Skip to content

Latest commit

 

History

History
208 lines (150 loc) · 5.6 KB

README.md

File metadata and controls

208 lines (150 loc) · 5.6 KB
Raw

Sift Log - JSON logging adapter for Python (now in color)

Features

  • Tag log statements with arbitrary values for easier grouping and analysis
  • Add keyword arguments that are converted to JSON values
  • Variable substitution
  • Specifies where log calls are made from
  • Colorized logs on a console (POSIX only)
  • TRACE log level built-in

Examples

Setup (with no bells and whistles yet)

import logging
from siftlog import SiftLog

core_logger = logging.getLogger()

logger = SiftLog(
    core_logger
)

logger.info('Hello')

{"msg": "Hello", "time": "12-12-14 10:12:01 EST", "level": "INFO"}

Logging with tags

log.debug('Creating new user', 'MONGO', 'STORAGE')

{"msg": "Creating new user", "tags": ["MONGO", "STORAGE"], "level": "DEBUG",}

Appending more data

log.debug('Some key', is_admin=True, username='papito')

{"msg": "Some key", "is_admin": true, "username": "papito", "level": "DEBUG"}

String substitution

log.debug('User "$username" admin? $is_admin', is_admin=False, username='fez')

{"msg": "User \"fez\" admin? False", "username": "fez", "is_admin": false, "level": "DEBUG"}

Note that this is more performant than regular logging. String interpolation is expensive, possibly causing significant drag in a system with lots of log statements. With deferred string substituion, a log statement will never get expanded if it's not being logged at a given level in the first place.

Constants (re-occurring values)

Logging is more noise than useful if you do not know the context of a log message. Which container is it running on? What environment is it in? Constants can be set up once, and they will stick throughout the lifecycle of a logger:

import os
from siftlog import SiftLog

log = SiftLog(logger, pid=os.getpid(), env='INTEGRATION')

{"msg": "Extremely hardcore", "pid": 37463, "env": "INTEGRATION", "level": "INFO"}

Dynamic logging context - callbacks

Often you need to add dynamic contextual data to log statements as opposed to one-time constants. SiftLog can accept methods that will provide dynamic information depending on the context.

Logging request IDs or user IDs is a very common use case. To log a thread-local property with Flask, for example, we can do the following:

import flask

def get_user_id():
    if flask.has_request_context():
        return flask.g.user_id

user_aware_logger = SiftLog(u_id=get_user_id)

{"msg": "Logged in", u_id="200713391", "level": "INFO"}

Setup

Logging to console

import sys
import logging
from siftlog import SiftLog

logger = logging.getLogger()
logger.setLevel(logging.INFO)
handler = logging.StreamHandler(sys.stdout)
logger.addHandler(handler)

log = SiftLog(logger)

In this fashion, you can direct the JSON logs to any logging handler.

Color

For enhanced flamboyancy, attach the ColorStreamHandler to your logger. The output will not have color if the logs are being output to a file, or on systems that are not POSIX (will not work on Windows for now).

from siftlog import SiftLog, ColorStreamHandler

logger = logging.getLogger()
handler = ColorStreamHandler(sys.stdout)
logger.addHandler(handler)

log = SiftLog(logger)

For development, you can opt in to use ColorPlainTextStreamHandler, for logs that are easier to parse visually.

Performance

While the above should play, it's highly recommended that the color handler is only attached conditionally for local development.

Different colors

You can change font background, text color, and boldness:

from siftlog import ColorStreamHandler

handler = ColorStreamHandler(sys.stdout)
handler.set_color(
    logging.DEBUG, bg=handler.WHITE, fg=handler.BLUE, bold=True
)

Supported colors

  • ColorStreamHandler.BLACK
  • ColorStreamHandler.RED
  • ColorStreamHandler.GREEN
  • ColorStreamHandler.YELLOW
  • ColorStreamHandler.BLUE
  • ColorStreamHandler.MAGENTA
  • ColorStreamHandler.CYAN
  • ColorStreamHandler.WHITE

Custom time format

log = SiftLog(logger)
SiftLog.TIME_FORMAT = '%Y/%m/%d %H:%M:%S.%f'

Define the format as accepted by strftime()

Custom location format

log = SiftLog(logger)
SiftLog.LOCATION_FORMAT = '$module:$method:$line_no'

The format should be a string containing any of the following variables:

  • $file
  • $line_no
  • $method
  • $module

Custom core key names

Core keys, such as msg and level can be overridden, if they clash with common keys you might be using.

The following can be redefined:

  • SiftLog.MESSAGE (default msg)
  • SiftLog.LEVEL (default level)
  • SiftLog.LOCATION (default loc)
  • SiftLog.TAGS (default tags)
  • SiftLog.TIME (default time)

As in:

log = SiftLog(logger)
SiftLog.log.MESSAGE = "MESSAGE"

Development flow

Poetry is used to manage the dependencies.

Most things can be accessed via the Makefile, if you have Make installed. Without Make, just inspect the Makefile for the available commands.

# use the right Python
poetry use path/to/python/3.8-ish

# make sure correct Python is used
make info

# install dependencies
make install

# run tests
make test

# run visual tests (same as tests but with output)
make visual

# formatting, linting, and type checking
make lint

Running a single test

In the standard Nose tests way:

poetry run nosetests siftlog/tests/test_log.py:TestLogger.test_tags