UP | HOME

Why Logic Programming?

Why? Always a good question, to understand logic programming, need to realize what we gain from it. Normally, if we want to solve a computational problem, we make a sequential command to get an answer. For example, what is Fibonacci's number at 3? We can have a racket program for this:

#lang racket

(define (fib n)
  (match n
    [0 1]
    [1 1]
    [n (+ (fib (- n 1))
          (fib (- n 2)))]))

(fib 3)

However, some questions aren't that easy to be resolved since need some synthesis, but we can take a look at how to resolve the Fibonacci problem via logic programming(use Datalog):

#lang datalog

(racket/base).

fib(0, 0).
fib(1, 1).

fib(N, F) :- N != 1,
            N != 0,
            N1 :- -(N, 1),
            N2 :- -(N, 2),
            fib(N1, F1),
            fib(N2, F2),
            F :- +(F1, F2).

fib(3, F)?

In Fibonacci this example they actually the same thing, but if I ask: g(x) is under f(x) follows the Big-O definition.

C : Real(\(\gt 0\)), N : Integer(\(\ge 0\)), \(\forall n \ge N, \exists C, g(N) < C \cdot f(N)\)

Does that still easy to answer? Solve this problem in Racket is really hard, but simple in Rosette:

#lang rosette/safe
(define-symbolic C N integer?)

(define (O f g)
  (solve
   (begin (assert (>= N 0))
          (assert (positive? C))
          (assert (< (g N)
                     (* C (f N)))))))

Can see all need to do is point out constraints. The only problem is we cannot use C : Real this definition since Real is not constructible. However, the power of logic programming already shows there. Hope you also like it ^{_}.

Date: 2020-07-15 Wed 00:00
Author: Lîm Tsú-thuàn