我如何区分两个numpy FloatingPointError异常？(How can I distinguish between two numpy FloatingPointError except

我如何区分两个numpy FloatingPointError异常？(How can I distinguish between two numpy FloatingPointError exceptions?)

我对python中的异常并不熟悉，所以我试图在这里解决问题。 numpy中有两个不同的FloatingPointError异常：

import numpy as np import sys np.seterr(divide='raise', invalid='raise') try: np.float64(0.0) / np.float64(0.0) except FloatingPointError: tb = sys.exc_info() print(tb) (<type 'exceptions.FloatingPointError'>, FloatingPointError('invalid value encountered in double_scalars',), traceback object at 0x10bd50cb0>) try: np.float64(1.0) / np.float64(0.0) except FloatingPointError: tb = sys.exc_info() print(tb) (<type 'exceptions.FloatingPointError'>, FloatingPointError('divide by zero encountered in double_scalars',), traceback object at 0x10faede60>)

我需要在我的程序中将0/0操作定义为0，这样我应该区分第一个和第二个异常。 这是我解决问题的方法：

if str(tb[1]) == 'invalid value encountered in double_scalars': print('first exception found') else: print('second exception found')

但我想我可以通过比较错误代码或类似的东西来做到这一点。 那么我怎样才能在Python中做到这一点？

I'm not really familiar with exceptions in python, so that I'm trying to sort out the problem here. There are two different FloatingPointError exceptions in numpy:

import numpy as np import sys np.seterr(divide='raise', invalid='raise') try: np.float64(0.0) / np.float64(0.0) except FloatingPointError: tb = sys.exc_info() print(tb) (<type 'exceptions.FloatingPointError'>, FloatingPointError('invalid value encountered in double_scalars',), traceback object at 0x10bd50cb0>) try: np.float64(1.0) / np.float64(0.0) except FloatingPointError: tb = sys.exc_info() print(tb) (<type 'exceptions.FloatingPointError'>, FloatingPointError('divide by zero encountered in double_scalars',), traceback object at 0x10faede60>)

I need to define 0/0 operation as 0 in my program, so that I should differ between first and second exceptions. Here is my solution of the problem:

if str(tb[1]) == 'invalid value encountered in double_scalars': print('first exception found') else: print('second exception found')

But I guess I can do it simply with comparing error code or something like that. So how can I do it in python?

最满意答案

最好通过numpy.seterr（）设置错误回调函数

class InvalidValueError(Exception): pass class DivideByZeroError(Exception): pass def err_handler(err, flag): if flag == 8: raise InvalidValueError(err) if flag == 1: raise DivideByZeroError(err) np.seterrcall(err_handler) np.seterr(divide='call', invalid='call')

在这种情况下，您可以根据错误类型TODO事物。

try: np.float64(0.0) / np.float64(0.0) # or `np.float64(1.0) / np.float64(0.0)` except InvalidValueError: # TODO except DivideByZeroError: # TODO

It's better to set a error callback function via numpy.seterr()

class InvalidValueError(Exception): pass class DivideByZeroError(Exception): pass def err_handler(err, flag): if flag == 8: raise InvalidValueError(err) if flag == 1: raise DivideByZeroError(err) np.seterrcall(err_handler) np.seterr(divide='call', invalid='call')

In this case, you can TODO things according to the error type.

try: np.float64(0.0) / np.float64(0.0) # or `np.float64(1.0) / np.float64(0.0)` except InvalidValueError: # TODO except DivideByZeroError: # TODO