輪子|Python2異常鏈

monkeysayhi發表於2017-12-21
Python

介紹一個自己造的輪子，Python2異常鏈。

需求

習慣了用java擼碼，雖說膠水程式碼多，但能比較好的用程式碼表達思路；而Python則簡潔到了簡陋的地步——各種雞肋的語法糖，各種不完善的機制。比如錯誤處理。

Python2沒有異常鏈，讓問題排查變得非常困難：

# coding=utf-8
import sys


class UnexpectedError(StandardError):
    pass


def divide(division, divided):
    if division == 0:
        raise ValueError("illegal input: %s, %s" % (division, divided))
    ans = division / divided
    return ans


a = 0
b = 0
try:
    print divide(a, b)
except ValueError as e:
    # m_counter.inc("err", 1)
    raise UnexpectedError("illegal input: %s, %s" % (a, b))
except ZeroDivisionError as e:
    # m_counter.inc("err", 1)
    raise UnexpectedError("divide by zero")
except StandardError as e:
    # m_counter.inc("err", 1)
    raise UnexpectedError("other error...")
複製程式碼

列印異常如下：

Traceback (most recent call last):
  File "/Users/monkeysayhi/PycharmProjects/Wheel/utils/tmp/tmp.py", line 22, in <module>
    raise UnexpectedError("illegal input: %s, %s" % (a, b))
__main__.UnexpectedError: illegal input: 0, 0
複製程式碼

不考慮程式碼風格，是標準的Python2異常處理方式：分別捕獲異常，再統一成一個異常，只有msg不同，重新丟擲。這種寫法又醜又冗餘，頂多可以改成這樣：

try:
    print divide(a, b)
except StandardError as e:
    # m_counter.inc("err", 1)
    raise UnexpectedError(e.message)
複製程式碼

即便如此，也無法解決一個最嚴重的問題：明明是11行丟擲的異常，但列印出來的異常棧卻只能追蹤到22行重新丟擲異常的raise語句。重點在於沒有記錄cause，使我們追蹤到22行之後，不知道為什麼會丟擲cause，也就無法定位到實際發生問題的程式碼。

異常鏈

最理想的方式，還是在異常棧中列印異常鏈：

try:
    print divide(a, b)
except StandardError as cause:
    # m_counter.inc("err", 1)
    raise UnexpectedError("some msg", cause)
複製程式碼

就像Java的異常棧，區分“要丟擲的異常UnexpectedError和引起該異常的原因cause”：

java.lang.RuntimeException: level 2 exception
	at com.msh.demo.exceptionStack.Test.fun2(Test.java:17)
	at com.msh.demo.exceptionStack.Test.main(Test.java:24)
	at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
	at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
	at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
	at java.lang.reflect.Method.invoke(Method.java:498)
	at com.intellij.rt.execution.application.AppMain.main(AppMain.java:147)
Caused by: java.io.IOException: level 1 exception
	at com.msh.demo.exceptionStack.Test.fun1(Test.java:10)
	at com.msh.demo.exceptionStack.Test.fun2(Test.java:15)
	... 6 more
複製程式碼

上述異常棧表示，RuntimeException由IOException導致；1行與9行下是各異常的呼叫路徑trace。不熟悉Java異常棧的可參考你真的會閱讀Java的異常資訊嗎？。

輪子

調研讓我們拒絕重複造輪子

Python3已經支援了異常鏈，通過from關鍵字即可記錄cause。

Python2 future包提供的所謂異常鏈raise_from我是完全沒明白到哪裡列印了cause:

from future.utils import raise_from


class DatabaseError(Exception):
    pass


class FileDatabase:
    def __init__(self, filename):
        try:
            self.file = open(filename)
        except IOError as exc:
            raise_from(DatabaseError('failed to open'), exc)


# test
fd = FileDatabase('non_existent_file.txt')
複製程式碼

那麼，11行丟擲的IOError呢？？？似乎僅僅多了一句無效資訊（future包裡的raise e）。

Traceback (most recent call last):
  File "/Users/mobkeysayhi/PycharmProjects/Wheel/utils/tmp/tmp.py", line 17, in <module>
    fd = FileDatabase('non_existent_file.txt')
  File "/Users/mobkeysayhi/PycharmProjects/Wheel/utils/tmp/tmp.py", line 13, in __init__
    raise_from(DatabaseError('failed to open'), exc)
  File "/Library/Python/2.7/site-packages/future/utils/__init__.py", line 454, in raise_from
    raise e
__main__.DatabaseError: failed to open
複製程式碼

有知道正確姿勢的求點破。

沒找到重複輪子真是極好的

非常簡單：

import traceback


class TracedError(BaseException):
    def __init__(self, msg="", cause=None):
        trace_msg = msg
        if cause is not None:
            _spfile = SimpleFile()
            traceback.print_exc(file=_spfile)
            _cause_tm = _spfile.read()
            trace_msg += "\n" \
                         + "\nCaused by:\n\n" \
                         + _cause_tm
        super(TracedError, self).__init__(trace_msg)


class ErrorWrapper(TracedError):
    def __init__(self, cause):
        super(ErrorWrapper, self).__init__("Just wrapping cause", cause)


class SimpleFile(object):
    def __init__(self, ):
        super(SimpleFile, self).__init__()
        self.buffer = ""

    def write(self, str):
        self.buffer += str

    def read(self):
        return self.buffer
複製程式碼

目前只支援單執行緒模型，github上有doc和測試用例，戳我戳我。

一個測試輸出如下：

Traceback (most recent call last):
  File "/Users/monkeysayhi/PycharmProjects/Wheel/utils/exception_chain/traced_error.py", line 68, in <module>
    __test()
  File "/Users/monkeysayhi/PycharmProjects/Wheel/utils/exception_chain/traced_error.py", line 64, in __test
    raise MyError("test MyError", e)
__main__.MyError: test MyError

Caused by:

Traceback (most recent call last):
  File "/Users/monkeysayhi/PycharmProjects/Wheel/utils/exception_chain/traced_error.py", line 62, in __test
    zero_division()
  File "/Users/monkeysayhi/PycharmProjects/Wheel/utils/exception_chain/traced_error.py", line 58, in zero_division
    a = 1 / 0
ZeroDivisionError: integer division or modulo by zero
複製程式碼

另外，為方便處理後重新丟擲某些異常，還提供了ErrorWrapper，僅接收一個cause作為引數。用法如下：

for pid in pids:
    # process might have died before getting to this line
    # so wrap to avoid OSError: no such process
    try:
        os.kill(pid, signal.SIGKILL)
    except OSError as os_e:
        if os.path.isdir("/proc/%d" % int(pid)):
            logging.warn("Timeout but fail to kill process, still exist: %d, " % int(pid))
            raise ErrorWrapper(os_e)
        logging.debug("Timeout but no need to kill process, already no such process: %d" % int(pid))
複製程式碼

參考：

本文連結：輪子|Python2異常鏈
作者：猴子007
出處：monkeysayhi.github.io
本文基於 知識共享署名-相同方式共享 4.0 國際許可協議釋出，歡迎轉載，演繹或用於商業目的，但是必須保留本文的署名及連結。

相關文章