XCSP3Output

cpbenchy run instances/ -s ortools -t 60 -p cpbenchy.observers:XCSP3Output

What it does

Writes what a solver in the XCSP3 competition prints, for each run on an XCSP3 instance, with the solution as an XML <instantiation> of the instance’s variables:

o 12
o 9
s OPTIMUM FOUND
v <instantiation type="optimum" cost="9">
v   <list>x[0] x[1] x[2] y</list>
v   <values>2 1 0 3</values>
v </instantiation>

Each competition specifies what a solver prints:

• o lines with each better objective value, while it solves
• an s line with its answer: OPTIMUM FOUND, SATISFIABLE, UNSATISFIABLE or UNKNOWN
• v lines with the solution

The output goes to each run’s logs/<run_id>.xcsp3.out, written inside the measured run, as a competition solver would print it, so it costs the run the time it would cost in the competition. A run that is killed keeps its o lines. Under cpbenchy solve, the output goes to stdout, where competitions read it.

Given the competition’s solution checker, it also checks every solution with it, after the run and outside its measured time.

Use it

import cpbenchy
from cpbenchy import observers

cpbenchy.run("xcsp3/", solvers=["ortools"], time_limit=60, plugins=[observers.XCSP3Output()])
cpbenchy.run("xcsp3/", solvers=["ortools"], time_limit=60, plugins=[observers.XCSP3Output(checker="checker.jar")])

cpbenchy run xcsp3/ -s ortools -t 60 -p cpbenchy.observers:XCSP3Output
cpbenchy run xcsp3/ -s ortools -t 60 -p "cpbenchy.observers:XCSP3Output(checker='checker.jar')"

The XCSP3 rules enable it, with the competition’s limits.

Options

What it records

Implementation

The XCSP3Output in src/cpbenchy/observers.py, lines 74–92 of 222, as of this version of the docs.

class XCSP3Output(CompetitionOutput):
    """XCSP3 competition output. With `checker` (the path of the competition's checker jar, needs java),
    the parent also checks every solution after its run, outside the measured time, and records the
    verdict as `extra["xcsp3_check"]`: "OK", or what is wrong."""

    name = "xcsp3"
    formats = ("xcsp3",)

    def __init__(self, checker=None):
        self.checker = checker

    def solution(self, ctx):
        cost = ctx.objective if ctx.model.has_objective() else None
        return fmt.xcsp3_instantiation(fmt.instance_variables(ctx.model), cost).splitlines()

    def on_result(self, run, result):
        output = self.output(run)
        if self.checker and result.status in ("optimal", "feasible") and output.exists():
            result.extra["xcsp3_check"] = xcsp3_check(self.checker, run.spec.instance.path, output)