This is really beautiful! I can think of a ton of use cases; so often I'm writing scraps of code or setting up quick DBs just to parse some specific dataset. I

This is really beautiful! I can think of a ton of use cases; so often I'm writing scraps of code or setting up quick DBs just to parse some specific dataset. I have a couple (naive) questions: 1. How are column types inferred? Does it try to avoid FLOAT types? What about blobs or JSON? (It would be a pretty stunning trick if it could notice valid JSON in a CSV file)... [edit: I see now from other columns that JSON is supported; nifty =)] 2. Does it try to infer charsets or just go with UTF8-mb4 or something else? and... any chance we'll see an NPM version?