Skip to main content

Wipe clean your Jupyter Notebooks!

Project description

nbwipers

Test License:MIT PyPI - Version Crates.io Conda codecov

nbwipers is a command line tool to wipe clean jupyter notebooks, written in Rust.

The interface and functionality are based on nbstripout and the idea to implement it in rust comes from nbstripout-fast.

Usage

nbwipers has a few subcommands that provide functionality related to cleaning Jupyter notebooks.

  • clean: clean a single notebook. This is more-or-less equivalent to nbstripout.
  • check: check notebooks in a given path for elements that would be removed by clean. This could be used in a CI context to enforce clean notebooks.
  • clean-all clean all notebooks in a given path. This one should be used carefully!
  • install register nbwipers as a git filter for ipynb files. Equivalent to nbstripout --install
  • uninstall remove nbwipers as a git filter.
  • check-install check that nbwipers or nbstripout is installed in the local repo. This is used in the pre-commit hook.

The full options can be found in CommandLineHelp.md.

Examples

To set up nbwipers as a git filter in your repository, use

nbwipers install local

If this step is performed on a pre-existing repo, you can touch your notebooks so that git can detect the changes. In bash:

for f in $(git ls-files '*.ipynb'); do touch $f; done

To check the notebooks in your folder, you can run the following

nbwipers check .

pre-commit

You can add the following to your pre-commit-config.yaml file to ensure that nbwipers or nbstripout is installed in your repo, in order to prevent Jupyter notebook outputs from being committed to version control.

  - repo: https://github.com/felixgwilliams/nbwipers-pre-commit
    rev: v0.3.4
    hooks:
      - id: nbwipers-check-install

Alternatively, you can use the URL for this repo in your config, but this will compile nbwipers from source, rather than retrieving the binary from PyPI, and is therefore not recommended.

If you are using your pre-commit configuration as part of CI, you should set the environment variable NBWIPERS_CHECK_INSTALL_EXIT_ZERO which forces this check to pass, since you do not need nbwipers configured in your CI environment.

Motivation

A working copy of a Jupyter notebook contains:

  1. Code written by the author.
  2. Notebook outputs: tables, logs, tracebacks, images, widgets and so on...
  3. Execution counts.
  4. Metadata, such as whether cells are collapsed, scrollable etc.

Of these categories of data, only the first — code written by the author — should definitely be tracked by version control, since it is the product of the author's intention and hard work. The other categories of data are subject to change outside of the explicit intentions of the author and are generally noisy from a version control perspective.

Moreover, including notebook outputs in version control

  • makes diffs harder to interpret, as they will contain lots of unintended changes.
  • increases the risk of a tricky merge conflict if different users run the same cell and get a slightly different result.
  • increases the amount of data committed, which can degrade repository performance.
  • risks leaking sensitive data.

An effective way to ensure you do not commit problematic parts of your notebooks is to use nbwipers or nbstripout as a git filter.

A git filter sits between your actual files and what git sees when you stage and commit your changes. This way, git only sees the transformed version of the file without the problematic elements. At the same time, you do not have to lose them from your local copy.

An exception is when you checkout a branch or do a git pull, which results in changes to the notebook. In this case, your local copy will be replaced by the clean version and you will lose your cell outputs.

Configuration

Configuration is currently done via the tool.nbwipers section of the pyproject.toml file. Most of the command line options can be set per-project in the pyproject.toml, nbwipers.toml or .nbwipers.toml file. If you use pyroject.toml, you need to put the configuration under [tool.nbwipers]. If you use nbwipers.toml or .nbwipers.toml, the configuration needs to be at the top level.

For example you can use extra-keys to specify additional notebook elements you want to ignore. If you don't need the python version or the details about the Jupyter Kernel, you can include the following in your pyproject.toml file:

[tool.nbwipers]
extra-keys = ["metadata.kernelspec", "metadata.language_info.version"]

The equivalent for nbwipers.toml or .nbwipers.toml is just

extra-keys = ["metadata.kernelspec", "metadata.language_info.version"]

This can be useful when collaborating, as the precise python version and the name assigned to the kernel are ephemeral and can change from person to person.

Testing Coverage

To test coverage, use the command:

cargo tarpaulin -o stdout -o html -o lcov --engine llvm

Using the llvm engine means that integration tests contribute to coverage.

Acknowledgements

nbwipers relies on inspiration and code from several projects. For the projects, whose code was used please see LICENSE for the third-party notices.

nbstripout

strip output from Jupyter and IPython notebooks

nbstripout is an invaluable tool for working with Jupyter Notebooks in the context of version control. This project forms the basis of the interface and logic of this project and is also the source of the testing examples.

nbstripout-fast

A much faster version of nbstripout by writing it in rust (of course).

nbstripout-fast, like this project, implements the functionality of nbstripout in Rust, while also allowing repo-level configuration in a YAML file.

With nbwipers I hoped to recreate the idea of nbstripout-fast, but with the ability to install as a git filter, and configuration via pyproject.toml.

ruff

An extremely fast Python linter and code formatter, written in Rust.

Ruff is quickly becoming the linter for python code, thanks to its performance, extensive set of rules and its ease of use. It was a definite source of knowledge for the organisation of the configuration and the file discovery. The schema for Jupyter Notebooks, and some of the file discovery code was adapted from Ruff.

pre-commit

A framework for managing and maintaining multi-language pre-commit hooks.

This repo contains a version of the check-large-files hook, that will not flag notebook files whose clean size is less that the threshold, even if the size on-disk including outputs is greater than the threshold. The logic and interface of the hook was adapted from the pre-commit-hooks repository.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

nbwipers-0.5.0.tar.gz (63.4 kB view hashes)

Uploaded Source

Built Distributions

nbwipers-0.5.0-py3-none-win_amd64.whl (1.9 MB view hashes)

Uploaded Python 3 Windows x86-64

nbwipers-0.5.0-py3-none-win32.whl (1.7 MB view hashes)

Uploaded Python 3 Windows x86

nbwipers-0.5.0-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (2.3 MB view hashes)

Uploaded Python 3 manylinux: glibc 2.17+ x86-64

nbwipers-0.5.0-py3-none-manylinux_2_17_s390x.manylinux2014_s390x.whl (2.5 MB view hashes)

Uploaded Python 3 manylinux: glibc 2.17+ s390x

nbwipers-0.5.0-py3-none-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl (2.4 MB view hashes)

Uploaded Python 3 manylinux: glibc 2.17+ ppc64le

nbwipers-0.5.0-py3-none-manylinux_2_17_i686.manylinux2014_i686.whl (2.5 MB view hashes)

Uploaded Python 3 manylinux: glibc 2.17+ i686

nbwipers-0.5.0-py3-none-manylinux_2_17_armv7l.manylinux2014_armv7l.whl (2.2 MB view hashes)

Uploaded Python 3 manylinux: glibc 2.17+ ARMv7l

nbwipers-0.5.0-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (2.2 MB view hashes)

Uploaded Python 3 manylinux: glibc 2.17+ ARM64

nbwipers-0.5.0-py3-none-macosx_11_0_arm64.whl (2.1 MB view hashes)

Uploaded Python 3 macOS 11.0+ ARM64

nbwipers-0.5.0-py3-none-macosx_10_12_x86_64.whl (2.2 MB view hashes)

Uploaded Python 3 macOS 10.12+ x86-64

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page