Missing final objective with OR-Tools CP-SAT solver

[AllenZzw]

model and data: cable-tree-wiring.zip

Running the schedule function with the above instance using CP-SAT results in inconsistency between stat.yml file and sol.yml.

The objective of the last solution in sol.yml is 1524056:

- configuration: CPSAT_free_all
  data_file: challenge/mznc2024_probs/cable-tree-wiring/A041.dzn
  model: challenge/mznc2024_probs/cable-tree-wiring/ctw.mzn
  problem: cable-tree-wiring
  solution:
    cfp: [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21,
      22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40,
      41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 1]
    objective: 1524056
    pfc: [24, 32, 34, 42, 43, 45, 4, 2, 6, 8, 10, 26, 29, 5, 40, 41, 44, 46, 49, 7,
      11, 14, 15, 20, 19, 30, 33, 35, 23, 27, 50, 3, 1, 36, 38, 9, 25, 28, 37, 31,
      22, 21, 47, 48, 39, 12, 13, 16, 17, 18]
  status: SATISFIED
  time: 21.35
- {configuration: CPSAT_free_all, data_file: challenge/mznc2024_probs/cable-tree-wiring/A041.dzn,
  model: challenge/mznc2024_probs/cable-tree-wiring/ctw.mzn, problem: cable-tree-wiring,
  status: OPTIMAL_SOLUTION}

while the stat.yml shows a different objective 1524060.0:

{boolVariables: 5270, configuration: CPSAT_free_all, data_file: challenge/mznc2024_probs/cable-tree-wiring/A041.dzn,
  evaluatedHalfReifiedConstraints: 78, evaluatedReifiedConstraints: 2450, failures: 9488,
  flatBoolConstraints: 2439, flatBoolVars: 4928, flatIntConstraints: 5700, flatIntVars: 2708,
  flatTime: 0.457866, method: minimize, model: challenge/mznc2024_probs/cable-tree-wiring/ctw.mzn,
  nSolutions: 77, objective: 1524060.0, objectiveBound: 1524060.0, paths: 0, problem: cable-tree-wiring,
  propagations: 19294008, solveTime: 21.2494, status: OPTIMAL_SOLUTION, time: 22.248722480027936}

[Dekker1]

Just to confirm, 1524060.0 is the objective of the solution just before the last one?

[AllenZzw]

Just to confirm, 1524060.0 is the objective of the solution just before the last one?

No, it is not. Here are the yaml files log.zip. Actually, 1524060.0 does not correspond to an objective of any solution found.

[Dekker1]

Looking at the code, it might be the expectation that we add a objective to the statistics, but I think OR-Tools might be outputting its own objective statistic and overriding it. (And the internal or FlatZinc objective value might be scaled or offset somehow from what the MiniZinc objective is).

I think I know how to resolve this and will get back to you with something to test.