This page will guide you through the process of setting up MiniZinc Python. MiniZinc Python requires the following software to be installed on you machine:
Note
MiniZinc is expected to be in its default location. If you are on a Linux machine or changed this location, then you will have to ensure that the minizinc
executable is located in a folder in the $PATH
environmental variable. When MiniZinc cannot be located, the following warning will be shown: MiniZinc was not found on the system: no default driver could be initialised. The path can manually be provided using find_driver
function.
MiniZinc Python can be found on PyPI. If you have the pip
package manager installed, then the simplest way of installing MiniZinc Python is using the following command:
$ pip install minizinc
Note
On machines that have both Python 2 and Python 3 installed you might have to use pip3
instead of pip
Note
If you require the parsed information of .dzn
files within your python environment, then you have to install the dzn
extra with the MiniZinc package: pip install minizinc[dzn]
To test everything is working let's run a basic example. The n-Queens problem is a famous problem within the constraint programming community. In the MiniZinc Examples we can find the following model for this problem:
int: n; % The number of queens.
array [1..n] of var 1..n: q;
include "alldifferent.mzn";
constraint alldifferent(q);
constraint alldifferent(i in 1..n)(q[i] + i);
constraint alldifferent(i in 1..n)(q[i] - i);
The following Python code will use MiniZinc Python to:
- Load the model from a file (
nqueens.mzn
) - Create an instance of the model for the Gecode solver
- Assign the value 4 to
n
in the instance - Print the positions of the Queens store in the array
q
from minizinc import Instance, Model, Solver
# Load n-Queens model from file
nqueens = Model("./nqueens.mzn")
# Find the MiniZinc solver configuration for Gecode
gecode = Solver.lookup("gecode")
# Create an Instance of the n-Queens model for Gecode
instance = Instance(gecode, nqueens)
# Assign 4 to n
instance["n"] = 4
result = instance.solve()
# Output the array q
print(result["q"])
One of MiniZinc's key features is the ability to use multiple solvers. MiniZinc Python allows you to use all of MiniZinc's solver using a solver configuration <https://www.minizinc.org/doc-latest/en/fzn-spec.html#solver-configuration-files>. Solver configurations were introduces in MiniZinc 2.2. In MiniZinc Python there are three ways of accessing a solver using solver configurations:
- You can
lookup
a solver configuration that is known to MiniZinc. These are solver configurations that are placed on standard locations or in a folder included in the$MZN_SOLVER_PATH
environmental variable. This is the most common way of accessing solvers. - You can
load
a solver configuration directly from a solver configuration file,.msc
. A description of the formatting of such files can be found in the MiniZinc documentation. Theminizinc.Solver.output_configuration
method can be used to generate a valid solver configuration. - You can create a new solver configuration,
Solver
.
Note
Solver loaded from file (2) or created in MiniZinc Python (3). Cannot share the combination of identifier and version with a solver known to MiniZinc (1). In these cases the solver configuration as known to MiniZinc will be used.
The following example shows an example of each method. It will lookup the Chuffed solver, then load a solver configuration from a file located at ./solvers/or-tools.msc
, and, finally, create a new solver configuration for a solver named "My Solver".
from minizinc import Solver
from pathlib import Path
# Lookup Chuffed among MiniZinc solver configurations.
# The argument can be a solver tag, its full identifier, or the last part of
# its identifier
chuffed = Solver.lookup("chuffed")
# Load solver configuration from file
or_tools = Solver.load(Path("./solvers/or-tools.msc"))
# Create a new solver configuration
# Arguments: name, version, identifier, executable
my_solver = Solver(
"My Solver",
"0.7",
"com.example.mysolver",
"/usr/local/bin/fzn-my-solver",
)
# You can now change other options in the solver created configuration
my_solver.mznlib = "/usr/local/share/mysolver/mznlib"
my_solver.stdFlags = ["-a", "-t", "-s"]
Sometimes we don't just require one solution for the given MiniZinc instance, but all possible solutions. The following variation of the previous example uses the all_solutions=True
parameter to ask for all solutions to the problem instance.
from minizinc import Instance, Model, Solver
gecode = Solver.lookup("gecode")
nqueens = Model("./nqueens.mzn")
instance = Instance(gecode, nqueens)
instance["n"] = 4
# Find and print all possible solutions
result = instance.solve(all_solutions=True)
for i in range(len(result)):
print(result[i, "q"])
The use of the all_solutions=True
parameter is limited to satisfaction models (solve satisfy
). MiniZinc currently does not support looking for all solutions for an optimisation model.
Similarly, in a optimisation model (solve maximize
or solve minimize
) we could want access to the intermediate solutions created by the solver during the optimisation process. (This could provide insight into the progress the solver makes). In this case the intermediate_solutions=True
parameter can be used. The following example prints the intermediate solutions that Gecode found to the trivial problem of find the highest uneven number between 1 and 10, but trying smaller values first.
from minizinc import Instance, Model, Solver
gecode = Solver.lookup("gecode")
trivial = Model()
trivial.add_string(
"""
var 1..10: x;
constraint (x mod 2) = 1;
solve ::int_search([x], input_order, indomain_min) maximize x;
"""
)
instance = Instance(gecode, trivial)
# Find and print all intermediate solutions
result = instance.solve(intermediate_solutions=True)
for i in range(len(result)):
print(result[i, "x"])
Note
Not all solver support the finding of all solutions and the printing of intermediate solutions. Solvers that support these functionalities will have -a
among the standard flags supported by the solvers. MiniZinc Python will automatically check if this flag is available. If this is not the case, then an exception will be thrown when the requesting all or intermediate solutions.
For information about other parameters that are available when solving a model instance, see minizinc.Instance.solve