How to Automatically Clear Outputs from Jupyter Notebooks before Committing

This post explains how to automatically clear Jupyter or iPython notebook output cells every time you commit or switch branches in a particular git repo. Enabling this in a repository allows for easier collaboration on Jupyter notebooks and prevents the repository’s size from ballooning due to embedded plots and data printouts!

1. Clone the script (into, say, ~/GitHub) from https://github.com/toobaz/ipynb_output_filter via

   git clone https://github.com/toobaz/ipynb_output_filter.git
   

2. Make the script executable

   chmod +x ~/GitHub/ipynb_output_filter/ipynb_output_filter.py
   

3. Create a ~/.gitattributes file

   touch ~/.gitattributes
   

   and add the following hook:

   echo "*.ipynb    filter=dropoutput_ipynb" >> ~/.gitattributes
   

4. Configure git to find your ~/.gitattributes file:

   git config --global core.attributesfile ~/.gitattributes
   

5. Add the dropout_ipynb section (that will trigger the script above) to your repo’s git config by pasting the following into path/to/repo/.git/config:

[filter "dropoutput_ipynb"]
     clean = ~/GitHub/ipynb_output_filter/ipynb_output_filter.py
     smudge = cat

From then on, committing any *.ipynb file should lead to the output being stripped.

If you found this post from Google, I hope this solved your problem. Don’t hesitate to get in touch with us if you need help setting up a data sourcing, analysis or reporting infrastructure.