PyConstraints

A simple, constraints satisfaction problem solver. Used for the YACS course scheduler project.

Usage

The Problem is the primary interface:

>>> from pyconstraints import Problem

And then specify your problem to solve with various constraints:

>>> p = Problem()
>>> p.add_variable('x', range(4)) # variable-name, domain
>>> p.add_variable('y', range(4))
# give constraint function and list of variables used
>>> p.add_constraint(lambda x, y: x != y, ['x', 'y'])
>>> p.add_constraint(lambda x: x % 2 == 0)

Then get your solutions:

>>> p.get_solutions()
# => ({'y': 0, 'x': 2},
#     {'y': 1, 'x': 0},
#     {'y': 1, 'x': 2},
#     {'y': 2, 'x': 0},
#     {'y': 3, 'x': 0},
#     {'y': 3, 'x': 2})

Or iteratively:

>>> p.iter_solutions().next()
# => {'y': 0, 'x': 2}

And that's it!

Using Another Solver

Simply pass the solver to the Problem constructor:

>>> from pyconstraints import BruteForceSolver, BacktrackingSolver
>>> p = Problem(BacktrackingSolver()) # BruteForceSolver is default

Because the BruteForceSolver uses itertools, there may be cases where it is faster than the BacktrackingSolver.

Writing Your Own Solver

For convinence, there is a pyconstraints.SolverInterface Abstract-Base Class if you want to implement all the features manually:

@abstractproperty
def solutions_seen(self):
    "Returns the number of solutions currently seen by the solver."

@abstractproperty
def solutions_at_points(self):
    """Returns a dictionary of {iteration_index: solution} of all known
    solutions while iterating.
    """

@abstractmethod
def set_conditions(self, variables, constraints):
    """Called by the Problem class to assign the variables and constraints
    for the problem.

        variables = {variable-name: list-of-domain-values}
        constraints = [(constraint_function,
                        variable-names,
                        default-variable-values)]
    """

@abstractmethod
def restore_point(self, starting_point=None):
    "Restores the iteration state to a given starting point."

@abstractmethod
def save_point(self):
    """Returns data to indicate a way to restore to the current iteration
    point.
    """

@abstractmethod
def __iter__(self):
    "Yields solutions."

But for convinence, you can inherit from the pyconstraints.SolverBase class which provides a primitive implementation for all the interface methods except for __iter__ and set_conditions.

Todo

• Speed up backtracking solver
• Add more solvers?