5 Best Ways to Perform Breadth-First Search on a Matrix in Python

Rate this post

πŸ’‘ Problem Formulation: Breadth-First Search (BFS) is a traversal algorithm that processes nodes level by level, making it ideal for finding the shortest path on graphs and matrices. In this article, we tackle the application of BFS on a matrix to find the shortest path from a given start position to a goal, representing the matrix as a grid where each cell is a node. For instance, given a 2D matrix with obstacles and open paths, we want to identify the shortest path from the top-left corner (start) to the bottom-right corner (goal).

Method 1: Using a Queue

The use of a queue is fundamental in implementing the BFS algorithm. This method involves initializing a queue with the starting node and processing each node by expanding its neighbors and adding them to the queue until the goal is reached.

Here’s an example:

from collections import deque

def bfs_matrix(matrix, start, goal):
    rows, cols = len(matrix), len(matrix[0])
    visited = set()
    queue = deque([start])

    while queue:
        x, y = queue.popleft()
        if (x, y) == goal:
            return True
        for dx, dy in [(-1, 0), (1, 0), (0, -1), (0, 1)]:
            nx, ny = x + dx, y + dy
            if 0 <= nx < rows and 0 <= ny < cols and (nx, ny) not in visited and matrix[nx][ny] != 0:
                visited.add((nx, ny))
                queue.append((nx, ny))
    return False

matrix = [
    [1, 1, 0, 1],
    [0, 1, 1, 1],
    [0, 0, 1, 0],
    [1, 1, 1, 1]
print(bfs_matrix(matrix, (0, 0), (3, 3)))



This code sets up a matrix and defines the BFS function bfs_matrix(), which uses a queue to explore the matrix nodes. The matrix is represented such that ‘1’ indicates a passable cell and ‘0’ indicates an impassable cell. It returns True when the goal is reached.

Method 2: Using Recursion

BFS can be achieved recursively by simulating a queue with function calls. This is less conventional and might be harder to understand but can be a novel approach in environments where a queue structure isn’t readily available.

Here’s an example:

# This method is not recommended due to Python's recursion limit and is just conceptual.
def recursive_bfs(matrix, visited, queue, goal):
    if not queue:
        return False
    (x, y) = queue.pop(0)
    if (x, y) == goal:
        return True
    visited.add((x, y))
    for dx, dy in [(-1, 0), (1, 0), (0, -1), (0, 1)]:
        nx, ny = x + dx, y + dy
        if (nx, ny) not in visited and valid_position(matrix, nx, ny):
            queue.append((nx, ny))
    return recursive_bfs(matrix, visited, queue, goal)

def valid_position(matrix, x, y):
    rows, cols = len(matrix), len(matrix[0])
    return 0 <= x < rows and 0 <= y < cols and matrix[x][y] == 1

# Assumes a predefined matrix and a call to recursive_bfs(matrix, set(), [(0, 0)], (3, 3))



This snippet is a conceptual demonstration of a recursive approach to BFS, which uses Python function calls as a substitute for an explicit queue. Note that due to Python’s recursion limit, this method may not be practical for large matrices or deep recursion.

Method 3: Using a Custom Queue Class

Creating a custom queue class can provide more control over the queue operations and can be optimized for BFS on a matrix. This method showcases object-oriented programming skills and can encapsulate the BFS logic neatly.

Here’s an example:

class Queue:
    def __init__(self):
        self.elements = []

    def enqueue(self, val):

    def dequeue(self):
        return self.elements.pop(0)

    def is_empty(self):
        return len(self.elements) == 0

def bfs_custom_queue(matrix, start, goal):
    # Similar BFS function as Method 1, using the custom Queue class for queue operations.

# Assumes a predefined matrix and a call to bfs_custom_queue(matrix, (0, 0), (3, 3))

The output will be similar to that of Method 1.

This code snippet introduces a custom Queue class for handling BFS operations. The BFS function is similar, but it uses instances of the Queue class to manage the frontier of nodes to be explored.

Method 4: Using Multithreading

For larger matrices, performing BFS in parallel using multithreading can potentially reduce execution time. This advanced method splits the matrix into segments and performs BFS on each concurrently. Caution is advised due to Python’s Global Interpreter Lock (GIL).

Here’s an example:

# Multithreading for BFS is a complex operation and often not ideal due to Python's GIL.
# Conceptual pseudo-code for multithreaded BFS is provided, actual implementation would require careful synchronization and state management.
def bfs_segment(matrix, segment, goal):
    # Perform BFS on a given matrix segment.

# Assume splitting the matrix into segments and invoking bfs_segment(matrix, segment, (3, 3)) in separate threads.

This approach is more theoretical, as implementing BFS with threads in Python comes with significant challenges due to the GIL which can limit the performance benefits of threading. This example provides a concept without a full implementation.

Bonus One-Liner Method 5: Using Python’s Powerful List Comprehensions

Although Python list comprehensions aren’t designed for BFS, you can write a dense line of code that captures the essence of BFS. This is more of an academic exercise than a practical solution.

Here’s an example:

# Pseudo-one-liner for BFS using list comprehensions. It does not handle all BFS nuances.
# The actual implementation of BFS cannot be reduced to a one-liner for all practical purposes.

# Assumes a predefined matrix and a partial implementation using list comprehensions.

This is not an actual implementation but illustrates that complex operations sometimes can be expressed in a concise way in Python, albeit without taking care of all the details and practical considerations.


  • Method 1: Using a Queue. Standard and widely used. It’s intuitive and mirrors the theoretical BFS algorithm closely. Can be memory intensive for large matrices.
  • Method 2: Using Recursion. More a theoretical concept than a practical method due to Python’s recursion limit. Not recommended for actual use but good for understanding the depth-first nature of Python function calls.
  • Method 3: Using a Custom Queue Class. Demonstrates custom implementation and encapsulation but might not provide significant advantages over using Python’s built-in data structures like deque.
  • Method 4: Using Multithreading. Not ideal due to Python’s GIL. Multithreading may complicate implementation without offering benefits in the context of BFS on a matrix.
  • Bonus Method 5: One-Liner (Conceptual). More academic than practical. It showcases Python’s expressive power but is not a proper BFS implementation.