A few years ago, I tried out Anaconda for package management. I don’t know whether it was due to my inexperience, or if things have really changed since then, but I hated it. And I that meant I had tp spend a good amount of time trying to purge it from my system.

Recently, I say Ryan Dale give a presentation on where conda is now, and I was really impressed. It promises to mangage all your dependencies for a project, regardless of language or OS, and that sounded awesome.

So, here is what I did.

Sources

https://conda.io/docs/build_tutorials/pkgs.html#optional-convert-conda-package-for-other-platforms http://conda-test.pydata.org/docs/build_tutorials/pkgs.html https://conda.io/docs/travis.html

Before you start

• Instal miniconda
• Have your package on PyPi
• Install anaconda-client (conda install anaconda-client) What not to do —

Here’s the thing: I went through the whole process of trying to manually make the meta.yaml file, completely misunderstanding the skeleton command. So, don’t try to make it yourself. If your package is on PyPi, just do the following

# make the meta.yaml file given the metadata on pypi
# ~/miniconda3/bin/conda skeleton pypi your_package
conda skeleton pypi pyutilsnrw

# build the package
conda build pyutilsnrw

# build versions for all OS's
conda convert --platform all /home/wherever/your/build/is/conda-bld/linux-64/pyutilsnrw-0.0.7694-py36h14194ac_0.tar.bz2

# login to anaconda (this starts an interactive Q&A for your anacond acredentials
anaconda login

# upload your build(s)
anaconda upload /home/wherever/your/build/is/conda-bld/linux-64/pyutilsnrw-0.0.7694-py36h14194ac_0.tar.bz2

@

Pretty straightforward, right? Now, your package will be available on your personal conda page.