In public repositories*

It’s an interesting view but it seems like the metrics themselves may be pretty error-prone with biases, and definitely cannot be used to draw conclusions.