First draft attempt at benchmarking logging vs. sys.stderr.write'ing. See: https://pypi.python.org/pypi/benchmark/

logging_test.py

import logging
import sys

logger = logging.getLogger('simple_example')
logger.setLevel(logging.DEBUG)
ch = logging.StreamHandler()
ch.setLevel(logging.DEBUG)
formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
ch.setFormatter(formatter)
logger.addHandler(ch)

verbosity = 2


def debug(message):
    if verbosity > 1:
        sys.stderr.write(message)


import benchmark


class BenchmarkLogging(benchmark.Benchmark):

    def test_debug(self):
        logging.debug("i am a message")


class BenchmarkPrint(benchmark.Benchmark):

    def test_debug(self):
        debug("i am a message")

if __name__ == "__main__":
    benchmark.main(format="markdown", numberFormat="%.4g", each=1000000)

output

Benchmark Report
================

BenchmarkLogging
----------------

 name | rank |  runs |      mean |       sd | timesBaseline
------|------|-------|-----------|----------|--------------
debug |    1 | 1e+06 | 2.293e-06 | 1.15e-06 |           1.0

BenchmarkPrint
--------------

 name | rank |  runs |      mean |       sd | timesBaseline
------|------|-------|-----------|----------|--------------
debug |    1 | 1e+06 | 1.362e-06 | 8.85e-07 |           1.0

Each of the above 2000000 runs were run in random, non-consecutive order by
`benchmark` v0.1.5 (http://jspi.es/benchmark) with Python 2.7.3
Linux-3.8.11-200.fc18.x86_64-x86_64 on 2013-08-20 14:01:51.