stacklog

Stack log messages


Keywords
stacklog, logging, python
License
MIT
Install
pip install stacklog==1.1.0

Documentation

PyPI Shield Downloads Travis CI Shield

stacklog

Stack log messages

Overview

Stacklog is a tiny Python library to stack log messages.

A stack-structured log is an approach to logging in which log messages are (conceptually) pushed onto a stack and emitted only when the corresponding block returns. Stacklog provides a single method, stacklog, which serves as either a decorator or a context manager. This is exceptionally useful in small projects or one-off scripts.

This is illustrated best with an example:

with stacklog(print, 'Running some code'):
    with stacklog(print, 'Running some other code'):
        pass

This produces the following logging output:

Running some code...
Running some other code...
Running some other code...DONE
Running some code...DONE

When the code within a stacklog context completes, the provided message is echoed along with the return status, one of DONE or FAILURE. That's pretty much it. Customization and advanced features are available through callbacks.

Install

stacklog has been developed and tested on Python 2.7 and 3.5+.

pip install stacklog

Quickstart

How often do you find yourself using the following logging anti-pattern in Python?

import logging

def a():
    logging.info('Running a')
    do_something()
    logging.info('Done with a')

def b():
    logging.info('Running b')
    a()
    logging.info('Done with b')

try:
    b()
except:
    logging.info('There was an error running b')

The intention here is to log the beginning and end of procedure calls for use in debugging or user monitoring. I call this an anti-pattern because:

  • it requires excessive manual attention to writing/updating logging calls at entry/exit sites
  • it results in redundant exception handling logic
  • the resulting log messages can be misleading if errors occur

Instead, the approach taken by stacklog is to accomplish this using only decorators and context managers.

Usage as decorator

Here is the above example using the stacklog as a decorator:

@stacklog(logging.info, 'Running a')
def a():
    raise Exception

@stacklog(logging.info, 'Running b')
def b():
    a()

b()

This produces logging output:

INFO:root:Running b...
INFO:root:Running a...
INFO:root:Running a...FAILURE
INFO:root:Running b...FAILURE

Usage as context manager

Here is another example using stacklog as a context manager:

>>> with stacklog(logging.info, 'Running some code'):
...     do_something()
...
INFO:root:Running some code...
INFO:root:Running some code...DONE

Providing custom conditions

A condition is a tuple exception, status. If the provided exception is raised during the execution of the provided code, the provided status is logged instead of the default FAILURE.

>>> with stacklog(logging.info, 'Running some code', conditions=[(NotImplementedError,
'SKIPPED')]):
...     raise NotImplementedError
...
INFO:root:Running some code...
INFO:root:Running some code...SKIPPED

Customization with callbacks

The behavior of stacklog is fully customizable with callbacks.

The main thing that a callback will do is call the passed stacklog instance's log method with some custom suffix.

First, there are three callbacks to customize the behavior of logging at the beginning of the block, at successful completion of the block, and at failure of the block. Only one function can be registered at a time for each of these events.

  • on_begin(func: stacklog -> None)
  • on_success(func: stacklog -> None)
  • on_failure(func: stacklog -> None)

Second, one can customize failure behavior given different possible exceptions that are raised, by passing a pair of functions, the first to match an exception that was raised during block execution and the second to respond to the exception. Many pairs of functions can be registered, but only the most recent one to be registered will be executed in the case that multiple functions match.

  • on_condition(match: *exc_info -> bool, func: stacklog, *exc_info -> None)

Third, one can initialize and dispose of resources before and after the block's execution. This is relevant for starting/stopping timers, etc. Many functions can be registered and they will all be executed.

  • on_enter(func: stacklog -> None)
  • on_exit(func: stacklog -> None)

See the implementation of stacktime for an example.

Adding timing information

One can customize stacklog with callbacks to, for example, add information on the duration of block execution.

>>> with stacktime(print, 'Running some code', unit='ms'):
...     time.sleep(1e-2)
...
Running some code...
Running some code...DONE in 11.11 ms