Rat in Maze

T
p
from __future__ import annotations


def solve_maze(
    maze: list[list[int]],
    source_row: int,
    source_column: int,
    destination_row: int,
    destination_column: int,
) -> list[list[int]]:
    """
    This method solves the "rat in maze" problem.
    Parameters :
        - maze: A two dimensional matrix of zeros and ones.
        - source_row: The row index of the starting point.
        - source_column: The column index of the starting point.
        - destination_row: The row index of the destination point.
        - destination_column: The column index of the destination point.
    Returns:
        - solution: A 2D matrix representing the solution path if it exists.
    Raises:
        - ValueError: If no solution exists or if the source or
            destination coordinates are invalid.
    Description:
        This method navigates through a maze represented as an n by n matrix,
        starting from a specified source cell and
        aiming to reach a destination cell.
        The maze consists of walls (1s) and open paths (0s).
        By providing custom row and column values, the source and destination
        cells can be adjusted.
    >>> maze = [[0, 1, 0, 1, 1],
    ...         [0, 0, 0, 0, 0],
    ...         [1, 0, 1, 0, 1],
    ...         [0, 0, 1, 0, 0],
    ...         [1, 0, 0, 1, 0]]
    >>> solve_maze(maze,0,0,len(maze)-1,len(maze)-1)    # doctest: +NORMALIZE_WHITESPACE
    [[0, 1, 1, 1, 1],
    [0, 0, 0, 0, 1],
    [1, 1, 1, 0, 1],
    [1, 1, 1, 0, 0],
    [1, 1, 1, 1, 0]]

    Note:
        In the output maze, the zeros (0s) represent one of the possible
        paths from the source to the destination.

    >>> maze = [[0, 1, 0, 1, 1],
    ...         [0, 0, 0, 0, 0],
    ...         [0, 0, 0, 0, 1],
    ...         [0, 0, 0, 0, 0],
    ...         [0, 0, 0, 0, 0]]
    >>> solve_maze(maze,0,0,len(maze)-1,len(maze)-1)    # doctest: +NORMALIZE_WHITESPACE
    [[0, 1, 1, 1, 1],
    [0, 1, 1, 1, 1],
    [0, 1, 1, 1, 1],
    [0, 1, 1, 1, 1],
    [0, 0, 0, 0, 0]]

    >>> maze = [[0, 0, 0],
    ...         [0, 1, 0],
    ...         [1, 0, 0]]
    >>> solve_maze(maze,0,0,len(maze)-1,len(maze)-1)    # doctest: +NORMALIZE_WHITESPACE
    [[0, 0, 0],
    [1, 1, 0],
    [1, 1, 0]]

    >>> maze = [[1, 0, 0],
    ...         [0, 1, 0],
    ...         [1, 0, 0]]
    >>> solve_maze(maze,0,1,len(maze)-1,len(maze)-1)    # doctest: +NORMALIZE_WHITESPACE
    [[1, 0, 0],
    [1, 1, 0],
    [1, 1, 0]]

    >>> maze = [[1, 1, 0, 0, 1, 0, 0, 1],
    ...         [1, 0, 1, 0, 0, 1, 1, 1],
    ...         [0, 1, 0, 1, 0, 0, 1, 0],
    ...         [1, 1, 1, 0, 0, 1, 0, 1],
    ...         [0, 1, 0, 0, 1, 0, 1, 1],
    ...         [0, 0, 0, 1, 1, 1, 0, 1],
    ...         [0, 1, 0, 1, 0, 1, 1, 1],
    ...         [1, 1, 0, 0, 0, 0, 0, 1]]
    >>> solve_maze(maze,0,2,len(maze)-1,2)  # doctest: +NORMALIZE_WHITESPACE
    [[1, 1, 0, 0, 1, 1, 1, 1],
    [1, 1, 1, 0, 0, 1, 1, 1],
    [1, 1, 1, 1, 0, 1, 1, 1],
    [1, 1, 1, 0, 0, 1, 1, 1],
    [1, 1, 0, 0, 1, 1, 1, 1],
    [1, 1, 0, 1, 1, 1, 1, 1],
    [1, 1, 0, 1, 1, 1, 1, 1],
    [1, 1, 0, 1, 1, 1, 1, 1]]
    >>> maze = [[1, 0, 0],
    ...         [0, 1, 1],
    ...         [1, 0, 1]]
    >>> solve_maze(maze,0,1,len(maze)-1,len(maze)-1)
    Traceback (most recent call last):
        ...
    ValueError: No solution exists!

    >>> maze = [[0, 0],
    ...         [1, 1]]
    >>> solve_maze(maze,0,0,len(maze)-1,len(maze)-1)
    Traceback (most recent call last):
        ...
    ValueError: No solution exists!

    >>> maze = [[0, 1],
    ...         [1, 0]]
    >>> solve_maze(maze,2,0,len(maze)-1,len(maze)-1)
    Traceback (most recent call last):
        ...
    ValueError: Invalid source or destination coordinates

    >>> maze = [[1, 0, 0],
    ...         [0, 1, 0],
    ...         [1, 0, 0]]
    >>> solve_maze(maze,0,1,len(maze),len(maze)-1)
    Traceback (most recent call last):
        ...
    ValueError: Invalid source or destination coordinates
    """
    size = len(maze)
    # Check if source and destination coordinates are Invalid.
    if not (0 <= source_row <= size - 1 and 0 <= source_column <= size - 1) or (
        not (0 <= destination_row <= size - 1 and 0 <= destination_column <= size - 1)
    ):
        raise ValueError("Invalid source or destination coordinates")
    # We need to create solution object to save path.
    solutions = [[1 for _ in range(size)] for _ in range(size)]
    solved = run_maze(
        maze, source_row, source_column, destination_row, destination_column, solutions
    )
    if solved:
        return solutions
    else:
        raise ValueError("No solution exists!")


def run_maze(
    maze: list[list[int]],
    i: int,
    j: int,
    destination_row: int,
    destination_column: int,
    solutions: list[list[int]],
) -> bool:
    """
    This method is recursive starting from (i, j) and going in one of four directions:
    up, down, left, right.
    If a path is found to destination it returns True otherwise it returns False.
    Parameters
        maze: A two dimensional matrix of zeros and ones.
        i, j : coordinates of matrix
        solutions: A two dimensional matrix of solutions.
    Returns:
        Boolean if path is found True, Otherwise False.
    """
    size = len(maze)
    # Final check point.
    if i == destination_row and j == destination_column and maze[i][j] == 0:
        solutions[i][j] = 0
        return True

    lower_flag = (not i < 0) and (not j < 0)  # Check lower bounds
    upper_flag = (i < size) and (j < size)  # Check upper bounds

    if lower_flag and upper_flag:
        # check for already visited and block points.
        block_flag = (solutions[i][j]) and (not maze[i][j])
        if block_flag:
            # check visited
            solutions[i][j] = 0

            # check for directions
            if (
                run_maze(maze, i + 1, j, destination_row, destination_column, solutions)
                or run_maze(
                    maze, i, j + 1, destination_row, destination_column, solutions
                )
                or run_maze(
                    maze, i - 1, j, destination_row, destination_column, solutions
                )
                or run_maze(
                    maze, i, j - 1, destination_row, destination_column, solutions
                )
            ):
                return True

            solutions[i][j] = 1
            return False
    return False


if __name__ == "__main__":
    import doctest

    doctest.testmod(optionflags=doctest.NORMALIZE_WHITESPACE)