Working on abstracting the block store in Bitcoin Core. Currently evrything just goes straight to a file. I want to introduce abstract readers and writers that could optionally manage the data in memory, or potentially even a proper database. Then tests could optionally run without touching the filesystem.
Discussion
Ok, got the undo files abstracted. Now on to an in-memory implementation and the block files.
Ok, all done. But it is slow af. Wonder what I'm doing wrong. Just growing a 2d vector to hold the data should be blazing fast.
It's about the same speed now as writing to the file. Kind of sad that it is not faster, but at the same this means Bitcoin Core is already doing a decent job at it.