Is adding project root directory to sys.path a good practice?

Question

I have a question about adding project path to python, for facilitating import effort.

Situation

When I write code in Python, I usually add necessary path to sys.path by using

import sys
sys.path.append("/path/to/dir/") # almost every `.py` need this

Sometimes, when my project gets bigger with many levels of directories, this approach seems bulky and error-prone (especially when I re-organize my files)

Recently, I start using a bash script (located at project root directory) that adding the sys.path.append with project root argument to .py file in the project. With this approach, I hardly have to manually care about importing a module.

Question

My question is: Is that a good practice? I find it convenient for myself, compared to my old method, but since the bash script is a separated file, I need 2 command to run any script in my project (one for the bash and one for the .py). I can include the command calling .py to the bash, but it far less flexible than directly call it from terminal.

Really want to hear some advices! Thanks in advance. Any suggestion will be gratefully appreciated!

Answer 1

It is generally not good practice to use manipulate sys.path within a python library or program. You should add the relevant paths to the PYTHONPATH in the calling environment for your python program:

PYTHONPATH="/path/to/other/projects/directory:$PYTHONPATH" python ...

or

export PYTHONPATH="/path/to/other/projects/directory:$PYTHONPATH"
python ...

This allows you to easily manipulate the paths that your program or library will search for dependencies easily without modifying your code.

It is also very easy to manage this in your personal development environment by modifying your bashrc or in your production environments in your init script (or other wrapper script) and provides you with one place to update each time you add or modify your project paths.

Given that you mention that you have almost one directory per .py file, you should also consider how your code might be organized into packages to further simplify your setup.