Advanced Usage of Pipenv

This document covers some of Pipenv’s more glorious and advanced features.

☤ Caveats

  • Dependencies of wheels provided in a Pipfile will not be captured by $ pipenv lock.
  • There are some known issues with using private indexes, related to hashing. We’re actively working to solve this problem. You may have great luck with this, however.
  • Installation is intended to be as determinstic as possible — use the --sequential flag to increase this, if experiencing issues.

☤ Specifying Package Indexes

If you’d like a specific package to be installed with a specific package index, you can do the following:

url = ""
verify_ssl = true
name = "pypi"

url = ""
verify_ssl = false
name = "home"


requests = {version="*", index="home"}
maya = {version="*", index="pypi"}
records = "*"

Very fancy.

☤ Specifying Basically Anything

If you’d like to specify that a specific package only be installed on certain systems, you can use PEP 508 specifiers to accomplish this.

Here’s an example Pipfile, which will only install pywinusb on Windows systems:

url = ""
verify_ssl = true
name = "pypi"

requests = "*"
pywinusb = {version = "*", sys_platform = "== 'win32'"}


Here’s a more complex example:

url = ""
verify_ssl = true

unittest2 = {version = ">=1.0,<3.0", markers="python_version < '2.7.9' or (python_version >= '3.0' and python_version < '3.4')"}

Magic. Pure, unadulterated magic.

☤ Deploying System Dependencies

You can tell Pipenv to install a Pipfile’s contents into its parent system with the --system flag:

$ pipenv install --system

This is useful for Docker containers, and deployment infrastructure (e.g. Heroku does this).

Also useful for deployment is the --deploy flag:

$ pipenv install --system --deploy

This will fail a build if the Pipfile.lock is out–of–date, instead of generating a new one.

pipenv and conda

To use Pipenv with a Conda–provided Python, you simply provide the path to the Python binary:

$ pipenv install --python=/path/to/anaconda/python

To reuse Conda–installed Python packages, use the --site-packages flag:

$ pipenv --python=/path/to/anaconda/python --site-packages

☤ Generating a requirements.txt

You can convert a Pipfile and Pipfile.lock into a requirements.txt file very easily, and get all the benefits of extras and other goodies we have included.

Let’s take this Pipfile:

url = ""
verify_ssl = true

requests = {version="*"}

And generate a requirements.txt out of it:

$ pipenv lock -r

If you wish to generate a requirements.txt with only the development requirements you can do that too! Let’s take the following Pipfile:

url = ""
verify_ssl = true

pytest = {version="*"}

And generate a requirements.txt out of it:

$ pipenv lock -r --dev

Very fancy.

☤ Detection of Security Vulnerabilities

Pipenv includes the safety package, and will use it to scan your dependency graph for known security vulnerabilities!


$ cat Pipfile
django = "==1.10.1"

$ pipenv check
Checking PEP 508 requirements…
Checking installed package safety…

33075: django >=1.10,<1.10.3 resolved (1.10.1 installed)!
Django before 1.8.x before 1.8.16, 1.9.x before 1.9.11, and 1.10.x before 1.10.3, when settings.DEBUG is True, allow remote attackers to conduct DNS rebinding attacks by leveraging failure to validate the HTTP Host header against settings.ALLOWED_HOSTS.

33076: django >=1.10,<1.10.3 resolved (1.10.1 installed)!
Django 1.8.x before 1.8.16, 1.9.x before 1.9.11, and 1.10.x before 1.10.3 use a hardcoded password for a temporary database user created when running tests with an Oracle database, which makes it easier for remote attackers to obtain access to the database server by leveraging failure to manually specify a password in the database settings TEST dictionary.

33300: django >=1.10,<1.10.7 resolved (1.10.1 installed)!
CVE-2017-7233: Open redirect and possible XSS attack via user-supplied numeric redirect URLs

Django relies on user input in some cases  (e.g.
:func:`django.contrib.auth.views.login` and :doc:`i18n </topics/i18n/index>`)
to redirect the user to an "on success" URL. The security check for these
redirects (namely ``django.utils.http.is_safe_url()``) considered some numeric
URLs (e.g. ``http:999999999``) "safe" when they shouldn't be.

Also, if a developer relies on ``is_safe_url()`` to provide safe redirect
targets and puts such a URL into a link, they could suffer from an XSS attack.

CVE-2017-7234: Open redirect vulnerability in ``django.views.static.serve()``

A maliciously crafted URL to a Django site using the
:func:`~django.views.static.serve` view could redirect to any other domain. The
view no longer does any redirects as they don't provide any known, useful

Note, however, that this view has always carried a warning that it is not
hardened for production use and should be used only as a development aid.



In order to enable this functionality while maintaining its permissive copyright license, pipenv embeds an API client key for the backend Safety API operated by rather than including a full copy of the CC-BY-NC-SA licensed Safety-DB database. This embedded client key is shared across all pipenv check users, and hence will be subject to API access throttling based on overall usage rather than individual client usage.

☤ Community Integrations

There are a range of community-maintained plugins and extensions available for a range of editors and IDEs, as well as different products which integrate with Pipenv projects:

Works in progress:

  • Sublime Text (Editor Integration)
  • PyCharm (Editor Integration)
  • Mysterious upcoming Google Cloud product (Cloud Hosting)

☤ Open a Module in Your Editor

Pipenv allows you to open any Python module that is installed (including ones in your codebase), with the $ pipenv open command:

$ pipenv install -e git+
Installing -e git+…
Updated Pipfile.lock!

$ pipenv open background
Opening '/Users/kennethreitz/.local/share/virtualenvs/hmm-mGOawwm_/src/background/' in your EDITOR.

This allows you to easily read the code you’re consuming, instead of looking it up on GitHub.


The standard EDITOR environment variable is used for this. If you’re using VS Code, for example, you’ll want to export EDITOR=code (if you’re on macOS you will want to install the command on to your PATH first).

☤ Automatic Python Installation

If you have pyenv installed and configured, Pipenv will automatically ask you if you want to install a required version of Python if you don’t already have it available.

This is a very fancy feature, and we’re very proud of it:

$ cat Pipfile
url = ""
verify_ssl = true


requests = "*"

python_version = "3.6"

$ pipenv install
Warning: Python 3.6 was not found on your system…
Would you like us to install latest CPython 3.6 with pyenv? [Y/n]: y
Installing CPython 3.6.2 with pyenv (this may take a few minutes)…
Making Python installation global…
Creating a virtualenv for this project…
Using /Users/kennethreitz/.pyenv/shims/python3 to create virtualenv…
No package provided, installing all dependencies.
Installing dependencies from Pipfile.lock…
🐍   ❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒❒ 5/5 — 00:00:03
To activate this project's virtualenv, run the following:
 $ pipenv shell

Pipenv automatically honors both the python_full_version and python_version PEP 508 specifiers.


☤ Automatic Loading of .env

If a .env file is present in your project, $ pipenv shell and $ pipenv run will automatically load it, for you:

$ cat .env

$ pipenv run python
Loading .env environment variables…
Python 2.7.13 (default, Jul 18 2017, 09:17:00)
[GCC 4.2.1 Compatible Apple LLVM 8.1.0 (clang-802.0.42)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import os
>>> os.environ['HELLO']

This is very useful for keeping production credentials out of your codebase. We do not recommend committing .env files into source control!

If your .env file is located in a different path or has a different name you may set the PIPENV_DOTENV_LOCATION environment variable:

$ PIPENV_DOTENV_LOCATION=/path/to/.env pipenv shell

To prevent pipenv from loading the .env file, set the PIPENV_DONT_LOAD_ENV environment variable:

$ PIPENV_DONT_LOAD_ENV=1 pipenv shell

☤ Support for Environment Variables

pipenv supports the usage of environment variables in values. For example:

[[source]] url = “https://${PYPI_USERNAME}:${PYPI_PASSWORD}” verify_ssl = true name = “pypi”


[packages] requests = {version=”*”, index=”home”} maya = {version=”*”, index=”pypi”} records = “*”

Environment variables may be specified as ${MY_ENVAR} or $MY_ENVAR. On Windows, %MY_ENVAR% is supported in addition to ${MY_ENVAR} or $MY_ENVAR.

☤ Configuration With Environment Variables

pipenv comes with a handful of options that can be enabled via shell environment variables. To activate them, simply create the variable in your shell and pipenv will detect it.

  • PIPENV_DEFAULT_PYTHON_VERSION — Use this version of Python when creating new virtual environments, by default (e.g. 3.6).
  • PIPENV_SHELL_FANCY — Always use fancy mode when invoking pipenv shell.
  • PIPENV_VENV_IN_PROJECT — If set, use .venv in your project directory instead of the global virtualenv manager pew.
  • PIPENV_COLORBLIND — Disable terminal colors, for some reason.
  • PIPENV_NOSPIN — Disable terminal spinner, for cleaner logs. Automatically set in CI environments.
  • PIPENV_MAX_DEPTH — Set to an integer for the maximum number of directories to recursively search for a Pipfile.
  • PIPENV_TIMEOUT — Set to an integer for the max number of seconds Pipenv will wait for virtualenv creation to complete. Defaults to 120 seconds.
  • PIPENV_IGNORE_VIRTUALENVS — Set to disable automatically using an activated virtualenv over the current project’s own virtual environment.
  • PIPENV_PIPFILE — When running pipenv from a $PWD other than the same directory where the Pipfile is located, instruct pipenv to find the Pipfile in the location specified by this environment variable.

If you’d like to set these environment variables on a per-project basis, I recommend utilizing the fantastic direnv project, in order to do so.

Also note that pip itself supports environment variables, if you need additional customization.

For example:

$ PIP_INSTALL_OPTION="-- -DCMAKE_BUILD_TYPE=Release" pipenv install -e .

☤ Custom Virtual Environment Location

Pipenv’s underlying pew dependency will automatically honor the WORKON_HOME environment variable, if you have it set — so you can tell pipenv to store your virtual environments wherever you want, e.g.:

export WORKON_HOME=~/.venvs

In addition, you can also have Pipenv stick the virtualenv in project/.venv by setting the PIPENV_VENV_IN_PROJECT environment variable.

☤ Testing Projects

Pipenv is being used in projects like Requests for declaring development dependencies and running the test suite.

We’ve currently tested deployments with both Travis-CI and tox with success.

Travis CI

An example Travis CI setup can be found in Requests. The project uses a Makefile to define common functions such as its init and tests commands. Here is a stripped down example .travis.yml:

language: python
    - "2.6"
    - "2.7"
    - "3.3"
    - "3.4"
    - "3.5"
    - "3.6"
    - "3.7-dev"

# command to install dependencies
install: "make"

# command to run tests
    - make test

and the corresponding Makefile:

    pip install pipenv
    pipenv install --dev

    pipenv run py.test tests

Tox Automation Project

Alternatively, you can configure a tox.ini like the one below for both local and external testing:

envlist = flake8-py3, py26, py27, py33, py34, py35, py36, pypy

deps = pipenv
    pipenv install --dev
    pipenv run py.test tests

basepython = python3.4
    pipenv install --dev
    pipenv run flake8 --version
    pipenv run flake8 docs project test

pipenv will automatically use the virtualenv provided by tox.

You might also want to add --ignore-pipfile to pipenv install, as to not accidentally modify the lock-file on each test run. This causes pipenv to ignore changes to the Pipfile and (more importantly) prevents it from adding the current environment to Pipfile.lock. This might be important as the current environment (i.e. the virtualenv provisioned by tox) will usually contain the current project (which may or may not be desired) and additional dependencies from tox’s deps directive. The initial provisioning may alternatively be disabled by adding skip_install = True to tox.ini.

This method requires you to be explicit about updating the lock-file, which is probably a good idea in any case.

A 3rd party plugin, tox-pipenv is also available to use Pipenv natively with tox.

☤ Shell Completion

To enable completion in fish, add this to your config:

eval (pipenv --completion)

Alternatively, with bash or zsh, add this to your config:

eval "$(pipenv --completion)"

Magic shell completions are now enabled!


☤ Working with Platform-Provided Python Components

It’s reasonably common for platform specific Python bindings for operating system interfaces to only be available through the system package manager, and hence unavailable for installation into virtual environments with pip. In these cases, the virtual environment can be created with access to the system site-packages directory:

$ pipenv --three --site-packages

To ensure that all pip-installable components actually are installed into the virtual environment and system packages are only used for interfaces that don’t participate in Python-level dependency resolution at all, use the PIP_IGNORE_INSTALLED setting:

$ PIP_IGNORE_INSTALLED=1 pipenv install --dev

☤ Pipfile vs

There is a subtle but very important distinction to be made between applications and libraries. This is a very common source of confusion in the Python community.

Libraries provide reusable functionality to other libraries and applications (let’s use the umbrella term projects here). They are required to work alongside other libraries, all with their own set of subdependencies. They define abstract dependencies. To avoid version conflicts in subdependencies of different libraries within a project, libraries should never ever pin dependency versions. Although they may specifiy lower or (less frequently) upper bounds, if they rely on some specific feature/fix/bug. Library dependencies are specified via install_requires in

Libaries are ultimately meant to be used in some application. Applications are different in that they usually are not depended on by other projects. They are meant to be deployed into some specific environment and only then should the exact versions of all their dependencies and subdependencies be made concrete. To make this process easier is currently the main goal of pipenv.

To summarize:

  • For libraries, define abstract dependencies via install_requires in The decision of which version exactly to be installed and where to obtain that dependency is not yours to make!
  • For applications, define dependencies and where to get them in the Pipfile and use this file to update the set of concrete dependencies in Pipfile.lock. This file defines a specific idempotent environment that is known to work for your project. The Pipfile.lock is your source of truth. The Pipfile is a convenience for you to create that lock-file, in that it allows you to still remain somewhat vague about the exact version of a dependency to be used. pipenv is there to help you define a working conflict-free set of specific dependency-versions, which would otherwise be a very tedious task.
  • Of course, Pipfile and pipenv are still useful for library developers, as they can be used to define a development or test environment.
  • And, of course, there are projects for which the distinction between library and application isn’t that clear. In that case, use install_requires alongside pipenv and Pipfile.

You can also do this:

$ pipenv install -e .

This will tell Pipenv to lock all your–declared dependencies.