gitrisky

Predict code bug risk with git metadata

Installation

Installation with pip is recommended:

pip install gitrisky

Note that gitrisky requires numpy. If you don't already have it pip will try to install it for you, but this can result in a suboptimal build, see e.g. here.

For development a few additional dependencies are required:

pip install -r requirements-dev.txt

Usage

gitrisky is installed as a command line tool.

Usage: gitrisky [OPTIONS] COMMAND [ARGS]...

Options:
  --help  Show this message and exit.

Commands:
  predict  Score a git commit bug risk model.
  train    Train a git commit bug risk model.

The typical workflow is to first train a model on the existing commit history of a repository:

$ cd repo/
$ gitrisky train
Model trained on 69 training examples with 14 positive cases

and then use the trained model to score subsequent commits:

$ gitrisky predict
Commit 910cdb3c has a bug score of 0.2 / 1.0

When invoked without any extra arguments gitrisky predict will score the most recent commit. You can also score a particular commit with the -c flag:

$ gitrisky predict -c 470741f
Commit 470741f has a bug score of 0.7 / 1.0

How does it work?

See this PyData talk for an explanation of how gitrisky works.

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for information about contributing to this project.

License

The code in this project is licensed under the MIT license. See LICENSE for details.

Acknowledgements

The initial prototype of gitrisky was developed at Civis Analytics during my 'Hack Time' (time explicitly allotted to explore offbeat ideas) .