Restrictions on clp(q/r) (Division)

General
1

I have been trying to apply library(clpq) (same with clpr) to this problem: https://www.youtube.com/watch?v=ABj3IC9pYlQ

You don’t need to watch this, the question which is posed is this:

Alice and Bob take 2 hours to finish a job. Alice and Charlie take three hours. Bob and Charlie take four hours. How long does it take when they work all together?

In the end, I found out that it worked with the formulation of the equations that Presh uses in the video, namely:

?- {1=2*(A+B), 3*(A+C)=1, 4*(B+C)=1, X*(A+B+C)=1}.
A = 7r24,
B = 5r24,
C = 1r24,
X = 24r13.

However, I had originally written the equations like 1/(A+B)=2 instead of 1=2*(A+B). This doesn’t work at all:

?- {1/(A+B)=2, 3*(A+C)=1, 4*(B+C)=1, X*(A+B+C)=1}.
{B=1r4-C, A=1r3-C, -1+A*X+B*X+C*X=0, -2+1/(A+B)=0},
{-1+A*X+B*X+C*X=0, -2+1/(A+B)=0},
{-1+A*X+B*X+C*X=0, -2+1/(A+B)=0},
{-1+A*X+B*X+C*X=0, -2+1/(A+B)=0},
{-1+A*X+B*X+C*X=0, -2+1/(A+B)=0}.

?- {1/(A+B)=2, 3*(A+C)=1, 4*(B+C)=1, X*(A+B+C)=1}, minimize(X).
false.

Being unexperienced with this library, that surprised me, because I would have thought that constraints would be converted to some canonical form anyway. Should this be expected? Perhaps some kind of warning in the documentation would be appropriate?

2

Just as a (hopefully interesting) aside, clpBNR works for both:

?- {1==2*(A+B), 3*(A+C)==1, 4*(B+C)==1, X*(A+B+C)==1}, solve([A,B,C,X]).
A:: 0.2916666...,
B:: 0.2083333...,
C:: 0.0416666...,
X:: 1.846153... .

?- {1/(A+B)==2, 3*(A+C)==1, 4*(B+C)==1, X*(A+B+C)==1}, solve([A,B,C,X]).
A:: 0.2916666...,
B:: 0.2083333...,
C:: 0.0416666...,
X:: 1.846153... .