% % (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, mkRecordSelId, mkPrimOpId, mkFCallId, mkFExportId, mkTickBoxOpId, mkReboxingAlt, wrapNewTypeBody, unwrapNewTypeBody, mkUnpackCase, mkProductBox, -- And some particular Ids; see below for why they are wired in wiredInIds, ghcPrimIds, unsafeCoerceId, realWorldPrimId, voidPrimId, voidArgId, nullPtrId, seqId, lazyId, lazyIdUnfolding, lazyIdKey, mkBoolId, mkRuntimeErrorApp, 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, unsafeCoerceName ) where #include "HsVersions.h" import Rules import TysPrim import TysWiredIn import Type import TcGadt import HsBinds import Coercion import TcType import CoreUtils import CoreUnfold import Literal import TyCon import Class import VarSet import Name import OccName import PrimOp import ForeignCall import DataCon import Id import Var ( Var, TyVar) import IdInfo import NewDemand import DmdAnal import CoreSyn import Unique import Maybes import PrelNames import BasicTypes hiding ( SuccessFlag(..) ) import Util import Outputable import FastString import ListSetOps import Module \end{code} %************************************************************************ %* * \subsection{Wired in Ids} %* * %************************************************************************ \begin{code} wiredInIds = [ -- These error-y things are 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 -- error-reporting functions that they have an 'open' -- result type. -- sof 1/99] eRROR_ID, -- This one isn't used anywhere else in the compiler -- But we still need it in wiredInIds so that when GHC -- compiles a program that mentions 'error' we don't -- import its type from the interface file; we just get -- the Id defined here. Which has an 'open-tyvar' type. rUNTIME_ERROR_ID, iRREFUT_PAT_ERROR_ID, nON_EXHAUSTIVE_GUARDS_ERROR_ID, nO_METHOD_BINDING_ERROR_ID, pAT_ERROR_ID, rEC_CON_ERROR_ID, lazyId ] ++ ghcPrimIds -- These Ids are exported from GHC.Prim ghcPrimIds = [ -- These can't be defined in Haskell, but they have -- perfectly reasonable unfoldings in Core realWorldPrimId, voidPrimId, unsafeCoerceId, nullPtrId, seqId, mkBoolId ] \end{code} %************************************************************************ %* * \subsection{Data constructors} %* * %************************************************************************ The wrapper for a constructor is an ordinary top-level 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 --- a context in a data decl is used to make sure 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 let-to-case 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. [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). The wrapper and work of MapPair get the types $WMapPair :: forall a b v. Map a (Map a b v) -> Map (a, b) v $wMapPair :: forall a b v. Map a (Map a b v) -> :R123Map a b v which implies that the wrapper code will have to apply the coercion moving between representation and family type. It is accessible via tyConFamilyCoercion_maybe and has kind Co123Map a b v :: {Map (a, b) v :=: :R123Map a b v} This coercion is conditionally applied by wrapFamInstBody. \begin{code} mkDataConIds :: Name -> Name -> DataCon -> DataConIds mkDataConIds wrap_name wkr_name data_con | isNewTyCon tycon = DCIds Nothing nt_work_id -- Newtype, only has a worker | any isMarkedStrict all_strict_marks -- Algebraic, needs wrapper || not (null eq_spec) -- NB: LoadIface.ifaceDeclSubBndrs || isFamInstTyCon tycon -- depends on this test = DCIds (Just alg_wrap_id) wrk_id | otherwise -- Algebraic, no wrapper = DCIds Nothing wrk_id where (univ_tvs, ex_tvs, eq_spec, theta, orig_arg_tys) = dataConFullSig data_con tycon = dataConTyCon data_con ----------- Wrapper -------------- -- We used to include the stupid theta in the wrapper's args -- but now we don't. Instead the type checker just injects these -- extra constraints where necessary. wrap_tvs = (univ_tvs `minusList` map fst eq_spec) ++ ex_tvs subst = mkTopTvSubst eq_spec famSubst = ASSERT( length (tyConTyVars tycon ) == length (mkTyVarTys univ_tvs) ) zipTopTvSubst (tyConTyVars tycon) (mkTyVarTys univ_tvs) -- substitution mapping the type constructor's type -- arguments to the universals of the data constructor -- (crucial when type checking interfaces) dict_tys = mkPredTys theta result_ty_args = substTyVars subst univ_tvs result_ty = case tyConFamInst_maybe tycon of -- ordinary constructor Nothing -> mkTyConApp tycon result_ty_args -- family instance constructor Just (familyTyCon, instTys) -> mkTyConApp familyTyCon ( substTys subst . substTys famSubst $ instTys) wrap_ty = mkForAllTys wrap_tvs $ mkFunTys dict_tys $ mkFunTys orig_arg_tys $ result_ty -- NB: watch out here if you allow user-written equality -- constraints in data constructor signatures ----------- Worker (algebraic data types only) -------------- -- The *worker* for the data constructor is the function that -- takes the representation arguments and builds the constructor. 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 -- Record that it's evaluated, -- even if arity = 0 wkr_sig = mkStrictSig (mkTopDmdType (replicate wkr_arity topDmd) cpr_info) -- Note [Data-con worker strictness] -- Notice that we do *not* say the worker is strict -- even if the data constructor is declared strict -- e.g. data T = MkT !(Int,Int) -- Why? Because the *wrapper* is strict (and its unfolding has case -- expresssions that do the evals) but the *worker* itself is not. -- If we pretend it is strict then when we see -- case x of y -> $wMkT y -- the simplifier thinks that y is "sure to be evaluated" (because -- $wMkT is strict) and drops the case. No, $wMkT is not strict. -- -- When the simplifer sees a pattern -- case e of MkT x -> ... -- it uses the dataConRepStrictness of MkT to mark x as evaluated; -- but that's fine... dataConRepStrictness comes from the data con -- not from the worker Id. cpr_info | isProductTyCon tycon && isDataTyCon tycon && wkr_arity > 0 && wkr_arity <= mAX_CPR_SIZE = retCPR | otherwise = TopRes -- RetCPR is only true for products that are real data types; -- that is, not unboxed tuples or [non-recursive] newtypes ----------- Workers for newtypes -------------- nt_work_id = mkGlobalId (DataConWrapId data_con) wkr_name wrap_ty nt_work_info nt_work_info = noCafIdInfo -- The NoCaf-ness is set by noCafIdInfo `setArityInfo` 1 -- Arity 1 `setUnfoldingInfo` newtype_unf newtype_unf = ASSERT( isVanillaDataCon data_con && isSingleton orig_arg_tys ) -- No existentials on a newtype, but it can have a context -- e.g. newtype Eq a => T a = MkT (...) mkCompulsoryUnfolding $ mkLams wrap_tvs $ Lam id_arg1 $ wrapNewTypeBody tycon result_ty_args (Var id_arg1) id_arg1 = mkTemplateLocal 1 (head orig_arg_tys) ----------- Wrappers for algebraic data types -------------- alg_wrap_id = mkGlobalId (DataConWrapId data_con) wrap_name wrap_ty alg_wrap_info alg_wrap_info = noCafIdInfo -- The NoCaf-ness is set by noCafIdInfo `setArityInfo` alg_arity -- It's important to specify the arity, so that partial -- applications are treated as values `setUnfoldingInfo` alg_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 -- The Cpr info can be important inside INLINE rhss, where the -- wrapper constructor isn't inlined. -- And the argument strictness can be important too; we -- may not inline a contructor when it is partially applied. -- For example: -- data W = C !Int !Int !Int -- ...(let w = C x in ...(w p q)...)... -- we want to see that w is strict in its two arguments alg_unf = mkTopUnfolding $ Note InlineMe $ mkLams wrap_tvs $ 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 result_ty_args $ Var wrk_id `mkTyApps` result_ty_args `mkVarApps` ex_tvs `mkTyApps` map snd eq_spec `mkVarApps` reverse rep_ids (dict_args,i2) = mkLocals 1 dict_tys (id_args,i3) = mkLocals i2 orig_arg_tys alg_arity = i3-1 mk_case :: (Id, StrictnessMark) -- Arg, strictness -> (Int -> [Id] -> CoreExpr) -- Body -> Int -- Next rep arg id -> [Id] -- Rep args so far, reversed -> 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 result_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 -- We do not treat very big tuples as CPR-ish: -- a) for a start we get into trouble because there aren't -- "enough" unboxed tuple types (a tiresome restriction, -- but hard to fix), -- b) more importantly, big unboxed tuples get returned mainly -- on the stack, and are often then allocated in the heap -- by the caller. So doing CPR for them may in fact make -- things worse. mkLocals i tys = (zipWith mkTemplateLocal [i..i+n-1] tys, i+n) where n = length tys -- If the type constructor is a representation type of a data instance, wrap -- the expression into a cast adjusting the expression type, which is an -- instance of the representation type, to the corresponding instance of the -- family instance type. -- wrapFamInstBody :: TyCon -> [Type] -> CoreExpr -> CoreExpr wrapFamInstBody tycon args result_expr | Just co_con <- tyConFamilyCoercion_maybe tycon = mkCoerce (mkSymCoercion (mkTyConApp co_con args)) result_expr | otherwise = result_expr \end{code} %************************************************************************ %* * \subsection{Record selectors} %* * %************************************************************************ We're going to build a record selector unfolding that looks like this: data T a b c = T1 { ..., op :: a, ...} | T2 { ..., op :: a, ...} | T3 sel = /\ a b c -> \ d -> case d of T1 ... x ... -> x T2 ... x ... -> x other -> error "..." Similarly for newtypes newtype N a = MkN { unN :: a->a } unN :: N a -> a -> a unN n = coerce (a->a) n We need to take a little care if the field has a polymorphic type: data R = R { f :: forall a. a->a } Then we want f :: forall a. R -> a -> a f = /\ a \ r = case r of R f -> f a (not f :: R -> forall a. a->a, which gives the type inference mechanism problems at call sites) Similarly for (recursive) newtypes newtype N = MkN { unN :: forall a. a->a } unN :: forall b. N -> b -> b unN = /\b -> \n:N -> (coerce (forall a. a->a) n) Note [Naughty record selectors] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ A "naughty" field is one for which we can't define a record selector, because an existential type variable would escape. For example: data T = forall a. MkT { x,y::a } We obviously can't define x (MkT v _) = v Nevertheless we *do* put a RecordSelId into the type environment so that if the user tries to use 'x' as a selector we can bleat helpfully, rather than saying unhelpfully that 'x' is not in scope. Hence the sel_naughty flag, to identify record selectors that don't really exist. In general, a field is naughty if its type mentions a type variable that isn't in the result type of the constructor. Note [GADT record selectors] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ For GADTs, we require that all constructors with a common field 'f' have the same result type (modulo alpha conversion). [Checked in TcTyClsDecls.checkValidTyCon] E.g. data T where T1 { f :: a } :: T [a] T2 { f :: a, y :: b } :: T [a] and now the selector takes that type as its argument: f :: forall a. T [a] -> a f t = case t of T1 { f = v } -> v T2 { f = v } -> v Note the forall'd tyvars of the selector are just the free tyvars of the result type; there may be other tyvars in the constructor's type (e.g. 'b' in T2). \begin{code} -- Steps for handling "naughty" vs "non-naughty" selectors: -- 1. Determine naughtiness by comparing field type vs result type -- 2. Install naughty ones with selector_ty of type _|_ and fill in mzero for info -- 3. If it's not naughty, do the normal plan. mkRecordSelId :: TyCon -> FieldLabel -> Id mkRecordSelId tycon field_label -- Assumes that all fields with the same field label have the same type | is_naughty = naughty_id | otherwise = sel_id where is_naughty = not (tyVarsOfType field_ty `subVarSet` res_tv_set) sel_id_details = RecordSelId tycon field_label is_naughty -- Escapist case here for naughty construcotrs -- We give it no IdInfo, and a type of forall a.a (never looked at) naughty_id = mkGlobalId sel_id_details field_label forall_a_a noCafIdInfo forall_a_a = mkForAllTy alphaTyVar (mkTyVarTy alphaTyVar) -- Normal case starts here sel_id = mkGlobalId sel_id_details field_label selector_ty info data_cons = tyConDataCons tycon data_cons_w_field = filter has_field data_cons -- Can't be empty! has_field con = field_label `elem` dataConFieldLabels con con1 = head data_cons_w_field res_tys = dataConResTys con1 res_tv_set = tyVarsOfTypes res_tys res_tvs = varSetElems res_tv_set data_ty = mkTyConApp tycon res_tys field_ty = dataConFieldType con1 field_label -- *Very* tiresomely, the selectors are (unnecessarily!) overloaded over -- just the dictionaries in the types of the constructors that contain -- the relevant field. [The Report says that pattern matching on a -- constructor gives the same constraints as applying it.] Urgh. -- -- However, not all data cons have all constraints (because of -- BuildTyCl.mkDataConStupidTheta). So we need to find all the data cons -- involved in the pattern match and take the union of their constraints. stupid_dict_tys = mkPredTys (dataConsStupidTheta data_cons_w_field) n_stupid_dicts = length stupid_dict_tys (field_tyvars,pre_field_theta,field_tau) = tcSplitSigmaTy field_ty field_theta = filter (not . isEqPred) pre_field_theta field_dict_tys = mkPredTys field_theta n_field_dict_tys = length field_dict_tys -- If the field has a universally quantified type we have to -- be a bit careful. Suppose we have -- data R = R { op :: forall a. Foo a => a -> a } -- Then we can't give op the type -- op :: R -> forall a. Foo a => a -> a -- because the typechecker doesn't understand foralls to the -- right of an arrow. The "right" type to give it is -- op :: forall a. Foo a => R -> a -> a -- But then we must generate the right unfolding too: -- op = /\a -> \dfoo -> \ r -> -- case r of -- R op -> op a dfoo -- Note that this is exactly the type we'd infer from a user defn -- op (R op) = op selector_ty :: Type selector_ty = mkForAllTys res_tvs $ mkForAllTys field_tyvars $ mkFunTys stupid_dict_tys $ mkFunTys field_dict_tys $ mkFunTy data_ty field_tau arity = 1 + n_stupid_dicts + n_field_dict_tys (strict_sig, rhs_w_str) = dmdAnalTopRhs sel_rhs -- Use the demand analyser to work out strictness. -- With all this unpackery it's not easy! info = noCafIdInfo `setCafInfo` caf_info `setArityInfo` arity `setUnfoldingInfo` mkTopUnfolding rhs_w_str `setAllStrictnessInfo` Just strict_sig -- Allocate Ids. We do it a funny way round because field_dict_tys is -- almost always empty. Also note that we use max_dict_tys -- rather than n_dict_tys, because the latter gives an infinite loop: -- n_dict tys depends on the_alts, which depens on arg_ids, which depends -- on arity, which depends on n_dict tys. Sigh! Mega sigh! stupid_dict_ids = mkTemplateLocalsNum 1 stupid_dict_tys max_stupid_dicts = length (tyConStupidTheta tycon) field_dict_base = max_stupid_dicts + 1 field_dict_ids = mkTemplateLocalsNum field_dict_base field_dict_tys dict_id_base = field_dict_base + n_field_dict_tys data_id = mkTemplateLocal dict_id_base data_ty arg_base = dict_id_base + 1 the_alts :: [CoreAlt] the_alts = map mk_alt data_cons_w_field -- Already sorted by data-con no_default = length data_cons == length data_cons_w_field -- No default needed default_alt | no_default = [] | otherwise = [(DEFAULT, [], error_expr)] -- The default branch may have CAF refs, because it calls recSelError etc. caf_info | no_default = NoCafRefs | otherwise = MayHaveCafRefs sel_rhs = mkLams res_tvs $ mkLams field_tyvars $ mkLams stupid_dict_ids $ mkLams field_dict_ids $ Lam data_id $ mk_result sel_body -- NB: A newtype always has a vanilla DataCon; no existentials etc -- res_tys will simply be the dataConUnivTyVars sel_body | isNewTyCon tycon = unwrapNewTypeBody tycon res_tys (Var data_id) | otherwise = Case (Var data_id) data_id field_ty (default_alt ++ the_alts) mk_result poly_result = mkVarApps (mkVarApps poly_result field_tyvars) field_dict_ids -- We pull the field lambdas to the top, so we need to -- apply them in the body. For example: -- data T = MkT { foo :: forall a. a->a } -- -- foo :: forall a. T -> a -> a -- foo = /\a. \t:T. case t of { MkT f -> f a } mk_alt data_con = ASSERT2( res_ty `tcEqType` field_ty, ppr data_con $$ ppr res_ty $$ ppr field_ty ) mkReboxingAlt rebox_uniqs data_con (ex_tvs ++ co_tvs ++ arg_vs) rhs where -- get pattern binders with types appropriately instantiated arg_uniqs = map mkBuiltinUnique [arg_base..] (ex_tvs, co_tvs, arg_vs) = dataConOrigInstPat arg_uniqs data_con res_tys rebox_base = arg_base + length ex_tvs + length co_tvs + length arg_vs rebox_uniqs = map mkBuiltinUnique [rebox_base..] -- data T :: *->* where T1 { fld :: Maybe b } -> T [b] -- Hence T1 :: forall a b. (a=[b]) => b -> T a -- fld :: forall b. T [b] -> Maybe b -- fld = /\b.\(t:T[b]). case t of -- T1 b' (c : [b]=[b']) (x:Maybe b') -- -> x `cast` Maybe (sym (right c)) -- Generate the refinement for b'=b, -- and apply to (Maybe b'), to get (Maybe b) Succeeded refinement = gadtRefine emptyRefinement ex_tvs co_tvs the_arg_id_ty = idType the_arg_id (rhs, res_ty) = case refineType refinement the_arg_id_ty of Just (co, res_ty) -> (Cast (Var the_arg_id) co, res_ty) Nothing -> (Var the_arg_id, the_arg_id_ty) field_vs = filter (not . isPredTy . idType) arg_vs the_arg_id = assoc "mkRecordSelId:mk_alt" (field_lbls `zip` field_vs) field_label field_lbls = dataConFieldLabels data_con error_expr = mkRuntimeErrorApp rEC_SEL_ERROR_ID field_ty full_msg full_msg = showSDoc (sep [text "No match in record selector", ppr sel_id]) -- unbox a product type... -- we will recurse into newtypes, casting along the way, and unbox at the -- first product data constructor we find. e.g. -- -- data PairInt = PairInt Int Int -- newtype S = MkS PairInt -- newtype T = MkT S -- -- If we have e = MkT (MkS (PairInt 0 1)) and some body expecting a list of -- ids, we get (modulo int passing) -- -- case (e `cast` CoT) `cast` CoS of -- PairInt a b -> body [a,b] -- -- The Ints passed around are just for creating fresh locals 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 x e args Con body) -- returns -- case (e `cast` ...) of bndr { Con args -> body } -- -- the type of the bndr passed in is irrelevent 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) -- ...and the dual reboxProduct :: [Unique] -- uniques to create new local binders -> Type -- type of product to box -> ([Unique], -- remaining uniques CoreExpr, -- boxed product [Id]) -- Ids being boxed into product 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 us con xs rhs) basically constructs the case -- alternative (con, xs, rhs) -- but it does the reboxing necessary to construct the *source* -- arguments, xs, from the representation arguments ys. -- For example: -- data T = MkT !(Int,Int) Bool -- -- mkReboxingAlt MkT [x,b] r -- = (DataAlt MkT, [y::Int,z::Int,b], let x = (y,z) in r) -- -- mkDataAlt should really be in DataCon, but it can't because -- it manipulates CoreSyn. mkReboxingAlt :: [Unique] -- Uniques for the new Ids -> DataCon -> [Var] -- Source-level args, including existential dicts -> CoreExpr -- RHS -> 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 = ([], []) -- Type variable case go (arg:args) stricts us | isTyVar arg = let (binds, args') = go args stricts us in (binds, arg:args') -- Term variable case 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') \end{code} %************************************************************************ %* * \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 forall-types. Thus: class Foo a where op :: forall b. Ord b => a -> b -> b Then the top-level 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 for-alls 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 :: Name -> Class -> Id mkDictSelId name clas = mkGlobalId (ClassOpId clas) name sel_ty info where sel_ty = mkForAllTys tyvars (mkFunTy (idType dict_id) (idType the_arg_id)) -- We can't just say (exprType rhs), because that would give a type -- C a -> C a -- for a single-op class (after all, the selector is the identity) -- But it's type must expose the representation of the dictionary -- to gat (say) C a -> (a -> a) info = noCafIdInfo `setArityInfo` 1 `setUnfoldingInfo` mkTopUnfolding rhs `setAllStrictnessInfo` Just strict_sig -- We no longer use 'must-inline' on record selectors. They'll -- inline like crazy if they scrutinise a constructor -- The strictness signature is of the form U(AAAVAAAA) -> T -- where the V depends on which item we are selecting -- It's worth giving one, so that absence info etc is generated -- even if the selector isn't inlined 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 = ASSERT( isVanillaDataCon data_con ) dataConRepArgTys data_con the_arg_id = assoc "MkId.mkDictSelId" (map idName (classSelIds clas) `zip` arg_ids) name pred = mkClassPred clas (mkTyVarTys tyvars) (dict_id:arg_ids) = mkTemplateLocals (mkPredTy pred : arg_tys) 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, arg_ids, Var the_arg_id)] wrapNewTypeBody :: TyCon -> [Type] -> CoreExpr -> CoreExpr -- The wrapper for the data constructor for a newtype looks like this: -- newtype T a = MkT (a,Int) -- MkT :: forall a. (a,Int) -> T a -- MkT = /\a. \(x:(a,Int)). x `cast` sym (CoT a) -- where CoT is the coercion TyCon assoicated with the newtype -- -- The call (wrapNewTypeBody T [a] e) returns the -- body of the wrapper, namely -- e `cast` (CoT [a]) -- -- If a coercion constructor is prodivided in the newtype, then we use -- it, otherwise the wrap/unwrap are both no-ops -- -- If the we are dealing with a newtype instance, we have a second coercion -- identifying the family instance with the constructor of the newtype -- instance. This coercion is applied in any case (ie, composed with the -- coercion constructor of the newtype or applied by itself). -- 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 -- When unwrapping, we do *not* apply any family coercion, because this will -- be done via a CoPat by the type checker. We have to do it this way as -- computing the right type arguments for the coercion requires more than just -- a spliting operation (cf, TcPat.tcConPat). -- 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 \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 `setArityInfo` arity `setAllStrictnessInfo` Just strict_sig -- For each ccall we manufacture a separate CCallOpId, giving it -- a fresh unique, a type that is correct for this particular ccall, -- and a CCall structure that gives the correct details about calling -- convention etc. -- -- The *name* of this Id is a local name whose OccName gives the full -- details of the ccall, type and all. This means that the interface -- file reader can reconstruct a suitable Id mkFCallId :: Unique -> ForeignCall -> Type -> Id mkFCallId uniq fcall ty = ASSERT( isEmptyVarSet (tyVarsOfType ty) ) -- A CCallOpId should have no free type variables; -- when doing substitutions won't substitute over it mkGlobalId (FCallId fcall) name ty info where occ_str = showSDoc (braces (ppr fcall <+> ppr ty)) -- The "occurrence name" of a ccall is the full info about the -- ccall; it is encoded, but may have embedded spaces etc! 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) mkFExportId :: Unique -> ForeignExport -> Type -> Id mkFExportId uniq fexport ty = ASSERT( isEmptyVarSet (tyVarsOfType ty) ) -- A CCallOpId should have no free type variables; -- when doing substitutions won't substitute over it mkGlobalId (FExportId fexport) name ty info where name = mkSystemVarName uniq FSLIT("fexport") info = vanillaIdInfo `setArityInfo` arity (_, tau) = tcSplitForAllTys ty (arg_tys, _) = tcSplitFunTys tau arity = length arg_tys mkTickBoxOpId :: Unique -> Module -> TickBoxId -> Id mkTickBoxOpId uniq mod ix = mkGlobalId (TickBoxOpId tickbox) name ty info where tickbox = TickBox mod ix occ_str = showSDoc (braces (ppr tickbox)) name = mkTickBoxOpName uniq occ_str info = noCafIdInfo ty = realWorldStatePrimTy \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 user-written 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 free-variable 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 instance-env to pass on to the next module (md_insts) in CoreTidy, afer tidying and globalising the top-level 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 -- Name to use for the dict fun; -> [TyVar] -> ThetaType -> Class -> [Type] -> Id mkDictFunId dfun_name inst_tyvars dfun_theta clas inst_tys = mkExportedLocalId dfun_name dfun_ty where dfun_ty = mkSigmaTy inst_tyvars dfun_theta (mkDictTy clas inst_tys) {- 1 dec 99: disable the Mark Jones optimisation for the sake of compatibility with Hugs. See `types/InstEnv' for a discussion related to this. (class_tyvars, sc_theta, _, _) = classBigSig clas not_const (clas, tys) = not (isEmptyVarSet (tyVarsOfTypes tys)) sc_theta' = substClasses (zipTopTvSubst class_tyvars inst_tys) sc_theta dfun_theta = case inst_decl_theta of [] -> [] -- If inst_decl_theta is empty, then we don't -- want to have any dict arguments, so that we can -- expose the constant methods. other -> nub (inst_decl_theta ++ filter not_const sc_theta') -- Otherwise we pass the superclass dictionaries to -- the dictionary function; the Mark Jones optimisation. -- -- NOTE the "nub". I got caught by this one: -- class Monad m => MonadT t m where ... -- instance Monad m => MonadT (EnvT env) m where ... -- Here, the inst_decl_theta has (Monad m); but so -- does the sc_theta'! -- -- NOTE the "not_const". I got caught by this one too: -- class Foo a => Baz a b where ... -- instance Wob b => Baz T b where.. -- Now sc_theta' has Foo T -} \end{code} %************************************************************************ %* * \subsection{Un-definable} %* * %************************************************************************ These Ids can't be defined in Haskell. They could be defined in unfoldings in the wired-in 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 built-in 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 nullPtrName = mkWiredInIdName gHC_PRIM FSLIT("nullPtr#") nullPtrIdKey nullPtrId seqName = mkWiredInIdName gHC_PRIM FSLIT("seq") seqIdKey seqId realWorldName = mkWiredInIdName gHC_PRIM FSLIT("realWorld#") realWorldPrimIdKey realWorldPrimId voidName = mkWiredInIdName gHC_PRIM FSLIT("void#") voidPrimIdKey voidPrimId mkBoolName = mkWiredInIdName gHC_PRIM FSLIT("b#") mkBoolIdKey mkBoolId lazyIdName = mkWiredInIdName gHC_BASE FSLIT("lazy") lazyIdKey lazyId errorName = mkWiredInIdName gHC_ERR FSLIT("error") errorIdKey eRROR_ID recSelErrorName = mkWiredInIdName gHC_ERR FSLIT("recSelError") recSelErrorIdKey rEC_SEL_ERROR_ID runtimeErrorName = mkWiredInIdName gHC_ERR FSLIT("runtimeError") runtimeErrorIdKey rUNTIME_ERROR_ID irrefutPatErrorName = mkWiredInIdName gHC_ERR FSLIT("irrefutPatError") irrefutPatErrorIdKey iRREFUT_PAT_ERROR_ID recConErrorName = mkWiredInIdName gHC_ERR FSLIT("recConError") recConErrorIdKey rEC_CON_ERROR_ID patErrorName = mkWiredInIdName gHC_ERR FSLIT("patError") patErrorIdKey pAT_ERROR_ID noMethodBindingErrorName = mkWiredInIdName gHC_ERR FSLIT("noMethodBindingError") noMethodBindingErrorIdKey nO_METHOD_BINDING_ERROR_ID nonExhaustiveGuardsErrorName = mkWiredInIdName gHC_ERR FSLIT("nonExhaustiveGuardsError") nonExhaustiveGuardsErrorIdKey nON_EXHAUSTIVE_GUARDS_ERROR_ID \end{code} \begin{code} -- unsafeCoerce# :: forall a b. a -> b 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) -- nullPtr# :: forall (a :: !) . a -- The reason is is here is because we don't provide -- a way to write this literal in Haskell. nullPtrId = pcMiscPrelId nullPtrName (literalType MachNullPtr) info where info = noCafIdInfo `setUnfoldingInfo` mkCompulsoryUnfolding (Lit MachNullPtr) -- seq :: forall (a::*) (b::??) . a -> b -> b 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)]) -- 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 built-in 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 the worker/wrapper pass -- (see WorkWrap.wwExpr) -- We could use inline phases to do this, but that would be vulnerable to changes in -- phase numbering....we must inline precisely after strictness analysis. lazyId = pcMiscPrelId lazyIdName ty info where info = noCafIdInfo ty = mkForAllTys [alphaTyVar] (mkFunTy alphaTy alphaTy) lazyIdUnfolding :: CoreExpr -- Used to expand 'lazyId' after strictness anal lazyIdUnfolding = mkLams [openAlphaTyVar,x] (Var x) where [x] = mkTemplateLocals [openAlphaTy] \end{code} @realWorld#@ used to be a magic literal, \tr{void#}. If things get nasty as-is, 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 :: Void# -> (# p, q #) This comes up in strictness analysis \begin{code} realWorldPrimId -- :: State# RealWorld = pcMiscPrelId realWorldName realWorldStatePrimTy (noCafIdInfo `setUnfoldingInfo` evaldUnfolding) -- The evaldUnfolding makes it look that realWorld# is evaluated -- which in turn makes Simplify.interestingArg return True, -- which in turn makes INLINE things applied to realWorld# likely -- to be inlined voidPrimId -- :: Void# = pcMiscPrelId voidName voidPrimTy (noCafIdInfo `setUnfoldingInfo` evaldUnfolding) voidArgId -- :: Void# = mkSysLocal FSLIT("void") voidArgIdKey voidPrimTy \end{code} The bool pseudo constructor turns an unboxed bool into a boxed one This can't be written in haskell because it uses Bool# literals \begin{code} -- b# :: Bool# -> Bool mkBoolId = pcMiscPrelId mkBoolName (mkFunTy boolPrimTy boolTy) info where info = noCafIdInfo `setUnfoldingInfo` mkCompulsoryUnfolding rhs [x] = mkTemplateLocals [boolPrimTy] rhs = Lam x $ Case (Var x) (mkWildId boolPrimTy) boolTy [(LitAlt (MachInteger 0 boolPrimTyCon),[],Var falseDataConId) ,(LitAlt (MachInteger 1 boolPrimTyCon),[],Var trueDataConId)] \end{code} %************************************************************************ %* * \subsection[PrelVals-error-related]{@error@ and friends; @trace@} %* * %************************************************************************ GHC randomly injects these into the code. @patError@ is just a version of @error@ for pattern-matching failures. It knows various ``codes'' which expand to longer strings---this saves space! @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 -- Should be of type (forall a. Addr# -> a) -- where Addr# points to a UTF8 encoded string -> Type -- The type to instantiate 'a' -> String -- The string to print -> CoreExpr mkRuntimeErrorApp err_id res_ty err_msg = mkApps (Var err_id) [Type res_ty, err_string] where err_string = Lit (mkStringLit err_msg) 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 -- The runtime error Ids take a UTF8-encoded string as argument mkRuntimeErrorId name = pc_bottoming_Id name runtimeErrorTy runtimeErrorTy = mkSigmaTy [openAlphaTyVar] [] (mkFunTy stringPrimTy openAlphaTy) \end{code} \begin{code} eRROR_ID = pc_bottoming_Id errorName errorTy errorTy :: Type errorTy = mkSigmaTy [openAlphaTyVar] [] (mkFunTys [mkListTy charTy] openAlphaTy) -- Notice the openAlphaTyVar. It says that "error" can be applied -- to unboxed as well as boxed types. This is OK because it never -- returns, so the return type is irrelevant. \end{code} %************************************************************************ %* * \subsection{Utilities} %* * %************************************************************************ \begin{code} pcMiscPrelId :: Name -> Type -> IdInfo -> Id pcMiscPrelId name ty info = mkVanillaGlobal name ty info -- We lie and say the thing is imported; otherwise, we get into -- a mess with dependency analysis; e.g., core2stg may heave in -- random calls to GHCbase.unpackPS__. If GHCbase is the module -- being compiled, then it's just a matter of luck if the definition -- will be in "the right place" to be in scope. pc_bottoming_Id name ty = pcMiscPrelId name ty bottoming_info where bottoming_info = vanillaIdInfo `setAllStrictnessInfo` Just strict_sig -- Do *not* mark them as NoCafRefs, because they can indeed have -- CAF refs. For example, pAT_ERROR_ID calls GHC.Err.untangle, -- which has some CAFs -- In due course we may arrange that these error-y things are -- regarded by the GC as permanently live, in which case we -- can give them NoCaf info. As it is, any function that calls -- any pc_bottoming_Id will itself have CafRefs, which bloats -- SRTs. strict_sig = mkStrictSig (mkTopDmdType [evalDmd] BotRes) -- These "bottom" out, no matter what their arguments \end{code}