Up to [cvs.NetBSD.org] / pkgsrc / devel / hs-vector
Request diff between arbitrary revisions
Keyword substitution: kv
Default branch: MAIN
devel/hs-vector: update to vector-0.13.2.0
# Changes in version 0.13.2.0
* Strict boxed vector `Data.Vector.Strict` and `Data.Vector.Strict.Mutable` is
added (#488). it ensures that all values in the vector are evaluated to WHNF.
* `DoNotUnboxStrict`, `DoNotUnboxLazy`, and `DoNotUnboxNormalForm` wrapper are
added for defining unbox instances for types that contain not unboxable fields.
[#503](https://github.com/haskell/vector/issues/506),
[#508](https://github.com/haskell/vector/pull/508)
* `spanR` and `breakR` were added [#476](https://github.com/haskell/vector/pull/476).
They allow parsing vector from the right.
* We had some improvements on `*.Mutable.{next,prev}Permutation{,By}`
[#498](https://github.com/haskell/vector/pull/498):
* Add `*.Mutable.prevPermutation{,By}` and `*.Mutable.nextPermutationBy`
* Improve time performance. We may now expect good specialization supported by inlining.
The implementation has also been algorithmically updated: in the previous implementation
the full enumeration of all the permutations of `[1..n]` took Omega(n*n!), but it now takes O(n!).
* Add tests for `{next,prev}Permutation`
* Add benchmarks for `{next,prev}Permutation`
* Cabal >= 3.0 is now required for building package (#481).
* `vector:benchmarks-O2` public sublibrary containing benchmarks is added (#481).
* Type family `Mutable` provides instances for arrays from `primitive`.
* Various documentation improvements.
devel/hs-vector: Fix build with GHC 9.8 This breaks build with the currently default GHC 9.6. Please bear with me until I switch the default compiler.
devel/hs-vector: Update to 0.13.1.0 Changes in version 0.13.1.0 * Specialized variants of findIndexR are reexported for all vector types. #469 * UnboxViaPrim could be used for deriving Unbox instances (V_UnboxViaPrim constructor is exported) #450 * Fields of Data.Vector.Fusion.Bundle.Size are now strict #456 * Compatibility with future GHC 9.10 release #462 * Test suite no longer fails when built with QuickCheck-2.14 #461 * Doctests now work with current versions of GHC #465 * Various documentation improvements
devel/hs-vector: Fix build with GHC 9.4
hs-vector: update to 0.13.0.0.
# Changes in version 0.13.0.0
* `mkType` from `Data.Vector.Generic` is deprecated in favor of
`Data.Data.mkNoRepType`
* The role signatures on several `Vector` types were too permissive, so they
have been tightened up:
* The role signature for `Data.Vector.Mutable.MVector` is now
`type role MVector nominal representational` (previously, both arguments
were `phantom`). [#224](https://github.com/haskell/vector/pull/224)
* The role signature for `Data.Vector.Primitive.Vector` is now
`type role Vector nominal` (previously, it was `phantom`).
The role signature for `Data.Vector.Primitive.Mutable.MVector` is now
`type role MVector nominal nominal` (previously, both arguments were
`phantom`). [#316](https://github.com/haskell/vector/pull/316)
* The role signature for `Data.Vector.Storable.Vector` is now
`type role Vector nominal` (previous, it was `phantom`), and the signature
for `Data.Vector.Storable.Mutable.MVector` is now
`type role MVector nominal nominal` (previous, both arguments were
`phantom`). [#235](https://github.com/haskell/vector/pull/235)
We pick `nominal` for the role of the last argument instead of
`representational` since the internal structure of a `Storable` vector is
determined by the `Storable` instance of the element type, and it is not
guaranteed that the `Storable` instances between two representationally
equal types will preserve this internal structure. One consequence of this
choice is that it is no longer possible to `coerce` between
`Storable.Vector a` and `Storable.Vector b` if `a` and `b` are nominally
distinct but representationally equal types. We now provide
`unsafeCoerce{M}Vector` and `unsafeCast` functions to allow this (the onus
is on the user to ensure that no `Storable` invariants are broken when
using these functions).
* Methods of type classes `Data.Vector.Generic.Mutable.MVector` and
`Data.Vector.Generic.Vector` use concrete monads (`ST`, etc) istead of being
polymorphic (`PrimMonad`, etc). [#335](https://github.com/haskell/vector/pull/335).
This makes it possible to derive `Unbox` with:
* `GeneralizedNewtypeDeriving`
* via `UnboxViaPrim` and `Prim` instance
* via `As` and `IsoUnbox` instance: [#378](https://github.com/haskell/vector/pull/378)
* Add `MonadFix` instance for boxed vectors: [#312](https://github.com/haskell/vector/pull/312)
* Re-export `PrimMonad` and `RealWorld` from mutable vectors:
[#320](https://github.com/haskell/vector/pull/320)
* Add `maximumOn` and `minimumOn`: [#356](https://github.com/haskell/vector/pull/356)
* The functions `scanl1`, `scanl1'`, `scanr1`, and `scanr1'` for immutable
vectors are now defined when given empty vectors as arguments,
in which case they return empty vectors. This new behavior is consistent
with the one of the corresponding functions in `Data.List`.
Prior to this change, applying an empty vector to any of those functions
resulted in an error. This change was introduced in:
[#382](https://github.com/haskell/vector/pull/382)
* Change allocation strategy for `unfoldrN`: [#387](https://github.com/haskell/vector/pull/387)
* Remove `CPP` driven error reporting in favor of `HasCallStack`:
[#397](https://github.com/haskell/vector/pull/397)
* Remove redundant `Storable` constraints on to/from `ForeignPtr` conversions:
[#394](https://github.com/haskell/vector/pull/394)
* Add `unsafeCast` to `Primitive` vectors: [#401](https://github.com/haskell/vector/pull/401)
* Make `(!?)` operator strict: [#402](https://github.com/haskell/vector/pull/402)
* Add `readMaybe`: [#425](https://github.com/haskell/vector/pull/425)
* Add `groupBy` and `group` for `Data.Vector.Generic` and the specialized
version in `Data.Vector`, `Data.Vector.Unboxed`, `Data.Vector.Storable` and
`Data.Vector.Primitive`. [#427](https://github.com/haskell/vector/pull/427)
* Add `toArraySlice` and `unsafeFromArraySlice` functions for conversion to and
from the underlying boxed `Array`: [#434](https://github.com/haskell/vector/pull/434)
Fix build with GHC 9
hs-*: add PLIST files These PLIST files have been autogenerated by mk/haskell.mk using HS_UPDATE_PLIST=yes during a bulk build. They will help to track changes to the packages. The Haskell packages didn't have PLIST files because their paths contained package hashes. These hashes are now determined by mk/haskell.mk, which makes it easy to generate easy to read PLIST files.
Update to vector-0.12.0.3 Changes in version 0.12.0.3 * Add support for ghc >=8.8 monad fail Changes in version 0.12.0.2 * Fixes issue #220, compact heap operations crashing on boxed vectors constructed using traverse. * remove usage of Data.Primitive.Address and clarify the memset Prim Storable smuggling trick in Vector.Storable.Mutable * backport injective type family support Changes in version 0.12.0.1 * Make sure `length` can be inlined * Include modules that test-suites depend on in other-modules Changes in version 0.12.0.0 * Documentation fixes/additions * New functions: createT, iscanl/r, iterateNM, unfoldrM, uniq * New instances for various vector types: Semigroup, MonadZip * Made `Storable` vectors respect memory alignment * Changed some macros to ConstraintKinds * Dropped compatibility with old GHCs to support this * Add `Eq1`, `Ord1`, `Show1`, and `Read1` `Vector` instances, and related helper functions. * Relax context for `Unbox (Complex a)`.
Update to 0.11.0.0
changelog:
Changes in version 0.11.0.0
* Define `Applicative` instances for `Data.Vector.Fusion.Util.{Box,Id}`
* Define non-bottom `fail` for `instance Monad Vector`
* New generalized stream fusion framework
* Various safety fixes
- Various overflows due to vector size have been eliminated
- Memory is initialized on creation of unboxed vectors
* Changes to SPEC usage to allow building under more conditions
Changes in version 0.10.12.3
* Allow building with `primtive-0.6`
Import vector-0.10.9.2 as devel/hs-vector,
packaged for wip by pho.
An efficient implementation of Int-indexed arrays (both mutable and
immutable), with a powerful loop fusion optimization framework.
It is structured as follows:
* Data.Vector
Boxed vectors of arbitrary types.
* Data.Vector.Unboxed
Unboxed vectors with an adaptive representation based on data type
families.
* Data.Vector.Storable
Unboxed vectors of Storable types.
* Data.Vector.Primitive
Unboxed vectors of primitive types as defined by the primitive
package. Data.Vector.Unboxed is more flexible at no performance
cost.
* Data.Vector.Generic
Generic interface to the vector types.