Question
write a Python program that traverses graphs in BFS and DFS manner. BFS will determine the shortest path distance (number of edges) from the root
write a Python program that traverses graphs in BFS and DFS manner. BFS will determine the shortest path distance (number of edges) from the root for each node reachable from the root. DFS will find cycles in the graph of nodes reachable from the root. Study the lecture on graphs, in particular graph traversals.
Some helper code is provided. Don't change it. Don't change your main, it is used to check your code's correctness.
It is your job to implement dfs and bfs. In both dfs and bfs, visit children of a node in left to right order, i.e., if adj is the adjacency list of a node, visit the children as follows: for nxt in adj
Given an input file in:
a b b c c a d d c
and root a
python dfbf.py in a produces:
dfbf.py BFS Input graph: nodeName (color, [adj list]) dictionary a ('white', ['b']) b ('white', ['c']) c ('white', ['a', 'd']) d ('white', ['c']) Root node: a BFS queue: (node name, distance) pairs [('a', 0), ('b', 1), ('c', 2), ('d', 3)] END BFS
DFS Input graph: nodeName (color, [adj list]) dictionary a ('white', ['b']) b ('white', ['c']) c ('white', ['a', 'd']) d ('white', ['c']) Root node a graph with root a is cyclic END DFS
'''
import sys
cyclic = False #keeping track in dfs whether a cycle was found
def read(fnm): """ read file fnm into dictionary each line has a nodeName followed by its adjacent nodeNames """ f = open(fnm) gr = {} #graph represented by dictionary for line in f: l =line.strip().split(" ") # ignore empty lines if l==['']:continue # dictionary: key: nodeName value: (color, adjList of names) gr[l[0]]= ('white',l[1:]) return gr
def dump(gr): print("Input graph: nodeName (color, [adj list]) dictionary ") for e in gr: print(e, gr[e])
def white(gr) : """ paint all gr nodes white """ for e in gr : gr[e] = ('white',gr[e][1])
''' return bfs queue with (node, distance) pairs ''' def dfs(visited, graph, node): if node not in visited: print (node) visited.add(node) for neighbour in graph[node]: dfs(visited, graph, neighbour)
visited = [] queue = []
def bfs (visited, graph, node): visited.append(node) queue.append(node)
while queue: s = queue.pop(0) print (s, end = " ")
for neighbour in graph[s]: if neighbour not in visited: visited.append(neighbour) queue.append(neighbour)
if __name__ == "__main__": print(sys.argv[0]) gr = read(sys.argv[1]) # file name root = sys.argv[2] # root node db = len(sys.argv)>3 # debug? print("BFS") dump(gr) print("Root node:", root) gr[root] = ('black',gr[root][1]) q = bfs(gr,[(root,0)]) print("BFS queue: (node name, distance) pairs") print(q) print("END BFS") print()
print("DFS") white(gr) dump(gr) print("Root node", root) dfsInit(gr,root) if cyclic: print("graph with root",root,"is cyclic") else: print("graph with root",root,"is not cyclic") print("END DFS")
Step by Step Solution
There are 3 Steps involved in it
Step: 1
Get Instant Access to Expert-Tailored Solutions
See step-by-step solutions with expert insights and AI powered tools for academic success
Step: 2
Step: 3
Ace Your Homework with AI
Get the answers you need in no time with our AI-driven, step-by-step assistance
Get Started