Amazon Ion Python

An implementation of Amazon Ion for Python.

This package is designed to work with Python 2.6+ and Python 3.3+

This package is an early work in progress under active development, and is not (yet) considered a complete implementation of Ion.

Development

It is recommended to use virtualenv to create a clean environment to build/test Ion Python.

$ virtualenv venv
...
$ . venv/bin/activate
$ pip install -r requirements.txt
$ pip install -e .

You can also run the tests through setup.py or py.test directly.

$ python setup.py test

Tox Setup

In order to verify that all platforms we support work with Ion Python, we use a combination of tox with pyenv.

Install relevant versions of Python:

$ for V in 2.6.9 2.7.12 3.3.6 3.4.5 3.5.2 pypy-5.3.1; do pyenv install $V; done

Note that on Mac OS X, you may need to change the CFLAGS:

$ for V in 2.6.9 2.7.12 3.3.6 3.4.5 3.5.2 pypy-5.3.1; do
    CFLAGS="-I$(xcrun --show-sdk-path)/usr/include" pyenv install $V; done

Once you have these installations, add them as a local pyenv configuration

$ pyenv local 2.6.9 2.7.12 3.3.6 3.4.5 3.5.2 pypy-5.3.1

At the time of this writing, on Mac OS X, you may have problems with pyenv and pypy. On this platform, it is probably easier to have pypy not managed by pyenv and install and use it directly from brew.

Assuming you have pyenv properly set up (making sure pyenv init is evaluated into your shell), you can now run tox:

# Run tox for all versions of python which executes py.test.
$ tox

# Run tox for just Python 2.7 and 3.5.
$ tox -e py27,py35

# Run tox for a specific version and run py.test with high verbosity
$ tox -e py27 -- py.test -vv

# Run tox for a specific version and just the virtual env REPL.
$ tox -e py27 -- python

TODO

Here are some, rather critical, things that need to be implemented:

• Design and implement an event oriented text and binary pull-parser with coroutines.
• Provide a simplejson style API for ease-of-use.
  • Provide proper type mappings that retains Ion data model metadata (e.g. Ion type and annotations).

In addition, there are build, deployment, or release tasks that are required:

• Add support for code coverage reporting.
  • Publish coverage to something like Coverage.io
• Provide proper documentation generation via Sphinx.
  • Add good surrounding documentation around setup/development/contribution/getting started.
  • Publish documentation to Read the Docs.
• Follow Python Packaging Guide best practices as appropriate.
  • Distribute into PyPI
• Consider using something like PyPy.js to build an interactive shell for playing with Ion python and provide a client-side Ion playground.