NameError: name '__file__' is not defined when rendering .qmd document

[laguill]

Description

Issue description

I can successfully run the code using the Jupyter extension in VS Code. However, when I attempt to render the document with Quarto, I encounter the following error:

NameError: name '__file__' is not defined

I have the following .qmd document:

---
title: "Data science "
subtitle: "Pandas"
format: 
  html:
    toc: true
    toc-depth: 4
    toc-expand: true
engine: jupyter
jupyter: python3
---
#### Load dataset

```{python}
import pandas as pd
from pathlib import Path

cwd = Path(__file__)
print(cwd)
data_path = cwd.parent / "data" / "Dataset.csv"

edStatsData = pd.read_csv(data_path)
```

Environment Details

• quarto version : 1.6.39
• OS: ubuntu

Additional Context

• __file__ works in regular Python scripts, but I suspect its behavior might differ in a Jupyter or Quarto context.
• Is there a workaround for this, such as an alternative to __file__ in Quarto?

[mcanouil]

Could you properly format your post using code blocks for code and terminal outputs? Thanks.
If your code contains code blocks, you need to enclose it using more backticks, i.e., usually four ````.
See https://quarto.org/bug-reports.html#formatting-make-githubs-markdown-work-for-us.

[mcanouil]

Why not use the fact that / means root of the project (meaning there is _quarto.yml) or location of the input file?

[laguill]

Thank you for your answer!

I think my code should be now better formatted

I used this method because it’s what I typically use when working with standalone Python scripts outside a structured project. However, you’re right that relying on the project root or input file location might be better in Quarto.

What I don’t understand is why this works in Jupyter but not when rendering with Quarto. Does Quarto’s rendering engine modify the environment so that __file__ isn’t defined?

Do you have any resources or insights on this difference?

[mcanouil]

Do you have any resources or insights on this difference?

I don't. I don't have deep technical knowledge of how Jupyter / nbconvert works.

[cscheid]

Just as a statement of facts:

• we don't modify the environment
• __file__ is not a requirement: https://docs.python.org/3/reference/datamodel.html#module.__file__

This can't be a Jupyter thing, because Jupyter doesn't know anything about the internal behavior of kernels. I don't know how IPython handles it, but clearly something different is happening from the way we're calling nbconvert

[laguill]

Should I open an issue for that ?

[s2t2]

Running into this issue as well.

Desired code, using dynamic and robust filepath compilation that will work no matter which directory the file is run from, no matter which operating system is being used:

import os

DATA_DIR = os.path.join(os.path.dirname(__file__), "..", "..", "data") 
#> "`NameError: name '__file__' is not defined`"

csv_filepath = os.path.join(DATA_DIR, "my_data.csv")
df.to_csv(csv_filepath, index=False)

Runs into error.

Alternative attempt, based on earlier comment in this thread proposing a reference to the root directory of the repo:

DATA_DIR = "/docs/data"
#> "`OSError: Cannot save file into a non-existent directory: '/docs/data'`" 
# even though that exists in the repo (its probably referencing the absolute system root filepath instead of the repo root directory).

csv_filepath = os.path.join(DATA_DIR, "my_data.csv")
df.to_csv(csv_filepath, index=False)

Runs into error.

Hard-coding the relative reference does work, but isn't an os agnostic solution, as windows slashes go the other way:

DATA_DIR = "../../data"

csv_filepath = os.path.join(DATA_DIR, "my_data.csv")
df.to_csv(csv_filepath, index=False)

[gordonwoodhull]

@s2t2, surely

DATA_DIR = os.path.join(“..”, “..”, “data")

should work?