jnbv 2024.11.18

Validation of Jupyter notebooks and kernels

JUPYTER NOTEBOOK VALIDATION

1. Project Overview
2. Install
3. Usage
4. Contributing

PROJECT OVERVIEW

This repository contains scripts used to validate ipynb notebooks and Jupyter kernels.

There is another related repository which contains a large number of test ipynb notebooks and routines to help with the execution of the validation tests via docker images and slurm batch jobs:

• jupyter-notebook-validation

PACKAGE

A python package has been created which can be installed via pip or conda:

• conda package:
• PyPi package:

GOALS

The goals of this project are to be able to:

1. Execute validation of Jupyter kernels using ipynb notebooks in the terminal
2. Execute validations non-interactively in a CI pipeline
3. Check execution output for errors
4. Compare output of an execution to known output
5. Select different Jupyter kernels to use
6. Log results of notebook executions and tests

EXAMPLE OUTPUT

Partial output from validation tests run in the terminal:

Output from an ipynb notebook created from failed execution of a notebook:

INSTALL

The jnbv module can be installed with pip or conda using one of the following:

• conda package:
• PyPi package:
• gitlab repository:

INSTALL WITH PIP

Create a virtual environment if you don't already have one:

python3 -m venv venv

Activate your environment, in this example it's in venv/:

source venv/bin/activate

Install the code from PyPi using pip:

pip install jnbv

Or install from gitlab using pip and git:

pip install git+https://gitlab.com/MAXIV-SCISW/JUPYTERHUB/jnbv.git

With either method, dependencies will also be installed.

INSTALL WITH CONDA

Activate and install directly into your environment:

source venv/bin/activate
conda install jnbv -c conda-forge

Or, you can add jnbv via your conda environment yaml file, by adding conda-forge to your list of channels, and jnbv to your list of dependencies, as in this example file:

---
name: my-conda-env
channels:
  - conda-forge
dependencies:
  - jnbv

Then update your conda environment, for example the base conda environment which is in the directory venv/:

venv/bin/conda env update --name base --file my-conda-env.yml

USAGE

jnbv -h

EXECUTE NOTEBOOK

For this, you will need to have:

1. A base environment in which jnbv has been installed (see above)
2. An ipynb notebook file
3. A Jupyter kernel that will be used to execute the file

Once you have all these items, the base environment should be activated:

source venv/bin/activate

Check to see what kernels you have available:

jupyter kernelspec list
    Available kernels:
      hdf5-kernel    /var/www/jupyterhub/jnbv/venv/share/jupyter/kernels/hdf5-kernel
      python3        /var/www/jupyterhub/jnbv/venv/share/jupyter/kernels/python3

If you don't have any kernels, then install ipykernel into your environment with either pip or conda:

pip install ipykernel
conda install ipykernel

and then you should have the default kernel "python3" available.

If you don't have an ipynb notebook handy, you can get an example notebook file here:

wget https://gitlab.com/MAXIV-SCISW/JUPYTERHUB/jnbv/-/raw/master/development/the-meaning-of-life.ipynb

And then the ipynb notebook can be executed in the terminal, using the default kernel python3 for example:

jnbv the-meaning-of-life.ipynb \
    --kernel python3 \
    --execute

READ NOTEBOOK

By defualt, the result of executing an ipynb file is a new ipynb file named output.ipynb. It can be read in the terminal with:

jnbv output.ipynb --read

TEST NOTEBOOK

The same file can be checked for errors:

jnbv output.ipynb --test

COMPARE NOTEBOOKS

And two ipynb notebooks can be compared:

jnbv output.ipynb --compare the-meaning-of-life.ipynb

EXECUTE, TEST, COMPARE, SAVE

All steps can be made to run in succession with one command:

jnbv the-meaning-of-life.ipynb \
    --kernel python3 \
    --execute --read --test --compare --save

Or, more simply using the --validate option, which is a combination of 5 other options:

jnbv the-meaning-of-life.ipynb \
    --kernel python3 \
    --validate

Note that above the option --save was also added, which then creates the output directory test-results/, and within that creates subdirectories with kernel names, date stamps, and finally log files and new ipynb files.
For example:

test-results/
└── python3/
    └── 2021-06-11_14-39-40/
        ├── the-meaning-of-life.ipynb
        └── the-meaning-of-life.log

Example output from executing a simple notebook using the default kernel:

CONTRIBUTING

Would you like to contribute to this project? See: CONTRIBUTING.md