Treat kind/type variables identically, demolish FKTV
authorVladislav Zavialov <vlad.z.4096@gmail.com>
Wed, 13 Feb 2019 14:15:49 +0000 (17:15 +0300)
committerMarge Bot <ben+marge-bot@smart-cactus.org>
Wed, 27 Feb 2019 14:53:52 +0000 (09:53 -0500)
commit5bc195b1fe788e9a900a15fbe473967850517c3e
tree83844589096791edb049f719a990004756e02114
parent4dbacba5d2831bc80c48f3986e59b1a1c91cc620
Treat kind/type variables identically, demolish FKTV

Implements GHC Proposal #24: .../ghc-proposals/blob/master/proposals/0024-no-kind-vars.rst
Fixes Trac #16334, Trac #16315

With this patch, scoping rules for type and kind variables have been
unified: kind variables no longer receieve special treatment. This
simplifies both the language and the implementation.

User-facing changes
-------------------

* Kind variables are no longer implicitly quantified when an explicit
  forall is used:

    p ::             Proxy (a :: k)    -- still accepted
    p :: forall k a. Proxy (a :: k)    -- still accepted
    p :: forall   a. Proxy (a :: k)    -- no longer accepted

  In other words, now we adhere to the "forall-or-nothing" rule more
  strictly.

  Related function: RnTypes.rnImplicitBndrs

* The -Wimplicit-kind-vars warning has been deprecated.

* Kind variables are no longer implicitly quantified in constructor
  declarations:

    data T a        = T1 (S (a :: k) | forall (b::k). T2 (S b)  -- no longer accepted
    data T (a :: k) = T1 (S (a :: k) | forall (b::k). T2 (S b)  -- still accepted

  Related function: RnTypes.extractRdrKindSigVars

* Implicitly quantified kind variables are no longer put in front of
  other variables:

    f :: Proxy (a :: k) -> Proxy (b :: j)

    f :: forall k j (a :: k) (b :: j). Proxy a -> Proxy b   -- old order
    f :: forall k (a :: k) j (b :: j). Proxy a -> Proxy b   -- new order

  This is a breaking change for users of TypeApplications. Note that
  we still respect the dpendency order: 'k' before 'a', 'j' before 'b'.
  See "Ordering of specified variables" in the User's Guide.

  Related function: RnTypes.rnImplicitBndrs

* In type synonyms and type family equations, free variables on the RHS
  are no longer implicitly quantified unless used in an outermost kind
  annotation:

    type T = Just (Nothing :: Maybe a)         -- no longer accepted
    type T = Just Nothing :: Maybe (Maybe a)   -- still accepted

  The latter form is a workaround due to temporary lack of an explicit
  quantification method. Ideally, we would write something along these
  lines:

    type T @a = Just (Nothing :: Maybe a)

  Related function: RnTypes.extractHsTyRdrTyVarsKindVars

* Named wildcards in kinds are fixed (Trac #16334):

    x :: (Int :: _t)    -- this compiles, infers (_t ~ Type)

  Related function: RnTypes.partition_nwcs

Implementation notes
--------------------

* One of the key changes is the removal of FKTV in RnTypes:

  - data FreeKiTyVars = FKTV { fktv_kis    :: [Located RdrName]
  -                          , fktv_tys    :: [Located RdrName] }
  + type FreeKiTyVars = [Located RdrName]

  We used to keep track of type and kind variables separately, but
  now that they are on equal footing when it comes to scoping, we
  can put them in the same list.

* extract_lty and family are no longer parametrized by TypeOrKind,
  as we now do not distinguish kind variables from type variables.

* PatSynExPE and the related Note [Pattern synonym existentials do not scope]
  have been removed (Trac #16315). With no implicit kind quantification,
  we can no longer trigger the error.

* reportFloatingKvs and the related Note [Free-floating kind vars]
  have been removed. With no implicit kind quantification,
  we can no longer trigger the error.
59 files changed:
compiler/main/DynFlags.hs
compiler/rename/RnSource.hs
compiler/rename/RnTypes.hs
compiler/typecheck/TcHsType.hs
compiler/typecheck/TcPatSyn.hs
compiler/typecheck/TcRnTypes.hs
compiler/typecheck/TcSplice.hs
compiler/typecheck/TcTyClsDecls.hs
docs/users_guide/8.10.1-notes.rst
docs/users_guide/glasgow_exts.rst
docs/users_guide/using-warnings.rst
testsuite/tests/codeGen/should_fail/T13233.hs
testsuite/tests/dependent/should_compile/T15264.stderr [deleted file]
testsuite/tests/dependent/should_compile/all.T
testsuite/tests/dependent/should_fail/BadTelescope2.hs
testsuite/tests/dependent/should_fail/BadTelescope2.stderr
testsuite/tests/dependent/should_fail/T15264.hs [moved from testsuite/tests/dependent/should_compile/T15264.hs with 82% similarity]
testsuite/tests/dependent/should_fail/T15264.stderr [new file with mode: 0644]
testsuite/tests/dependent/should_fail/T15825.hs
testsuite/tests/dependent/should_fail/T15825.stderr
testsuite/tests/dependent/should_fail/all.T
testsuite/tests/ghci/scripts/T7873.stderr
testsuite/tests/ghci/scripts/T9293.stdout
testsuite/tests/ghci/scripts/ghci024.stdout
testsuite/tests/ghci/scripts/ghci057.stdout
testsuite/tests/indexed-types/should_compile/ExplicitForAllFams1.hs
testsuite/tests/parser/should_compile/DumpRenamedAst.stderr
testsuite/tests/partial-sigs/should_compile/T15039a.stderr
testsuite/tests/partial-sigs/should_compile/T15039b.stderr
testsuite/tests/partial-sigs/should_compile/T15039c.stderr
testsuite/tests/partial-sigs/should_compile/T15039d.stderr
testsuite/tests/partial-sigs/should_compile/T16334.hs [new file with mode: 0644]
testsuite/tests/partial-sigs/should_compile/T16334.stderr [new file with mode: 0644]
testsuite/tests/partial-sigs/should_compile/all.T
testsuite/tests/patsyn/should_compile/T14498.hs [moved from testsuite/tests/patsyn/should_fail/T14498.hs with 90% similarity]
testsuite/tests/patsyn/should_compile/all.T
testsuite/tests/patsyn/should_fail/T14507.hs
testsuite/tests/patsyn/should_fail/all.T
testsuite/tests/polykinds/BadKindVar.hs
testsuite/tests/polykinds/BadKindVar.stderr
testsuite/tests/polykinds/KindVarOrder.script [new file with mode: 0644]
testsuite/tests/polykinds/KindVarOrder.stdout [new file with mode: 0644]
testsuite/tests/polykinds/T10503.hs
testsuite/tests/polykinds/T10503.stderr
testsuite/tests/polykinds/T13985.stderr
testsuite/tests/polykinds/T14561.hs
testsuite/tests/polykinds/T14563.hs
testsuite/tests/polykinds/T14846.stderr
testsuite/tests/polykinds/T14887a.hs
testsuite/tests/polykinds/T8616.hs
testsuite/tests/polykinds/T8616.stderr
testsuite/tests/polykinds/T9144.hs
testsuite/tests/polykinds/all.T
testsuite/tests/typecheck/should_compile/T13343.hs
testsuite/tests/typecheck/should_fail/T12973.hs
testsuite/tests/typecheck/should_fail/T13983.stderr
testsuite/tests/typecheck/should_fail/T15629.hs
testsuite/tests/typecheck/should_fail/T15629.stderr
testsuite/tests/typecheck/should_fail/T15797.hs