nbgitconvert

Automatically render jupyter notebooks to html, markdown or pdf on git commits. Works with Bitbucket, no plugins necessary.

Install/Setup

To install, move the nbgitconvert file to the root of your repo and run:

$ ./nbgitconvert --setup

This does two things:

1. creates a nbgitconvert.ini file that tells nbgitconvert which output types should be generated
2. create a git hook such that nbgitconvert runs everytime a *.ipynb file is committed.

the config file

in the [output_types] section there is a line for every supported output type and whether to convert to that output type. nbgitconvert will convert all committed *.ipynb files to all output types specified in the nbgitconvert.ini file.By default, only html is marked for output.

The output path can also be specified in the config. Include {output_type} in the path to save each output type to a different directory.

Update the config generated during the setup step as needed. Note that in the sample nbgitconver.ini file in this repo all formats have been set to "yes". The rendered notebooks in the jupyter notebooks/rendered/ directory were created from this config.

Usage

After running the above setup steps, nbgitconvert will convert any committed *.ipynb files to the output formats specified in nbgitconvert.ini, adding them to the renders/{output_type} directories, and commiting those new/updated files.

Assume you've already run the setup steps and are using the default values from nbgitconvert.ini (only converts to html). Here are two scenarios:

you create a new jupyter notebook

1. you created a new jupyter notebook named new_notebook.ipynb
2. you git add the notebook $ git add new_notebook.ipynb
3. you commit: $ git commit -m "added a new notebook"
4. the prehook runs, generating an html version of your file at rendered/html/new_notebook.html and git adds that new file before commiting.
5. Now when you push your changes to the branch, they'll include the rendered html version of your notebook.

you make a change to an existing notebook

Since this notebook already existed, let's assume the html version of the notebook also already existed.

1. you made a change to an existing notebook named existing_notebook.ipynb. Now your jupyter notebook is out of sync with the rendered html version of the notebook.
2. you git add the notebook $ git add existing_notebook.ipynb
3. you commit: $ git commit -m "updated existing notebook"
4. the prehook runs, regenerating the html version of your file at rendered/html/existing_notebook.html and git adds that updated file before commiting.
5. Now when you push your changes to the branch, they'll include the changes to your jupyter notebook and the updated rendered html version of your notebook.

Notes

1. nbconvert has issues converting notebooks with spaces in their names to pdf. To be safe, avoid putting spaces in your notebooks' names.