=================================
Environments and Package Managers
=================================

Pip and venv
------------

Throughout this tutorial, we used ``pip``, a Python package manager, to install
our software and other Python software we needed. In :doc:`preliminaries`, we
used ``venv`` to create a "virtual environment", a sandbox separate from the
general system Python where we could install and use software without
interfering with any system Python tools or other projects.

This tool chain is perfectly valid and effective, but you made find that as your
project evolves, other tools might be necessary for managing packages and
virtual environments (e.g. if you need to work with multiple versions of
Python). There are many alternatives to ``venv``.  `This post of Stack Overflow
<https://stackoverflow.com/a/41573588/1221924>`_ is a good survey of the field.

Conda
-----

Conda, which is both a package manager (like ``pip``) and an virtual
environment manager (like ``venv``) was developed specifically for the
scientific Python community. Conda is not limited to Python packages: it has
grown into a general-purpose user-space package manager. Many important
scientific Python packages have important dependencies that aren't Python
packages and can't be installed using ``pip``. Before conda, the user had to
sort out how to install these extra dependencies using the system package
manager, and it was a common pain point. Conda can install all of the
dependencies.

Conda is conceptually a heavier lift that ``pip`` and ``venv``, which is why we
still with the basics in :doc:`preliminaries`, but it is extremely popular
among both beginners and experts, and we recommend becoming familiar with it.

Check whether ``conda`` is already installed:

.. code-block:: bash

   conda

If that is not found, follow the steps in the
`conda installation guide <https://conda.io/docs/user-guide/install/index.html>`_,
which has instructions for Windows, OSX, and Linux. It offers both miniconda (a
minimal installation) and Anaconda. For our purposes, either is suitable.
Miniconda is a faster installation.

Now, create an environment.

.. code-block:: bash

   conda create -n my-env python=3.6

The term ``my-env`` can be anything. It names the new environment.

Every time you open up a new Terminal / Command Prompt to work on your
project, activate that environment. This means that when you type ``python3``
or, equivalently, ``python`` you will be getting a specific installation of
Python and Python packages, separate from any default installation on your
machine.

.. code-block:: bash

   conda activate my-env

The use of virtual environments leads to multiple instances of ``python``,
``pip``, ``ipython``, ``pytest``, ``flake8`` and other executables on your
machine. If you encounter unexpected behavior, use ``which ____`` to see which
environment a given command is coming  from. (Linux and OSX only.)