Python context-log library

About

context-log is a simple library to emit contextual information in structured logs (JSON).

It works particularly well in a Docker or Serverless (e.g. AWS Lambda) environment where a single thread executes a request and produces a response.

The library uses python threading to store contextual information that is automatically added to all subsequent logs in a contextMap field.

Because the library uses the Python thread local context it works across packages and modules in a given project.

The approach is loosely based on the Log4j 2 API Thread Context.

Usage

Structured logging can be achieved with the python-json-logger library.

Simply add project dependencies to requirements.txt:

python_json_logger
PyYAML
context-log

Add the code below to the main code module.

Add the following YAML configuration in the resources/logging.yaml file, which outputs JSON structured logs to stdout.

version: 1
formatters:
  json:
    class:  .jsonlogger.JsonFormatter
    format: '%(asctime)s %(name)s %(levelname)s %(message)s %(filename)s'
handlers:
  console:
    class: logging.StreamHandler
    level: DEBUG
    formatter: json
root:
  level: DEBUG
  handlers:
    - console

Use the context_log library to emit logs. Example below.

import logging.config
import yaml

with open('resources/logging.yaml', 'r') as log_config_file:
  logging.config.dictConfig(yaml.safe_load(log_config_file))

from context_log import ContextLog

def handler(event, context):
    # Clear context (e.g. re-use) and get logger
    log = ContextLog.get_logger('handler', True)
    log.info('start')

    ContextLog.put('ip', '1.2.3.4')

    # Helper to add start time in ISO and epoch time
    ContextLog.put_request_start_time()

    # Process request
    sleep(0.1)

    # Helper to add end time in ISO and epoch time
    # as well as duration in milliseconds
    ContextLog.put_request_end_time()

    log.info('end')

First log info event:

{
    "asctime": "2019-09-19 11:53:20,479",
    "name": "handler",
    "levelname": "INFO",
    "message": "start",
    "filename": "test_example.py",
    "contextMap": {}
}

Second log info event:

{
  "asctime": "2019-09-19 11:53:20,580",
  "name": "handler",
  "levelname": "INFO",
  "message": "end",
  "filename": "test_example.py",
  "contextMap": {
      "ip": "1.2.3.4",
      "start-time": "2019-09-19T11:53:20.480085",
      "epoch-start-time": 1568890400.480085,
      "end-time": "2019-09-19T11:53:20.580513",
      "epoch-end-time": 1568890400.580513,
      "duration": 100.428}
}

The Detail

The standard logger is wrapped by a LoggerAdapter. It is therefore imperative that the ContextLog.get_logger(name='<name>', clear=True|False) call is made to get the logger to emit contextual logs.

Use clear=True when starting a new request in order to clear the previous context if the thread is re-used. This is typically the case in thread pools and in AWS Lambda's.

To manipulate or retrieve the contextMap use the following methods:

• clear()
• put(key, value)
• get(key)
• get_map()

There are also a number of helpers in an attempt to standardise log output contextMap fields:

• put_request_id(request_id)
• put_request_method(request_method)
• put_request_path(request_path)
• put_response_status(response_status)
• put_start_time()
• put_end_time()
• put_request_user_id(request_user_id)
• put_request_client_id(request_client_id)
• put_request_primary_ip(primary_ip)
• put_request_client_ip(client_ip)
• put_request_viewer_country(viewer_country)
• put_trigger_source(trigger_source)

Contributing

Pull requests are more than welcome.

Running pytest

Create virtualenv, download dependencies and run tests:

python3 -m venv .venv
source .venv/bin/activate
pip3 install -r tests/requirements.txt
pip3 install -e .
pytest

Running tox

pip3 install --user --upgrade tox
tox

Releasing library to PyPI

Short version from the Packaging Python Projects site.

Install the release tools:

python3 -m pip install --user --upgrade setuptools wheel twine

Remove old distribution(s):

rm -rf dist/

Build the context-log package:

python3 setup.py sdist bdist_wheel

Upload context-log first to Test PyPI:

python3 -m twine upload --repository-url https://test.pypi.org/legacy/ dist/*