On: We saved $50k/year with a Go microservice coded in...

[dbrgn]

Sounds like a great case for Postgres hstore (like OpenStreetMap does it)?

[stu432 → dbrgn]

I mean, not really, theres no magic in Postgres' implementation of the pattern. They even spell it out for you:

"can be useful in various scenarios, such as rows with many attributes that are rarely examined"

They are querying these quite heavily, they aren't just random attributes they need to retrieve.

[dbrgn → stu432]

You're right, I think most OSM databases extract the most commonly used keys into dedicated, potentially indexed columns.

[bpicolo → dbrgn]

hstore querying is quite slow (and GIN indexes on hstores are pretty massive). I'd always go jsonb over hstores these days, but jsonb has the same indexing problem. JSON has a well-optimized/spec compliant serializer/deserializer in every language you can imagine as a baseline, whereas hstore does not.

[rickycook → dbrgn]

i think that’s basically what it was designed for :p