This message was deleted.

You could try to use Nominatim for the geocoding, but I suspect you'll find ready-made datasets for the post codes.

Thanks @Fabian Iwand. I made a full set of UK postcodes and their longlat but the CSV is 70Mb so I couldn't upload it. And it seemed like a heavy way to do it.

That seems unreasonably large - how many post codes are there? You could also try a different format like parquet for the encoding: https://observablehq.com/@observablehq/csv-to-parquet

1.7 million 🙂 Parquet is another good suggestion. I've never tried using that format so it's a good excuse to get familiar with it.

Euclidian distance will work, but if you'd like to be a pedant you could use a Haversine distance measure - Turf exports one https://turfjs.org/docs/#distance

@Max Bo I'm definitely open to some advanced geospatial pedantry. Thanks for the tip, I've never used Haversine before.

If we're optimizin' then you might also wanna take a look at https://github.com/mourner/geokdbush 🙂

Oh wow, that looks great too! So many fun new things to learn! 😃 However, discipline demands I put my Product Owner trousers on and start with the MVP before checking out all these fancy new things.