Contents

Notebook Cleaning

One of the powerful features of the .ipynb notebook format is it’s ability to serialise rich code cell outputs as part of the document format, as well as storing other artefacts of cell execution, such as the cell execution count number.

When releasing notebooks, or committing notebooks to a code repository, there may be a requirement to clear notebooks of cell outputs, and potentially other elements.

Various tools exist that can be used to clean notebooks, including:

  • nbconvert : the official nbconvert tool can be used to convert notebook documents to other document types. The tool can also be run with a cell pre-processor to rewrite a notebook with cell outputs cleared: jupyter nbconvert NOTEBOOK.ipynb --to notebook --ClearOutputPreprocessor.enabled=True --output NOTEBOOK

  • choldgraf/nbclean: a Pyhton package for scripting the cleaning of notebooks, including clearing cell outputs , removing cells based on tags and cell content and removing text within a cell between delimited start and end strings.

  • srstevenson/nb-clean: a command line utility, and Github filters, for removing cell execution counts, cell outputs, and (optionally) empty cells and cell metadata;

  • innovationOUtside/nb_workflow_tools: various command-line utilities for working with notebooks, including the ability to create compressed (zipped) archive files with automatically cleaned notebooks.

Clearing All Notebooks in a Directory

The above packages provide several differnt methods for clearing multiple notebooks. For example:

# innovationOUtside/nb_workflow_tools

# Clear all notebooks in a specified directory
tm351nbrun -r clearOutput DIRECTORY/
# Clear all notebooks in current directory and child directories
tm351nbrun -r clearOutput .


# srstevenson/nb-clean

# Clear all notebooks in a particular directory (preseve metadata)
nb-clean clean -m PATH/*.ipynb
# Clean notebooks in pattern matching subdirectories (preseve metadata)
nb-clean clean -m Part\ */*.ipynb