Friedrich is getting some much-deserved Reddit love for Dataset, his python library providing “databases for lazy people“. The idea is to allow you to build an SQL table from Python, with columns being auto-created as needed. It gives you all the power of SQL for free, without having to think about your data until you’ve got it in place.
It’s one of my favourite tools in the under-appreciated world of “small data”. I use it for exploratory data analysis, small scripts, and proof-of-concept applications. Most of the time I’m dealing with no more than a few million records, so I don’t need to think about optimizations. But I like the power and simplicity of SQL, and I’d much rather have my data in postgres than mongodb. Not least because I know that if I ever need to improve performance, I can easily add a few indexes and change some column types, and I’ll near-immediately be at a decently-performing database for most applications.