MiniZinc: type error: identifier `friend' already defined groupphoto.ozn:8

[brudgers]

Command $ minizinc groupphoto.mzn data/groupphoto1.dzn.

Code:

 include "alldifferent.mzn";
 array[Person] of var 0..4: obj =
    [left(position[i]) +
     right(position[i]) +
     front(position[i]) +
     rear(position[i])
     | i in Person];

output [show(obj),
    show(position)];

solve maximize sum(obj);
int: n;
array[Person, Person] of bool: friend;

set of int: Person = 1..n;
array[Person] of var Person: position;
constraint alldifferent(position);
int: m = n div 2;
constraint assert(n mod 2 = 0, "n must be even");

function var 0..1: left(var Person: i) =
  if i > 1 /\ i != m + 1
  then bool2int(friend[position[i], position[i-1]])
  else 0 endif;

function var 0..1: right(var Person: i) =
  if i < n /\ i != m
  then bool2int(friend[position[i], position[i+1]])
  else 0 endif;

function var 0..1: front(var Person: i) =
  if i > m
  then bool2int(friend[position[i], position[i-m]])
  else 0 endif;

function var 0..1: rear(var Person: i) =
  if i < m
  then bool2int(friend[position[i], position[i+m]])
  else 0 endif;

[schutta]

Using the command to flatten the model

$ mzn2fzn -D "n = 2; friend = [| false, false | false, false |];" bug_44.mzn

results in the following ozn-file

output [show(obj),show(position)];
set of int: Person = 1..2;
array [Person] of int: position;
function var 0..1 : left(var Person: i) = if i > 1 /\ i!=m+1 then bool2int(friend[position[i],position[i-1]]) else 0 endif;
array [Person,Person] of bool: friend = [|false,false|false,false|];
int: m = 1;
array [Person] of int: obj;
array [Person,Person] of bool: friend = [|false,false|false,false|];
int: m = 1;

in that the declaration of the parameters friend and m are duplicated. MiniZinc 2.0.{0,1} crashes with a segmentation fault and MiniZinc 2.0.2+ leads to that erroneous ozn file.

[schutta]

A work-around for the current version is to use the command mzn2fzn to flatten the model, which generates a FlatZinc file with extension "fzn" and the erroneous output (ozn file), and then to use your preferred FlatZinc solver to solve the flatten model. The solver output won't conform with the specified one in the MiniZinc model, but if you need the "well-formed" output then remove duplicates from the ozn file (by hand or script) and pipe the solver's output to solns2out, e.g.,

$ cat <solver output file> | solns2out <corrected ozn file>

[guidotack]

Thanks for the report. I've added a fix, it will be part of the next release.