[GHC] #10697: Change template-haskell API to allow NOUNPACK, lazy annotations

[GHC]

#10697: Change template-haskell API to allow NOUNPACK, lazy annotations
-------------------------------------+-------------------------------------
        Reporter:  RyanGlScott       |                Owner:
            Type:  feature request   |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Template Haskell  |              Version:  7.10.1
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #5290, #8347      |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by RyanGlScott):

 What I'm still very unsure about is how we should handle strictness in
 `reify` versus handling strictness during splicing.

 When spicing in a data declaration, we should probably interpret it as
 literally as possible. For example,

 {{{#!hs
 [d| data U = U {-# UNPACK #-} Int |]
 }}}

 should splice just that, even though GHC would notice that a field is
 marked `{-# UNPACK #-}` without a strictness annotation, emit a warning,
 and then change that field to lazy afterwards. (On a side note, TH
 currently doesn't parse `{-# UNPACK #-}` annotations, but that would be
 easy to fix.)

 On the other hand, if we were to define a datatype
 [http://git.haskell.org/ghc.git/blob/688069ca83e949b9bde9883af7df26114e2f9bc0:/compiler/basicTypes/DataCon.hs#l499
 like this]:

 {{{#!hs
 data T = MkT !Int {-# UNPACK #-} !Int Bool
 }}}

 and call `reify ''T`, the strictness annotations would probably be
 meaningless by themselves. You'd need to know whether `-O`, `-funbox-
 strict-fields`, and/or `-XStrictData` were enabled to get the full story.
 I believe this is why internally, GHC distinguishes between
 [http://git.haskell.org/ghc.git/blob/688069ca83e949b9bde9883af7df26114e2f9bc0:/compiler/basicTypes/DataCon.hs#l457
 HsSrcBangs] (strictness annotations as the user wrote) and
 [http://git.haskell.org/ghc.git/blob/688069ca83e949b9bde9883af7df26114e2f9bc0:/compiler/basicTypes/DataCon.hs#l470
 HsImplBangs] (strictness annotations as GHC interprets them).

 With TH, we can cheat a little bit by using the
 [https://phabricator.haskell.org/D1200 forthcoming] `isExtEnabled`
 functionality to query whether `StrictData` is turned on, but for a
 [https://ghc.haskell.org/trac/ghc/ticket/10716 similar ticket] involving
 the ability to lookup strictness via generics, we have no such escape
 hatch. We should probably come up with a satisfying answer to this
 question so we can have a uniform treatment of strictness reification.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10697#comment:6>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler