%
% (c) The University of Glasgow 2006
% (c) The AQUA Project, Glasgow University, 1998
%
This module contains definitions for the IdInfo for things that
have a standard form, namely:
data constructors
record selectors
method and superclass selectors
primitive operations
\begin{code}
module MkId (
mkDictFunId, mkDefaultMethodId,
mkDictSelId,
mkDataConIds,
mkPrimOpId, mkFCallId, mkTickBoxOpId, mkBreakPointOpId,
mkReboxingAlt, wrapNewTypeBody, unwrapNewTypeBody,
wrapFamInstBody, unwrapFamInstScrut,
mkUnpackCase, mkProductBox,
wiredInIds, ghcPrimIds,
unsafeCoerceId, realWorldPrimId, voidArgId, nullAddrId, seqId,
lazyId, lazyIdKey,
mkRuntimeErrorApp, mkImpossibleExpr,
rEC_CON_ERROR_ID, iRREFUT_PAT_ERROR_ID, rUNTIME_ERROR_ID,
nON_EXHAUSTIVE_GUARDS_ERROR_ID, nO_METHOD_BINDING_ERROR_ID,
pAT_ERROR_ID, eRROR_ID, rEC_SEL_ERROR_ID,
unsafeCoerceName
) where
#include "HsVersions.h"
import Rules
import TysPrim
import TysWiredIn
import PrelRules
import Type
import Coercion
import TcType
import CoreUtils ( exprType, mkCoerce )
import CoreUnfold
import Literal
import TyCon
import Class
import VarSet
import Name
import PrimOp
import ForeignCall
import DataCon
import Id
import Var ( Var, TyVar, mkCoVar, mkExportedLocalVar )
import IdInfo
import NewDemand
import CoreSyn
import Unique
import PrelNames
import BasicTypes hiding ( SuccessFlag(..) )
import Util
import Outputable
import FastString
import ListSetOps
import Module
\end{code}
%************************************************************************
%* *
\subsection{Wired in Ids}
%* *
%************************************************************************
Note [Wiredin Ids]
~~~~~~~~~~~~~~~~~~~
There are several reasons why an Id might appear in the wiredInIds:
(1) The ghcPrimIds are wired in because they can't be defined in
Haskell at all, although the can be defined in Core. They have
compulsory unfoldings, so they are always inlined and they have
no definition site. Their home module is GHC.Prim, so they
also have a description in primops.txt.pp, where they are called
'pseudoops'.
(2) The 'error' function, eRROR_ID, is wired in because we don't yet have
a way to express in an interface file that the result type variable
is 'open'; that is can be unified with an unboxed type
[The interface file format now carry such information, but there's
no way yet of expressing at the definition site for these
errorreporting functions that they have an 'open'
result type.
(3) Other error functions (rUNTIME_ERROR_ID) are wired in (a) because
the desugarer generates code that mentiones them directly, and
(b) for the same reason as eRROR_ID
(4) lazyId is wired in because the wiredin version overrides the
strictness of the version defined in GHC.Base
In cases (24), the function has a definition in a library module, and
can be called; but the wiredin version means that the details are
never read from that module's interface file; instead, the full definition
is right here.
\begin{code}
wiredInIds :: [Id]
wiredInIds
= [
eRROR_ID,
rUNTIME_ERROR_ID,
iRREFUT_PAT_ERROR_ID,
nON_EXHAUSTIVE_GUARDS_ERROR_ID,
nO_METHOD_BINDING_ERROR_ID,
pAT_ERROR_ID,
rEC_CON_ERROR_ID,
rEC_SEL_ERROR_ID,
lazyId
] ++ ghcPrimIds
ghcPrimIds :: [Id]
ghcPrimIds
= [
realWorldPrimId,
unsafeCoerceId,
nullAddrId,
seqId
]
\end{code}
%************************************************************************
%* *
\subsection{Data constructors}
%* *
%************************************************************************
The wrapper for a constructor is an ordinary toplevel binding that evaluates
any strict args, unboxes any args that are going to be flattened, and calls
the worker.
We're going to build a constructor that looks like:
data (Data a, C b) => T a b = T1 !a !Int b
T1 = /\ a b ->
\d1::Data a, d2::C b ->
\p q r -> case p of { p ->
case q of { q ->
Con T1 [a,b] [p,q,r]}}
Notice that
* d2 is thrown away
one *could* construct dictionaries at the site the constructor
is used, but the dictionary isn't actually used.
* We have to check that we can construct Data dictionaries for
the types a and Int. Once we've done that we can throw d1 away too.
* We use (case p of q -> ...) to evaluate p, rather than "seq" because
all that matters is that the arguments are evaluated. "seq" is
very careful to preserve evaluation order, which we don't need
to be here.
You might think that we could simply give constructors some strictness
info, like PrimOps, and let CoreToStg do the lettocase transformation.
But we don't do that because in the case of primops and functions strictness
is a *property* not a *requirement*. In the case of constructors we need to
do something active to evaluate the argument.
Making an explicit case expression allows the simplifier to eliminate
it in the (common) case where the constructor arg is already evaluated.
Note [Wrappers for data instance tycons]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In the case of data instances, the wrapper also applies the coercion turning
the representation type into the family instance type to cast the result of
the wrapper. For example, consider the declarations
data family Map k :: * -> *
data instance Map (a, b) v = MapPair (Map a (Pair b v))
The tycon to which the datacon MapPair belongs gets a unique internal
name of the form :R123Map, and we call it the representation tycon.
In contrast, Map is the family tycon (accessible via
tyConFamInst_maybe). A coercion allows you to move between
representation and family type. It is accessible from :R123Map via
tyConFamilyCoercion_maybe and has kind
Co123Map a b v :: {Map (a, b) v ~ :R123Map a b v}
The wrapper and worker of MapPair get the types
$WMapPair :: forall a b v. Map a (Map a b v) -> Map (a, b) v
$WMapPair a b v = MapPair a b v `cast` sym (Co123Map a b v)
MapPair :: forall a b v. Map a (Map a b v) -> :R123Map a b v
This coercion is conditionally applied by wrapFamInstBody.
It's a bit more complicated if the data instance is a GADT as well!
data instance T [a] where
T1 :: forall b. b -> T [Maybe b]
Hence
Co7T a :: T [a] ~ :R7T a
Now we want
$WT1 :: forall b. b -> T [Maybe b]
$WT1 b v = T1 (Maybe b) b (Maybe b) v
`cast` sym (Co7T (Maybe b))
T1 :: forall c b. (c ~ Maybe b) => b -> :R7T c
\begin{code}
mkDataConIds :: Name -> Name -> DataCon -> DataConIds
mkDataConIds wrap_name wkr_name data_con
| isNewTyCon tycon
= DCIds Nothing nt_work_id
| any isMarkedStrict all_strict_marks
|| not (null eq_spec)
|| isFamInstTyCon tycon
= DCIds (Just alg_wrap_id) wrk_id
| otherwise
= DCIds Nothing wrk_id
where
(univ_tvs, ex_tvs, eq_spec,
eq_theta, dict_theta, orig_arg_tys, res_ty) = dataConFullSig data_con
tycon = dataConTyCon data_con
wrk_id = mkGlobalId (DataConWorkId data_con) wkr_name
(dataConRepType data_con) wkr_info
wkr_arity = dataConRepArity data_con
wkr_info = noCafIdInfo
`setArityInfo` wkr_arity
`setAllStrictnessInfo` Just wkr_sig
`setUnfoldingInfo` evaldUnfolding
wkr_sig = mkStrictSig (mkTopDmdType (replicate wkr_arity topDmd) cpr_info)
cpr_info | isProductTyCon tycon &&
isDataTyCon tycon &&
wkr_arity > 0 &&
wkr_arity <= mAX_CPR_SIZE = retCPR
| otherwise = TopRes
nt_work_id = mkGlobalId (DataConWrapId data_con) wkr_name wrap_ty nt_work_info
nt_work_info = noCafIdInfo
`setArityInfo` 1
`setUnfoldingInfo` newtype_unf
id_arg1 = mkTemplateLocal 1 (head orig_arg_tys)
newtype_unf = ASSERT2( isVanillaDataCon data_con &&
isSingleton orig_arg_tys, ppr data_con )
mkCompulsoryUnfolding $
mkLams wrap_tvs $ Lam id_arg1 $
wrapNewTypeBody tycon res_ty_args (Var id_arg1)
wrap_tvs = (univ_tvs `minusList` map fst eq_spec) ++ ex_tvs
res_ty_args = substTyVars (mkTopTvSubst eq_spec) univ_tvs
eq_tys = mkPredTys eq_theta
dict_tys = mkPredTys dict_theta
wrap_ty = mkForAllTys wrap_tvs $ mkFunTys eq_tys $ mkFunTys dict_tys $
mkFunTys orig_arg_tys $ res_ty
alg_wrap_id = mkGlobalId (DataConWrapId data_con) wrap_name wrap_ty alg_wrap_info
alg_wrap_info = noCafIdInfo
`setArityInfo` wrap_arity
`setUnfoldingInfo` wrap_unf
`setAllStrictnessInfo` Just wrap_sig
all_strict_marks = dataConExStricts data_con ++ dataConStrictMarks data_con
wrap_sig = mkStrictSig (mkTopDmdType arg_dmds cpr_info)
arg_dmds = map mk_dmd all_strict_marks
mk_dmd str | isMarkedStrict str = evalDmd
| otherwise = lazyDmd
wrap_unf = mkImplicitUnfolding $ Note InlineMe $
mkLams wrap_tvs $
mkLams eq_args $
mkLams dict_args $ mkLams id_args $
foldr mk_case con_app
(zip (dict_args ++ id_args) all_strict_marks)
i3 []
con_app _ rep_ids = wrapFamInstBody tycon res_ty_args $
Var wrk_id `mkTyApps` res_ty_args
`mkVarApps` ex_tvs
`mkTyApps` map snd eq_spec
`mkVarApps` eq_args
`mkVarApps` reverse rep_ids
(dict_args,i2) = mkLocals 1 dict_tys
(id_args,i3) = mkLocals i2 orig_arg_tys
wrap_arity = i31
(eq_args,_) = mkCoVarLocals i3 eq_tys
mkCoVarLocals i [] = ([],i)
mkCoVarLocals i (x:xs) = let (ys,j) = mkCoVarLocals (i+1) xs
y = mkCoVar (mkSysTvName (mkBuiltinUnique i) (fsLit "dc_co")) x
in (y:ys,j)
mk_case
:: (Id, StrictnessMark)
-> (Int -> [Id] -> CoreExpr)
-> Int
-> [Id]
-> CoreExpr
mk_case (arg,strict) body i rep_args
= case strict of
NotMarkedStrict -> body i (arg:rep_args)
MarkedStrict
| isUnLiftedType (idType arg) -> body i (arg:rep_args)
| otherwise ->
Case (Var arg) arg res_ty [(DEFAULT,[], body i (arg:rep_args))]
MarkedUnboxed
-> unboxProduct i (Var arg) (idType arg) the_body
where
the_body i con_args = body i (reverse con_args ++ rep_args)
mAX_CPR_SIZE :: Arity
mAX_CPR_SIZE = 10
mkLocals i tys = (zipWith mkTemplateLocal [i..i+n1] tys, i+n)
where
n = length tys
\end{code}
Note [Newtype datacons]
~~~~~~~~~~~~~~~~~~~~~~~
The "data constructor" for a newtype should always be vanilla. At one
point this wasn't true, because the newtype arising from
class C a => D a
looked like
newtype T:D a = D:D (C a)
so the data constructor for T:C had a single argument, namely the
predicate (C a). But now we treat that as an ordinary argument, not
part of the thetatype, so all is well.
%************************************************************************
%* *
\subsection{Dictionary selectors}
%* *
%************************************************************************
Selecting a field for a dictionary. If there is just one field, then
there's nothing to do.
Dictionary selectors may get nested foralltypes. Thus:
class Foo a where
op :: forall b. Ord b => a -> b -> b
Then the toplevel type for op is
op :: forall a. Foo a =>
forall b. Ord b =>
a -> b -> b
This is unlike ordinary record selectors, which have all the foralls
at the outside. When dealing with classes it's very convenient to
recover the original type signature from the class op selector.
\begin{code}
mkDictSelId :: Bool
-> Name -> Class -> Id
mkDictSelId no_unf name clas
= mkGlobalId (ClassOpId clas) name sel_ty info
where
sel_ty = mkForAllTys tyvars (mkFunTy (idType dict_id) (idType the_arg_id))
info = noCafIdInfo
`setArityInfo` 1
`setAllStrictnessInfo` Just strict_sig
`setUnfoldingInfo` (if no_unf then noUnfolding
else mkImplicitUnfolding rhs)
strict_sig = mkStrictSig (mkTopDmdType [arg_dmd] TopRes)
arg_dmd | isNewTyCon tycon = evalDmd
| otherwise = Eval (Prod [ if the_arg_id == id then evalDmd else Abs
| id <- arg_ids ])
tycon = classTyCon clas
[data_con] = tyConDataCons tycon
tyvars = dataConUnivTyVars data_con
arg_tys = dataConRepArgTys data_con
eq_theta = dataConEqTheta data_con
the_arg_id = assoc "MkId.mkDictSelId" (map idName (classSelIds clas) `zip` arg_ids) name
pred = mkClassPred clas (mkTyVarTys tyvars)
dict_id = mkTemplateLocal 1 $ mkPredTy pred
(eq_ids,n) = mkCoVarLocals 2 $ mkPredTys eq_theta
arg_ids = mkTemplateLocalsNum n arg_tys
mkCoVarLocals i [] = ([],i)
mkCoVarLocals i (x:xs) = let (ys,j) = mkCoVarLocals (i+1) xs
y = mkCoVar (mkSysTvName (mkBuiltinUnique i) (fsLit "dc_co")) x
in (y:ys,j)
rhs = mkLams tyvars (Lam dict_id rhs_body)
rhs_body | isNewTyCon tycon = unwrapNewTypeBody tycon (map mkTyVarTy tyvars) (Var dict_id)
| otherwise = Case (Var dict_id) dict_id (idType the_arg_id)
[(DataAlt data_con, eq_ids ++ arg_ids, Var the_arg_id)]
\end{code}
%************************************************************************
%* *
Boxing and unboxing
%* *
%************************************************************************
\begin{code}
unboxProduct :: Int -> CoreExpr -> Type -> (Int -> [Id] -> CoreExpr) -> CoreExpr
unboxProduct i arg arg_ty body
= result
where
result = mkUnpackCase the_id arg con_args boxing_con rhs
(_tycon, _tycon_args, boxing_con, tys) = deepSplitProductType "unboxProduct" arg_ty
([the_id], i') = mkLocals i [arg_ty]
(con_args, i'') = mkLocals i' tys
rhs = body i'' con_args
mkUnpackCase :: Id -> CoreExpr -> [Id] -> DataCon -> CoreExpr -> CoreExpr
mkUnpackCase bndr arg unpk_args boxing_con body
= Case cast_arg (setIdType bndr bndr_ty) (exprType body) [(DataAlt boxing_con, unpk_args, body)]
where
(cast_arg, bndr_ty) = go (idType bndr) arg
go ty arg
| (tycon, tycon_args, _, _) <- splitProductType "mkUnpackCase" ty
, isNewTyCon tycon && not (isRecursiveTyCon tycon)
= go (newTyConInstRhs tycon tycon_args)
(unwrapNewTypeBody tycon tycon_args arg)
| otherwise = (arg, ty)
reboxProduct :: [Unique]
-> Type
-> ([Unique],
CoreExpr,
[Id])
reboxProduct us ty
= let
(_tycon, _tycon_args, _pack_con, con_arg_tys) = deepSplitProductType "reboxProduct" ty
us' = dropList con_arg_tys us
arg_ids = zipWith (mkSysLocal (fsLit "rb")) us con_arg_tys
bind_rhs = mkProductBox arg_ids ty
in
(us', bind_rhs, arg_ids)
mkProductBox :: [Id] -> Type -> CoreExpr
mkProductBox arg_ids ty
= result_expr
where
(tycon, tycon_args, pack_con, _con_arg_tys) = splitProductType "mkProductBox" ty
result_expr
| isNewTyCon tycon && not (isRecursiveTyCon tycon)
= wrap (mkProductBox arg_ids (newTyConInstRhs tycon tycon_args))
| otherwise = mkConApp pack_con (map Type tycon_args ++ map Var arg_ids)
wrap expr = wrapNewTypeBody tycon tycon_args expr
mkReboxingAlt
:: [Unique]
-> DataCon
-> [Var]
-> CoreExpr
-> CoreAlt
mkReboxingAlt us con args rhs
| not (any isMarkedUnboxed stricts)
= (DataAlt con, args, rhs)
| otherwise
= let
(binds, args') = go args stricts us
in
(DataAlt con, args', mkLets binds rhs)
where
stricts = dataConExStricts con ++ dataConStrictMarks con
go [] _stricts _us = ([], [])
go (arg:args) stricts us
| isTyVar arg
= let (binds, args') = go args stricts us
in (binds, arg:args')
go (arg:args) (str:stricts) us
| isMarkedUnboxed str
=
let (binds, unpacked_args') = go args stricts us'
(us', bind_rhs, unpacked_args) = reboxProduct us (idType arg)
in
(NonRec arg bind_rhs : binds, unpacked_args ++ unpacked_args')
| otherwise
= let (binds, args') = go args stricts us
in (binds, arg:args')
go (_ : _) [] _ = panic "mkReboxingAlt"
\end{code}
%************************************************************************
%* *
Wrapping and unwrapping newtypes and type families
%* *
%************************************************************************
\begin{code}
wrapNewTypeBody :: TyCon -> [Type] -> CoreExpr -> CoreExpr
wrapNewTypeBody tycon args result_expr
= wrapFamInstBody tycon args inner
where
inner
| Just co_con <- newTyConCo_maybe tycon
= mkCoerce (mkSymCoercion (mkTyConApp co_con args)) result_expr
| otherwise
= result_expr
unwrapNewTypeBody :: TyCon -> [Type] -> CoreExpr -> CoreExpr
unwrapNewTypeBody tycon args result_expr
| Just co_con <- newTyConCo_maybe tycon
= mkCoerce (mkTyConApp co_con args) result_expr
| otherwise
= result_expr
wrapFamInstBody :: TyCon -> [Type] -> CoreExpr -> CoreExpr
wrapFamInstBody tycon args body
| Just co_con <- tyConFamilyCoercion_maybe tycon
= mkCoerce (mkSymCoercion (mkTyConApp co_con args)) body
| otherwise
= body
unwrapFamInstScrut :: TyCon -> [Type] -> CoreExpr -> CoreExpr
unwrapFamInstScrut tycon args scrut
| Just co_con <- tyConFamilyCoercion_maybe tycon
= mkCoerce (mkTyConApp co_con args) scrut
| otherwise
= scrut
\end{code}
%************************************************************************
%* *
\subsection{Primitive operations}
%* *
%************************************************************************
\begin{code}
mkPrimOpId :: PrimOp -> Id
mkPrimOpId prim_op
= id
where
(tyvars,arg_tys,res_ty, arity, strict_sig) = primOpSig prim_op
ty = mkForAllTys tyvars (mkFunTys arg_tys res_ty)
name = mkWiredInName gHC_PRIM (primOpOcc prim_op)
(mkPrimOpIdUnique (primOpTag prim_op))
(AnId id) UserSyntax
id = mkGlobalId (PrimOpId prim_op) name ty info
info = noCafIdInfo
`setSpecInfo` mkSpecInfo (primOpRules prim_op name)
`setArityInfo` arity
`setAllStrictnessInfo` Just strict_sig
mkFCallId :: Unique -> ForeignCall -> Type -> Id
mkFCallId uniq fcall ty
= ASSERT( isEmptyVarSet (tyVarsOfType ty) )
mkGlobalId (FCallId fcall) name ty info
where
occ_str = showSDoc (braces (ppr fcall <+> ppr ty))
name = mkFCallName uniq occ_str
info = noCafIdInfo
`setArityInfo` arity
`setAllStrictnessInfo` Just strict_sig
(_, tau) = tcSplitForAllTys ty
(arg_tys, _) = tcSplitFunTys tau
arity = length arg_tys
strict_sig = mkStrictSig (mkTopDmdType (replicate arity evalDmd) TopRes)
mkTickBoxOpId :: Unique -> Module -> TickBoxId -> Id
mkTickBoxOpId uniq mod ix = mkTickBox' uniq mod ix realWorldStatePrimTy
mkBreakPointOpId :: Unique -> Module -> TickBoxId -> Id
mkBreakPointOpId uniq mod ix = mkTickBox' uniq mod ix ty
where ty = mkSigmaTy [openAlphaTyVar] [] openAlphaTy
mkTickBox' uniq mod ix ty = mkGlobalId (TickBoxOpId tickbox) name ty info
where
tickbox = TickBox mod ix
occ_str = showSDoc (braces (ppr tickbox))
name = mkTickBoxOpName uniq occ_str
info = noCafIdInfo
\end{code}
%************************************************************************
%* *
\subsection{DictFuns and default methods}
%* *
%************************************************************************
Important notes about dict funs and default methods
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Dict funs and default methods are *not* ImplicitIds. Their definition
involves userwritten code, so we can't figure out their strictness etc
based on fixed info, as we can for constructors and record selectors (say).
We build them as LocalIds, but with External Names. This ensures that
they are taken to account by freevariable finding and dependency
analysis (e.g. CoreFVs.exprFreeVars).
Why shouldn't they be bound as GlobalIds? Because, in particular, if
they are globals, the specialiser floats dict uses above their defns,
which prevents good simplifications happening. Also the strictness
analyser treats a occurrence of a GlobalId as imported and assumes it
contains strictness in its IdInfo, which isn't true if the thing is
bound in the same module as the occurrence.
It's OK for dfuns to be LocalIds, because we form the instanceenv to
pass on to the next module (md_insts) in CoreTidy, afer tidying
and globalising the toplevel Ids.
BUT make sure they are *exported* LocalIds (mkExportedLocalId) so
that they aren't discarded by the occurrence analyser.
\begin{code}
mkDefaultMethodId dm_name ty = mkExportedLocalId dm_name ty
mkDictFunId :: Name
-> [TyVar]
-> ThetaType
-> Class
-> [Type]
-> Id
mkDictFunId dfun_name inst_tyvars dfun_theta clas inst_tys
= mkExportedLocalVar DFunId dfun_name dfun_ty vanillaIdInfo
where
dfun_ty = mkSigmaTy inst_tyvars dfun_theta (mkDictTy clas inst_tys)
\end{code}
%************************************************************************
%* *
\subsection{Undefinable}
%* *
%************************************************************************
These Ids can't be defined in Haskell. They could be defined in
unfoldings in the wiredin GHC.Prim interface file, but we'd have to
ensure that they were definitely, definitely inlined, because there is
no curried identifier for them. That's what mkCompulsoryUnfolding
does. If we had a way to get a compulsory unfolding from an interface
file, we could do that, but we don't right now.
unsafeCoerce# isn't so much a PrimOp as a phantom identifier, that
just gets expanded into a type coercion wherever it occurs. Hence we
add it as a builtin Id with an unfolding here.
The type variables we use here are "open" type variables: this means
they can unify with both unlifted and lifted types. Hence we provide
another gun with which to shoot yourself in the foot.
\begin{code}
mkWiredInIdName mod fs uniq id
= mkWiredInName mod (mkOccNameFS varName fs) uniq (AnId id) UserSyntax
unsafeCoerceName = mkWiredInIdName gHC_PRIM (fsLit "unsafeCoerce#") unsafeCoerceIdKey unsafeCoerceId
nullAddrName = mkWiredInIdName gHC_PRIM (fsLit "nullAddr#") nullAddrIdKey nullAddrId
seqName = mkWiredInIdName gHC_PRIM (fsLit "seq") seqIdKey seqId
realWorldName = mkWiredInIdName gHC_PRIM (fsLit "realWorld#") realWorldPrimIdKey realWorldPrimId
lazyIdName = mkWiredInIdName gHC_BASE (fsLit "lazy") lazyIdKey lazyId
errorName = mkWiredInIdName gHC_ERR (fsLit "error") errorIdKey eRROR_ID
recSelErrorName = mkWiredInIdName cONTROL_EXCEPTION_BASE (fsLit "recSelError") recSelErrorIdKey rEC_SEL_ERROR_ID
runtimeErrorName = mkWiredInIdName cONTROL_EXCEPTION_BASE (fsLit "runtimeError") runtimeErrorIdKey rUNTIME_ERROR_ID
irrefutPatErrorName = mkWiredInIdName cONTROL_EXCEPTION_BASE (fsLit "irrefutPatError") irrefutPatErrorIdKey iRREFUT_PAT_ERROR_ID
recConErrorName = mkWiredInIdName cONTROL_EXCEPTION_BASE (fsLit "recConError") recConErrorIdKey rEC_CON_ERROR_ID
patErrorName = mkWiredInIdName cONTROL_EXCEPTION_BASE (fsLit "patError") patErrorIdKey pAT_ERROR_ID
noMethodBindingErrorName = mkWiredInIdName cONTROL_EXCEPTION_BASE (fsLit "noMethodBindingError")
noMethodBindingErrorIdKey nO_METHOD_BINDING_ERROR_ID
nonExhaustiveGuardsErrorName
= mkWiredInIdName cONTROL_EXCEPTION_BASE (fsLit "nonExhaustiveGuardsError")
nonExhaustiveGuardsErrorIdKey nON_EXHAUSTIVE_GUARDS_ERROR_ID
\end{code}
\begin{code}
unsafeCoerceId
= pcMiscPrelId unsafeCoerceName ty info
where
info = noCafIdInfo `setUnfoldingInfo` mkCompulsoryUnfolding rhs
ty = mkForAllTys [openAlphaTyVar,openBetaTyVar]
(mkFunTy openAlphaTy openBetaTy)
[x] = mkTemplateLocals [openAlphaTy]
rhs = mkLams [openAlphaTyVar,openBetaTyVar,x] $
Cast (Var x) (mkUnsafeCoercion openAlphaTy openBetaTy)
nullAddrId :: Id
nullAddrId = pcMiscPrelId nullAddrName addrPrimTy info
where
info = noCafIdInfo `setUnfoldingInfo`
mkCompulsoryUnfolding (Lit nullAddrLit)
seqId :: Id
seqId = pcMiscPrelId seqName ty info
where
info = noCafIdInfo `setUnfoldingInfo` mkCompulsoryUnfolding rhs
ty = mkForAllTys [alphaTyVar,openBetaTyVar]
(mkFunTy alphaTy (mkFunTy openBetaTy openBetaTy))
[x,y] = mkTemplateLocals [alphaTy, openBetaTy]
rhs = mkLams [alphaTyVar,openBetaTyVar,x,y] (Case (Var x) x openBetaTy [(DEFAULT, [], Var y)])
lazyId :: Id
lazyId = pcMiscPrelId lazyIdName ty info
where
info = noCafIdInfo
ty = mkForAllTys [alphaTyVar] (mkFunTy alphaTy alphaTy)
\end{code}
Note [seqId magic]
~~~~~~~~~~~~~~~~~~
'seq' is special in several ways.
a) Its second arg can have an unboxed type
x `seq` (v +# w)
b) Its fixity is set in LoadIface.ghcPrimIface
c) It has quite a bit of desugaring magic.
See DsUtils.lhs Note [Desugaring seq (1)] and (2) and (3)
d) There is some special rule handing: Note [RULES for seq]
Note [Rules for seq]
~~~~~~~~~~~~~~~~~~~~
Roman found situations where he had
case (f n) of _ -> e
where he knew that f (which was strict in n) would terminate if n did.
Notice that the result of (f n) is discarded. So it makes sense to
transform to
case n of _ -> e
Rather than attempt some general analysis to support this, I've added
enough support that you can do this using a rewrite rule:
RULE "f/seq" forall n. seq (f n) e = seq n e
You write that rule. When GHC sees a case expression that discards
its result, it mentally transforms it to a call to 'seq' and looks for
a RULE. (This is done in Simplify.rebuildCase.) As usual, the
correctness of the rule is up to you.
To make this work, we need to be careful that the magical desugaring
done in Note [seqId magic] item (c) is *not* done on the LHS of a rule.
Or rather, we arrange to undo it, in DsBinds.decomposeRuleLhs.
Note [lazyId magic]
~~~~~~~~~~~~~~~~~~~
lazy :: forall a?. a? -> a? (i.e. works for unboxed types too)
Used to lazify pseq: pseq a b = a `seq` lazy b
Also, no strictness: by being a builtin Id, all the info about lazyId comes from here,
not from GHC.Base.hi. This is important, because the strictness
analyser will spot it as strict!
Also no unfolding in lazyId: it gets "inlined" by a HACK in CorePrep.
It's very important to do this inlining *after* unfoldings are exposed
in the interface file. Otherwise, the unfolding for (say) pseq in the
interface file will not mention 'lazy', so if we inline 'pseq' we'll totally
miss the very thing that 'lazy' was there for in the first place.
See Trac #3259 for a real world example.
lazyId is defined in GHC.Base, so we don't *have* to inline it. If it
appears unapplied, we'll end up just calling it.
@realWorld#@ used to be a magic literal, \tr{void#}. If things get
nasty asis, change it back to a literal (@Literal@).
voidArgId is a Local Id used simply as an argument in functions
where we just want an arg to avoid having a thunk of unlifted type.
E.g.
x = \ void :: State# RealWorld -> (# p, q #)
This comes up in strictness analysis
\begin{code}
realWorldPrimId
= pcMiscPrelId realWorldName realWorldStatePrimTy
(noCafIdInfo `setUnfoldingInfo` evaldUnfolding)
voidArgId :: Id
voidArgId
= mkSysLocal (fsLit "void") voidArgIdKey realWorldStatePrimTy
\end{code}
%************************************************************************
%* *
\subsection[PrelValserrorrelated]{@error@ and friends; @trace@}
%* *
%************************************************************************
GHC randomly injects these into the code.
@patError@ is just a version of @error@ for patternmatching
failures. It knows various ``codes'' which expand to longer
strings
@absentErr@ is a thing we put in for ``absent'' arguments. They jolly
well shouldn't be yanked on, but if one is, then you will get a
friendly message from @absentErr@ (rather than a totally random
crash).
@parError@ is a special version of @error@ which the compiler does
not know to be a bottoming Id. It is used in the @_par_@ and @_seq_@
templates, but we don't ever expect to generate code for it.
\begin{code}
mkRuntimeErrorApp
:: Id
-> Type
-> String
-> CoreExpr
mkRuntimeErrorApp err_id res_ty err_msg
= mkApps (Var err_id) [Type res_ty, err_string]
where
err_string = Lit (mkMachString err_msg)
mkImpossibleExpr :: Type -> CoreExpr
mkImpossibleExpr res_ty
= mkRuntimeErrorApp rUNTIME_ERROR_ID res_ty "Impossible case alternative"
rEC_SEL_ERROR_ID = mkRuntimeErrorId recSelErrorName
rUNTIME_ERROR_ID = mkRuntimeErrorId runtimeErrorName
iRREFUT_PAT_ERROR_ID = mkRuntimeErrorId irrefutPatErrorName
rEC_CON_ERROR_ID = mkRuntimeErrorId recConErrorName
pAT_ERROR_ID = mkRuntimeErrorId patErrorName
nO_METHOD_BINDING_ERROR_ID = mkRuntimeErrorId noMethodBindingErrorName
nON_EXHAUSTIVE_GUARDS_ERROR_ID = mkRuntimeErrorId nonExhaustiveGuardsErrorName
mkRuntimeErrorId :: Name -> Id
mkRuntimeErrorId name = pc_bottoming_Id name runtimeErrorTy
runtimeErrorTy :: Type
runtimeErrorTy = mkSigmaTy [openAlphaTyVar] [] (mkFunTy addrPrimTy openAlphaTy)
\end{code}
\begin{code}
eRROR_ID = pc_bottoming_Id errorName errorTy
errorTy :: Type
errorTy = mkSigmaTy [openAlphaTyVar] [] (mkFunTys [mkListTy charTy] openAlphaTy)
\end{code}
%************************************************************************
%* *
\subsection{Utilities}
%* *
%************************************************************************
\begin{code}
pcMiscPrelId :: Name -> Type -> IdInfo -> Id
pcMiscPrelId name ty info
= mkVanillaGlobalWithInfo name ty info
pc_bottoming_Id :: Name -> Type -> Id
pc_bottoming_Id name ty
= pcMiscPrelId name ty bottoming_info
where
bottoming_info = vanillaIdInfo `setAllStrictnessInfo` Just strict_sig
`setArityInfo` 1
strict_sig = mkStrictSig (mkTopDmdType [evalDmd] BotRes)
\end{code}