mstksg / nonempty-containers Goto Github PK

Hey @mstksg - loving the library! Would you mind a PR for non-empty vectors, or do you have a particular reason why this was not included? I'd be happy to take this on.

When I try and build with GHC 9.0, I get an error like this:

[ 1 of 12] Compiling Data.IntMap.NonEmpty.Internal ( src/Data/IntMap/NonEmpty/Internal.hs, dist/build/Data/IntMap/NonEmpty/Internal.o, dist/build/Data/IntMap/NonEmpty/Internal.dyn_o )

src/Data/IntMap/NonEmpty/Internal.hs:184:20: error:
    • Couldn't match type: c (t0 a)
                     with: forall d. Data d => c (t d)
      Expected: (forall d. Data d => c (t d)) -> Maybe (c (NEIntMap a))
        Actual: c (t0 a) -> Maybe (c (NEIntMap a))
    • In the expression: gcast1
      In an equation for ‘dataCast1’: dataCast1 = gcast1
      In the instance declaration for ‘Data (NEIntMap a)’
    • Relevant bindings include
        dataCast1 :: (forall d. Data d => c (t d))
                     -> Maybe (c (NEIntMap a))
          (bound at src/Data/IntMap/NonEmpty/Internal.hs:184:3)
    |
184 |   dataCast1      = gcast1
    |                    ^^^^^^

I know it opens a can of worms in terms of naming, but I missed a union version of NEMap.insertMap which still gets the O(m*log(n/m + 1)), m <= n complexity.

I'm using something like:

unionMap1 :: Ord a => Map a b -> NEMap a b -> NEMap a b
unionMap1 m n = NEMap.withNonEmpty n (<> n) m

unionMap2 :: Ord a => NEMap a b -> Map a b -> NEMap a b
unionMap2 n m = NEMap.withNonEmpty n (n <>) m

I'm happy to open a PR if you like these, but could use some help choosing names.

I recently found myself in need of the equivalent of Data.HashSet.NonEmpty, but this doesn't exist in this package. Could you add it?

Given that NESet guarantees the presence of at least one element, it would be nice to have something like this:

uncons :: NESet a -> (a, Set a)

This could be a pattern instead. The idea is that it gives back a pair of the smallest element (by Ord) and whatever's left. This could be very useful for what I'm currently trying to do.

I know already that IsList NESet is not safe because of [ ] :: NESet a.

But considering that there exists already in the base an IsList NonEmpty instance,
it might come as handy especially in a ghci REPL, where I guess people have OverloadedLists enabled to inputs sets,maps,lists,nonempty's more fast.

I was writing a Ixed instance for NESeq and glancing over the source for adjust'' it uses Seq.adjust instead of Seq.adjust'.

I'm pretty sure that will still result in the memory leak described in the adjust docs!

I think it would be handy to have a standard conversion from nonempty list-like containers to nonempty lists:

class (HasNonEmpty s, IsList s) => IsNonEmptyList s where
  fromNonEmptyList :: NonEmpty (Item s) -> NE s
  toNonEmptyList :: NE s -> NonEmpty (Item s)

where Item is the associated type for IsList. Thoughts?