CODEWITHSUNDEEP
pythonarraysnumpydeep-copy
bob.sacamento
bob.sacamento

Reputation: 6651

One liner for creating several deep copies of a numpy array?

In initializing my code, I need to create several numpy arrays all of the same shape. It's straightforward to do

>>> nx=10
>>> ny=10
>>> a = np.zeros((ny,nx))
>>> b = np.copy(a)
>>> c = np.copy(a)
>>> d = np.copy(a)
>>> etc.

but it is certainly tedious. I was hoping there might be a one-liner to do this. I tried

>>> (b,c,d,e,f,g) = 6*[np.copy(a)]

but that gives me several references to a, not independent copies.

Is there something similar that will give independent copies?

Upvotes: 1

Views: 97

Answers (1)

Will Da Silva
Will Da Silva

Reputation: 7040

We can take advantage of tuple unpacking here. If you're creating a standard Numpy array (zeros, ones, eye, etc.) then you can do it by setting the outermost value of the shape to the number of copies you'd like:

a, b, c, d, e, f, g = np.zeros((7, ny, nx))

Be aware that if you create your "copies" this way, they're actually all slices into the same array.

If you would actually like to make copies of a particular array (and not operate on slices of one larger array), you should unpack a generator expression of calls to np.copy

# a is the numpy array to be copied
b, c, d, e, f, g = (np.copy(a) for _ in range(6))

Make sure that the number of copies being made (6 or 7 in the examples above) is accurate.

Upvotes: 2

Related Questions