Pylint’s enforcement of percent style string formatting is not about style, it’s a matter of correctness. Formatting strings before passing them to a logger may produce undesirable results with error reporting services like Sentry and may impact application performance/reliability.

Why does Pylint complain?

Let’s look at logging using various string formatting techniques.

import logging

username = 'jdoe'"{username} logged in")         # (1) bad"{} logged in".format(username)) # (2) bad"%s logged in" % username)       # (3) bad"%s logged in", username)        # (4) good

Pylint, with the proper logging lints enabled, will only accept #4 as valid. Let’s see why by looking at the logging record as it hits a filter.

record: logging.LogRecord ...

# record attributes for our log messages
record.msg, record.args == "jdoe logged in", ()        # (1)
record.msg, record.args == "jdoe logged in", ()        # (2)
record.msg, record.args == "jode logged in", ()        # (3)
record.msg, record.args == "%s logged in", ("jdoe", )  # (4)

We can see that the message #4 hasn’t been formatted, while our pre-formatted values obviously have. This lack of formatting makes the logging module especially useful.

Benefits of delayed log formatting

  1. Error reporting. Services that install a custom log handler can aggregate logs into single events when the msg has not been formatted.
  2. Robustness. If there is a problem with formatting a message it will be logged instead of raising an exception.
  3. Performance. Arguments aren’t evaluated unless logging is enabled
  4. It’s good practice according to the the Python docs.

Here are some pylints to enable to insure that logging is correct.