It can also be a good idea to split the RNG into a tree for different areas (i.e. seed multiple RNGs from the main one), so that adjusting the generation for one aspect doesn't shift around everything else in a seed (especially in something like Minecraft where different parts of the world might be generated on different versions of the game).
reply(Note this is roughly what slay the spire did, but if they were to use a 'master' RNG output as the source of these sub-seeds then these correlations would also not be a problem. With a custom implementation they could save the RNG state directly as opposed to hacking around calling the RNG X times on loading a save)
What's burned me before is iterating over hash maps. B-tree maps (or hash maps that are guaranteed to iterate in insertion order, or any fixed order) are your friend.
replyI've been burned by this outside games as well. Computation heavy process, in production it writes the inputs into the output as well for reproducibility, but some of them used unordered maps so it would slightly differ when you loaded it back up. Long term solution was to stop using unordered maps in general, but short term we had to make sure the inputs got sorted before being inserted so they would have the same insertion order every time...
replyI'm so happy most modern language stdlibs decided having a guaranteed order for maps and other collections is a good idea.
replyAlthough I can definitely respect Go's decision to always iterate over maps in a random order.