ModuleNotFoundError: Error when trying to create a Python Data Loader

[alessandrofajr]

I am trying to create a Data Loader using Python, but I keep encountering the following error: ModuleNotFoundError: No module named 'pandas'.

Here is the full traceback:

load /data/data.csv → [missing] Traceback (most recent call last):
  File "C:\Users\observableframework\src\data\data.csv.py", line 1, in <module>
    import pandas as pd
ModuleNotFoundError: No module named 'pandas'
error in 636ms: loader exited with code 1

Details:

In the file src\data\data.csv.py, here is my code:

import pandas as pd
import numpy as np

import sys

# Simple random DataFrame
df = pd.DataFrame({
    'A': np.random.randint(1, 10, 5),
    'B': np.random.rand(5)             
})

df.to_csv(sys.stdout, index=False)

My Setup:

OS: Windows 11
Python: 3.12.2
Node: 20.11.0
Observable Framework: 1.11.0

Steps I followed:

• I created a new project using the Observable Framework with npx "@observablehq/framework@latest" create in an empty folder.
• Then, I created a Python virtual environment in the same folder using python -m virtualenv .venv.
• I activated the virtual environment with .venv\Scripts\activate.
• I added pandas and numpy to a requirements.txt file located at the same folder as package.json files and ran pip install -r requirements.txt

I tried calling my Data Loader from a .md file like this:

const myData = FileAttachment("data/data.csv").csv({typed: true});

The error occurs for any Python package that isn't part of the standard library. I’ve tried creating new projects with clean installs, but nothing seems to work.

If I run something like python src\data\data.csv.py, the code runs just fine.

[mythmon]

Are you activating the the virtual environment? I don't know what that looks like on Windows, but on other platforms you would run source .venv/bin/activate. You need to do this in each terminal where you are going to run npm run build so that it knows to use the virtualenv.

Apologies, I see that you already said you did this.

[mbostock]

Did you start the preview server within the virtual environment as described in Data loaders: Execution?

[alessandrofajr]

Yes, with the virtualenv activated, I ran npm run dev

[mythmon]

Another thing to check out, by default Framework uses the binary name python3 to run the Python interpreter, not python. If you run python3 src\data\data.csv.py, does the code work?

If you want to use just python, you could set the interpreters option to that:

{
  interpreters: {
    ".py": ["python"],
  }
}

[alessandrofajr]

This did the trick! Thank you