Why numpy.empty_like() return an array fill with zero is unsafe?

Question

I am confused by what I was reading in Python for data analysis on Page 82.

In [25]: np.empty((2, 3, 2))
Out[25]:
array([[[  4.94065646e-324,   4.94065646e-324],
        [  3.87491056e-297,   2.46845796e-130],
        [  4.94065646e-324,   4.94065646e-324]],
       [[  1.90723115e+083,   5.73293533e-053],
        [ -2.33568637e+124,  -6.70608105e-012],
        [  4.42786966e+160,   1.27100354e+025]]])

It’s not safe to assume that np.empty will return an array of all zeros. In many cases, as previously shown, it will return uninitialized garbage values.

numpy.zeros() returns a array with zeros, but why is it unsafe to assume that numpy.empty() won't be an array of zeros?

All they are saying is that sometimes such an array is all zeros, other times it will be odd looking numbers such as that example. So only use empty when you will assign ALL elements new values. — hpaulj
– hpaulj, Commented Jun 14, 2017 at 4:49
If you feel the answer here has solved the problem, please mark it as 'accepted' by clicking the green check mark next to it. Otherwise, please add a comment, or edit your question to clarify it. — Will Da Silva
– Will Da Silva, Commented Jun 20, 2019 at 15:00

Will Da Silva · Accepted Answer · 2020-09-04 16:52:54Z

4

The implementation of numpy.empty() does not write the value of zero to the addresses of memory it reserves for the array. This makes it faster, but leaves behind whatever values were previously occupying that space in memory.

edited Sep 4, 2020 at 16:52

answered Jun 14, 2017 at 2:34

Will Da Silva

7,1512 gold badges35 silver badges55 bronze badges

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

3 Comments

l0o0 Over a year ago

As you say, the values in the array produced by numpy.empty() is existed before the array created. Then I think the values may be string or int, but they will be transfered to float64 at last. Am I right?

Will Da Silva Over a year ago

@l0o0 yes, they are just remnants of what was previously in memory at those locations. When memory is deallocated, it is not overridden with zeros, as that would be a costly procedure. Instead, the data is just left as is.

l0o0 Over a year ago

Thank you for your help, @Will Da Silva

Collectives™ on Stack Overflow

Why numpy.empty_like() return an array fill with zero is unsafe?

1 Answer 1

3 Comments

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Related