imclab / communityanalyzer

Experimenting with automated analysis of GitHub community behaviors.

CommunityAnalyzer

Experimenting with automated analysis of GitHub community behaviors.

Motivation

The idea is to take a data-driven approach to understanding what works and doesn't work in a community.

1. What causes issues to be closed faster?
2. What causes issues to linger for a long time?
3. What style of working creates positive sentiment?

Process

1. Download all issues.
2. Visualize issues opening/closing the way you would the lineup of Grateful Dead.
3. Extract metrics from issues.
4. Determine impact of metrics on time to completion.

Issue Metrics

• Number of comments.
• Number of participants.
• Comments per participant.
• Average, minimum, and maximum length of comments.
• Presence of code in comments.
• Number of labels, and which labels.
• Whether someone is assigned to the issue.
• Whether the issue is milestoned.
• The size of the diff in the commit that closes the issue.
• LIWC and sentiment analysis of the comments.

API Notes

• 5000 requests / hour (stick with one a second and you're safe)
• https://developer.github.com/v3/oauth/
• https://api.github.com/repos/openframeworks/openFrameworks/issues/1/comments
• https://api.github.com/repos/openframeworks/openFrameworks/issues/1/events