% % (c) The University of Glasgow 2006 % (c) The GRASP/AQUA Project, Glasgow University, 1992-1998 % Desugaring exporessions. \begin{code} {-# LANGUAGE PatternGuards #-} {-# OPTIONS -fno-warn-tabs #-} module Language.Haskell.Liquid.Desugar.DsExpr ( dsExpr, dsLExprWithLoc, dsLocalBinds, dsValBinds, dsLit ) where -- #include "HsVersions.h" import Language.Haskell.Liquid.GhcMisc (srcSpanTick) import Language.Haskell.Liquid.Desugar.Match import Language.Haskell.Liquid.Desugar.DsBinds import Language.Haskell.Liquid.Desugar.DsGRHSs import Language.Haskell.Liquid.Desugar.DsListComp import Language.Haskell.Liquid.Desugar.DsUtils import Language.Haskell.Liquid.Desugar.DsArrows import DsMonad import Language.Haskell.Liquid.Desugar.MatchLit import Name import NameEnv import HsSyn -- NB: The desugarer, which straddles the source and Core worlds, sometimes -- needs to see source types import TcType import TcEvidence import TcRnMonad import Type import CoreSyn import CoreUtils import CoreFVs import MkCore import DynFlags import StaticFlags import CostCentre import Id import VarSet import VarEnv import DataCon import TysWiredIn import BasicTypes import PrelNames import Maybes import SrcLoc import Util import Bag import Outputable import FastString import Control.Monad \end{code} %************************************************************************ %* * dsLocalBinds, dsValBinds %* * %************************************************************************ \begin{code} dsLocalBinds :: HsLocalBinds Id -> CoreExpr -> DsM CoreExpr dsLocalBinds EmptyLocalBinds body = return body dsLocalBinds (HsValBinds binds) body = dsValBinds binds body dsLocalBinds (HsIPBinds binds) body = dsIPBinds binds body ------------------------- dsValBinds :: HsValBinds Id -> CoreExpr -> DsM CoreExpr dsValBinds (ValBindsOut binds _) body = foldrM ds_val_bind body binds dsValBinds (ValBindsIn _ _) _ = panic "dsValBinds ValBindsIn" ------------------------- dsIPBinds :: HsIPBinds Id -> CoreExpr -> DsM CoreExpr dsIPBinds (IPBinds ip_binds ev_binds) body = do { ds_binds <- dsTcEvBinds ev_binds ; let inner = mkCoreLets ds_binds body -- The dict bindings may not be in -- dependency order; hence Rec ; foldrM ds_ip_bind inner ip_binds } where ds_ip_bind (L _ (IPBind ~(Right n) e)) body = do e' <- dsLExprWithLoc e return (Let (NonRec n e') body) ------------------------- ds_val_bind :: (RecFlag, LHsBinds Id) -> CoreExpr -> DsM CoreExpr -- Special case for bindings which bind unlifted variables -- We need to do a case right away, rather than building -- a tuple and doing selections. -- Silently ignore INLINE and SPECIALISE pragmas... ds_val_bind (NonRecursive, hsbinds) body | [L loc bind] <- bagToList hsbinds, -- Non-recursive, non-overloaded bindings only come in ones -- ToDo: in some bizarre case it's conceivable that there -- could be dict binds in the 'binds'. (See the notes -- below. Then pattern-match would fail. Urk.) strictMatchOnly bind = putSrcSpanDs loc (dsStrictBind bind body) -- Ordinary case for bindings; none should be unlifted ds_val_bind (_is_rec, binds) body = do { prs <- dsLHsBinds binds ; -- ASSERT2( not (any (isUnLiftedType . idType . fst) prs), ppr _is_rec $$ ppr binds ) case prs of [] -> return body _ -> return (Let (Rec prs) body) } -- Use a Rec regardless of is_rec. -- Why? Because it allows the binds to be all -- mixed up, which is what happens in one rare case -- Namely, for an AbsBind with no tyvars and no dicts, -- but which does have dictionary bindings. -- See notes with TcSimplify.inferLoop [NO TYVARS] -- It turned out that wrapping a Rec here was the easiest solution -- -- NB The previous case dealt with unlifted bindings, so we -- only have to deal with lifted ones now; so Rec is ok ------------------ dsStrictBind :: HsBind Id -> CoreExpr -> DsM CoreExpr dsStrictBind (AbsBinds { abs_tvs = [], abs_ev_vars = [] , abs_exports = exports , abs_ev_binds = ev_binds , abs_binds = binds }) body = do { let body1 = foldr bind_export body exports bind_export export b = bindNonRec (abe_poly export) (Var (abe_mono export)) b ; body2 <- foldlBagM (\body bind -> dsStrictBind (unLoc bind) body) body1 binds ; ds_binds <- dsTcEvBinds ev_binds ; return (mkCoreLets ds_binds body2) } dsStrictBind (FunBind { fun_id = L _ fun, fun_matches = matches, fun_co_fn = co_fn , fun_tick = tick, fun_infix = inf }) body -- Can't be a bang pattern (that looks like a PatBind) -- so must be simply unboxed = do { (args, rhs) <- matchWrapper (FunRhs (idName fun ) inf) matches -- ; MASSERT( null args ) -- Functions aren't lifted -- ; MASSERT( isIdHsWrapper co_fn ) ; let rhs' = mkOptTickBox tick rhs ; return (bindNonRec fun rhs' body) } dsStrictBind (PatBind {pat_lhs = pat, pat_rhs = grhss, pat_rhs_ty = ty }) body = -- let C x# y# = rhs in body -- ==> case rhs of C x# y# -> body do { rhs <- dsGuarded grhss ty ; let upat = unLoc pat eqn = EqnInfo { eqn_pats = [upat], eqn_rhs = cantFailMatchResult body } ; var <- selectMatchVar upat ; result <- matchEquations PatBindRhs [var] [eqn] (exprType body) ; return (bindNonRec var rhs result) } dsStrictBind bind body = pprPanic "dsLet: unlifted" (ppr bind $$ ppr body) ---------------------- strictMatchOnly :: HsBind Id -> Bool strictMatchOnly (AbsBinds { abs_binds = binds }) = anyBag (strictMatchOnly . unLoc) binds strictMatchOnly (PatBind { pat_lhs = lpat, pat_rhs_ty = ty }) = isUnLiftedType ty || isBangLPat lpat || any (isUnLiftedType . idType) (collectPatBinders lpat) strictMatchOnly (FunBind { fun_id = L _ id }) = isUnLiftedType (idType id) strictMatchOnly _ = False -- I hope! Checked immediately by caller in fact \end{code} %************************************************************************ %* * \subsection[DsExpr-vars-and-cons]{Variables, constructors, literals} %* * %************************************************************************ \begin{code} -- dsLExpr :: LHsExpr Id -> DsM CoreExpr -- dsLExpr = dsLExprWithLoc dsLExprWithLoc :: LHsExpr Id -> DsM CoreExpr -- dsLExprWithLoc (L loc e) = putSrcSpanDs loc $ dsExpr e dsLExprWithLoc (L loc e) -- = error "DIED in dsLExprWithLoc" = do ce <- putSrcSpanDs loc $ dsExpr e m <- getModuleDs return $ Tick (srcSpanTick m loc) ce dsExpr :: HsExpr Id -> DsM CoreExpr dsExpr (HsPar e) = dsLExprWithLoc e dsExpr (ExprWithTySigOut e _) = dsLExprWithLoc e dsExpr (HsVar var) = return (varToCoreExpr var) -- See Note [Desugaring vars] dsExpr (HsIPVar _) = panic "dsExpr: HsIPVar" dsExpr (HsLit lit) = dsLit lit dsExpr (HsOverLit lit) = dsOverLit lit dsExpr (HsWrap co_fn e) = do { e' <- dsExpr e ; wrapped_e <- dsHsWrapper co_fn e' ; warn_id <- woptM Opt_WarnIdentities ; when warn_id $ warnAboutIdentities e' wrapped_e ; return wrapped_e } dsExpr (NegApp expr neg_expr) = App <$> dsExpr neg_expr <*> dsLExprWithLoc expr dsExpr (HsLam a_Match) = uncurry mkLams <$> matchWrapper LambdaExpr a_Match dsExpr (HsLamCase arg matches@(MatchGroup _ rhs_ty)) | isEmptyMatchGroup matches -- A Core 'case' is always non-empty = -- So desugar empty HsLamCase to error call mkErrorAppDs pAT_ERROR_ID (funResultTy rhs_ty) (ptext (sLit "\\case")) | otherwise = do { arg_var <- newSysLocalDs arg ; ([discrim_var], matching_code) <- matchWrapper CaseAlt matches ; return $ Lam arg_var $ bindNonRec discrim_var (Var arg_var) matching_code } dsExpr (HsApp fun arg) = mkCoreAppDs <$> dsLExprWithLoc fun <*> dsLExprWithLoc arg \end{code} Note [Desugaring vars] ~~~~~~~~~~~~~~~~~~~~~~ In one situation we can get a *coercion* variable in a HsVar, namely the support method for an equality superclass: class (a~b) => C a b where ... instance (blah) => C (T a) (T b) where .. Then we get $dfCT :: forall ab. blah => C (T a) (T b) $dfCT ab blah = MkC ($c$p1C a blah) ($cop a blah) $c$p1C :: forall ab. blah => (T a ~ T b) $c$p1C ab blah = let ...; g :: T a ~ T b = ... } in g That 'g' in the 'in' part is an evidence variable, and when converting to core it must become a CO. Operator sections. At first it looks as if we can convert \begin{verbatim} (expr op) \end{verbatim} to \begin{verbatim} \x -> op expr x \end{verbatim} But no! expr might be a redex, and we can lose laziness badly this way. Consider \begin{verbatim} map (expr op) xs \end{verbatim} for example. So we convert instead to \begin{verbatim} let y = expr in \x -> op y x \end{verbatim} If \tr{expr} is actually just a variable, say, then the simplifier will sort it out. \begin{code} dsExpr (OpApp e1 op _ e2) = -- for the type of y, we need the type of op's 2nd argument mkCoreAppsDs <$> dsLExprWithLoc op <*> mapM dsLExprWithLoc [e1, e2] dsExpr (SectionL expr op) -- Desugar (e !) to ((!) e) = mkCoreAppDs <$> dsLExprWithLoc op <*> dsLExprWithLoc expr -- dsLExprWithLoc (SectionR op expr) -- \ x -> op x expr dsExpr (SectionR op expr) = do core_op <- dsLExprWithLoc op -- for the type of x, we need the type of op's 2nd argument let (x_ty:y_ty:_, _) = splitFunTys (exprType core_op) -- See comment with SectionL y_core <- dsLExprWithLoc expr x_id <- newSysLocalDs x_ty y_id <- newSysLocalDs y_ty return (bindNonRec y_id y_core $ Lam x_id (mkCoreAppsDs core_op [Var x_id, Var y_id])) dsExpr (ExplicitTuple tup_args boxity) = do { let go (lam_vars, args) (Missing ty) -- For every missing expression, we need -- another lambda in the desugaring. = do { lam_var <- newSysLocalDs ty ; return (lam_var : lam_vars, Var lam_var : args) } go (lam_vars, args) (Present expr) -- Expressions that are present don't generate -- lambdas, just arguments. = do { core_expr <- dsLExprWithLoc expr ; return (lam_vars, core_expr : args) } ; (lam_vars, args) <- foldM go ([], []) (reverse tup_args) -- The reverse is because foldM goes left-to-right ; return $ mkCoreLams lam_vars $ mkConApp (tupleCon (boxityNormalTupleSort boxity) (length tup_args)) (map (Type . exprType) args ++ args) } dsExpr (HsSCC cc expr@(L loc _)) = do mod_name <- getModuleDs count <- doptM Opt_ProfCountEntries uniq <- newUnique Tick (ProfNote (mkUserCC cc mod_name loc uniq) count True) <$> dsLExprWithLoc expr dsExpr (HsCoreAnn _ expr) = dsLExprWithLoc expr dsExpr (HsCase discrim matches@(MatchGroup _ rhs_ty)) | isEmptyMatchGroup matches -- A Core 'case' is always non-empty = -- So desugar empty HsCase to error call mkErrorAppDs pAT_ERROR_ID (funResultTy rhs_ty) (ptext (sLit "case")) | otherwise = do { core_discrim <- dsLExprWithLoc discrim ; ([discrim_var], matching_code) <- matchWrapper CaseAlt matches ; return (bindNonRec discrim_var core_discrim matching_code) } -- Pepe: The binds are in scope in the body but NOT in the binding group -- This is to avoid silliness in breakpoints dsExpr (HsLet binds body) = do body' <- dsLExprWithLoc body dsLocalBinds binds body' -- We need the `ListComp' form to use `deListComp' (rather than the "do" form) -- because the interpretation of `stmts' depends on what sort of thing it is. -- dsExpr (HsDo ListComp stmts res_ty) = dsListComp stmts res_ty dsExpr (HsDo PArrComp stmts _) = dsPArrComp (map unLoc stmts) dsExpr (HsDo DoExpr stmts _) = dsDo stmts dsExpr (HsDo GhciStmt stmts _) = dsDo stmts dsExpr (HsDo MDoExpr stmts _) = dsDo stmts dsExpr (HsDo MonadComp stmts _) = dsMonadComp stmts dsExpr (HsIf mb_fun guard_expr then_expr else_expr) = do { pred <- dsLExprWithLoc guard_expr ; b1 <- dsLExprWithLoc then_expr ; b2 <- dsLExprWithLoc else_expr ; case mb_fun of Just fun -> do { core_fun <- dsExpr fun ; return (mkCoreApps core_fun [pred,b1,b2]) } Nothing -> return $ mkIfThenElse pred b1 b2 } dsExpr (HsMultiIf res_ty alts) | null alts = mkErrorExpr | otherwise = do { match_result <- liftM (foldr1 combineMatchResults) (mapM (dsGRHS IfAlt res_ty) alts) ; error_expr <- mkErrorExpr ; extractMatchResult match_result error_expr } where mkErrorExpr = mkErrorAppDs nON_EXHAUSTIVE_GUARDS_ERROR_ID res_ty (ptext (sLit "multi-way if")) \end{code} \noindent \underline{\bf Various data construction things} % ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ \begin{code} dsExpr (ExplicitList elt_ty xs) = dsExplicitList elt_ty xs -- We desugar [:x1, ..., xn:] as -- singletonP x1 +:+ ... +:+ singletonP xn -- dsExpr (ExplicitPArr ty []) = do emptyP <- dsDPHBuiltin emptyPVar return (Var emptyP `App` Type ty) dsExpr (ExplicitPArr ty xs) = do singletonP <- dsDPHBuiltin singletonPVar appP <- dsDPHBuiltin appPVar xs' <- mapM dsLExprWithLoc xs return . foldr1 (binary appP) $ map (unary singletonP) xs' where unary fn x = mkApps (Var fn) [Type ty, x] binary fn x y = mkApps (Var fn) [Type ty, x, y] dsExpr (ArithSeq expr (From from)) = App <$> dsExpr expr <*> dsLExprWithLoc from dsExpr (ArithSeq expr (FromTo from to)) = mkApps <$> dsExpr expr <*> mapM dsLExprWithLoc [from, to] dsExpr (ArithSeq expr (FromThen from thn)) = mkApps <$> dsExpr expr <*> mapM dsLExprWithLoc [from, thn] dsExpr (ArithSeq expr (FromThenTo from thn to)) = mkApps <$> dsExpr expr <*> mapM dsLExprWithLoc [from, thn, to] dsExpr (PArrSeq expr (FromTo from to)) = mkApps <$> dsExpr expr <*> mapM dsLExprWithLoc [from, to] dsExpr (PArrSeq expr (FromThenTo from thn to)) = mkApps <$> dsExpr expr <*> mapM dsLExprWithLoc [from, thn, to] dsExpr (PArrSeq _ _) = panic "DsExpr.dsExpr: Infinite parallel array!" -- the parser shouldn't have generated it and the renamer and typechecker -- shouldn't have let it through \end{code} \noindent \underline{\bf Record construction and update} % ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ For record construction we do this (assuming T has three arguments) \begin{verbatim} T { op2 = e } ==> let err = /\a -> recConErr a T (recConErr t1 "M.lhs/230/op1") e (recConErr t1 "M.lhs/230/op3") \end{verbatim} @recConErr@ then converts its arugment string into a proper message before printing it as \begin{verbatim} M.lhs, line 230: missing field op1 was evaluated \end{verbatim} We also handle @C{}@ as valid construction syntax for an unlabelled constructor @C@, setting all of @C@'s fields to bottom. \begin{code} dsExpr (RecordCon (L _ data_con_id) con_expr rbinds) = do con_expr' <- dsExpr con_expr let (arg_tys, _) = tcSplitFunTys (exprType con_expr') -- A newtype in the corner should be opaque; -- hence TcType.tcSplitFunTys mk_arg (arg_ty, lbl) -- Selector id has the field label as its name = case findField (rec_flds rbinds) lbl of (rhs:rhss) -> -- ASSERT( null rhss ) dsLExprWithLoc rhs [] -> mkErrorAppDs rEC_CON_ERROR_ID arg_ty (ppr lbl) unlabelled_bottom arg_ty = mkErrorAppDs rEC_CON_ERROR_ID arg_ty empty labels = dataConFieldLabels (idDataCon data_con_id) -- The data_con_id is guaranteed to be the wrapper id of the constructor con_args <- if null labels then mapM unlabelled_bottom arg_tys else mapM mk_arg (zipEqual "dsExpr:RecordCon" arg_tys labels) return (mkApps con_expr' con_args) \end{code} Record update is a little harder. Suppose we have the decl: \begin{verbatim} data T = T1 {op1, op2, op3 :: Int} | T2 {op4, op2 :: Int} | T3 \end{verbatim} Then we translate as follows: \begin{verbatim} r { op2 = e } ===> let op2 = e in case r of T1 op1 _ op3 -> T1 op1 op2 op3 T2 op4 _ -> T2 op4 op2 other -> recUpdError "M.lhs/230" \end{verbatim} It's important that we use the constructor Ids for @T1@, @T2@ etc on the RHSs, and do not generate a Core constructor application directly, because the constructor might do some argument-evaluation first; and may have to throw away some dictionaries. Note [Update for GADTs] ~~~~~~~~~~~~~~~~~~~~~~~ Consider data T a b where T1 { f1 :: a } :: T a Int Then the wrapper function for T1 has type $WT1 :: a -> T a Int But if x::T a b, then x { f1 = v } :: T a b (not T a Int!) So we need to cast (T a Int) to (T a b). Sigh. \begin{code} dsExpr expr@(RecordUpd record_expr (HsRecFields { rec_flds = fields }) cons_to_upd in_inst_tys out_inst_tys) | null fields = dsLExprWithLoc record_expr | otherwise = -- ASSERT2( notNull cons_to_upd, ppr expr ) do { record_expr' <- dsLExprWithLoc record_expr ; field_binds' <- mapM ds_field fields ; let upd_fld_env :: NameEnv Id -- Maps field name to the LocalId of the field binding upd_fld_env = mkNameEnv [(f,l) | (f,l,_) <- field_binds'] -- It's important to generate the match with matchWrapper, -- and the right hand sides with applications of the wrapper Id -- so that everything works when we are doing fancy unboxing on the -- constructor aguments. ; alts <- mapM (mk_alt upd_fld_env) cons_to_upd ; ([discrim_var], matching_code) <- matchWrapper RecUpd (MatchGroup alts in_out_ty) ; return (add_field_binds field_binds' $ bindNonRec discrim_var record_expr' matching_code) } where ds_field :: HsRecField Id (LHsExpr Id) -> DsM (Name, Id, CoreExpr) -- Clone the Id in the HsRecField, because its Name is that -- of the record selector, and we must not make that a lcoal binder -- else we shadow other uses of the record selector -- Hence 'lcl_id'. Cf Trac #2735 ds_field rec_field = do { rhs <- dsLExprWithLoc (hsRecFieldArg rec_field) ; let fld_id = unLoc (hsRecFieldId rec_field) ; lcl_id <- newSysLocalDs (idType fld_id) ; return (idName fld_id, lcl_id, rhs) } add_field_binds [] expr = expr add_field_binds ((_,b,r):bs) expr = bindNonRec b r (add_field_binds bs expr) -- Awkwardly, for families, the match goes -- from instance type to family type tycon = dataConTyCon (head cons_to_upd) in_ty = mkTyConApp tycon in_inst_tys in_out_ty = mkFunTy in_ty (mkFamilyTyConApp tycon out_inst_tys) mk_alt upd_fld_env con = do { let (univ_tvs, ex_tvs, eq_spec, theta, arg_tys, _) = dataConFullSig con subst = mkTopTvSubst (univ_tvs `zip` in_inst_tys) -- I'm not bothering to clone the ex_tvs ; eqs_vars <- mapM newPredVarDs (substTheta subst (eqSpecPreds eq_spec)) ; theta_vars <- mapM newPredVarDs (substTheta subst theta) ; arg_ids <- newSysLocalsDs (substTys subst arg_tys) ; let val_args = zipWithEqual "dsExpr:RecordUpd" mk_val_arg (dataConFieldLabels con) arg_ids mk_val_arg field_name pat_arg_id = nlHsVar (lookupNameEnv upd_fld_env field_name `orElse` pat_arg_id) inst_con = noLoc $ HsWrap wrap (HsVar (dataConWrapId con)) -- Reconstruct with the WrapId so that unpacking happens wrap = mkWpEvVarApps theta_vars <.> mkWpTyApps (mkTyVarTys ex_tvs) <.> mkWpTyApps [ty | (tv, ty) <- univ_tvs `zip` out_inst_tys , not (tv `elemVarEnv` wrap_subst) ] rhs = foldl (\a b -> nlHsApp a b) inst_con val_args -- Tediously wrap the application in a cast -- Note [Update for GADTs] wrap_co = mkTcTyConAppCo tycon [ lookup tv ty | (tv,ty) <- univ_tvs `zip` out_inst_tys ] lookup univ_tv ty = case lookupVarEnv wrap_subst univ_tv of Just co' -> co' Nothing -> mkTcReflCo ty wrap_subst = mkVarEnv [ (tv, mkTcSymCo (mkTcCoVarCo eq_var)) | ((tv,_),eq_var) <- eq_spec `zip` eqs_vars ] pat = noLoc $ ConPatOut { pat_con = noLoc con, pat_tvs = ex_tvs , pat_dicts = eqs_vars ++ theta_vars , pat_binds = emptyTcEvBinds , pat_args = PrefixCon $ map nlVarPat arg_ids , pat_ty = in_ty } ; let wrapped_rhs | null eq_spec = rhs | otherwise = mkLHsWrap (WpCast wrap_co) rhs ; return (mkSimpleMatch [pat] wrapped_rhs) } \end{code} Here is where we desugar the Template Haskell brackets and escapes \begin{code} -- Template Haskell stuff -- #ifdef GHCI -- dsExpr (HsBracketOut x ps) = dsBracket x ps -- #else dsExpr (HsBracketOut _ _) = panic "dsExpr HsBracketOut" -- #endif dsExpr (HsSpliceE s) = pprPanic "dsExpr:splice" (ppr s) -- Arrow notation extension dsExpr (HsProc pat cmd) = dsProcExpr pat cmd \end{code} Hpc Support \begin{code} dsExpr (HsTick tickish e) = do e' <- dsLExprWithLoc e return (Tick tickish e') -- There is a problem here. The then and else branches -- have no free variables, so they are open to lifting. -- We need someway of stopping this. -- This will make no difference to binary coverage -- (did you go here: YES or NO), but will effect accurate -- tick counting. dsExpr (HsBinTick ixT ixF e) = do e2 <- dsLExprWithLoc e do { -- ASSERT(exprType e2 `eqType` boolTy) mkBinaryTickBox ixT ixF e2 } \end{code} \begin{code} -- HsSyn constructs that just shouldn't be here: dsExpr (ExprWithTySig {}) = panic "dsExpr:ExprWithTySig" dsExpr (HsBracket {}) = panic "dsExpr:HsBracket" dsExpr (HsQuasiQuoteE {}) = panic "dsExpr:HsQuasiQuoteE" dsExpr (HsArrApp {}) = panic "dsExpr:HsArrApp" dsExpr (HsArrForm {}) = panic "dsExpr:HsArrForm" dsExpr (HsTickPragma {}) = panic "dsExpr:HsTickPragma" dsExpr (EWildPat {}) = panic "dsExpr:EWildPat" dsExpr (EAsPat {}) = panic "dsExpr:EAsPat" dsExpr (EViewPat {}) = panic "dsExpr:EViewPat" dsExpr (ELazyPat {}) = panic "dsExpr:ELazyPat" dsExpr (HsType {}) = panic "dsExpr:HsType" dsExpr (HsDo {}) = panic "dsExpr:HsDo" findField :: [HsRecField Id arg] -> Name -> [arg] findField rbinds lbl = [rhs | HsRecField { hsRecFieldId = id, hsRecFieldArg = rhs } <- rbinds , lbl == idName (unLoc id) ] \end{code} %-------------------------------------------------------------------- Note [Desugaring explicit lists] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Explicit lists are desugared in a cleverer way to prevent some fruitless allocations. Essentially, whenever we see a list literal [x_1, ..., x_n] we: 1. Find the tail of the list that can be allocated statically (say [x_k, ..., x_n]) by later stages and ensure we desugar that normally: this makes sure that we don't cause a code size increase by having the cons in that expression fused (see later) and hence being unable to statically allocate any more 2. For the prefix of the list which cannot be allocated statically, say [x_1, ..., x_(k-1)], we turn it into an expression involving build so that if we find any foldrs over it it will fuse away entirely! So in this example we will desugar to: build (\c n -> x_1 `c` x_2 `c` .... `c` foldr c n [x_k, ..., x_n] If fusion fails to occur then build will get inlined and (since we defined a RULE for foldr (:) []) we will get back exactly the normal desugaring for an explicit list. This optimisation can be worth a lot: up to 25% of the total allocation in some nofib programs. Specifically Program Size Allocs Runtime CompTime rewrite +0.0% -26.3% 0.02 -1.8% ansi -0.3% -13.8% 0.00 +0.0% lift +0.0% -8.7% 0.00 -2.3% Of course, if rules aren't turned on then there is pretty much no point doing this fancy stuff, and it may even be harmful. =======> Note by SLPJ Dec 08. I'm unconvinced that we should *ever* generate a build for an explicit list. See the comments in GHC.Base about the foldr/cons rule, which points out that (foldr k z [a,b,c]) may generate *much* less code than (a `k` b `k` c `k` z). Furthermore generating builds messes up the LHS of RULES. Example: the foldr/single rule in GHC.Base foldr k z [x] = ... We do not want to generate a build invocation on the LHS of this RULE! We fix this by disabling rules in rule LHSs, and testing that flag here; see Note [Desugaring RULE left hand sides] in Desugar To test this I've added a (static) flag -fsimple-list-literals, which makes all list literals be generated via the simple route. \begin{code} dsExplicitList :: PostTcType -> [LHsExpr Id] -> DsM CoreExpr -- See Note [Desugaring explicit lists] dsExplicitList elt_ty xs = do { dflags <- getDynFlags ; xs' <- mapM dsLExprWithLoc xs ; let (dynamic_prefix, static_suffix) = spanTail is_static xs' ; if opt_SimpleListLiterals -- -fsimple-list-literals || not (dopt Opt_EnableRewriteRules dflags) -- Rewrite rules off -- Don't generate a build if there are no rules to eliminate it! -- See Note [Desugaring RULE left hand sides] in Desugar || null dynamic_prefix -- Avoid build (\c n. foldr c n xs)! then return $ mkListExpr elt_ty xs' else mkBuildExpr elt_ty (mkSplitExplicitList dynamic_prefix static_suffix) } where is_static :: CoreExpr -> Bool is_static e = all is_static_var (varSetElems (exprFreeVars e)) is_static_var :: Var -> Bool is_static_var v | isId v = isExternalName (idName v) -- Top-level things are given external names | otherwise = False -- Type variables mkSplitExplicitList prefix suffix (c, _) (n, n_ty) = do { let suffix' = mkListExpr elt_ty suffix ; folded_suffix <- mkFoldrExpr elt_ty n_ty (Var c) (Var n) suffix' ; return (foldr (App . App (Var c)) folded_suffix prefix) } spanTail :: (a -> Bool) -> [a] -> ([a], [a]) spanTail f xs = (reverse rejected, reverse satisfying) where (satisfying, rejected) = span f $ reverse xs \end{code} Desugar 'do' and 'mdo' expressions (NOT list comprehensions, they're handled in DsListComp). Basically does the translation given in the Haskell 98 report: \begin{code} dsDo :: [LStmt Id] -> DsM CoreExpr dsDo stmts = goL stmts where goL [] = panic "dsDo" goL (L loc stmt:lstmts) = putSrcSpanDs loc (go loc stmt lstmts) go _ (LastStmt body _) stmts = -- ASSERT( null stmts ) dsLExprWithLoc body -- The 'return' op isn't used for 'do' expressions go _ (ExprStmt rhs then_expr _ _) stmts = do { rhs2 <- dsLExprWithLoc rhs ; warnDiscardedDoBindings rhs (exprType rhs2) ; then_expr2 <- dsExpr then_expr ; rest <- goL stmts ; return (mkApps then_expr2 [rhs2, rest]) } go _ (LetStmt binds) stmts = do { rest <- goL stmts ; dsLocalBinds binds rest } go _ (BindStmt pat rhs bind_op fail_op) stmts = do { body <- goL stmts ; rhs' <- dsLExprWithLoc rhs ; bind_op' <- dsExpr bind_op ; var <- selectSimpleMatchVarL pat ; let bind_ty = exprType bind_op' -- rhs -> (pat -> res1) -> res2 res1_ty = funResultTy (funArgTy (funResultTy bind_ty)) ; match <- matchSinglePat (Var var) (StmtCtxt DoExpr) pat res1_ty (cantFailMatchResult body) ; match_code <- handle_failure pat match fail_op ; return (mkApps bind_op' [rhs', Lam var match_code]) } go loc (RecStmt { recS_stmts = rec_stmts, recS_later_ids = later_ids , recS_rec_ids = rec_ids, recS_ret_fn = return_op , recS_mfix_fn = mfix_op, recS_bind_fn = bind_op , recS_rec_rets = rec_rets, recS_ret_ty = body_ty }) stmts = -- ASSERT( length rec_ids > 0 ) goL (new_bind_stmt : stmts) where new_bind_stmt = L loc $ BindStmt (mkBigLHsPatTup later_pats) mfix_app bind_op noSyntaxExpr -- Tuple cannot fail tup_ids = rec_ids ++ filterOut (`elem` rec_ids) later_ids tup_ty = mkBigCoreTupTy (map idType tup_ids) -- Deals with singleton case rec_tup_pats = map nlVarPat tup_ids later_pats = rec_tup_pats rets = map noLoc rec_rets mfix_app = nlHsApp (noLoc mfix_op) mfix_arg mfix_arg = noLoc $ HsLam (MatchGroup [mkSimpleMatch [mfix_pat] body] (mkFunTy tup_ty body_ty)) mfix_pat = noLoc $ LazyPat $ mkBigLHsPatTup rec_tup_pats body = noLoc $ HsDo DoExpr (rec_stmts ++ [ret_stmt]) body_ty ret_app = nlHsApp (noLoc return_op) (mkBigLHsTup rets) ret_stmt = noLoc $ mkLastStmt ret_app -- This LastStmt will be desugared with dsDo, -- which ignores the return_op in the LastStmt, -- so we must apply the return_op explicitly go _ (ParStmt {}) _ = panic "dsDo ParStmt" go _ (TransStmt {}) _ = panic "dsDo TransStmt" handle_failure :: LPat Id -> MatchResult -> SyntaxExpr Id -> DsM CoreExpr -- In a do expression, pattern-match failure just calls -- the monadic 'fail' rather than throwing an exception handle_failure pat match fail_op | matchCanFail match = do { fail_op' <- dsExpr fail_op ; dflags <- getDynFlags ; fail_msg <- mkStringExpr (mk_fail_msg dflags pat) ; extractMatchResult match (App fail_op' fail_msg) } | otherwise = extractMatchResult match (error "It can't fail") mk_fail_msg :: DynFlags -> Located e -> String mk_fail_msg dflags pat = "Pattern match failure in do expression at " ++ showPpr dflags (getLoc pat) \end{code} %************************************************************************ %* * Warning about identities %* * %************************************************************************ Warn about functions like toInteger, fromIntegral, that convert between one type and another when the to- and from- types are the same. Then it's probably (albeit not definitely) the identity \begin{code} warnAboutIdentities :: CoreExpr -> CoreExpr -> DsM () warnAboutIdentities (Var v) wrapped_fun | idName v `elem` conversionNames , let fun_ty = exprType wrapped_fun , Just (arg_ty, res_ty) <- splitFunTy_maybe fun_ty , arg_ty `eqType` res_ty -- So we are converting ty -> ty = warnDs (vcat [ ptext (sLit "Call of") <+> ppr v <+> dcolon <+> ppr fun_ty , nest 2 $ ptext (sLit "can probably be omitted") , parens (ptext (sLit "Use -fno-warn-identities to suppress this messsage)")) ]) warnAboutIdentities _ _ = return () conversionNames :: [Name] conversionNames = [ toIntegerName, toRationalName , fromIntegralName, realToFracName ] -- We can't easily add fromIntegerName, fromRationalName, -- becuase they are generated by literals \end{code} %************************************************************************ %* * \subsection{Errors and contexts} %* * %************************************************************************ \begin{code} -- Warn about certain types of values discarded in monadic bindings (#3263) warnDiscardedDoBindings :: LHsExpr Id -> Type -> DsM () warnDiscardedDoBindings rhs rhs_ty | Just (m_ty, elt_ty) <- tcSplitAppTy_maybe rhs_ty = do { -- Warn about discarding non-() things in 'monadic' binding ; warn_unused <- woptM Opt_WarnUnusedDoBind ; if warn_unused && not (isUnitTy elt_ty) then warnDs (unusedMonadBind rhs elt_ty) else -- Warn about discarding m a things in 'monadic' binding of the same type, -- but only if we didn't already warn due to Opt_WarnUnusedDoBind do { warn_wrong <- woptM Opt_WarnWrongDoBind ; case tcSplitAppTy_maybe elt_ty of Just (elt_m_ty, _) | warn_wrong, m_ty `eqType` elt_m_ty -> warnDs (wrongMonadBind rhs elt_ty) _ -> return () } } | otherwise -- RHS does have type of form (m ty), which is wierd = return () -- but at lesat this warning is irrelevant unusedMonadBind :: LHsExpr Id -> Type -> SDoc unusedMonadBind rhs elt_ty = ptext (sLit "A do-notation statement discarded a result of type") <+> ppr elt_ty <> dot $$ ptext (sLit "Suppress this warning by saying \"_ <- ") <> ppr rhs <> ptext (sLit "\",") $$ ptext (sLit "or by using the flag -fno-warn-unused-do-bind") wrongMonadBind :: LHsExpr Id -> Type -> SDoc wrongMonadBind rhs elt_ty = ptext (sLit "A do-notation statement discarded a result of type") <+> ppr elt_ty <> dot $$ ptext (sLit "Suppress this warning by saying \"_ <- ") <> ppr rhs <> ptext (sLit "\",") $$ ptext (sLit "or by using the flag -fno-warn-wrong-do-bind") \end{code}