Bash completion support for the conda command

We currently support completion for commands, options, sub-commands, sub-command options. For example:

$ conda <tab>
activate   convert     develop   info      list         render   skeleton    verify
build      create      env       init      metapackage  run      uninstall
clean      deactivate  help      inspect   package      search   update
config     debug       index     install   remove       server   upgrade
$ conda

$ conda d<tab>
deactivate  debug  develop
$ conda de

$ conda --<tab>
--help  --version
$ conda --

$ conda develop --<tab>
 --build_ext  --clean  --help  --name  --no-pth-file  --prefix  --uninstall
$ conda develop --

$ conda env <tab>
create  export  list    remove  update
$ conda env

$ conda env create --<tab>
--file  --force  --help  --json --name  --prefix  --quiet  --verbose
$ conda env create --

Note: on some systems you have to hit <tab> twice to trigger completion.

Completions are dynamically determined based on the plugins you have loaded and the help strings for the various commands. This makes completion a little slower than it could be, however, it does allows the code to automatically adjust to future changes, and hence, minimising the need for maintenance.

Note: MacOS has bash-3.2 (from 2007) installed by default. The upstream bash-completion library, which we depend on, requires a bash >= 4.2, so you will need to upgrade your bash if you want to use this script. Upgrading bash is possible, either via homebrew or conda, or other methods, but you probably want to search for how to change the location of your shell in MacOS.

Installation

There are two supported ways to install this feature:

Method 1: Using conda

The easiest way to install this feature is via the conda-bash-completion package:

conda install -c conda-forge conda-bash-completion

This installs the completion code and a specially patched version of the bash-completion library into you default environment. If you have already set up conda shell initialization via conda init bash, then simply restarting your shell should be sufficient to complete the integration.

Some uses prefer to disable the automatic activation of their base environment by setting auto_activate_base: false in their ~/.condarc file. These users will need to append something like the following

CONDA_ROOT=~/anaconda3   # <- set to your Anaconda/Miniconda installation directory
if [[ -r $CONDA_ROOT/etc/profile.d/bash_completion.sh ]]; then
    source $CONDA_ROOT/etc/profile.d/bash_completion.sh
else
    echo "WARNING: could not find conda-bash-completion setup script"
fi

to their ~/.bashrc script in order to have the completion code loaded. Note, the above block must be run after any existing bash-completion setup and after the conda initialization block.

Method 2: Manually

If you already have the bash-completion library installed system-wide (see Repology for a comprehensive list of operating system distributions, package names, and available versions), and already have your shell set up to load completions, then you could simply copy the conda completion script to the ~/.local/share/bash-completion/completions/ directory.

Testing

The tests require bash-completion, pytest, pexpect installed in an active conda environment.

You could run something like:

conda activate base
conda install -c tartansandal pytest pexpect bash-completion
pytest

Ideally would like a dev enironment to be set up and used.

Seem to have some problems with activation and changing the prompt, which confuses pexpect.

At some point I'll figure out how to do that, but for release testing I'm just using my current environment.

Thanks

Thanks to Kale Franz (@kalefranz) for encouranging me to work on this in the first place 😄.

Thanks to Simon Frei (@imsodin) for raising and providing extensive help in debugging some difficult integration cases.

Thanks to Mike Sarahan (@msarahan) for encourangment and discusions that lead to packaging this feature rather than hard wiring it into the bash initialization.

Special thanks to all users who have provided feeback on issues and bugs. 😄