7

The sample program enumerates and counts the number of 8-queen solutions. (sorry if the code is hard to read; this is machine-generated from an S-expression. The original code is https://www.cpp.edu/~jrfisher/www/prolog_tutorial/2_11.html)

rules:

[user].
(perm([X|Y],Z) :- (perm(Y,W),takeout(X,Z,W))).
perm([],[]).
takeout(X,[X|R],R).
(takeout(X,[F|R],[F|S]) :- (takeout(X,R,S))).
(solve(P) :- (perm([1,2,3,4,5,6,7,8],P),combine([1,2,3,4,5,6,7,8],P,S,D),alldiff(S),alldiff(D))).
(combine([X1|X],[Y1|Y],[S1|S],[D1|D]) :- (is(S1,+(X1,Y1)),is(D1,-(X1,Y1)),combine(X,Y,S,D))).
combine([],[],[],[]).
(alldiff([X|Y]) :- (\+ member(X,Y),alldiff(Y))).
alldiff([X]).
end_of_file.

query:

(setof(P,solve(P),Set),length(Set,L),write(L),write('\n'),fail).

swipl returns 92; while yap returns 40320. Also, when I query solve(P), swipl only returns two solutions (which also contradicts 92); yap returns much more (possibly 40320 of them). So why the difference? Is there such a serious compatibility issue?

Versions:

  • YAP 6.2.2 (x86_64-linux): Sat Sep 17 13:59:03 UTC 2016
  • SWI-Prolog version 7.2.3 for amd64
Erik Kaplun
  • 37,128
  • 15
  • 99
  • 111
Asai Masataro
  • 184
  • 5

1 Answers1

8

In older versions of YAP, a query for an undefined predicate simply failed. In the case above, it is member/2 which is not defined in YAP. And for this reason your test alldif/1 always succeeds - thus the large number you get.

The behavior for this is governed by the Prolog flag unknown whose default value should be error. In YAP 6.2 the default was (incorrectly) fail. This was corrected in 6.3. Say

:- set_prolog_flag(unknown, error).

to get a clean error for undefined predicates. Then, you would need to define member/2.

false
  • 10,264
  • 13
  • 101
  • 209
  • 2
    thank you, so `member` is not the built-ins... but those "prologues" (compatibility layer, sort of) sounds good. I can make the lisp-to-prolog library to load it whenever the process is launched. – Asai Masataro Nov 10 '17 at 13:51
  • hmm, by the way, just cloned yap 6.3.3, after the `.configure` it doesn't build because of `make: *** No rule to make target 'H/dglobals.h', needed by 'yap'. Stop.` ... – Asai Masataro Nov 10 '17 at 14:31
  • Wondering why the standardization is going so bad compared to the other CSP community. Minizinc is a common format in CSP and is doing the competition annually. SAT has DIMACS and SMT has one as well. – Asai Masataro Nov 10 '17 at 14:35
  • 3
    @Asai: The transition from 6.2->6.3 is rather a case of confluence. – false Nov 10 '17 at 14:39