btreesets

[konsumlamm]

Fusion has a btreetables module, which mimics the tables module from the Nim standard library. Are there any plans to also add a btreesets module (which would possibly mimic the sets module from the standard library)? An implementation could probably use btreetables internally or just copy the implementation.

[timotheecour]

An implementation could probably use btreetables internally or just copy the implementation.

same comment as nim-lang/RFCs#200 (comment)

wrapper is not re-implementation
My main problem with nim's Table/HashSet/Intests/etc apis is not the number of unique table types, but the fact that these are more or less full re-implementations (with some partial code reuse). MultiTable as proposed here is just a thin wrapper, and in fact has a Table as a field. So all the complex aspects (optimizing performance for collisions, handling tombstones etc) is already taken care of. IMO other table types should be thin wrappers, eg HashSet is a prime candidate; and I believe performance would be at least equal to current HashSet performance; there are other cases.
The only justification for re-implementation is a benchmark showing performance of a wrapper is not as good as a re-implementation in at least some meaningful cases; I haven't seen such benchmark.

[konsumlamm]

I also like the API of std/critbits, which allows to set the value type to void, making the mapping/table behave like a set. This way, we can avoid code duplication at a very low cost, we'd just need a type alias and maybe some set-specific wrapper procs.