what's the difference between these 2 python codes? why different results?

Question

I'm writing a dfs code like below:

def dfs(self, graph, node, path):
    if node==len(graph)-1:
        self.res.append(path)
    else:
        for i in graph[node]:
            path.append(i)
            self.dfs(graph, i, path)
            print(path.pop())

but got undesired result, while I changed the for loop code in dfs to:

self.dfs(graph, i, path+[i])

the result is what I want. But I could not figure out what's the difference of these 2 piece codes. Thanks

Answer 1

Algorithm-wise, I need more information about how you represent the graph to give more insight. But code-wise, the difference is caused by the fact that List in Python is mutable . That means in your original code, when you pop the path list, the entry in the res will change accordingly. Check this by the following.

a = [1,2,3]
b = []

b.append(a)  # b = [[1,2,3]]
a.pop()      # now b = [[1,2]]

But when you change the argument to path + [i] , a new List is constructed, ie the path inside the next recursive call of dfs is decoupled from that in the previous one. Check this by the following.

a = [1,2]
b = []

b.append(a+[3])  # b = [[1,2,3]]
a.pop()          
# b is still [[1,2,3]], because the expression a + [3] will be evaluated to another List

Answer 2

Te original version mutates path in each iteration of each recursion. The second version just makes a copy and adds an extra element. This copy is not bound to a name, it just get passed to the next level of recursion.