CODEWITHSUNDEEP
pythonnumpy
A.J.
A.J.

Reputation: 9015

Python numpy upgrade failing test for percentile method

When I have updated the numpy version from 1.11.3 to 1.16.5, my tests are failing because of of fraction difference between the values calculated by the percentile method.

numpy = 1.11.3

import numpy as np
print (np.percentile([1.0, 0.0, 0.0, 0.0], [15.0, 85.0])[1])
> 0.54999999999999982

numpy = 1.16.5 

import numpy as np
print (np.percentile([1.0, 0.0, 0.0, 0.0], [15.0, 85.0])[1])
> 0.5499999999999998

I am looking for an answer that what has changed that is causing the failure, I wanted to understand the upgrade so that I do not get unexpected results once released to prod.

Thanks.

Upvotes: 3

Views: 180

Answers (1)

Fakher Mokadem
Fakher Mokadem

Reputation: 1099

Actually your test is fine, you are just printing two different precisions 

import numpy as np
print (np.percentile([1.0, 0.0, 0.0, 0.0], [15.0, 85.0])[1])
> 0.54999999999999982 # has 17 decimal places

import numpy as np
print (np.percentile([1.0, 0.0, 0.0, 0.0], [15.0, 85.0])[1])
> 0.5499999999999998 # has only 16

the actual value is indeed 0.54999999999999982236. So your test is fine.

To make sure your numeric tests with numpy work as expected , you can setup the precision of printed values with np.set_printoptions(precision=20) .

Upvotes: 1

Related Questions