Python shallow copy and deep copy in using append method

Question

Some problems come out when using append method in python3.5. The code is presented

# generate boson basis in lexicographic order
def boson_basis(L,N):
basis=[]
state=[0 for i in range(1,L+1)]
pos=0
# initialize the state to |N,0,...,0>
state[0]=N
basis.append(state)
# find the first non-zero position in reverse order
while state[L-1]<N:
    for i in range(-2,-L-1,-1):
        if state[i]>0:
            pos=L+i
            break
    sum=0
    for i in range(0,pos):
        sum=sum+state[i]
    state[pos]=state[pos]-1
    state[pos+1]=N-sum-state[pos]
    basis.append(state)
return basis        

result=boson_basis(3,3)

the expected result should be [[3,0,0],[2,1,0],...,[0,0,3]], but this code generates wrong results with all elements are the same as the last one, i.e. [[0,0,3],...,[0,0,3]]. I use the pdb to debug it and I find that once the state is modified, the former state that has been appended into basis is also changed simultaneously. It implies that append uses deepcopy automatically which is beyond my understanding. In fact, this error can be fixed if we use basis(state.copy()) explicitly.

On the other hand, the following simple code shows no error in using append

x=3
b=[]
b.append(x)
x=x+2

after x is changed to x=5, b remains unchanged b=[3]. It really puzzles me and seems contradictory with the former example.

Answer 1

As revealed in the comments already, there's no copy whatsoever involved in an append operation. So you'll have to explicitly take care of this yourself, e.g. by replacing

basis.append(state)

with

basis.append(state[:])

The slicing operation with : creates a copy of state. Mind: it does not copy the lists elements - which as long as you're keeping only plain numbers and not objects in your list should be fine though.