Update array while inside for loop over arrays

Question

I have a Numpy array and can successfully update all its elements with one line:

array_1 = np.array([1, 2, 3, 4])
array_1 = array_1 / 10.0

print(array_1)
# [0.1 0.2 0.3 0.4] -- Success!

However, when I have a list of Numpy arrays and iterate over them with a for in loop, I cannot apply the same operation and get back the desired results.

array_1 = np.array([1, 2, 3, 4])
array_2 = np.array([5, 6, 7, 8])
array_3 = np.array([9, 10, 11, 12])

print(array_1) # [1 2 3 4]
print(array_2) # [5 6 7 8]
print(array_3) # [ 9 10 11 12]

for array in [array_1, array_2, array_3]:
    array = array / 10.0

print(array_1) # [1 2 3 4] -- No changes??
print(array_2) # [5 6 7 8]
print(array_3) # [ 9 10 11 12]

Why am I unable to update these arrays inside a loop? My understanding is that in the line

for array in [array_1, array_2, array_3]:

array will be a pointer to each of the three Numpy arrays.

How can I fix this problem?

You have to distinguish "variable" from "value". In the first example you are creating a new value, and changing the binding of the variable array_1 to the new value. In the second example, you are creating a new value, and changing the binding of array - but array_1 still contains the old value. — Amadan
– Amadan, Commented Jul 16, 2021 at 21:44
Assigning to array makes it point to the new array, it doesn't update the array that it points to. — Barmar
– Barmar, Commented Jul 16, 2021 at 21:44

0x263A · Accepted Answer · 2021-07-16 22:29:56Z

2

You can make a shallow copy of the target array inside the for loop to edit the original.

for array in [array_1,array_2,array_3]:
    array[:] = array / 10.0

EDIT With Explanation---

In the for loop the control variable is its own object that deep copies the item being iterated over. We can use the [:] operation to make a shallow copy of the target item that references the original object. The following code demonstrates this concept:

array_1 = ['foo']
print(id(array_1)) # Original object id
for array in [array_1]:
    array = [1]
    print(id(array)) # Deep copy id
for array in [array_1]:
    array[:] = [1]
    print(id(array)) # Original object id

edited Jul 16, 2021 at 22:29

answered Jul 16, 2021 at 22:11

0x263A

1,85912 silver badges24 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

stackoverflowuser2010 Over a year ago

Can you please explain a bit more? Why do we need to use array[:] = ?

0x263A Over a year ago

Yeah no problem, should've been clearer @stackoverflowuser2010

Max Shouman · Accepted Answer · 2021-07-16 22:08:05Z

1

Becuse you're only assigning new values to the control variable of the loop, so you need to address the actual elements by indexing:

arrays = [array_1, array_2, array_3]

for i in range(len(arrays)):
    arrays[i] = arrays[i] / 10.0

or even more comprehensively:

arrays = [array / 10.0 for array in arrays]

or even by filtering:

arrays = list(filter(lambda x: x / 10.0, arrays))

EDIT:

As @ForceBru noted in the comments section, in-place division would force the return of a float type array whereas the original is int. So the following was changed in the first code snippet: arrays[i] /= 10.0 -> arrays[i] = arrays[i] / 10.0

edited Jul 16, 2021 at 22:08

answered Jul 16, 2021 at 21:45

Max Shouman

1,3311 gold badge6 silver badges12 bronze badges

6 Comments

stackoverflowuser2010 Over a year ago

This doesn't work for me. I get this error:

TypeError: ufunc 'true_divide' output (typecode 'd') could not be coerced to provided output parameter (typecode 'l') according to the casting rule ''same_kind''

pointing to the arrays[i] /= 10.0 line. I'm using Python 3.7.

ForceBru Over a year ago

@stackoverflowuser2010, actually, I got the same error. Looks like the arrays are of integer type, but in-place division by a double (10.0) would require an array of type double, yet you can't write a double into an array of integers, hence the error. Division like np.array([9, 10, 11, 12]) / 10.0 works fine because it creates a new array.

stackoverflowuser2010 Over a year ago

@ForceBru: Is this a bug or a feature?

ForceBru Over a year ago

@stackoverflowuser2010, well, it didn't let you write a floating-point number into an array of integers - I guess that makes sense, since you didn't request any casting. However, NumPy 1.8 doesn't raise any errors in the same code and silently casts the resulting floats to integers. I think it's a feature that enforces better typing

stackoverflowuser2010 Over a year ago

@ForceBru: Hold on. Why does this work? arrays[0] = arrays[0] / 10.0. The LHS is an array of int, but the RHS is an array of double.

|

Collectives™ on Stack Overflow

Update array while inside for loop over arrays

2 Answers 2

2 Comments

EDIT:

6 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

2 Comments

EDIT:

6 Comments

Your Answer

Sign up or log in

Post as a guest

Related