Changelog for generic-data-1.1.0.2

1.1.0.2 -- 2024-12-06

Remove Eq and Ord for Generically1 instances. They are now available in base 4.18, and base-orphans 0.8.8 for backwards compatibility.
(Revisions) Also compatible with base 4.20 (GHC 9.10) and 4.19 (GHC 9.8)

Generically and Generically1 are in base 4.17 (GHC 9.4.1)!
- generic-data reexports Generically and Generically1 if using base >= 4.17. The following instances remain as orphans: Eq, Ord, Read, Show, Enum, Ix, Bounded, Foldable, Traversable, Read1, Show1.
- base 4.17 includes instances for the non-stock-derivable classes: Semigroup and Monoid for Generically; Eq1, Ord1, Functor, Applicative, and Alternative for Generically1.
- Note: the Semigroup and Monoid instances of base's Generically are those of generic-data's GenericProduct (which is subtly different from Generically's previous instance in generic-data).
- Generically and Generically1 are no longer defined using record syntax, so the unGenerically(1) field accessor no longer exists. The field accessors for FiniteEnumeration and GenericProduct were also removed for uniformity.

Fix conIdToString (it was completely broken)
Add conIdMin and conIdMax representing the leftmost and rightmost constructors of a data type.
Add NonEmptyType and IsEmptyType to express the constraint that a generic type must or must not be empty.
Reexport Generic and Generic1 for convenience.

Improved definition of gfoldMap, gtraverse, and sequenceA. The optimized Core of Traversable instances eliminates all GHC.Generic instance boilerplate. In many cases, it is identical to the result of GHC's DeriveFoldable and DeriveTraversable extensions (note: this was already not a problem for gfmap).

It's worth noting that there are currently issues with inlining which prevent optimizations that generic-data would ideally rely on.
- The biggest issue is that GHC will not even inline the to and from methods of the Generic instance it derives for large types (this shows up at around 5 constructors and 10 fields, which is indeed not really big). This will be fixed by a patch for GHC (WIP): https://gitlab.haskell.org/ghc/ghc/-/merge_requests/2965
- There appear to be some more inlining issues beyond that (issue #40).

Add microsurgery CopyRep.
Improve documentation of Microsurgery module.
Fix a bug where gshowsPrec would incorrectly display prefix uses of symbol data constructors or record selectors (e.g., data R = (:!:) Int Int or data S = MkS { (##) :: Int -> Int }). Thanks to RyanGlScott.
Fix a bug where gshowsPrec would incorrectly display infix uses of alphanumeric data constructors (e.g., data T = Int `MkT` Int). Thanks to RyanGlScott.

Add GenericProduct, for deriving via GenericProduct B when B is not the type A you want the derived instance for. Note this used to be Generically's behavior for Monoid before 0.7.0.0.
Add generic implementations for Ix. Thanks to Topsii.
Add conIdNamed, to get a ConId by its type-level name
Add instance Show (ConId a)
Improve type errors for deriving Semigroup and Monoid via Generically. Thanks to yairchu.

Change Monoid instance for Generically, to be compatible with users' non-generic instances of Semigroup. Thanks to yairchu.
Add gcoerce, gcoerceBinop.

Add Surgery newtype for DerivingVia
Derecordify, Typeage, RenameFields, RenameConstrs, OnFields are no longer type families, but defunctionalized symbols to be applied using GSurgery.

Created Microsurgery module. Initial set of surgeries:
- Derecordify
- Typeage
- RenameFields, RenameConstrs
- Some doc about using generic-lens for surgeries

Add generic implementations of enumFrom, enumFromThen, enumFromTo, enumFromThenTo. They are actually required to be explicit for correct Enum instances. Thanks to Topsii.
Parameterize GEnum by a type-level option, and add FiniteEnum option to allow Enum to be derived for composite types. Thanks to Topsii.

Released generic-data