ghstat

Statistical multi-criteria decision-making comparator for GitHub projects. Combines GitHub REST API data with local git clone analysis for accurate historical commit statistics.

Background

The tool was built out of a frustration familiar to most engineers: choosing between open source dependencies by star count alone. Stars measure marketing, not health.

The methodology behind ghstat was presented at Linux Foundation Open Source Summit Europe 2018 in Edinburgh, UK — "Methodology of Multi-Criteria Comparison and Typology of Open Source Projects". The core argument: open source project selection should be treated like engineering — systematic, multi-dimensional, reproducible. Commit velocity, contributor retention, issue resolution, and code churn together tell a story that no single metric can.

A companion blog post from 2018 walking through the analysis on Python frameworks: Using ghstat for open source project statistics and ratings.

Seven years later the tool has been rebuilt with hybrid analysis (GitHub API + local git clone) and run across ~200 repositories in 12 categories. The methodology holds.

Getting started

1. Generate a GitHub token

Go to https://github.com/settings/tokens and create a token with repo scope.

2. Clone and configure

git clone https://github.com/fedir/ghstat
cd ghstat
cp .env.sample .env
# edit .env and set your token

3. Build and run

make build
make run-go

Output is written to stats/go_frameworks.csv.

Usage

make help                 # list all available commands
make rate-limit           # check GitHub API quota
make cache-clear          # wipe HTTP response cache (preserves local clones)
make clone-clear          # remove local git clones in tmp/projects/
make run-go               # compare Go frameworks
make run-go-microservices # compare Go microservice toolkits
make run-rust-crates      # top 25 Rust crates
make run-cncf             # 50 CNCF cloud native projects
make run-devops           # 40 DevOps tools
make run-all              # run all comparisons and regenerate ratings.md
make clean-data-<category># remove clones for a specific category (cms, databases, langs, go, rust, js, python, ruby, java, cncf)
make clean-data-all       # remove all clones
make test                 # run tests with coverage

Custom comparison:

./ghstat -r angular/angular,facebook/react,vuejs/vue -f stats/js.csv -t tmp

Flags

Flag	Default	Description
-r	Go frameworks	Comma-separated list of owner/repo
-f	(required)	Output CSV file path
-t	test_data	Cache folder
-l		Check GitHub rate limit
-cc		Clear HTTP cache
-ccdr		Dry-run cache clear
-d		Debug mode

How it works

Each repository is analysed from two sources:

• GitHub API — real-time data: stars, forks, issues, license, author profile, closed issues, tags, contributors
• Local git clone — authoritative history: commit count, additions/deletions, commit size, contribution period, returning contributors

On first run repos are cloned to tmp/projects/. On subsequent runs the clones are updated. Local stats override API stats when available, so repositories where GitHub's stats API returns 202 (inactive repos) still get accurate data.

Comparison methodology

Each repository is scored across these criteria (more is better unless noted):

• Stargazers — popularity
• Age — newest is better
• Total commits — activity (from local git)
• Closed issues % — maintenance quality
• Commits/day — development pace (from local git)
• Top 10 contributors followers — community notability
• Active forkers % — engagement
• Returning contributors — project retention (from local git)
• Average contribution period — contributor loyalty (from local git)
• Total releases — release cadence

A final overall placement is computed by summing individual rankings.

Ratings

Detailed statistics with ratings