Python logger.debug将参数转换为字符串而不进行日志记录 [英] Python logger.debug converting arguments to string without logging

问题描述

我正在优化执行某种计算的Python程序.它相当广泛地使用了NumPy.该代码中充满了 logger.debug 调用(logger是标准的Python日志对象).

I'm optimizing a Python program that performs some sort of calculation. It uses NumPy quite extensively. The code is sprinkled with logger.debug calls (logger is the standard Python log object).

当我运行cProfile时，我看到将数组转换为字符串的Numpy函数花费了50％的执行时间.这令人惊讶，因为没有处理程序将消息输出为DEBUG级别，只有INFO或更高级别.

When I run cProfile I see that Numpy's function that converts an array to string takes 50% of the execution time. This is surprising, since there is no handler that outputs messages as the DEBUG level, only INFO and above.

即使没有人使用该字符串，记录器为什么还要将其参数转换为字符串?有没有办法防止它发生(除了不执行记录程序调用)?

Why is the logger converting its arguments to string even though nobody is going to use this string? Is there a way to prevent it (other than not performing the logger calls)?

推荐答案

没有看到您的代码，很难分辨正在发生什么，但是要查看两者的 logging .__ init __.py 的源代码.Python2.7和Python3.4(我所躺在的地方)显示，例如，对 logger.debug(...)的调用看起来像:

Without seeing your code, it's hard to tell what's happening, but looking at the source code for logging.__init__.py for both Python2.7 and Python3.4 (what I had lying around) shows that a call to logger.debug(...) for example, looks like:

if self.isEnabledFor(DEBUG):
    self._log(DEBUG, msg, args, **kwargs)

这暗示着，如果未启用 DEBUG ，则 logger 不会对任何参数进行任何处理.

Which implies that if DEBUG isn't enabled, the logger isn't going to do any processing on any of the arguments.

一个简短的注释，尽管我之前很忙，但是，在调用 logging.debug(...)之前，您可能不小心构建了字符串.注意区别:

One quick note, that I've burned myself on before though, is that you may be accidentally building the string before ever calling logging.debug(...). Note the difference:

logger.debug("My big array: %s" % myBigArray) ## DON'T DO THIS
logger.debug("My big array: %s", myBigArray)  ## much better

不同之处在于第一行甚至在进入 logger.debug 之前就进行了字符串格式化-您传递的是一个(可能是大的)字符串参数.另一方面，仅在启用 DEBUG 的情况下，第二行才会进行字符串格式化.

The difference is the first line does the string-formatting before it ever even gets into logger.debug - you're passing in a single argument, which is a (potentially large) string. The second line, on the other hand, will do the string-formatting only if DEBUG is enabled.

这篇关于Python logger.debug将参数转换为字符串而不进行日志记录的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！