version >= 4.6.0
//solver.Set (RANDOM_SEED, z3RandomSeed);
Params solverParameter = z3Context.MkParams();
solverParameter.Add(RANDOM_SEED, z3RandomSeed);
solver.Parameters = solverParameter;
solver.Assert(constraints.ToArray());
if (solver.Check() != Status.SATISFIABLE)
{
return(new List <BinaryOption>());
}
else
{
model = solver.Model;
}
}
List <BinaryOption> result = RetrieveConfiguration(variables, model, termToOption);
return(result);
}
/// <summary>
/// Based on a given (partial) configuration and a variability, we aim at finding all optimally maximal or minimal (in terms of selected binary options) configurations.
/// </summary>
/// <param name="config">The (partial) configuration which needs to be expaned to be valid.</param>
/// <param name="vm">Variability model containing all options and their constraints.</param>
/// <param name="minimize">If true, we search for the smallest (in terms of selected options) valid configuration. If false, we search for the largest one.</param>
/// <param name="unwantedOptions">Binary options that we do not want to become part of the configuration. Might be part if there is no other valid configuration without them</param>
/// <returns>A list of configurations that satisfies the VM and the goal (or null if there is none).</returns>
public List <List <BinaryOption> > MaximizeConfig(List <BinaryOption> config, VariabilityModel vm, bool minimize, List <BinaryOption> unwantedOptions)
{
List <List <BinaryOption> > optimalConfigurations = new List <List <BinaryOption> >();
List <BoolExpr> variables;
Dictionary <BoolExpr, BinaryOption> termToOption;
Dictionary <BinaryOption, BoolExpr> optionToTerm;
Tuple <Context, BoolExpr> z3Tuple = Z3Solver.GetInitializedBooleanSolverSystem(out variables, out optionToTerm, out termToOption, vm, this.henard);
Context z3Context = z3Tuple.Item1;
BoolExpr z3Constraints = z3Tuple.Item2;
List <BoolExpr> constraints = new List <BoolExpr>();
constraints.Add(z3Constraints);
List <BoolExpr> requireConfigs = new List <BoolExpr>();
if (config != null)
{
foreach (BinaryOption option in config)
{
requireConfigs.Add(optionToTerm[option]);
}
constraints.Add(z3Context.MkAnd(requireConfigs.ToArray()));
}
ArithExpr[] optimizationGoals = new ArithExpr[variables.Count];
for (int r = 0; r < variables.Count; r++)
{
BinaryOption currOption = termToOption[variables[r]];
ArithExpr numericVariable = z3Context.MkIntConst(currOption.Name);
int weight;
if (minimize)
{
weight = 1;
}
else
{
weight = -1;
}
if (unwantedOptions != null && (unwantedOptions.Contains(termToOption[variables[r]]) && !config.Contains(termToOption[variables[r]])))
{
weight = 10000;
}
constraints.Add(z3Context.MkEq(numericVariable, z3Context.MkITE(variables[r], z3Context.MkInt(weight), z3Context.MkInt(0))));
optimizationGoals[r] = numericVariable;
}
Optimize optimizer = z3Context.MkOptimize();
optimizer.Assert(constraints.ToArray());
optimizer.MkMinimize(z3Context.MkAdd(optimizationGoals));
int bestSize = 0;
int currentSize = 0;
while (optimizer.Check() == Status.SATISFIABLE && currentSize >= bestSize)
{
Model model = optimizer.Model;
List <BinaryOption> solution = RetrieveConfiguration(variables, model, termToOption);
currentSize = solution.Count;
if (currentSize >= bestSize)
{
optimalConfigurations.Add(solution);
}
if (bestSize == 0)
{
bestSize = solution.Count;
}
currentSize = solution.Count;
optimizer.Assert(z3Context.MkNot(Z3Solver.ConvertConfiguration(z3Context, solution, optionToTerm, vm)));
}
return(optimalConfigurations);
}
/// <summary>
/// Generates up to n solutions of the given variability model.
/// Note that this method could also generate less than n solutions if the variability model does not contain sufficient solutions.
/// Moreover, in the case that <code>n < 0</code>, all solutions are generated.
/// </summary>
/// <param name="vm">The <see cref="VariabilityModel"/> to obtain solutions for.</param>
/// <param name="n">The number of solutions to obtain.</param>
/// <returns>A list of configurations, in which a configuration is a list of SELECTED binary options.</returns>
public List <List <BinaryOption> > GenerateUpToNFast(VariabilityModel vm, int n)
{
// Use the random seed to produce new random seeds
Random random = new Random(Convert.ToInt32(z3RandomSeed));
List <BoolExpr> variables;
Dictionary <BoolExpr, BinaryOption> termToOption;
Dictionary <BinaryOption, BoolExpr> optionToTerm;
Tuple <Context, BoolExpr> z3Tuple = Z3Solver.GetInitializedBooleanSolverSystem(out variables, out optionToTerm, out termToOption, vm, this.henard, random.Next());
Context z3Context = z3Tuple.Item1;
BoolExpr z3Constraints = z3Tuple.Item2;
List <BoolExpr> excludedConfigurations = new List <BoolExpr>();
List <BoolExpr> constraints = Z3Solver.lastConstraints;
List <List <BinaryOption> > configurations = new List <List <BinaryOption> >();
Microsoft.Z3.Solver s = z3Context.MkSolver();
// TODO: The following line works for z3Solver version >= 4.6.0
//solver.Set (RANDOM_SEED, z3RandomSeed);
Params solverParameter = z3Context.MkParams();
if (henard)
{
solverParameter.Add(RANDOM_SEED, NextUInt(random));
}
else
{
solverParameter.Add(RANDOM_SEED, z3RandomSeed);
}
s.Parameters = solverParameter;
s.Assert(z3Constraints);
s.Push();
Model model = null;
while (s.Check() == Status.SATISFIABLE && (configurations.Count < n || n < 0))
{
model = s.Model;
List <BinaryOption> config = RetrieveConfiguration(variables, model, termToOption);
configurations.Add(config);
if (henard)
{
BoolExpr newConstraint = Z3Solver.NegateExpr(z3Context, Z3Solver.ConvertConfiguration(z3Context, config, optionToTerm, vm));
excludedConfigurations.Add(newConstraint);
Dictionary <BoolExpr, BinaryOption> oldTermToOption = termToOption;
// Now, initialize a new one for the next configuration
z3Tuple = Z3Solver.GetInitializedBooleanSolverSystem(out variables, out optionToTerm, out termToOption, vm, this.henard, random.Next());
z3Context = z3Tuple.Item1;
z3Constraints = z3Tuple.Item2;
s = z3Context.MkSolver();
//s.Set (RANDOM_SEED, NextUInt (random));
solverParameter = z3Context.MkParams();
solverParameter.Add(RANDOM_SEED, NextUInt(random));
s.Parameters = solverParameter;
constraints = Z3Solver.lastConstraints;
excludedConfigurations = Z3Solver.ConvertConstraintsToNewContext(oldTermToOption, optionToTerm, excludedConfigurations, z3Context);
constraints.AddRange(excludedConfigurations);
s.Assert(z3Context.MkAnd(Z3Solver.Shuffle(constraints, new Random(random.Next()))));
s.Push();
}
else
{
s.Add(Z3Solver.NegateExpr(z3Context, Z3Solver.ConvertConfiguration(z3Context, config, optionToTerm, vm)));
}
}
return(configurations);
}
