gapt.provers.maxsat.MaxSAT.scala Maven / Gradle / Ivy

Go to download

Show more of this group Show more artifacts with this name
Show all versions of gapt_3 Show documentation

General Architecture for Proof Theory

The newest version!

package gapt.provers.maxsat

import gapt.expr._
import gapt.expr.formula.Formula
import gapt.expr.formula.hol._
import gapt.formats.dimacs._
import gapt.logic.hol.CNFp
import gapt.logic.hol.fastStructuralCNF
import gapt.models.PropositionalModel
import gapt.proofs.HOLClause
import gapt.utils.Logger

object MaxSATSolver {
  val logger = Logger("maxsat")
}

/**
 * Solver for Weighted Partial MaxSAT problems.
 */
abstract class MaxSATSolver {
  import MaxSATSolver.logger._

  /**
   * Solves a weighted partial MaxSAT problem.
   *
   * @param hard Hard constraints in CNF.
   * @param soft Soft constraints in CNF along with their weights.
   * @return None if hard is unsatisfiable, otherwise Some(model), where model is a model
   * of hard maximizing the sum of the weights of soft.
   */
  def solve(hard: DIMACS.CNF, soft: Seq[(DIMACS.Clause, Int)]): Option[DIMACS.Model]

  def solve(hard: Iterable[HOLClause], soft: Iterable[(HOLClause, Int)]): Option[PropositionalModel] = {
    val encoding = new DIMACSEncoding
    debug(s"${hard.size} hard clauses with ${hard.toSeq.map(_.size).sum} literals and ${hard.flatMap(_.elements).toSet.size} unique variables")
    solve(
      encoding.encodeCNF(hard),
      soft map { case (clause, weight) => encoding.encodeClause(clause) -> weight } toSeq
    ) map { dimacsModel =>
      encoding.decodeModel(dimacsModel)
    }
  }

  /**
   * @param hard Hard constraints.
   * @param soft Soft constraints along with their weights.
   * @return None if hard is unsatisfiable, otherwise Some(model), where model is a model
   * of hard maximizing the sum of the weights of soft.
   */
  def solve(hard: Formula, soft: Iterable[(Formula, Int)]): Option[PropositionalModel] = {
    solve(
      time("tseitin") { fastStructuralCNF()(hard)._1 },
      soft.map(s => CNFp(s._1).map(f => (f, s._2))).flatten
    )
  }
}