Predict code bug risk with git metadata
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
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
See this PyData talk for an explanation of how gitrisky works.
Contributions are welcome! Please see CONTRIBUTING.md for information about
contributing to this project.
The code in this project is licensed under the MIT license. See LICENSE for details.
The initial prototype of gitrisky was developed at
Civis Analytics during my 'Hack Time'
(time explicitly allotted to explore offbeat ideas) .