Cannot import Python modules from sibling directory in Jupyter Notebook running in VS Code (os.chdir and sys.path.append do not work)

[AliMoeen]

I am running Jupyter Notebooks inside Visual Studio Code using the official Jupyter extension.

My project structure is as follows:

my_project/
├── notebooks/
│   └── analysis.ipynb
├── libs/
│   ├── __init__.py
│   └── my_module.py

In analysis.ipynb, I would like to import my_module using either:

from libs import my_module

or:

import my_module

However, I consistently get the following error:

ModuleNotFoundError: No module named 'libs'

I have read the [official VS Code Jupyter documentation](https://code.visualstudio.com/docs/datascience/jupyter-notebooks), but it does not cover anything about import paths or how sys.path is handled.

Here are the things I have tried:

1. Changing the working directory:

import os
os.chdir("..")

This does not work. The import still fails.

2. Manually adding the path to sys.path:

import sys
import os
sys.path.append(os.path.abspath("../libs"))
import my_module

This also does not work. The import fails with the same ModuleNotFoundError.

In a normal Python script (outside of a notebook), these methods usually work for me. But in this case, inside a Jupyter Notebook running in VS Code, they are not working.

I would like to understand:

1. Why do these approaches not work in this environment?
2. How can I properly import modules from sibling or parent directories in this setup?
3. What is the recommended way to structure this in VS Code when using Jupyter Notebooks?

Any advice or clarification would be greatly appreciated.

[DonJayamanne]

Please could you enable logging as follows:

• Open settings & go into Jupyter->Logging
• Change the value to trace
• Reload VS Code,
• Attempt to repro the issue & then copy the output from the Jupyter output panel (use the command Jupyter: Show Output to get to the logs).

[AaronPedersen-QUT]

I do exactly what you're trying to do all the time, with a bit extra to allow me to run a python file located in a completely different directory, (from within any workspace) as if I were in that other file's directory.

I usually just copy and paste this block between projects, modifying the path, and it has been serving me well so far:

import os
import sys
os.system("clear")
RootDirectory = "/home/username/absolute/path/to/the/workspace/project/root/directory"
os.chdir(RootDirectory)
for SearchPath in [SP for SP in sys.path if (RootDirectory in SP)]: sys.path.remove(SearchPath)
sys.path.insert(0, f"{RootDirectory}/relative/path/to/your/modules/library/data")
sys.path.insert(0, f"{RootDirectory}/relative/path/to/your/second/directory/for/modules/library/data")
#(and just keep adding more relative paths as needed, in ascending order of priority)#
sys.path.insert(0, os.getcwd())

if ((1)):
	print("="*100)
	print(f"Current Working Directory (CWD):\n'{os.getcwd()}'\n")
	print("Search Paths:")
	for i, PathLine in enumerate(sys.path): print(F"[{i+1}]: '{PathLine}'")
	print("="*100)

The only real difference I'm seeing is that I'm inserting my new (abs) paths at the front (where they will get searched first), while you're appending them to the back (where they will be searched last).
I don't see why it would be, but perhaps it could be stopping its search early, after searching a specific directory/path, without being able to get to your new paths? (Literally just an uneducated guess.)

---

Also, I just generally wouldn't suggest changing the CWD using a relative path string in Jupyter NB, because without resetting the kernel, every time you run that cell, the CWD will take a further step back; it's not like a regular python file, where the CWD resets every time you run the code.

---

Another thing you could try I guess, is to write out the absolute path to the module yourself, or print out the sys.path to verify that os.path.abspath("../libs") is forming the correct path to your module.

I recon just printing out the code below should hopefully help you debug your pathing issues (that is if it is an issue with pathing):

if ((1)):
	print("="*100)
	print(f"Current Working Directory (CWD):\n'{os.getcwd()}'\n")
	print("Search Paths:")
	for i, PathLine in enumerate(sys.path): print(F"[{i+1}]: '{PathLine}'")
	print("="*100)

---

Another potential issue might be just that Pylance isn't seeing the module (prior to PYTHONPATH being changed during runtime), but that shouldn't throw an error; it should only just put a yellow line under the import and not give you the tooltips for the objects retrieved from that module.
(You can fix this using .envrc files to modify the PYTHONPATH environment variable as you enter you project directory; or by going into the VS-Code setting and permanently adding your module's path to the Pylance search paths.)

[AaronPedersen-QUT]

The if ((1)):, is just a manual toggle that I use to prevent blocks of Python code from being included in the Bytecode during interpretation, without having to comment out or entirely remove said blocks of code. (The closest thing python has to conditional compilation.)
You can just ignore it if you're gonna try out one of my code suggestions above.