The Python virtual environment with Pyenv & Pipenv

python virtual environment stars in space banner

There is a history of confusion around packaging and managing a python virtual environment. Do we use pyvenv? venv? virtualenvwrapper? Lately, other options have come out of the woodwork. The use of pyenv gives us the ability to manage python versions much like nvm and rvm, while Pipenv is the successor to pip itself and is slated to be merged in eventually.

Installing Pyenv

Pyenv allows you to easily install different versions of python along side each other on your system without conflicts. It is not supported on windows but python can easily be installed from their website. Just make sure to add it to your windows PATH.

Installation is pretty straight forward, simply run these commands to get it setup.

$ git clone https://github.com/pyenv/pyenv.git ~/.pyenv
$ echo 'export PYENV_ROOT="$HOME/.pyenv"' >> ~/.bash_profile
$ echo 'export PATH="$PYENV_ROOT/bin:$PATH"' >> ~/.bash_profile

There are some more installation details on their repo page if you have any trouble. They also have instructions for users of zsh. You will likely need to reload your terminal after installation to have the pyenv command available.

➜ pyenv
pyenv 1.2.12-2-geb68ec94
Usage: pyenv <command> [<args>]

Some useful pyenv commands are:
   commands    List all available pyenv commands
   local       Set or show the local application-specific Python version
   global      Set or show the global Python version
   shell       Set or show the shell-specific Python version
   install     Install a Python version using python-build
   uninstall   Uninstall a specific Python version
   rehash      Rehash pyenv shims (run this after installing executables)
   version     Show the current Python version and its origin
   versions    List all Python versions available to pyenv
   which       Display the full path to an executable
   whence      List all Python versions that contain the given executable

See `pyenv help <command>' for information on a specific command.
For full documentation, see: https://github.com/pyenv/pyenv#readme

The primary sub-commands we will be dealing with today are local, global, and install. First, let’s look and see what versions we have on our system.

➜ pyenv versions
* system

Installing New Versions

For most users, you will only have the system interpreter. Let’s install 3.7 as another global version of python.

➜ pyenv install 3.7
python-build: definition not found: 3.7

The following versions contain `3.7' in the name:
  2.3.7
  3.3.7
  3.7.0
  3.7-dev
  3.7.1
  3.7.2
  3.7.3
  miniconda-3.7.0
  miniconda3-3.7.0
  stackless-3.3.7

See all available versions with `pyenv install --list'.
...

If the version we ask for is only a partial match, it will show us which versions we can install. Let’s install the latest, which will take a few minutes.

➜ pyenv install 3.7.3
Downloading Python-3.7.3.tar.xz...
-> https://www.python.org/ftp/python/3.7.3/Python-3.7.3.tar.xz
Installing Python-3.7.3...
Installed Python-3.7.3 to /home/shawn/.pyenv/versions/3.7.3

Now let’s make 3.7.3 globally available so there is no messing with our system python:

➜ pyenv global      
system
➜ pyenv global 3.7.3
➜ pyenv global      
3.7.3
➜ pyenv versions
  system
* 3.7.3 (set by /home/shawn/.pyenv/version)

We are now defaulting to 3.7.3. For now, we are done with pyenv. It is a very simple tool to use.

Pipenv: the Python Virtual Environment

To install pipenv, you can install it through your systems package manager as described here on their repo page.

The setup

To get started we need to create a simple project. Let’s clone a repo I have that has a few basic files in it for tracking projects in git.

➜ git clone [email protected]:autoferrit/template.git demo
Cloning into 'demo'...
remote: Enumerating objects: 6, done.
remote: Counting objects: 100% (6/6), done.
remote: Compressing objects: 100% (6/6), done.
remote: Total 6 (delta 0), reused 0 (delta 0)
Receiving objects: 100% (6/6), done.
➜ cd demo
➜ ls -la
Permissions Size User  Date Modified Git Name
drwxr-xr-x     - shawn 27 Jun 12:10   -- .git
.rwxr-xr-x   474 shawn 27 Jun 12:09   -- .editorconfig
.rwxr-xr-x  2.2k shawn 27 Jun 12:09   -- .gitattributes
.rw-r--r--   138 shawn 27 Jun 12:09   -- Pipfile
.rw-r--r--    96 shawn 27 Jun 12:09   -- Readme.md

You can see the file contents at the gitlab project.

  • .editorconfig: helps tell the editors (with appropriate plugins) what formatting options should be used on files in this project.
  • .gitattributes: This keeps line endings in check when working on various platforms (I am looking at you, windows).
  • Pipfile This is where we will keep of application requirements rather than using requirements.txt

The Pipfile would also be generated from the following command.

➜ pipenv --python 3.7.3 install

This will create your python virtual environment, which you can then use by running pipenv shell. The best part is that Pipenv supports Pyenv and will use python versions installed using that tool making our system much cleaner.

A simple API

To show how to use it, we are going to create a very basic API using FastAPI. Install the requirements with the following command:

➜ pipenv install fastapi uvicorn

This will create a Pipfile.lock which is analogous to package-lock.json in node. Creating a repeatable environment when the code is run through a CI server or production is critical and this is where Pipenv shines. It makes sure we can actually deploy the same versions of packages AND their requirements. Let’s run this code.

➜ uvicorn main:app --reload

Now you can see your API by visiting http://127.0.0.1:8000/items/5?q=somequery. If you would like to follow along in this tutorial, you can read the fastapi tutorial in their documentation here.

The Pipfile

Let’s open up our Pipfile to see what has installed in our python virtual environment.

[[source]]
name = "pypi"
url = "https://pypi.org/simple"
verify_ssl = true

[dev-packages]

[packages]
fastapi = "*"
uvicorn = "*"

[requires]
python_version = "3.7"

By default, Pipenv will always install the newest versions of packages. Which is what we should all be doing. When new versions are released and we run pipenv install it will install the newer version that was released. To update a package, we can run pipenv update fastapi or if we want to update all we can run pipenv update. If we need to pin a specific version, you can do so in the same way if you were to use standard pip and requirements.txt.

How do I install Dev Packages in my environment?

To install development packages just pass the --dev argument.

➜ pipenv install --dev ptpython

And to install development packages:

➜ pipenv install --dev

By default, pipenv install will ONLY install the base packages that should live on production. This is what we want. Passing the --dev parameter will allow us to install base AND dev packages locally or during the CI process.

The pip in the $SHELL

To activate the environment, simply run

demo on writingcode master [!?] 
➜ pipenv shell        
Launching subshell in virtual environment…
 . /home/shawn/code/sandbox/demo/.venv/bin/activate

demo on writingcode master [!?] via demo 
➜ pip list    
Package    Version
---------- -------
Click      7.0    
fastapi    0.30.0 
h11        0.8.1  
httptools  0.0.13 
pip        19.1.1 
pydantic   0.28   
setuptools 41.0.1 
starlette  0.12.0 
uvicorn    0.8.2  
uvloop     0.12.2 
websockets 7.0    
wheel      0.33.4 

Note: I am showing my full terminal text here to show that it is now adding via demo to the prompt for the python environment. You will likely see this represented differently. Often times it looks like this:

(demo) $ pip list
...

Checking for vulnerabilities in your virtual environment

Possibly one of the best features is the ability to check for vulnerabilities.

➜ pipenv check
Checking PEP 508 requirements…
Passed!
Checking installed package safety…
All good!

We have no vulnerabilities in our python virtual environment. That’s a good thing. Let’s try to add one.

➜ pipenv install 'django==2.1'
➜ pipenv check
Checking PEP 508 requirements…
Passed!
Checking installed package safety…
36883: django <2.1.6,>=2.1.0 resolved (2.1 installed)!
Django 2.1.x before 2.1.6  allows Uncontrolled Memory Consumption via a malicious attacker-supplied value to the django.utils.numberformat.format() function.

36522: django <2.1.2,>=2.1 resolved (2.1 installed)!
An issue was discovered in Django 2.1 before 2.1.2, in which unprivileged users can read the password hashes of arbitrary accounts. The read-only password widget used by the Django Admin to display an obfuscated password hash was bypassed if a user has only the "view" permission (new in Django 2.1), resulting in display of the entire password hash to those users. This may result in a vulnerability for sites with legacy user accounts using insecure hashes.

36517: django <2.1.2,>=2.1.0 resolved (2.1 installed)!
django before 2.1.2 fixes a security bug in 2.1.x. 
If an admin user has the change permission to the user model, only part of the
password hash is displayed in the change form. Admin users with the view (but
not change) permission to the user model were displayed the entire hash.

WHOAH! Don’t install Django 2.1. Luckily it was fixed, let’s update it.

➜ pipenv update django
Locking [dev-packages] dependencies…
Locking [packages] dependencies…
✔ Success! 
Updated Pipfile.lock (9d4a23)!
Installing dependencies from Pipfile.lock (9d4a23)…
  🐍   ▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉ 11/11 — 00:00:01
All dependencies are now up-to-date!

➜ pip list            
Package    Version
---------- -------
...  
Django     2.1    
...

Wait why is the old version installed? This is GOOD! Why? It means that our setup is idempotent. Why? It didn’t work because we pinned version “2.1” of Django. So we should unpin that in the Pipfile by changing the version to "*". We should now be able to have the expected outcome.

➜ pipenv install            
Pipfile.lock (10cb8d) out of date, updating to (9d4a23)…
Locking [dev-packages] dependencies…
Locking [packages] dependencies…
✔ Success! 
Updated Pipfile.lock (10cb8d)!
Installing dependencies from Pipfile.lock (10cb8d)…
  🐍   ▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉▉ 12/12 — 00:00:04

➜ pip list      
Package    Version
---------- -------
...  
Django     2.2.2  
...

Conclusion

Pyenv and Pipenv make it much easier to keep your python virtual environment and versions in check. You can now make your environments consistently deployable across systems. Python has a history of being a headache when it comes to consistently installing requirements across systems. Using these tools will help you focus more on your actual code.

Unfortunately, Pipenv isn’t geared towards those making libraries to be uploaded to pypi which still requires setup.py. Hopefully, in the future, they will find a way to update the packaging system with these too

How do you manage your environments and dependencies in Python?