How to suppress the error message when dividing 0 by 0 using np.divide (alongside other floats)?

Question

My attempt:

import numpy as np
np.seterr(divide='ignore')
a=np.array([4/3,0,0])
b=np.array([1,0,0])
np.divide(a,b)

The ouput I get:

__main__:1: RuntimeWarning: invalid value encountered in true_divide
array(1.33333333,        nan, nan])

If I ask it again:

np.divide(a,b)

then no RuntimeWarning is displayed.

I'm happy with the array that is output, but I don't want the error message; that is what I thought seterr would fix. How can I get rid of the warning? (I would rather not write a for loop that makes exceptions for 0/0.)

Answer 1

Try:

IntentService

or

np.seterr(invalid='ignore')

From the np.seterr(all='ignore') docs:

seterr

- Invalid operation: result is not an expressible number, typically indicates that a NaN was produced. produces a 1/0 error/warning.

FloatingPointError: divide by zero encountered in true_divide produces a 0/0 error/warning.

Warning: invalid value encountered in true_divide will catch both.

all can be used to temporarily change errstate.

np.seterr

In [1472]: with np.errstate(invalid='print'): y=np.divide([1.2,0,0],[1,0,0]) ......: Warning: invalid value encountered in true_divide are normally issued the first time the problem arise in a run, and then are silent.