How to install Conda to use the group's tools and codes.
Conda is a package management system suitable for Python packages and also for many scientific codes, not necessarily written in Python (for example, Packmol and Gnuplot are available as Conda packages). Conda manages dependencies, meaning it installs all that is required for a given software to run and tries to keep packages updated while ensuring compatibility of versions.
Download a Miniconda installer archive
bash Miniconda3-latest-XXX-XXX.sh
conda activate
conda config --append channels conda-forge
conda install numpy matplotlib jupyter lmfit
You are ready to use (most of) the group's Python tools.
I recommend Miniconda, which starts as a smaller installation. Packages are then installed as needed.
In the Installing section choose the archive corresponding to your platform. On Mac I use the bash archive and then in a terminal do:
bash Miniconda3-latest-XXX-XXX.sh
This creates a folder named miniconda3 in your home dir, and also asks to modify your shell configuration files to enable activation of the conda environment (you'll have to close and reopen your terminal shell).
If Conda is not activated automatically you can do that with
conda activate
To exit the conda environment type
conda deactivate
(If you have other Python installations in your computer outside Conda, for example with Homebrew or Pip, it is safer to manage those installations while Conda is deactivated.)
Channels are repositories of packages. The defaults channel contains the large set of Anaconda packages and is sufficient for many uses. It is the channel configured by default.
For an even wider choice and for scientific computing, you can add the conda-forge channel:
conda config --append channels conda-forge
You can check your configuration of channels:
conda config --show channels
You can change the order of priority, see Managing channels.
The command to install packages is
conda install numpy
conda install matplotlib
To search for packages (in the available channels):
conda search lmfit
You can list the packages already installed with
conda list
To keep packages up-to-date issue the following command regularly:
conda upgrade --all
Environments allow different groups of packages to exist in separate spaces. This is useful for large packages that have many dependencies, which easily give rise to conflicts between versions of certain dependency packages.
The default environment is called base. I recommend using the base environment for the essential Python packages such as numpy, matplotlib, pandas or jupyter. These, together with lmfit and maybe a few more tools, are enough to run the group's Python scripts.
If you will not be using large computational codes, working only in the base environment should suffice and you can skip the rest of this section.
If you will be using large packages (OpenMM, Psi4) then I recommend creating a separate environment for each one. For example,
conda create --name omm
You can list your environments with
conda env list
To activate or deactivate the environment you want to work in:
conda activate omm
conda deactivate
(you will be back in base).
You can install packages in your active environment as usual.
Conda and Pip are two different package management systems, and a certain discipline is required to avoid confusions. It is possible to install Pip inside a Conda environment and then packages from the PyPi repository. However, this is not an optimal setup and should be avoided if possible (Pip doesn't know about Conda dependencies). Instead, it is advisable to install packages from conda-forge and only resort to Pip packages as a last option.
Pip can also reside outside Conda (this is common in a Linux system or in MacOS with Homebrew). In this case it is safer to deactivate Conda before installing or updating Pip packages.