如何逐步使用z3和没有命题值的模型?

时间:2012-12-19 19:17:57

标签: model z3

您能否告诉我如何逐步使用求解器Z3? 而且,当我使用v.name()时,如何才能获得没有命题值的模型? 例如,在调用程序cout<<v.name()<<m.get_const_interp(v);之后,我们可以得到模型 x = 3, p = true, y = 4,因为我不需要p = true,我可以从模型集中删除吗?

1 个答案:

答案 0 :(得分:6)

我添加了新的C ++示例,演示了如何使用Z3 C ++ API进行增量求解。新示例已经可用online。我在帖子的末尾复制了这些例子。

关于第二个问题,在Z3中,模型本质上是只读对象。您可以简单地忽略您不关心的值。您还可以为隐藏不需要的值的模型对象编写自己的包装器。

void incremental_example1() {
    std::cout << "incremental example1\n";
    context c;
    expr x = c.int_const("x");
    solver s(c);
    s.add(x > 0);
    std::cout << s.check() << "\n";
    // We can add more formulas to the solver
    s.add(x < 0);
    // and, invoke s.check() again...
    std::cout << s.check() << "\n";
}

void incremental_example2() {
    // In this example, we show how push() and pop() can be used
    // to remove formulas added to the solver.
    std::cout << "incremental example2\n";
    context c;
    expr x = c.int_const("x");
    solver s(c);
    s.add(x > 0);
    std::cout << s.check() << "\n";
    // push() creates a backtracking point (aka a snapshot).
    s.push();
    // We can add more formulas to the solver
    s.add(x < 0);
    // and, invoke s.check() again...
    std::cout << s.check() << "\n";
    // pop() will remove all formulas added between this pop() and the matching push()
    s.pop();
    // The context is satisfiable again
    std::cout << s.check() << "\n";
    // and contains only x > 0
    std::cout << s << "\n";
}

void incremental_example3() {
    // In this example, we show how to use assumptions to "remove" 
    // formulas added to a solver. Actually, we disable them.
    std::cout << "incremental example3\n";
    context c;
    expr x = c.int_const("x");
    solver s(c);
    s.add(x > 0);
    std::cout << s.check() << "\n";
    // Now, suppose we want to add x < 0 to the solver, but we also want
    // to be able to disable it later.
    // To do that, we create an auxiliary Boolean variable
    expr b = c.bool_const("b");
    // and, assert (b implies x < 0)
    s.add(implies(b, x < 0));
    // Now, we check whether s is satisfiable under the assumption "b" is true.
    expr_vector a1(c);
    a1.push_back(b);
    std::cout << s.check(a1) << "\n";
    // To "disable" (x > 0), we may just ask with the assumption "not b" or not provide any assumption.
    std::cout << s.check() << "\n";
    expr_vector a2(c);
    a2.push_back(!b);
    std::cout << s.check(a2) << "\n";
}