Skip to main content

Run experiments with Python - quick and clean.

Project description

Version Python Version Build Status Coverage Status

There is no shortage of great Python libraries for command line interfaces, logging, configuration file management, persistent state, or plotting. When writing small scripts for quick experiments though, it’s often too much effort to configure these components, and one ends up rewriting the same setup code over and over again.

Pyexperiment fixes this by providing a simple way to jump start a short experiment. Importing pyexperiment will give you:

  • A basic command line interface that allows calling arbitrary functions (and passing arguments) from the command prompt, providing help text derived from the functions’ docstrings and zsh/bash autocompletion (based on the standard library’s argparse and argcomplete).

  • A simple configuration management with an easy way to provide default values (based on the excellent configobj library).

  • A thread-safe logger with configurable logging levels, timing utilities with statistics, and rotating log files (based on the standard library’s logging module).

  • Persistent state with platform independent, configurable, (optionally rotating) state files that are compatible with many other programs (based on h5py).

  • A sensible setup for plotting with configurable defaults (based on matplotlib).

  • Many other bits and pieces that might come in handy…

As a design principle, pyexperiment’s components come ready to use without any further configuration. Inevitably then, the choices made in this setup are opinionated and may or may not fit your personal taste. Feel free to start a discussion on the issues page.

For more documentation, see the automatically generated pages here. For usage examples, check the examples folder.

Installation

The easiest way to install pyexperiment is from pypi, just call pip install pyexperiment (in a virtualenv, prepend sudo for system wide installation).

The pyexperiment package has a few external dependencies (as you can see in the requirements.txt):

  • six

  • configobj

  • numpy

  • h5py

  • matplotlib

  • lockfile

  • toolz

  • IPython (optional, adds –interactive command)

  • argcomplete (optional, adds activate_autocompletion command)

If you install (the h5py dependency) from pypi, you may need to install libhdf5 first, e.g., by running sudo apt-get install libhdf5-dev. You may also find that you need to install cython first, e.g., by running either sudo apt-get install Cython or pip install Cython.

Reproducible experiments

To keep your experiments reproducible and avoid dependency problems, it is a good idea to automate the setup of your development environment, e.g., using a Vagrant box, or - in many cases even better - a Docker image. To get started with pyexperiment using Vagrant or Docker, you can use the Vagrantfile and setup script here, or the Dockerfile and setup scripts here.

License

The pyexperiment package is licensed under an MIT licence (see the LICENSE).

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pyexperiment-0.2.21.tar.gz (20.7 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page