Uploading conda binaries to Anaconda Cloud

After a successful build, you typically want to upload the package to the Anaconda Cloud, where we currently host all packages. Please do not abuse the "unlimited" space we were given there as a result of good networking with Continuum Analytics. (and also finding a bug in their software :P) Before you do this, you need to install anaconda-client:

conda install anaconda-client

A unique package name is always identified as platform/name-version-build_string.tar.bz2. This is important to remember, if you don't want to overwrite an existing package by mistake. Packages can be given non-standard labels (standard is the main).

Adding the --label option tells Anaconda Cloud to make the upload visible only to users who specify that label. For example:

anaconda upload /path/to/conda-package-2.0.tar.bz2 --label test

If you do not put any label, the default one is main. You can put multiple labels on packages. You can also manage your package labels from the dashboard on the Anaconda Cloud UI.

Use --force to replace a package with a newer build. Anaconda first fully removes the existing package, before uploading your new one. If the upload process fails, the existing package is still gone. Beware, overwriting a package means the previous labels will be lost, in addition to the package being gone for good!

We suggest to always download/backup a package before uploading one on the Anaconda Cloud. Caveat: When you download a package from Anaconda Cloud, the platform name (e.g., linux-64, os-64) will be prepended to the original .tar.bz2 file. In case you need to re-upload the backed-up version, remove that prefix, otherwise you endup with checksum problems, and the package cannot be installed by users!