A simple python package to solve all your problems with pythonpath, working directory, file paths, module imports and environment variables.

Problem: I would like to be able to:

• Run my python scripts from anywhere
• Always import python modules relatively to the project root directory
• Always access files relatively to the project root so I don't have to specify a series of ../ to get to the data
• Always have access to environment variables from .env file without having to load them manually
• Have all the above benefits in notebooks even if they're nested in subdirectories

Solution: The rootutils package provides a flexible way to setup the python project with a simple one-liner. It finds the project root based on the location of specified file name, e.g. .project-root or .git.

The package is tiny and continuosly maintained.

pip install rootutils

import rootutils
# find absolute root path (searches for directory containing .project-root file)
# search starts from current file and recursively goes over parent directories
# returns pathlib object
path = rootutils.find_root(search_from=__file__, indicator=".project-root")
# find absolute root path (searches for directory containing any of the files on the list)
path = rootutils.find_root(search_from=__file__, indicator=[".git", "setup.cfg"])
# take advantage of the pathlib syntax
data_dir = path / "data"
assert data_dir.exists(), f"path doesn't exist: {data_dir}"
# set root directory
rootutils.set_root(
    path=path # path to the root directory
    project_root_env_var=True, # set the PROJECT_ROOT environment variable to root directory
    dotenv=True, # load environment variables from .env if exists in root directory
    pythonpath=True, # add root directory to the PYTHONPATH (helps with imports)
    cwd=True, # change current working directory to the root directory (helps with filepaths)
)

Simplest usage with one-liner (combines find_root() and set_root() into one method):

import rootutils
root = rootutils.setup_root(__file__, dotenv=True, pythonpath=True, cwd=False)

Default root indicators (used when you don't specify indicator arg):

[".project-root", "setup.cfg", "setup.py", ".git", "pyproject.toml"]

autoroot is an experimental package that reduces rootutils to single import, without the need to execute any setup calls. This means just the act of importing this dependency (import autorootcwd) causes execution of recurrent search for .project-root file.

Installation:

pip install autoroot autorootcwd

This adds root folder to pythonpath, sets PROJECT_ROOT env var, and loads variables from .env:

import autoroot # root setup, do not delete

This also changes working directory to root:

import autorootcwd # root setup, do not delete

Autoroot exist for convenience and speed. For example, it's faster to just add import autorootcwd at the beginning when creating new notebook.

Package page: https://github.com/ashleve/autoroot

This package is heavily inspired by:

https://github.com/chendaniely/pyprojroot

https://github.com/pashminacameron/py-repo-root

https://github.com/EduardKononov/from-root