No hidden formula. I had to download the 0.json (schema) and 1.json (data) files, evaluate these as JSON, then decode individual rows as base64 to find more JSON+flat buffers. I hacked together a very basic Node.js script for that:
Use: node stats.js ./path/to/file/1.json
I’ve heard this kind of stats should be added to Coda somewhere soon, so you’ll be able to see this in the UI and not the way I had to do it.