Lazy logger message string evaluation

Question

I\'m using standard python logging module in my python application:

import logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(\"log\")
whi         


        

Answer 1

As Shane points out, using

log.debug("Some message: a=%s b=%s", a, b)

... instead of this:

log.debug("Some message: a=%s b=%s" % (a, b))

saves some time by only performing the string formatting if the message is actually logged.

This does not completely solve the problem, though, as you may have to pre-process the values to format into the string, such as:

log.debug("Some message: a=%s b=%s", foo.get_a(), foo.get_b())

In that case, obj.get_a() and obj.get_b() will be computed even if no logging happens.

A solution to that would be to use lambda functions, but this requires some extra machinery:

class lazy_log_debug(object):
    def __init__(self, func):
        self.func = func
        logging.debug("%s", self)
    def __str__(self):
        return self.func()

... then you can log with the following:

lazy_log_debug(lambda: "Some message: a=%s b=%s" % (foo.get_a(), foo.get_b()))

In that case, the lambda function will only be called if log.debug decides to perform the formatting, hence calling the __str__ method.

Mind you: the overhead of that solution may very well exceed the benefit :-) But at least in theory, it makes it possible to do perfectly lazy logging.