3.8. Using MiniZinc in Jupyter Notebooks

You can use MiniZinc inside a Jupyter / IPython notebook using the iminizinc Python module. The module provides a “cell magic” extension that lets you solve MiniZinc models.

The module requires an existing installation of MiniZinc.

3.8.1. Installation

You can install or upgrade this module via pip:

pip install -U iminizinc

Consult your Python documentation to find out if you need any extra options (e.g. you may want to use the --user flag to install only for the current user, or you may want to use virtual environments).

Make sure that the minizinc binary are on the PATH environment variable when you start the notebook server. The easiest way to do that is to get the “bundled installation” that includes the MiniZinc IDE and a few solvers, available from GitHub here: https://github.com/MiniZinc/MiniZincIDE/releases/latest You then need to change your PATH environment variable to include the MiniZinc installation.

3.8.2. Basic usage

After installing the module, you have to load the extension using %load_ext iminizinc. This will enable the cell magic %%minizinc, which lets you solve MiniZinc models. Here is a simple example:

In[1]:  %load_ext iminizinc

In[2]:  n=8

In[3]:  %%minizinc

        include "globals.mzn";
        int: n;
        array[1..n] of var 1..n: queens;
        constraint all_different(queens);
        constraint all_different([queens[i]+i | i in 1..n]);
        constraint all_different([queens[i]-i | i in 1..n]);
        solve satisfy;
Out[3]: {u'queens': [4, 2, 7, 3, 6, 8, 5, 1]}

As you can see, the model binds variables in the environment (in this case, n) to MiniZinc parameters, and returns an object with fields for all declared decision variables.

Alternatively, you can bind the decision variables to Python variables:

In[1]:  %load_ext iminizinc

In[2]:  n=8

In[3]:  %%minizinc -m bind

        include "globals.mzn";
        int: n;
        array[1..n] of var 1..n: queens;
        constraint all_different(queens);
        constraint all_different([queens[i]+i | i in 1..n]);
        constraint all_different([queens[i]-i | i in 1..n]);
        solve satisfy;

In[4]:  queens

Out[4]: [4, 2, 7, 3, 6, 8, 5, 1]

If you want to find all solutions of a satisfaction problem, or all intermediate solutions of an optimisation problem, you can use the -a flag:

In[1]:  %load_ext iminizinc

In[2]:  n=6

In[3]:  %%minizinc -a

        include "globals.mzn";
        int: n;
        array[1..n] of var 1..n: queens;
        constraint all_different(queens);
        constraint all_different([queens[i]+i | i in 1..n]);
        constraint all_different([queens[i]-i | i in 1..n]);
        solve satisfy;

Out[3]: [{u'queens': [5, 3, 1, 6, 4, 2]},
         {u'queens': [4, 1, 5, 2, 6, 3]},
         {u'queens': [3, 6, 2, 5, 1, 4]},
         {u'queens': [2, 4, 6, 1, 3, 5]}]

The magic supports a number of additional options, in particular loading MiniZinc models and data from files. Some of these may only work with the development version of MiniZinc (i.e., not the one that comes with the bundled binary releases). You can take a look at the help using

In[1]:  %%minizinc?