Package Version management

modulefiles

Learn more: https://modules.readthedocs.io/en/latest/INSTALL.html

1. Use modulefiles to manage custom installed libraries. Run module avail to see list of available modules. Run module --help for more information.

2. MODULEPATH variable in your environment will be used by module command to search for available modules. Recommended MODULEPATH for a user is ~/Documents/opt/modulefiles

conda

Typically get conda from miniconda

Also read Python’s messy tools

Local Installation: user level, in $HOME/miniconda3

Global Installation: Somewhere in /opt/miniconda3 ,
needs sudo access, also need to change install location during installation
The global conda can then be setup individually for each user with a conda init something like command
Then each user can create conda envs in its local $HOME/.conda path

Steps to remove local install conda

Assume miniconda is install in $HOME/miniconda3 for the current user.

1. Backup the current installation directory i.e. ~/miniconda3. This will be handy later if you want to use the same envrionments without reinstalling them. Else at least backup the environments using following command:

   mkdir conda-env-yamls && cd conda-env-yamls
    
   conda env list | grep -v '#' | grep -v '^$' |  awk '{print $1}' | while read -r env_name ; do conda env export -n ${env_name} > ${env_name}-env.yaml ; done

2. Install anaconda-clean in the base environment, run anaconda-clean to clean up dotfiles. It will show which files will be removed. anaconda-clean is supposed to create a backup directory, but in both Ryzen and Intel WS, it did not create a backup dir at home. So do a manual backup first

3. To delete current installation in ~/miniconda3, simply run rm -r ~/miniconda3

4. Delete lines regarding “conda initialise” from all your shell dotfiles (.bashrc, .zshrc, any other)

5. Reload the shell (logout and login) to see if you can still access conda (you shouldn’t be able to).

Steps to install conda globally

1. Download the latest miniconda3 installer https://docs.conda.io/projects/miniconda/en/latest/. Check shasum.

2. Run installer with sudo.

3. Select installation dir as “/opt/miniconda3”.

4. After installation, “DO NOT” initialise the environment via the installation setup.

5. Run /opt/miniconda3/bin/conda init to initialise conda in your shell, which will add “conda initialise” section in your dotfiles.

6. Logout and login. Run conda info to see details of the installation.

Steps to get back previous envionments

1. If you backedup previous conda installation dir, say ”~/miniconda3” ; and your new dir is “/opt/miniconda3”, then to get back the previous envrionments, simply run the command

cp ~/miniconda3/envs/* /opt/miniconda3/envs/

apt (ubuntu)

How to manage alternate versions of the same app using OS pkg manager?

For ubuntu, uses apt , Using update-alternatives
It maintains symbolic links determining default commands

The RHEL, which uses dnf/yum package manager, has alternatives command to achive the same. Know more

e.g. To know how to install multiple gcc versions and manage them: https://phoenixnap.com/kb/install-gcc-ubuntu

This is how update-alternatives will create a symlink in the default gcc executable location /usr/bin/gcc , pointing to one of the installed gcc versions

sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-12 12 --slave /usr/bin/g++ g++ /usr/bin/g++-12
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-13 13 --slave /usr/bin/g++ g++ /usr/bin/g++-13

To change the gcc version, use
sudo update-alternatives --config gcc

Passing g++ as --slave ensures that when gcc is updated, g++ link is also updated