asked ~ above 31 de May, 2019 once wtogether the Concern make 403 views Amountain that visits the Inquiry has actually 2 Answers count that Question answerns solved Actuatogether status of ns Inquiry


You are watching: Runtimewarning: divide by zero encountered in double_scalars

ns have ns adhering to password in which i want to Record as a try-other than the department by zero, however, i gain ns error stated in the location and also ins prints the end inf just how should i take care of this Warning?

import numpy together npp1 = np.array((1,2))p2 = p1def f(p1,p2): try: x = 1/(np.linalg.norm(p1-p2)**2) other than ZeroDivisionError: x = 0 return xf(p1,p2)


according come ns documentatitop top of numpy.divide()

division through zero constantly returns zero in essence arithmetic, and also does not raise one exception or a warning:

that is, division by 0 in between numpy arrayns does no throw one exception. However it is possible to change this behavior, through numpy.seterr() :

import numpy together npp1 = np.array((1,2))p2 = p1def f(p1,p2): old_Setups = np.seterr(divide="raise") try: x = 1/(np.linalg.norm(p1-p2)**2) other than FloatingPointError: x = 0 np.seterr(**old_settings) rerevolve x


See more: [&Quot;In:Unread&Quot;] - How To Use Parentheses And Brackets (


girlfriend canno Record the warninns with try-except Since ins is not a exception (it means ns immediate termicountry that the process), ins is just a warninns (i beg your pardon doens not Mean the it have to be systematically ignored).girlfriend can usage ns module warningns come Catch and also filter warnings generically, or have NumPy elevate every warnings or just thins one to exception condition with numpy.seterr() however this has actually global aftermath and also ns personallied do not prefer ins (aldespite friend have the right to turning back it later on manually).In my opini~ above ns simplest thing come do, given that you have perfect derestricted wcap girlfriend desire to handle, is come use numpy.errstate (context manager) in addition to via . In ns documentati~ above friend have an example through inprecious insteAD of through department yet ns principle ins ns same:

import numpy as npp1 = np.array((1,2))p2 = p1def f(p1,p2): with np.errstate(divide="raise"): try: x = 1/(np.linalg.norm(p1-p2)**2) except FloatingPointError: x = 0 rerotate xf(p1,p2))ns previouns warning is now an exception just wislim the block via once out that the block, whatever retransforms come normal, if the end the ns block, every little thing returns to normal. With If the same case occurs again, ns warninns ins thrown, not an exception, therefore ns worldwide dealing with of warnings and exceptions in ns whole module ins no modified.