Reproduce - Notebooks

You can use nbreproduce to execute individual Jupyter notebooks locally on your machine inside a docker environment. Jupyter Notebooks can be distributed as individual files/URLs with the name of the docker image it needs to be reproduced consistently.

nbreproduce will create a new notebook with the suffix -reproduce.ipynb in the current directory. To check the outputs you can open the reproduced notebooks using your favourite notebook frontend, JupyterLab/nteract/VSCode/PyCharm...

Local Notebooks

You can reproduce local notebooks by linking a docker image which is requried to build the environment.

$ nbreproduce --docker DOCKER_IMAGE name_of_notebook.ipynb

You wouldn't have to provide the DOCKER_IMAGE option after the first run as it is saved inside the notebook metadata (look for the tag name docker_image). If you share this notebook users wouldn't need to use --docker DOCKER_IMAGE while reproducing the notebook.

To reproduce a notebook which already has a docker_image tag in the metadata of the notebook.

$ nbreproduce name_of_notebook.ipynb

URL Notebooks

NOTE: This currently only works with GitHub URL links and docker images on DockerHub.

nbreproduce can be used in combination with a GitHub link of a Jupyter Notebook with a linked docker image.

To fetch a notebook and run it inside a docker container:

$ nbreproduce --docker DOCKER_IMAGE --url LINK_TO_URL

As an example to test if everything works locally you can test with a sample notebook hosted at https://github.com/econ-ark/nbreproduce/blob/master/tests/hello_world.ipynb.

$ nbreproduce --docker jupyter/scipy-notebook:latest --url https://github.com/econ-ark/nbreproduce/blob/master/tests/hello_world.ipynb