Veure's Database
I recently wrote about Veure's test suite and today I'll write a bit about how we manage our database. Sadly, this will be a long post because it's a complicated problem and there's a lot to discuss.
When I first started Veure, I used SQLite to prototype, but it's so incredibly limited that I quickly switched to Postgres. It's been a critically important decision, but I want to take a moment to explain why.
All software effectively has four "phases" which amount to:
- Initialization
- Input
- Calculation
- Output
Note that we could rewrite the above as:
- Initialization of data
- Input of data
- Calculation of data
- Output of data
Notice a pattern?
Yeah, I thought so. There are all sorts of areas where we could get things wrong in software, but the further down the stack(s) you go, the more care you need to take because the more damaging bugs can be. Data storage is often pretty low in your stack and you don't want to get this wrong. So what happens?