All Versions
Latest Version
Avg Release Cycle
219 days
Latest Release
524 days ago

Changelog History
Page 1

  • v0.4.1 Changes

    May 11, 2020
    • โž• Add MonadFail instances for LFreshMT and FreshMT

    • ๐Ÿ— Builds with GHC 8.10

  • v0.4.0 Changes

    September 18, 2018
    • ๐Ÿ†• New binding specification type Ignore.

    Any two Ignore T terms will always be alpha-equivalent to each other, will be considered to contain no variables, and will not have any substitution apply beneath Ignore. Useful for attaching annotation terms to your AST.

        import Text.Parsec.Pos (SourcePos)
        data Expr =
           | Lambda (Ignore SourcePos) (Bind (Name Expr) Expr)

    As expected, any two Lambda expressions will be considered alpha-equivalent even if they differ in source position.

    Note that the Ignore will block operations on Name a for all a, which can be a little unexpected:

        data Ty =
          TyVar (Name Ty)
          | TyArr Ty Ty
        instance Subst Ty Ty where
        data Expr =
          | Var (Name Expr)
          | Lambda (Ignore Ty) (Bind (Name Expr) Expr)
         instance Subst Ty Expr

    Applying a substitution of a type for a free type variable to a Lambda will not descend into the Ignore Ty.

    Thanks Reed Mullanix (TOTWBF) for the new operation.

    • ๐Ÿ›  Fix an issue in substitution where traversal would not continue in an AST node for which isvar or isCoerceVar is defined to return non-Nothing but which had additional structure.

    For example, in a language with meta variables and explicit substitutions:

         data Expr =
             -- normal variables that stand for expressions
           | Var (Name Expr)
              -- a meta variable occurrence and an explicit substitution
              -- of expressions to substitute in for the free variables
           | MetaVar (Name Meta) [(Name Expr, Expr)]
         -- a meta variable stands for an expression with some free term vars
         data Meta = MetaVar Expr
         -- substitution for a meta in an expression
         instance Subst Expr Meta where
           isCoerceVar (MetaVar u sub) = Just (SubstCoerce u (Just . applyExplicitSubst sub))
         applyExplicitSubst :: [(Name Expr, Expr)] -> Meta -> Expr
         applyExplicitSubst s (MetaVar e) = substs s e

    Given an expression e1 defined as MetaVar "u" [("x", 10)], we may want to substitute a Meta ("x" + "x") for "u" to get 10 + 10 (that is, we replace "u" by the expression "x" + "x" and immediately apply the substitution 10 for "x").

    Now suppose we have an expression e2 defined as MetaVar "v" [("y", e1)] (that is, an occurrence of meta var "v" together with a substitution of e1 from above for "y"). If we again try to substitute Meta ("x" + "x") for "u" in e2, we would expect to get MetaVar "v" [("y", 10 + 10)] (that is, since "v" is not equal to "u", we leave the meta var alone, but substitute for any occurrences of "u" in the explicit substitution, so e1 becomes 10 + 10 as before).

    The bug in previous versions of unbound-generics was that we would incorrectly leave MetaVar "v" [("y", e1)] unchanged as soon as we saw that isCoerceVar (MetaVar "v" [("y", e1)]) returned Just (SubstCoerce "u" ...) where "u" /= "v".

    Thanks Reed Mullanix (TOTWBF) for finding and fixing this issue.

  • v0.3.4 Changes

    September 17, 2018
    • โฌ†๏ธ Bump containers upper bound to support 0.6. (GHC 8.6.1 support) Thanks Christiaan Baaij.
  • v0.3.3 Changes

    May 11, 2018

    โฌ†๏ธ Bump exceptions bound to >= 0.8 && < 0.11 to allow exceptions-0.10.0

  • v0.3.2 Changes

    March 09, 2018
    • โฌ†๏ธ Bump deepseq >= remove benchmark dependency on deepseq-generics
      ๐Ÿ— when building with GHC >= 7.10
    • โœ… Tested with GHC 8.4.1
    • โœ… Tested with GHC 8.2.2
    • Compile with -Wcompat
    • โž• Add Semigroup instances for all types that were previously Monoid instances
    • โž• Added more examples to the examples/ directory
    • โž• Added exceptions dependency and MonadThrow, MonadCatch, MonadMask instances for FreshMT and LFreshMT.
  • v0.3.1 Changes

    May 24, 2016
    • โœ… Tested with GHC 8.0.1
    • โœ‚ Removed Generic b constraint from Subst b (Name a) instance.
  • v0.3 Changes

    • โšก๏ธ Change types of open and close to take NthPatFind and NamePatFind instead of generic patterns, update call sites.
    • โž• Add newtype wrappers and Monoid instances for NthPatFind and NamePatFind
    • ๐Ÿ”„ Change isTerm to return All instead of Bool
  • v0.2 Changes

    August 01, 2015
    • Incorporating some of the extras/oversights from clash-lib Unbound.Generics.LocallyNameless.Extra

      • Make Embed an instance of Ord
      • NFData instances (see below)
    • ๐Ÿ†“ Re-implement freshen' and gfreshen using a free monad to give GHC a chance to inline it all away. This changes the type of gfreshen. Major version bump.

      • Expose FFM, liftFFM and retractFFM
    • Provide NFData instances for all the combinators. Depend on 'deepseq'

    • Start benchmarking some of the operations (particularly unbind).

  • v0.1.2 Changes

    May 08, 2015
    • โž• Added Shift type for lifting the scope of an Embed combinator up one binding level.
    • (in 0.1.1, which never went to Hackage) Added a makeClosedAlpha TH splice for making 'Alpha' instances for types that don't participate in Alpha in a non-trivial way. A type T is trivial with respect to Alpha if:
      • it cannot contain any names
      • thus its aeq is just (==)
      • and open and close are both identity operations.
  • v0.1.2.1 Changes

    • ๐Ÿ›  Fix ghc-7.10 build.
    • Haddock cleanup.