umm.... since I was mentioned, bear with me few sentences... I don't know much about sparse set, but after I read [this <https://www.geeksforgeeks.org/sparse-set/](https://forum.nim-lang.org/postActivity.xml#this-lthttpscolonslashslashwwwdotgeeksforgeeksdotorgslashsparseminussetslash) , it is obvious to me that the strength of sparse set is the O(1) clearing of all elements, which is rather different than a LRU cache. Also, from the data structure of sparse set, it can be said that it is very performant at its operations, though it has its own limitation on continuous range of integer as key and space requirement proportional to the size of range.
Btw, IMO, you don't have to think too much about the internal structure before hand. You can always improve the implementation as long as the interface unchanged. Maybe you can write a benchmark later to determine whether to use pairs or separate ones, though I think either case is likely fast enough to me and as a user, I wouldn't bother too much if one day I need a sparse set/map. Anyway, looking forward to your package~
