No canonical answer after permutation of goals

[notoria]

?- abs(#X) #= #Y #<==> #B, B = 0, Y = -1.
   Y = -1, B = 0, clpz:(#_A#=abs(#X)), clpz:(_A in 0..sup).
?- abs(#X) #= #Y #<==> #B, Y = -1, B = 0.
   Y = -1, B = 0, clpz:(X in inf..sup).
?-

The answers are equivalent but I expected a canonical answer after permutation of goals.

[UWN]

As a general remark, you cannot expect this from a constraint system. The way how this is internally implemented is entirely implementation dependent.

What you do get here is at least some representation of the constraint store such that reposting will produce a similar store admitting the very same solutions (but not necessarily the same leaf answers). Rare are the systems that provide that functionality at all (That's SICStus' clpfd and clpz).

Further, some redundant domains of the form inf..sup are not shown.

[notoria]

True but in this case there is a performance issue, the first query requires more resources.

This also happens without reification:

?- #D#= #A* #B^ #C, D= -3,A= -3,B=2.
   D = -3, A = -3, B = 2, clpz:(2^ #C#=1).
?- #D#= #A* #B^ #C, D= -3,B=2,A= -3.
   D = -3, A = -3, B = 2, clpz:(C in inf..1), clpz:(2^ #C#=1).
?- #D#= #A//(#B+ #C), A= -3,B= -3,D=0.
   D = 0, A = -3, B = -3, clpz:(C in inf..2\/4..sup), clpz:(-3+ #C#= #_A), clpz:(_A in inf.. -4\/4..sup), clpz:(-3// #_A#=0).
?- #D#= #A//(#B+ #C), A= -3,D=0,B= -3.
   D = 0, A = -3, B = -3, clpz:(C in inf.. -1\/7..sup), clpz:(-3+ #C#= #_A), clpz:(_A in inf.. -4\/4..sup), clpz:(-3// #_A#=0).
?-

[UWN]

Different answer constraints, describing exactly the same solution. In general you cannot expect more than that. Of course, there might always be a better implementation.