NCQ Dataset (NPM Package Data)

Dataset for use in Node Code Query, contains package information in a tab separated csv file. The unzipped size is ~700MB.

You do not need to manually download this file for use in NCQ, the setup scripts will handle this for you automatically.

The dataset contains the following fields:

Mined from the NPM registry:

• Package name
• Description
• Keywords
• License
• repositoryUrl
• timeModified

Derived from data on the NPM registry:

• Array of Node.js code snippets extracted from the package README using https://github.com/Brittany-Reid/npm-code-snippets
• Number of markdown code blocks in the README (number may be larger than node.js snippets, these are non-filtered)
• Number of lines in the README
• If an install example exists in the README (if a code block exists with `npm install` or a `install` header exists)
• If a run example exists in the README (if a code block exists with `npm run` or a usage header exists)

Mined from GitHub for packages with a GitHub repository (values will be 0 or false for packages missing this data)

• Number of stars
• Is a fork?
• Number of forks
• Number of watchers
• If a test directory exists (if the top level directory contains a folder called `test` or `tests`)

This updated version also contains:

• The confidence value of an 'able to install' prediction.