Deploying Pywr releases

This repository is used to deploy pywr releases to conda & pypi. Currently conda deployment is done using travis (Linux and OSX) and appveyor (Windows). Pypi releases are a more manual process at the current time.

Releasing to Anaconda.

Travis: https://travis-ci.org/pywr/pywr-deploy/ Appveyor: https://ci.appveyor.com/project/snorfalorpagus/pywr-deploy

Release process:

• Update version number in pywr/__init__.py and conda-recipe/meta.yaml to release.
• Tag release: git tag -a vX.X -m "Release X.X"
• Update version number in pywr/__init__.py and conda-recipe/meta.yaml to development.
• Push tag: git push origin vX.X
• Update version number and git_rev in pywr-deploy/meta.yaml
• Push to deploy!

Releasing to Pypi

Source distribution

Simply compile the source distribution locally from the appropriate tagged commit of Pywr. Then upload using twine.

python setup.py sdist

Linux wheels

Once the new release is tagged and available on Github the build-wheels.sh script can be used inside a specific docker container to build all of the Linux wheels. This process builds GLPK inside the manylinux1 container and then builds wheels for all the versions of Python required.

docker run -v `pwd`:/io quay.io/pypa/manylinux1_x86_64 /io/build-wheels.sh

After completing the wheels should be located in the wheelhouse folder of the pywr-deploy repository. These can then be uploaded to pypi using twine.

twine upload wheelhouse/*

Note there is no deployment of Python 2.x wheels. It's too much effort.