This Node/Python library creates ml features for Pull Requests by learning information about a Github Project. The aim of this library is to aid data scientists build ml models for predicting pull request behaviour

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

What things you need to install the software and how to install them

You will need the following:

• Node 8 or newer
• MongoDB 3.2 or newer
• Git
• A Github Access Token for using the Github API. This post explains how to get yours.

0. Choose a project to predict. In this document I will use https://github.com/Netflix/pygenie, because it is smaller, but you can use any, like the Node project

1. Clone this repository into your machine:

   git clone https://github.com/benny-hal/pullreq-ml.git

2. Install dependencies

   cd pullreq-ml # or pullreq-ml-master
   npm install

3. Run mongo

   docker run -p 27017:27017 --name some-mongo -d mongo

4. Replace the contents of config.js with the actual repo and database authentication. For example

    module.exports = {
        // Local Mongo DB
        MONGO_DB_URL: 'mongodb://github:github@localhost:27017/github',
        // Token
        GITHUB_ACCESS_TOKEN: '<your token here>',
        // Repo Information for example for https://github.com/Netflix/pygenie you should put
    }

5. Clone the target repos inside the targetrepo folder

   cd targetrepo
   git clone https://github.com/Netflix/pygenie.git 

6. Start fetching Repo information

   node fetch.js

7. Create a features df for the PRs

   node query.js

8. The jira-fetcher is a bit primitive: it uses jira client to query (jql) jira to find tickets that were reopened and have PRs, we assume that the first PR in those tickets had a bug. We use this to locate buggy PRs ibn order to train our algorithm and build a model. This is not ideal and should be improved. In pull_request.js we use an HTTP cal to scrap Jira online interface in order to get the PR use itself because I couldn't find a way to do it with the client. This should be improved as well!