nbresult 0.0.2

Extract results from Jupyter notebooks

nbresult

A simple package to test Jupyter notebook result for the Le Wagon's Data Science Bootcamp.

Installation

Installation with pip from Pypi:

pip install nbresult

Usage

Considering the default data challenge architecture:

.
├── challenge.ipynb
├── Makefile
├── README.md
├── data
│   └── data.csv
└── tests
    ├── __init__.py
    └── test_results.py

If you want to test a variable log_model_score from the challenge.ipynb notebook with pytest:

At the end of the notebook add a cell with the following code:

from nbresult.challenge_result import ChallengeResult


RESULT = ChallengeResult(
    score=log_model_score
)
RESULT.write()

This outputs a results.json file in the tests directory:

# tests/results.json

{
  "score": 0.829004329004329
}

The notebook results can be imported from anywhere with:

from nbresult.challenge_result import ChallengeResult


results = ChallengeResult().load('path/to/results.json')

So you can write test the log_model_score with pytest:

# test_results.py

import unittest
import os
from nbresult.challenge_result import ChallengeResult


class TestResults(unittest.TestCase):
    results = ChallengeResult().load(os.path.join(
        os.path.dirname(__file__),
        'results.json')
    )

    def test_model_score(self):
        self.assertEqual(self.results['score'] > 0.82, True)

Finally you can run your tests with pytest:

pytest tests/test_results.py

OR

Run the tests with make:

• Setup a Makefile

# Makefile

default: pytest

pytest:
  PYTHONDONTWRITEBYTECODE=1 pytest -v --color=yes

• Run make