[FIXED] A Python numpy array is being rewritten in a loop - shouldn't be doing this

Issue

Using Anaconda, I have written a simple code and I am at wit's end trying to figure out how an array keeps getting rewritten in a for loop after it's defined. I write the array (random 0 or 1s) named box and define a new array holder as a "copy" of the array I want to leave alone. However, box keeps getting rewritten and I just don't see how it should be doing this.

#FUNCTIONS

def initialize():
    config = np.zeros([n, n])
    for x in range(n):
        for y in range(n):
            if np.random.rand() < p:
                config[x, y] = 1
    return config

def roller(x):
    test1 = np.roll(x,1,axis=0)
    test2 = np.roll(x,1,axis=1)
    test3 = np.roll(x,-1,axis=1)
    test4 = np.roll(x,-1,axis=0)
    hold = np.sum([test1,test2,test3,test4],axis=0)
    return hold

def loop(steps,holder,store_config):
    for t in range(steps):
        tot = roller(holder)
        for x in range(n):
            for y in range(n):
                if tot[x,y] >= 4:
                    holder[x,y] = 1
                else: 
                    holder[x,y] = 0
        store_config[:,:,t] = holder

def func():
    start = time.time()
    time_steps = 20
    store_config = np.zeros([n,n,time_steps])
    loop(time_steps,holder,store_config)
    end = time.time()
    print(end - start, np.sum(store_config))

#CONSTANTS

n = 100
p = .2
box = initialize() #Array to leave ALONE
print(np.sum(box))

#Action
holder = box #Array to manipulate and redefine as box
func()
print(np.sum(box))

If you the value from the output of np.sum(box) should match before and after func() is ran , but they never do. The intention was that when I rerun func(), it spits out a value, but just iterates over the same defined "box" array but it keeps getting rewritten. I don't see how its possible. I thought arrays were treated like variables inside a function where they're not global? Each of the three sections #Functions, #Constants, and #Action would be in their own cells in a Conda Notebook.

Solution

I think the issue you are having is that arrays, just as lists and dicts, are assigned by reference. This happens no matter if it's a variable assignment or as an argument being passed to a function.

def f(x):
    x[0] = 0
    return x

a = array([1, 1])
b = f(a)

WILL result in a and b being equal, since x is being manipulated and then returned. If you want to retain a, you must copy the array first:

def f(x):
    x_ = x.copy()
    x_[0] = 0
    return x_

I hope this clarifies things a bit. :-)

Answered By - dalum