YouTube video by sparsevector

Recently, I've been enamored with the new falling block puzzle game Tetrachroma, a fiendishly mind-bending fusion of Tetris and Reversi. Questions about its...

YouTube video by WaterArko

A blog about video games. Game Reviews, analysis, thoughts, sentiments, stories, anecdotes.

YouTube video by sparsevector

It’s cool that there are mutable and unboxed variants, but it’s unclear where the boxed immutable sets stand relative to the standard (also sparse) sets provided by [_unordered-_]_containers_. I would guess that they’re more compactly represented with worse asymptotics for certain operations, but this could do to be documented. Actually, taking a look at the implementation, it’s using newtype SparseVector a = SparseVector {unSparseVector :: Vector (Bool, a)} which, contrary to its name, looks pretty dense to me? I expected something like data SparseVector a = SparseVector { entries :: !(V.Vector a) , indices :: !(U.Vector Int) -- Invariant: sorted } If it’s intended that these “sparse” vectors and sets are not sparsely represented but only sparse on some semantic level, then that absolutely merits big warnings somewhere.

Thanks for checking it out! Hmm that’s an interesting point as far as memory layout goes but I think this is close to the inspirations I got from `Bevy` and ECS back and forth to have this sparse set structure: data SparseSet i a = SparseSet { dense :: Vector a, sparse :: SparseVector i -- Invariant: sorted } (from https://skypjack.github.io/) EDIT: This will i.e. allocate slots for all elements up to index `1,000` in inserting there, which is good for game engines but probably not typical behavior so I should definitely note that in the docs As far as the `containers` packages go I think the closest structures I found were `Map` or `HashMap` since this would probably be closer to `Map`s than `Set`s

20 track album