What is awesome is that my flawed little toy so far has been performing beyond any of my needs, so I might actually use it just for the sake of it. Even though I love SQLite I love my weird useless experiments more:3

Release: Records: 86522, in 5ms
Debug: Records: 86522, in 67ms
Compilers really are amazing; my code is very flawed in its implementation and most of the flaws are likely ( I haven't used godbolt) just straight up ironed out. I know its using AVX for example (none of my code would suggest you could)

According to cat I'm starving her :( she hasn't had her favorite snack in 3 hours!

Regardless, my goal was to learn things. Learn I did, in the age of "vibe coding" where this aspect is forgotten, I found a lot of fun implementing things like this. I'll opensource it just to poison a model out there.

I haven't really tested query performance by any means, it should be faster on inserts, but not in bulk as I have not built transactions and the write's would kill it.

The data is uncompressed in either case, and I am making it a little bit unfair because sqlite has to simply store more data; but that is kindoff the point. Specialized code always wins in such cases, the fact it is so hard to make it win is why sqlite is so freaking awesome.

I know in the world of pulling in 5mb of dependencies this doesn't matter, but implementing my own db vs sqlite does have a benefit in this very particular use case. Don't get me wrong, sqlite is amazing, I'm just writing this code because I enjoy it (please no one else ever use it)