{-# LANGUAGE CPP #-} {-# LANGUAGE ConstraintKinds #-} {-# LANGUAGE FlexibleContexts #-} {-# LANGUAGE MultiWayIf #-} {-# LANGUAGE ScopedTypeVariables #-} {-# LANGUAGE TypeFamilies #-} {-# OPTIONS_GHC -Wno-incomplete-record-updates #-} {-# OPTIONS_GHC -Wno-incomplete-uni-patterns #-} {- (c) The GRASP/AQUA Project, Glasgow University, 1992-1998 Renaming of expressions Basically dependency analysis. Handles @Match@, @GRHSs@, @HsExpr@, and @Qualifier@ datatypes. In general, all of these functions return a renamed thing, and a set of free variables. -} module GHC.Rename.Expr ( rnLExpr, rnExpr, rnStmts, AnnoBody ) where #include "HsVersions.h" import GHC.Prelude import GHC.Rename.Bind ( rnLocalBindsAndThen, rnLocalValBindsLHS, rnLocalValBindsRHS , rnMatchGroup, rnGRHS, makeMiniFixityEnv) import GHC.Hs import GHC.Tc.Utils.Env ( isBrackStage ) import GHC.Tc.Utils.Monad import GHC.Unit.Module ( getModule ) import GHC.Rename.Env import GHC.Rename.Fixity import GHC.Rename.Utils ( HsDocContext(..), bindLocalNamesFV, checkDupNames , bindLocalNames , mapMaybeFvRn, mapFvRn , warnUnusedLocalBinds, typeAppErr , checkUnusedRecordWildcard ) import GHC.Rename.Unbound ( reportUnboundName ) import GHC.Rename.Splice ( rnBracket, rnSpliceExpr, checkThLocalName ) import GHC.Rename.HsType import GHC.Rename.Pat import GHC.Driver.Session import GHC.Builtin.Names import GHC.Types.FieldLabel import GHC.Types.Fixity import GHC.Types.Id.Make import GHC.Types.Name import GHC.Types.Name.Set import GHC.Types.Name.Reader import GHC.Types.Unique.Set import GHC.Types.SourceText import GHC.Utils.Misc import GHC.Data.List.SetOps ( removeDups ) import GHC.Utils.Error import GHC.Utils.Panic import GHC.Utils.Outputable as Outputable import GHC.Types.SrcLoc import GHC.Data.FastString import Control.Monad import GHC.Builtin.Types ( nilDataConName ) import qualified GHC.LanguageExtensions as LangExt import Data.List (unzip4, minimumBy) import Data.List.NonEmpty ( NonEmpty(..) ) import Data.Maybe (isJust, isNothing) import Control.Arrow (first) import Data.Ord import Data.Array import qualified Data.List.NonEmpty as NE {- Note [Handling overloaded and rebindable constructs] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ For overloaded constructs (overloaded literals, lists, strings), and rebindable constructs (e.g. if-then-else), our general plan is this, using overloaded labels #foo as an example: * In the RENAMER: transform HsOverLabel "foo" ==> XExpr (HsExpansion (HsOverLabel #foo) (fromLabel `HsAppType` "foo")) We write this more compactly in concrete-syntax form like this #foo ==> fromLabel @"foo" Recall that in (HsExpansion orig expanded), 'orig' is the original term the user wrote, and 'expanded' is the expanded or desugared version to be typechecked. * In the TYPECHECKER: typecheck the expansion, in this case fromLabel @"foo" The typechecker (and desugarer) will never see HsOverLabel In effect, the renamer does a bit of desugaring. Recall GHC.Hs.Expr Note [Rebindable syntax and HsExpansion], which describes the use of HsExpansion. RebindableSyntax: If RebindableSyntax is off we use the built-in 'fromLabel', defined in GHC.Builtin.Names.fromLabelClassOpName If RebindableSyntax if ON, we look up "fromLabel" in the environment to get whichever one is in scope. This is accomplished by lookupSyntaxName, and it applies to all the constructs below. Here are the constructs that we transform in this way. Some are uniform, but several have a little bit of special treatment: * HsIf (if-the-else) if b then e1 else e2 ==> ifThenElse b e1 e2 We do this /only/ if rebindable syntax is on, because the coverage checker looks for HsIf (see GHC.HsToCore.Coverage.addTickHsExpr) That means the typechecker and desugarer need to understand HsIf for the non-rebindable-syntax case. * OverLabel (overloaded labels, #lbl) #lbl ==> fromLabel @"lbl" As ever, we use lookupSyntaxName to look up 'fromLabel' See Note [Overloaded labels] * ExplicitList (explicit lists [a,b,c]) When (and only when) OverloadedLists is on [e1,e2] ==> fromListN 2 [e1,e2] NB: the type checker and desugarer still see ExplicitList, but to them it always means the built-in lists. * SectionL and SectionR (left and right sections) (`op` e) ==> rightSection op e (e `op`) ==> leftSection (op e) where `leftSection` and `rightSection` are levity-polymorphic wired-in Ids. See Note [Left and right sections] * It's a bit painful to transform `OpApp e1 op e2` to a `HsExpansion` form, because the renamer does precedence rearrangement after name resolution. So the renamer leaves an OpApp as an OpApp. The typechecker turns `OpApp` into a use of `HsExpansion` on the fly, in GHC.Tc.Gen.Head.splitHsApps. RebindableSyntax does not affect this. Note [Overloaded labels] ~~~~~~~~~~~~~~~~~~~~~~~~ For overloaded labels, note that we /only/ apply `fromLabel` to the Symbol argument, so the resulting expression has type fromLabel @"foo" :: forall a. IsLabel "foo" a => a Now ordinary Visible Type Application can be used to instantiate the 'a': the user may have written (#foo @Int). Notice that this all works fine in a kind-polymorphic setting (#19154). Suppose we have fromLabel :: forall {k1} {k2} (a:k1). blah Then we want to instantiate those inferred quantifiers k1,k2, before type-applying to "foo", so we get fromLabel @Symbol @blah @"foo" ... And those inferred kind quantifiers will indeed be instantiated when we typecheck the renamed-syntax call (fromLabel @"foo"). -} {- ************************************************************************ * * \subsubsection{Expressions} * * ************************************************************************ -} rnExprs :: [LHsExpr GhcPs] -> RnM ([LHsExpr GhcRn], FreeVars) rnExprs ls = rnExprs' ls emptyUniqSet where rnExprs' [] acc = return ([], acc) rnExprs' (expr:exprs) acc = do { (expr', fvExpr) <- rnLExpr expr -- Now we do a "seq" on the free vars because typically it's small -- or empty, especially in very long lists of constants ; let acc' = acc `plusFV` fvExpr ; (exprs', fvExprs) <- acc' `seq` rnExprs' exprs acc' ; return (expr':exprs', fvExprs) } -- Variables. We look up the variable and return the resulting name. rnLExpr :: LHsExpr GhcPs -> RnM (LHsExpr GhcRn, FreeVars) rnLExpr = wrapLocFstMA rnExpr rnExpr :: HsExpr GhcPs -> RnM (HsExpr GhcRn, FreeVars) finishHsVar :: LocatedA Name -> RnM (HsExpr GhcRn, FreeVars) -- Separated from rnExpr because it's also used -- when renaming infix expressions finishHsVar (L l name) = do { this_mod <- getModule ; when (nameIsLocalOrFrom this_mod name) $ checkThLocalName name ; return (HsVar noExtField (L (la2na l) name), unitFV name) } rnUnboundVar :: RdrName -> RnM (HsExpr GhcRn, FreeVars) rnUnboundVar v = if isUnqual v then -- Treat this as a "hole" -- Do not fail right now; instead, return HsUnboundVar -- and let the type checker report the error return (HsUnboundVar noExtField (rdrNameOcc v), emptyFVs) else -- Fail immediately (qualified name) do { n <- reportUnboundName v ; return (HsVar noExtField (noLocA n), emptyFVs) } rnExpr (HsVar _ (L l v)) = do { dflags <- getDynFlags ; let dup_fields_ok = xopt_DuplicateRecordFields dflags ; mb_name <- lookupExprOccRn dup_fields_ok v ; case mb_name of { Nothing -> rnUnboundVar v ; Just (UnambiguousGre (NormalGreName name)) | name == nilDataConName -- Treat [] as an ExplicitList, so that -- OverloadedLists works correctly -- Note [Empty lists] in GHC.Hs.Expr , xopt LangExt.OverloadedLists dflags -> rnExpr (ExplicitList noAnn []) | otherwise -> finishHsVar (L (na2la l) name) ; Just (UnambiguousGre (FieldGreName fl)) -> let sel_name = flSelector fl in return ( HsRecFld noExtField (Unambiguous sel_name (L l v) ), unitFV sel_name) ; Just AmbiguousFields -> return ( HsRecFld noExtField (Ambiguous noExtField (L l v) ), emptyFVs) } } rnExpr (HsIPVar x v) = return (HsIPVar x v, emptyFVs) rnExpr (HsUnboundVar _ v) = return (HsUnboundVar noExtField v, emptyFVs) -- HsOverLabel: see Note [Handling overloaded and rebindable constructs] rnExpr (HsOverLabel _ v) = do { (from_label, fvs) <- lookupSyntaxName fromLabelClassOpName ; return ( mkExpandedExpr (HsOverLabel noAnn v) $ HsAppType noExtField (genLHsVar from_label) hs_ty_arg , fvs ) } where hs_ty_arg = mkEmptyWildCardBndrs $ wrapGenSpan $ HsTyLit noExtField (HsStrTy NoSourceText v) rnExpr (HsLit x lit@(HsString src s)) = do { opt_OverloadedStrings <- xoptM LangExt.OverloadedStrings ; if opt_OverloadedStrings then rnExpr (HsOverLit x (mkHsIsString src s)) else do { ; rnLit lit ; return (HsLit x (convertLit lit), emptyFVs) } } rnExpr (HsLit x lit) = do { rnLit lit ; return (HsLit x(convertLit lit), emptyFVs) } rnExpr (HsOverLit x lit) = do { ((lit', mb_neg), fvs) <- rnOverLit lit -- See Note [Negative zero] ; case mb_neg of Nothing -> return (HsOverLit x lit', fvs) Just neg -> return (HsApp noComments (noLocA neg) (noLocA (HsOverLit x lit')) , fvs ) } rnExpr (HsApp x fun arg) = do { (fun',fvFun) <- rnLExpr fun ; (arg',fvArg) <- rnLExpr arg ; return (HsApp x fun' arg', fvFun `plusFV` fvArg) } rnExpr (HsAppType _ fun arg) = do { type_app <- xoptM LangExt.TypeApplications ; unless type_app $ addErr $ typeAppErr "type" $ hswc_body arg ; (fun',fvFun) <- rnLExpr fun ; (arg',fvArg) <- rnHsWcType HsTypeCtx arg ; return (HsAppType NoExtField fun' arg', fvFun `plusFV` fvArg) } rnExpr (OpApp _ e1 op e2) = do { (e1', fv_e1) <- rnLExpr e1 ; (e2', fv_e2) <- rnLExpr e2 ; (op', fv_op) <- rnLExpr op -- Deal with fixity -- When renaming code synthesised from "deriving" declarations -- we used to avoid fixity stuff, but we can't easily tell any -- more, so I've removed the test. Adding HsPars in GHC.Tc.Deriv.Generate -- should prevent bad things happening. ; fixity <- case op' of L _ (HsVar _ (L _ n)) -> lookupFixityRn n L _ (HsRecFld _ f) -> lookupFieldFixityRn f _ -> return (Fixity NoSourceText minPrecedence InfixL) -- c.f. lookupFixity for unbound ; lexical_negation <- xoptM LangExt.LexicalNegation ; let negation_handling | lexical_negation = KeepNegationIntact | otherwise = ReassociateNegation ; final_e <- mkOpAppRn negation_handling e1' op' fixity e2' ; return (final_e, fv_e1 `plusFV` fv_op `plusFV` fv_e2) } rnExpr (NegApp _ e _) = do { (e', fv_e) <- rnLExpr e ; (neg_name, fv_neg) <- lookupSyntax negateName ; final_e <- mkNegAppRn e' neg_name ; return (final_e, fv_e `plusFV` fv_neg) } ------------------------------------------ -- Record dot syntax rnExpr (HsGetField _ e f) = do { (getField, fv_getField) <- lookupSyntaxName getFieldName ; (e, fv_e) <- rnLExpr e ; let f' = rnHsFieldLabel f ; return ( mkExpandedExpr (HsGetField noExtField e f') (mkGetField getField e (fmap (unLoc . hflLabel) f')) , fv_e `plusFV` fv_getField ) } rnExpr (HsProjection _ fs) = do { (getField, fv_getField) <- lookupSyntaxName getFieldName ; circ <- lookupOccRn compose_RDR ; let fs' = fmap rnHsFieldLabel fs ; return ( mkExpandedExpr (HsProjection noExtField fs') (mkProjection getField circ (fmap (fmap (unLoc . hflLabel)) fs')) , unitFV circ `plusFV` fv_getField) } ------------------------------------------ -- Template Haskell extensions rnExpr e@(HsBracket _ br_body) = rnBracket e br_body rnExpr (HsSpliceE _ splice) = rnSpliceExpr splice --------------------------------------------- -- Sections -- See Note [Parsing sections] in GHC.Parser rnExpr (HsPar x (L loc (section@(SectionL {})))) = do { (section', fvs) <- rnSection section ; return (HsPar x (L loc section'), fvs) } rnExpr (HsPar x (L loc (section@(SectionR {})))) = do { (section', fvs) <- rnSection section ; return (HsPar x (L loc section'), fvs) } rnExpr (HsPar x e) = do { (e', fvs_e) <- rnLExpr e ; return (HsPar x e', fvs_e) } rnExpr expr@(SectionL {}) = do { addErr (sectionErr expr); rnSection expr } rnExpr expr@(SectionR {}) = do { addErr (sectionErr expr); rnSection expr } --------------------------------------------- rnExpr (HsPragE x prag expr) = do { (expr', fvs_expr) <- rnLExpr expr ; return (HsPragE x (rn_prag prag) expr', fvs_expr) } where rn_prag :: HsPragE GhcPs -> HsPragE GhcRn rn_prag (HsPragSCC x1 src ann) = HsPragSCC x1 src ann rnExpr (HsLam x matches) = do { (matches', fvMatch) <- rnMatchGroup LambdaExpr rnLExpr matches ; return (HsLam x matches', fvMatch) } rnExpr (HsLamCase x matches) = do { (matches', fvs_ms) <- rnMatchGroup CaseAlt rnLExpr matches ; return (HsLamCase x matches', fvs_ms) } rnExpr (HsCase _ expr matches) = do { (new_expr, e_fvs) <- rnLExpr expr ; (new_matches, ms_fvs) <- rnMatchGroup CaseAlt rnLExpr matches ; return (HsCase noExtField new_expr new_matches, e_fvs `plusFV` ms_fvs) } rnExpr (HsLet _ binds expr) = rnLocalBindsAndThen binds $ \binds' _ -> do { (expr',fvExpr) <- rnLExpr expr ; return (HsLet noExtField binds' expr', fvExpr) } rnExpr (HsDo _ do_or_lc (L l stmts)) = do { ((stmts', _), fvs) <- rnStmtsWithPostProcessing do_or_lc rnExpr postProcessStmtsForApplicativeDo stmts (\ _ -> return ((), emptyFVs)) ; return ( HsDo noExtField do_or_lc (L l stmts'), fvs ) } -- ExplicitList: see Note [Handling overloaded and rebindable constructs] rnExpr (ExplicitList _ exps) = do { (exps', fvs) <- rnExprs exps ; opt_OverloadedLists <- xoptM LangExt.OverloadedLists ; if not opt_OverloadedLists then return (ExplicitList noExtField exps', fvs) else do { (from_list_n_name, fvs') <- lookupSyntaxName fromListNName ; let rn_list = ExplicitList noExtField exps' lit_n = mkIntegralLit (length exps) hs_lit = wrapGenSpan (HsLit noAnn (HsInt noExtField lit_n)) exp_list = genHsApps from_list_n_name [hs_lit, wrapGenSpan rn_list] ; return ( mkExpandedExpr rn_list exp_list , fvs `plusFV` fvs') } } rnExpr (ExplicitTuple _ tup_args boxity) = do { checkTupleSection tup_args ; (tup_args', fvs) <- mapAndUnzipM rnTupArg tup_args ; return (ExplicitTuple noExtField tup_args' boxity, plusFVs fvs) } where rnTupArg (Present x e) = do { (e',fvs) <- rnLExpr e ; return (Present x e', fvs) } rnTupArg (Missing _) = return (Missing noExtField, emptyFVs) rnExpr (ExplicitSum _ alt arity expr) = do { (expr', fvs) <- rnLExpr expr ; return (ExplicitSum noExtField alt arity expr', fvs) } rnExpr (RecordCon { rcon_con = con_id , rcon_flds = rec_binds@(HsRecFields { rec_dotdot = dd }) }) = do { con_lname@(L _ con_name) <- lookupLocatedOccRn con_id ; (flds, fvs) <- rnHsRecFields (HsRecFieldCon con_name) mk_hs_var rec_binds ; (flds', fvss) <- mapAndUnzipM rn_field flds ; let rec_binds' = HsRecFields { rec_flds = flds', rec_dotdot = dd } ; return (RecordCon { rcon_ext = noExtField , rcon_con = con_lname, rcon_flds = rec_binds' } , fvs `plusFV` plusFVs fvss `addOneFV` con_name) } where mk_hs_var l n = HsVar noExtField (L (noAnnSrcSpan l) n) rn_field (L l fld) = do { (arg', fvs) <- rnLExpr (hsRecFieldArg fld) ; return (L l (fld { hsRecFieldArg = arg' }), fvs) } rnExpr (RecordUpd { rupd_expr = expr, rupd_flds = rbinds }) = case rbinds of Left flds -> -- 'OverloadedRecordUpdate' is not in effect. Regular record update. do { ; (e, fv_e) <- rnLExpr expr ; (rs, fv_rs) <- rnHsRecUpdFields flds ; return ( RecordUpd noExtField e (Left rs), fv_e `plusFV` fv_rs ) } Right flds -> -- 'OverloadedRecordUpdate' is in effect. Record dot update desugaring. do { ; unlessXOptM LangExt.RebindableSyntax $ addErr $ text "RebindableSyntax is required if OverloadedRecordUpdate is enabled." ; let punnedFields = [fld | (L _ fld) <- flds, hsRecPun fld] ; punsEnabled <-xoptM LangExt.RecordPuns ; unless (null punnedFields || punsEnabled) $ addErr $ text "For this to work enable NamedFieldPuns." ; (getField, fv_getField) <- lookupSyntaxName getFieldName ; (setField, fv_setField) <- lookupSyntaxName setFieldName ; (e, fv_e) <- rnLExpr expr ; (us, fv_us) <- rnHsUpdProjs flds ; return ( mkExpandedExpr (RecordUpd noExtField e (Right us)) (mkRecordDotUpd getField setField e us) , plusFVs [fv_getField, fv_setField, fv_e, fv_us] ) } rnExpr (ExprWithTySig _ expr pty) = do { (pty', fvTy) <- rnHsSigWcType ExprWithTySigCtx pty ; (expr', fvExpr) <- bindSigTyVarsFV (hsWcScopedTvs pty') $ rnLExpr expr ; return (ExprWithTySig noExtField expr' pty', fvExpr `plusFV` fvTy) } -- HsIf: see Note [Handling overloaded and rebindable constructs] -- Because of the coverage checker it is most convenient /not/ to -- expand HsIf; unless we are in rebindable syntax. rnExpr (HsIf _ p b1 b2) = do { (p', fvP) <- rnLExpr p ; (b1', fvB1) <- rnLExpr b1 ; (b2', fvB2) <- rnLExpr b2 ; let fvs_if = plusFVs [fvP, fvB1, fvB2] rn_if = HsIf noExtField p' b1' b2' -- Deal with rebindable syntax -- See Note [Handling overloaded and rebindable constructs] ; mb_ite <- lookupIfThenElse ; case mb_ite of Nothing -- Non rebindable-syntax case -> return (rn_if, fvs_if) Just ite_name -- Rebindable-syntax case -> do { let ds_if = genHsApps ite_name [p', b1', b2'] fvs = plusFVs [fvs_if, unitFV ite_name] ; return (mkExpandedExpr rn_if ds_if, fvs) } } rnExpr (HsMultiIf _ alts) = do { (alts', fvs) <- mapFvRn (rnGRHS IfAlt rnLExpr) alts ; return (HsMultiIf noExtField alts', fvs) } rnExpr (ArithSeq _ _ seq) = do { opt_OverloadedLists <- xoptM LangExt.OverloadedLists ; (new_seq, fvs) <- rnArithSeq seq ; if opt_OverloadedLists then do { ; (from_list_name, fvs') <- lookupSyntax fromListName ; return (ArithSeq noExtField (Just from_list_name) new_seq , fvs `plusFV` fvs') } else return (ArithSeq noExtField Nothing new_seq, fvs) } {- ************************************************************************ * * Static values * * ************************************************************************ For the static form we check that it is not used in splices. We also collect the free variables of the term which come from this module. See Note [Grand plan for static forms] in GHC.Iface.Tidy.StaticPtrTable. -} rnExpr e@(HsStatic _ expr) = do -- Normally, you wouldn't be able to construct a static expression without -- first enabling -XStaticPointers in the first place, since that extension -- is what makes the parser treat `static` as a keyword. But this is not a -- sufficient safeguard, as one can construct static expressions by another -- mechanism: Template Haskell (see #14204). To ensure that GHC is -- absolutely prepared to cope with static forms, we check for -- -XStaticPointers here as well. unlessXOptM LangExt.StaticPointers $ addErr $ hang (text "Illegal static expression:" <+> ppr e) 2 (text "Use StaticPointers to enable this extension") (expr',fvExpr) <- rnLExpr expr stage <- getStage case stage of Splice _ -> addErr $ sep [ text "static forms cannot be used in splices:" , nest 2 $ ppr e ] _ -> return () mod <- getModule let fvExpr' = filterNameSet (nameIsLocalOrFrom mod) fvExpr return (HsStatic fvExpr' expr', fvExpr) {- ********************************************************************* * * Arrow notation * * ********************************************************************* -} rnExpr (HsProc x pat body) = newArrowScope $ rnPat (ArrowMatchCtxt ProcExpr) pat $ \ pat' -> do { (body',fvBody) <- rnCmdTop body ; return (HsProc x pat' body', fvBody) } rnExpr other = pprPanic "rnExpr: unexpected expression" (ppr other) -- HsWrap {- ********************************************************************* * * Operator sections * * ********************************************************************* -} rnSection :: HsExpr GhcPs -> RnM (HsExpr GhcRn, FreeVars) -- See Note [Parsing sections] in GHC.Parser -- Also see Note [Handling overloaded and rebindable constructs] rnSection section@(SectionR x op expr) -- See Note [Left and right sections] = do { (op', fvs_op) <- rnLExpr op ; (expr', fvs_expr) <- rnLExpr expr ; checkSectionPrec InfixR section op' expr' ; let rn_section = SectionR x op' expr' ds_section = genHsApps rightSectionName [op',expr'] ; return ( mkExpandedExpr rn_section ds_section , fvs_op `plusFV` fvs_expr) } rnSection section@(SectionL x expr op) -- See Note [Left and right sections] = do { (expr', fvs_expr) <- rnLExpr expr ; (op', fvs_op) <- rnLExpr op ; checkSectionPrec InfixL section op' expr' ; postfix_ops <- xoptM LangExt.PostfixOperators -- Note [Left and right sections] ; let rn_section = SectionL x expr' op' ds_section | postfix_ops = HsApp noAnn op' expr' | otherwise = genHsApps leftSectionName [wrapGenSpan $ HsApp noAnn op' expr'] ; return ( mkExpandedExpr rn_section ds_section , fvs_op `plusFV` fvs_expr) } rnSection other = pprPanic "rnSection" (ppr other) {- Note [Left and right sections] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Dealing with left sections (x *) and right sections (* x) is surprisingly fiddly. We expand like this (`op` e) ==> rightSection op e (e `op`) ==> leftSection (op e) Using an auxiliary function in this way avoids the awkwardness of generating a lambda, esp if `e` is a redex, so we *don't* want to generate `(\x -> op x e)`. See Historical Note [Desugaring operator sections] Here are their definitions: leftSection :: forall r1 r2 n (a:TYPE r1) (b:TYPE r2). (a %n-> b) -> a %n-> b leftSection f x = f x rightSection :: forall r1 r2 r3 (a:TYPE r1) (b:TYPE r2) (c:TYPE r3). (a %n1 -> b %n2-> c) -> b %n2-> a %n1-> c rightSection f y x = f x y Note the wrinkles: * We do /not/ use lookupSyntaxName, which would make left and right section fall under RebindableSyntax. Reason: it would be a user- facing change, and there are some tricky design choices (#19354). Plus, infix operator applications would be trickier to make rebindable, so it'd be inconsistent to do so for sections. TL;DR: we still us the renamer-expansion mechanism for operator sections , but only to eliminate special-purpose code paths in the renamer and desugarer. * leftSection and rightSection must be levity-polymorphic, to allow (+# 4#) and (4# +#) to work. See GHC.Types.Id.Make. Note [Wired-in Ids for rebindable syntax] in * leftSection and rightSection must be multiplicity-polymorphic. (Test linear/should_compile/OldList showed this up.) * Because they are levity-polymorphic, we have to define them as wired-in Ids, with compulsory inlining. See GHC.Types.Id.Make.leftSectionId, rightSectionId. * leftSection is just ($) really; but unlike ($) it is levity polymorphic in the result type, so we can write `(x +#)`, say. * The type of leftSection must have an arrow in its first argument, because (x `ord`) should be rejected, because ord does not take two arguments * It's important that we define leftSection in an eta-expanded way, (i.e. not leftSection f = f), so that (True `undefined`) `seq` () = (leftSection (undefined True) `seq` ()) evaluates to () and not undefined * If PostfixOperators is ON, then we expand a left section like this: (e `op`) ==> op e with no auxiliary function at all. Simple! Historical Note [Desugaring operator sections] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ This Note explains some historical trickiness in desugaring left and right sections. That trickiness has completely disappeared now that we desugar to calls to 'leftSection` and `rightSection`, but I'm leaving it here to remind us how nice the new story is. Desugaring left sections with -XPostfixOperators is straightforward: convert (expr `op`) to (op expr). Without -XPostfixOperators it's a bit more tricky. At first it looks as if we can convert (expr `op`) naively to \x -> op expr x But no! expr might be a redex, and we can lose laziness badly this way. Consider map (expr `op`) xs for example. If expr were a redex then eta-expanding naively would result in multiple evaluations where the user might only have expected one. So we convert instead to let y = expr in \x -> op y x Also, note that we must do this for both right and (perhaps surprisingly) left sections. Why are left sections necessary? Consider the program (found in #18151), seq (True `undefined`) () according to the Haskell Report this should reduce to () (as it specifies desugaring via eta expansion). However, if we fail to eta expand we will rather bottom. Consequently, we must eta expand even in the case of a left section. If `expr` is actually just a variable, say, then the simplifier will inline `y`, eliminating the redundant `let`. Note that this works even in the case that `expr` is unlifted. In this case bindNonRec will automatically do the right thing, giving us: case expr of y -> (\x -> op y x) See #18151. -} {- ************************************************************************ * * Field Labels * * ************************************************************************ -} rnHsFieldLabel :: Located (HsFieldLabel GhcPs) -> Located (HsFieldLabel GhcRn) rnHsFieldLabel (L l (HsFieldLabel x label)) = L l (HsFieldLabel x label) rnFieldLabelStrings :: FieldLabelStrings GhcPs -> FieldLabelStrings GhcRn rnFieldLabelStrings (FieldLabelStrings fls) = FieldLabelStrings (map rnHsFieldLabel fls) {- ************************************************************************ * * Arrow commands * * ************************************************************************ -} rnCmdArgs :: [LHsCmdTop GhcPs] -> RnM ([LHsCmdTop GhcRn], FreeVars) rnCmdArgs [] = return ([], emptyFVs) rnCmdArgs (arg:args) = do { (arg',fvArg) <- rnCmdTop arg ; (args',fvArgs) <- rnCmdArgs args ; return (arg':args', fvArg `plusFV` fvArgs) } rnCmdTop :: LHsCmdTop GhcPs -> RnM (LHsCmdTop GhcRn, FreeVars) rnCmdTop = wrapLocFstM rnCmdTop' where rnCmdTop' :: HsCmdTop GhcPs -> RnM (HsCmdTop GhcRn, FreeVars) rnCmdTop' (HsCmdTop _ cmd) = do { (cmd', fvCmd) <- rnLCmd cmd ; let cmd_names = [arrAName, composeAName, firstAName] ++ nameSetElemsStable (methodNamesCmd (unLoc cmd')) -- Generate the rebindable syntax for the monad ; (cmd_names', cmd_fvs) <- lookupSyntaxNames cmd_names ; return (HsCmdTop (cmd_names `zip` cmd_names') cmd', fvCmd `plusFV` cmd_fvs) } rnLCmd :: LHsCmd GhcPs -> RnM (LHsCmd GhcRn, FreeVars) rnLCmd = wrapLocFstMA rnCmd rnCmd :: HsCmd GhcPs -> RnM (HsCmd GhcRn, FreeVars) rnCmd (HsCmdArrApp _ arrow arg ho rtl) = do { (arrow',fvArrow) <- select_arrow_scope (rnLExpr arrow) ; (arg',fvArg) <- rnLExpr arg ; return (HsCmdArrApp noExtField arrow' arg' ho rtl, fvArrow `plusFV` fvArg) } where select_arrow_scope tc = case ho of HsHigherOrderApp -> tc HsFirstOrderApp -> escapeArrowScope tc -- See Note [Escaping the arrow scope] in GHC.Tc.Types -- Before renaming 'arrow', use the environment of the enclosing -- proc for the (-<) case. -- Local bindings, inside the enclosing proc, are not in scope -- inside 'arrow'. In the higher-order case (-<<), they are. -- infix form rnCmd (HsCmdArrForm _ op _ (Just _) [arg1, arg2]) = do { (op',fv_op) <- escapeArrowScope (rnLExpr op) ; let L _ (HsVar _ (L _ op_name)) = op' ; (arg1',fv_arg1) <- rnCmdTop arg1 ; (arg2',fv_arg2) <- rnCmdTop arg2 -- Deal with fixity ; fixity <- lookupFixityRn op_name ; final_e <- mkOpFormRn arg1' op' fixity arg2' ; return (final_e, fv_arg1 `plusFV` fv_op `plusFV` fv_arg2) } rnCmd (HsCmdArrForm _ op f fixity cmds) = do { (op',fvOp) <- escapeArrowScope (rnLExpr op) ; (cmds',fvCmds) <- rnCmdArgs cmds ; return ( HsCmdArrForm noExtField op' f fixity cmds' , fvOp `plusFV` fvCmds) } rnCmd (HsCmdApp x fun arg) = do { (fun',fvFun) <- rnLCmd fun ; (arg',fvArg) <- rnLExpr arg ; return (HsCmdApp x fun' arg', fvFun `plusFV` fvArg) } rnCmd (HsCmdLam _ matches) = do { (matches', fvMatch) <- rnMatchGroup (ArrowMatchCtxt KappaExpr) rnLCmd matches ; return (HsCmdLam noExtField matches', fvMatch) } rnCmd (HsCmdPar x e) = do { (e', fvs_e) <- rnLCmd e ; return (HsCmdPar x e', fvs_e) } rnCmd (HsCmdCase _ expr matches) = do { (new_expr, e_fvs) <- rnLExpr expr ; (new_matches, ms_fvs) <- rnMatchGroup (ArrowMatchCtxt ArrowCaseAlt) rnLCmd matches ; return (HsCmdCase noExtField new_expr new_matches , e_fvs `plusFV` ms_fvs) } rnCmd (HsCmdLamCase x matches) = do { (new_matches, ms_fvs) <- rnMatchGroup (ArrowMatchCtxt ArrowCaseAlt) rnLCmd matches ; return (HsCmdLamCase x new_matches, ms_fvs) } rnCmd (HsCmdIf _ _ p b1 b2) = do { (p', fvP) <- rnLExpr p ; (b1', fvB1) <- rnLCmd b1 ; (b2', fvB2) <- rnLCmd b2 ; mb_ite <- lookupIfThenElse ; let (ite, fvITE) = case mb_ite of Just ite_name -> (mkRnSyntaxExpr ite_name, unitFV ite_name) Nothing -> (NoSyntaxExprRn, emptyFVs) ; return (HsCmdIf noExtField ite p' b1' b2', plusFVs [fvITE, fvP, fvB1, fvB2])} rnCmd (HsCmdLet _ binds cmd) = rnLocalBindsAndThen binds $ \ binds' _ -> do { (cmd',fvExpr) <- rnLCmd cmd ; return (HsCmdLet noExtField binds' cmd', fvExpr) } rnCmd (HsCmdDo _ (L l stmts)) = do { ((stmts', _), fvs) <- rnStmts ArrowExpr rnCmd stmts (\ _ -> return ((), emptyFVs)) ; return ( HsCmdDo noExtField (L l stmts'), fvs ) } --------------------------------------------------- type CmdNeeds = FreeVars -- Only inhabitants are -- appAName, choiceAName, loopAName -- find what methods the Cmd needs (loop, choice, apply) methodNamesLCmd :: LHsCmd GhcRn -> CmdNeeds methodNamesLCmd = methodNamesCmd . unLoc methodNamesCmd :: HsCmd GhcRn -> CmdNeeds methodNamesCmd (HsCmdArrApp _ _arrow _arg HsFirstOrderApp _rtl) = emptyFVs methodNamesCmd (HsCmdArrApp _ _arrow _arg HsHigherOrderApp _rtl) = unitFV appAName methodNamesCmd (HsCmdArrForm {}) = emptyFVs methodNamesCmd (HsCmdPar _ c) = methodNamesLCmd c methodNamesCmd (HsCmdIf _ _ _ c1 c2) = methodNamesLCmd c1 `plusFV` methodNamesLCmd c2 `addOneFV` choiceAName methodNamesCmd (HsCmdLet _ _ c) = methodNamesLCmd c methodNamesCmd (HsCmdDo _ (L _ stmts)) = methodNamesStmts stmts methodNamesCmd (HsCmdApp _ c _) = methodNamesLCmd c methodNamesCmd (HsCmdLam _ match) = methodNamesMatch match methodNamesCmd (HsCmdCase _ _ matches) = methodNamesMatch matches `addOneFV` choiceAName methodNamesCmd (HsCmdLamCase _ matches) = methodNamesMatch matches `addOneFV` choiceAName --methodNamesCmd _ = emptyFVs -- Other forms can't occur in commands, but it's not convenient -- to error here so we just do what's convenient. -- The type checker will complain later --------------------------------------------------- methodNamesMatch :: MatchGroup GhcRn (LHsCmd GhcRn) -> FreeVars methodNamesMatch (MG { mg_alts = L _ ms }) = plusFVs (map do_one ms) where do_one (L _ (Match { m_grhss = grhss })) = methodNamesGRHSs grhss ------------------------------------------------- -- gaw 2004 methodNamesGRHSs :: GRHSs GhcRn (LHsCmd GhcRn) -> FreeVars methodNamesGRHSs (GRHSs _ grhss _) = plusFVs (map methodNamesGRHS grhss) ------------------------------------------------- methodNamesGRHS :: Located (GRHS GhcRn (LHsCmd GhcRn)) -> CmdNeeds methodNamesGRHS (L _ (GRHS _ _ rhs)) = methodNamesLCmd rhs --------------------------------------------------- methodNamesStmts :: [LStmtLR GhcRn GhcRn (LHsCmd GhcRn)] -> FreeVars methodNamesStmts stmts = plusFVs (map methodNamesLStmt stmts) --------------------------------------------------- methodNamesLStmt :: LStmtLR GhcRn GhcRn (LHsCmd GhcRn) -> FreeVars methodNamesLStmt = methodNamesStmt . unLoc methodNamesStmt :: StmtLR GhcRn GhcRn (LHsCmd GhcRn) -> FreeVars methodNamesStmt (LastStmt _ cmd _ _) = methodNamesLCmd cmd methodNamesStmt (BodyStmt _ cmd _ _) = methodNamesLCmd cmd methodNamesStmt (BindStmt _ _ cmd) = methodNamesLCmd cmd methodNamesStmt (RecStmt { recS_stmts = L _ stmts }) = methodNamesStmts stmts `addOneFV` loopAName methodNamesStmt (LetStmt {}) = emptyFVs methodNamesStmt (ParStmt {}) = emptyFVs methodNamesStmt (TransStmt {}) = emptyFVs methodNamesStmt ApplicativeStmt{} = emptyFVs -- ParStmt and TransStmt can't occur in commands, but it's not -- convenient to error here so we just do what's convenient {- ************************************************************************ * * Arithmetic sequences * * ************************************************************************ -} rnArithSeq :: ArithSeqInfo GhcPs -> RnM (ArithSeqInfo GhcRn, FreeVars) rnArithSeq (From expr) = do { (expr', fvExpr) <- rnLExpr expr ; return (From expr', fvExpr) } rnArithSeq (FromThen expr1 expr2) = do { (expr1', fvExpr1) <- rnLExpr expr1 ; (expr2', fvExpr2) <- rnLExpr expr2 ; return (FromThen expr1' expr2', fvExpr1 `plusFV` fvExpr2) } rnArithSeq (FromTo expr1 expr2) = do { (expr1', fvExpr1) <- rnLExpr expr1 ; (expr2', fvExpr2) <- rnLExpr expr2 ; return (FromTo expr1' expr2', fvExpr1 `plusFV` fvExpr2) } rnArithSeq (FromThenTo expr1 expr2 expr3) = do { (expr1', fvExpr1) <- rnLExpr expr1 ; (expr2', fvExpr2) <- rnLExpr expr2 ; (expr3', fvExpr3) <- rnLExpr expr3 ; return (FromThenTo expr1' expr2' expr3', plusFVs [fvExpr1, fvExpr2, fvExpr3]) } {- ************************************************************************ * * \subsubsection{@Stmt@s: in @do@ expressions} * * ************************************************************************ -} {- Note [Deterministic ApplicativeDo and RecursiveDo desugaring] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Both ApplicativeDo and RecursiveDo need to create tuples not present in the source text. For ApplicativeDo we create: (a,b,c) <- (\c b a -> (a,b,c)) <$> For RecursiveDo we create: mfix (\ ~(a,b,c) -> do ...; return (a',b',c')) The order of the components in those tuples needs to be stable across recompilations, otherwise they can get optimized differently and we end up with incompatible binaries. To get a stable order we use nameSetElemsStable. See Note [Deterministic UniqFM] to learn more about nondeterminism. -} type AnnoBody body = ( Outputable (body GhcPs) , Anno (StmtLR GhcPs GhcPs (LocatedA (body GhcPs))) ~ SrcSpanAnnA , Anno (StmtLR GhcRn GhcPs (LocatedA (body GhcPs))) ~ SrcSpanAnnA , Anno (StmtLR GhcRn GhcRn (LocatedA (body GhcRn))) ~ SrcSpanAnnA ) -- | Rename some Stmts rnStmts :: AnnoBody body => HsStmtContext GhcRn -> (body GhcPs -> RnM (body GhcRn, FreeVars)) -- ^ How to rename the body of each statement (e.g. rnLExpr) -> [LStmt GhcPs (LocatedA (body GhcPs))] -- ^ Statements -> ([Name] -> RnM (thing, FreeVars)) -- ^ if these statements scope over something, this renames it -- and returns the result. -> RnM (([LStmt GhcRn (LocatedA (body GhcRn))], thing), FreeVars) rnStmts ctxt rnBody = rnStmtsWithPostProcessing ctxt rnBody noPostProcessStmts -- | like 'rnStmts' but applies a post-processing step to the renamed Stmts rnStmtsWithPostProcessing :: AnnoBody body => HsStmtContext GhcRn -> (body GhcPs -> RnM (body GhcRn, FreeVars)) -- ^ How to rename the body of each statement (e.g. rnLExpr) -> (HsStmtContext GhcRn -> [(LStmt GhcRn (LocatedA (body GhcRn)), FreeVars)] -> RnM ([LStmt GhcRn (LocatedA (body GhcRn))], FreeVars)) -- ^ postprocess the statements -> [LStmt GhcPs (LocatedA (body GhcPs))] -- ^ Statements -> ([Name] -> RnM (thing, FreeVars)) -- ^ if these statements scope over something, this renames it -- and returns the result. -> RnM (([LStmt GhcRn (LocatedA (body GhcRn))], thing), FreeVars) rnStmtsWithPostProcessing ctxt rnBody ppStmts stmts thing_inside = do { ((stmts', thing), fvs) <- rnStmtsWithFreeVars ctxt rnBody stmts thing_inside ; (pp_stmts, fvs') <- ppStmts ctxt stmts' ; return ((pp_stmts, thing), fvs `plusFV` fvs') } -- | maybe rearrange statements according to the ApplicativeDo transformation postProcessStmtsForApplicativeDo :: HsStmtContext GhcRn -> [(ExprLStmt GhcRn, FreeVars)] -> RnM ([ExprLStmt GhcRn], FreeVars) postProcessStmtsForApplicativeDo ctxt stmts = do { -- rearrange the statements using ApplicativeStmt if -- -XApplicativeDo is on. Also strip out the FreeVars attached -- to each Stmt body. ado_is_on <- xoptM LangExt.ApplicativeDo ; let is_do_expr | DoExpr{} <- ctxt = True | otherwise = False -- don't apply the transformation inside TH brackets, because -- GHC.HsToCore.Quote does not handle ApplicativeDo. ; in_th_bracket <- isBrackStage <$> getStage ; if ado_is_on && is_do_expr && not in_th_bracket then do { traceRn "ppsfa" (ppr stmts) ; rearrangeForApplicativeDo ctxt stmts } else noPostProcessStmts ctxt stmts } -- | strip the FreeVars annotations from statements noPostProcessStmts :: HsStmtContext GhcRn -> [(LStmt GhcRn (LocatedA (body GhcRn)), FreeVars)] -> RnM ([LStmt GhcRn (LocatedA (body GhcRn))], FreeVars) noPostProcessStmts _ stmts = return (map fst stmts, emptyNameSet) rnStmtsWithFreeVars :: AnnoBody body => HsStmtContext GhcRn -> ((body GhcPs) -> RnM ((body GhcRn), FreeVars)) -> [LStmt GhcPs (LocatedA (body GhcPs))] -> ([Name] -> RnM (thing, FreeVars)) -> RnM ( ([(LStmt GhcRn (LocatedA (body GhcRn)), FreeVars)], thing) , FreeVars) -- Each Stmt body is annotated with its FreeVars, so that -- we can rearrange statements for ApplicativeDo. -- -- Variables bound by the Stmts, and mentioned in thing_inside, -- do not appear in the result FreeVars rnStmtsWithFreeVars ctxt _ [] thing_inside = do { checkEmptyStmts ctxt ; (thing, fvs) <- thing_inside [] ; return (([], thing), fvs) } rnStmtsWithFreeVars mDoExpr@MDoExpr{} rnBody stmts thing_inside -- Deal with mdo = -- Behave like do { rec { ...all but last... }; last } do { ((stmts1, (stmts2, thing)), fvs) <- rnStmt mDoExpr rnBody (noLocA $ mkRecStmt noAnn (noLocA all_but_last)) $ \ _ -> do { last_stmt' <- checkLastStmt mDoExpr last_stmt ; rnStmt mDoExpr rnBody last_stmt' thing_inside } ; return (((stmts1 ++ stmts2), thing), fvs) } where Just (all_but_last, last_stmt) = snocView stmts rnStmtsWithFreeVars ctxt rnBody (lstmt@(L loc _) : lstmts) thing_inside | null lstmts = setSrcSpanA loc $ do { lstmt' <- checkLastStmt ctxt lstmt ; rnStmt ctxt rnBody lstmt' thing_inside } | otherwise = do { ((stmts1, (stmts2, thing)), fvs) <- setSrcSpanA loc $ do { checkStmt ctxt lstmt ; rnStmt ctxt rnBody lstmt $ \ bndrs1 -> rnStmtsWithFreeVars ctxt rnBody lstmts $ \ bndrs2 -> thing_inside (bndrs1 ++ bndrs2) } ; return (((stmts1 ++ stmts2), thing), fvs) } ---------------------- {- Note [Failing pattern matches in Stmts] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Many things desugar to HsStmts including monadic things like `do` and `mdo` statements, pattern guards, and list comprehensions (see 'HsStmtContext' for an exhaustive list). How we deal with pattern match failure is context-dependent. * In the case of list comprehensions and pattern guards we don't need any 'fail' function; the desugarer ignores the fail function of 'BindStmt' entirely. So, for list comprehensions, the fail function is set to 'Nothing' for clarity. * In the case of monadic contexts (e.g. monad comprehensions, do, and mdo expressions) we want pattern match failure to be desugared to the 'fail' function (from MonadFail type class). At one point we failed to make this distinction, leading to #11216. -} rnStmt :: AnnoBody body => HsStmtContext GhcRn -> (body GhcPs -> RnM (body GhcRn, FreeVars)) -- ^ How to rename the body of the statement -> LStmt GhcPs (LocatedA (body GhcPs)) -- ^ The statement -> ([Name] -> RnM (thing, FreeVars)) -- ^ Rename the stuff that this statement scopes over -> RnM ( ([(LStmt GhcRn (LocatedA (body GhcRn)), FreeVars)], thing) , FreeVars) -- Variables bound by the Stmt, and mentioned in thing_inside, -- do not appear in the result FreeVars rnStmt ctxt rnBody (L loc (LastStmt _ (L lb body) noret _)) thing_inside = do { (body', fv_expr) <- rnBody body ; (ret_op, fvs1) <- if isMonadCompContext ctxt then lookupStmtName ctxt returnMName else return (noSyntaxExpr, emptyFVs) -- The 'return' in a LastStmt is used only -- for MonadComp; and we don't want to report -- "non in scope: return" in other cases -- #15607 ; (thing, fvs3) <- thing_inside [] ; return (([(L loc (LastStmt noExtField (L lb body') noret ret_op), fv_expr)] , thing), fv_expr `plusFV` fvs1 `plusFV` fvs3) } rnStmt ctxt rnBody (L loc (BodyStmt _ (L lb body) _ _)) thing_inside = do { (body', fv_expr) <- rnBody body ; (then_op, fvs1) <- lookupQualifiedDoStmtName ctxt thenMName ; (guard_op, fvs2) <- if isComprehensionContext ctxt then lookupStmtName ctxt guardMName else return (noSyntaxExpr, emptyFVs) -- Only list/monad comprehensions use 'guard' -- Also for sub-stmts of same eg [ e | x<-xs, gd | blah ] -- Here "gd" is a guard ; (thing, fvs3) <- thing_inside [] ; return ( ([(L loc (BodyStmt noExtField (L lb body') then_op guard_op), fv_expr)] , thing), fv_expr `plusFV` fvs1 `plusFV` fvs2 `plusFV` fvs3) } rnStmt ctxt rnBody (L loc (BindStmt _ pat (L lb body))) thing_inside = do { (body', fv_expr) <- rnBody body -- The binders do not scope over the expression ; (bind_op, fvs1) <- lookupQualifiedDoStmtName ctxt bindMName ; (fail_op, fvs2) <- monadFailOp pat ctxt ; rnPat (StmtCtxt ctxt) pat $ \ pat' -> do { (thing, fvs3) <- thing_inside (collectPatBinders CollNoDictBinders pat') ; let xbsrn = XBindStmtRn { xbsrn_bindOp = bind_op, xbsrn_failOp = fail_op } ; return (( [( L loc (BindStmt xbsrn pat' (L lb body')), fv_expr )] , thing), fv_expr `plusFV` fvs1 `plusFV` fvs2 `plusFV` fvs3) }} -- fv_expr shouldn't really be filtered by the rnPatsAndThen -- but it does not matter because the names are unique rnStmt _ _ (L loc (LetStmt _ binds)) thing_inside = rnLocalBindsAndThen binds $ \binds' bind_fvs -> do { (thing, fvs) <- thing_inside (collectLocalBinders CollNoDictBinders binds') ; return ( ([(L loc (LetStmt noAnn binds'), bind_fvs)], thing) , fvs) } rnStmt ctxt rnBody (L loc (RecStmt { recS_stmts = L _ rec_stmts })) thing_inside = do { (return_op, fvs1) <- lookupQualifiedDoStmtName ctxt returnMName ; (mfix_op, fvs2) <- lookupQualifiedDoStmtName ctxt mfixName ; (bind_op, fvs3) <- lookupQualifiedDoStmtName ctxt bindMName ; let empty_rec_stmt = emptyRecStmtName { recS_ret_fn = return_op , recS_mfix_fn = mfix_op , recS_bind_fn = bind_op } -- Step1: Bring all the binders of the mdo into scope -- (Remember that this also removes the binders from the -- finally-returned free-vars.) -- And rename each individual stmt, making a -- singleton segment. At this stage the FwdRefs field -- isn't finished: it's empty for all except a BindStmt -- for which it's the fwd refs within the bind itself -- (This set may not be empty, because we're in a recursive -- context.) ; rnRecStmtsAndThen ctxt rnBody rec_stmts $ \ segs -> do { let bndrs = nameSetElemsStable $ foldr (unionNameSet . (\(ds,_,_,_) -> ds)) emptyNameSet segs -- See Note [Deterministic ApplicativeDo and RecursiveDo desugaring] ; (thing, fvs_later) <- thing_inside bndrs ; let (rec_stmts', fvs) = segmentRecStmts (locA loc) ctxt empty_rec_stmt segs fvs_later -- We aren't going to try to group RecStmts with -- ApplicativeDo, so attaching empty FVs is fine. ; return ( ((zip rec_stmts' (repeat emptyNameSet)), thing) , fvs `plusFV` fvs1 `plusFV` fvs2 `plusFV` fvs3) } } rnStmt ctxt _ (L loc (ParStmt _ segs _ _)) thing_inside = do { (mzip_op, fvs1) <- lookupStmtNamePoly ctxt mzipName ; (bind_op, fvs2) <- lookupStmtName ctxt bindMName ; (return_op, fvs3) <- lookupStmtName ctxt returnMName ; ((segs', thing), fvs4) <- rnParallelStmts (ParStmtCtxt ctxt) return_op segs thing_inside ; return (([(L loc (ParStmt noExtField segs' mzip_op bind_op), fvs4)], thing) , fvs1 `plusFV` fvs2 `plusFV` fvs3 `plusFV` fvs4) } rnStmt ctxt _ (L loc (TransStmt { trS_stmts = stmts, trS_by = by, trS_form = form , trS_using = using })) thing_inside = do { -- Rename the 'using' expression in the context before the transform is begun (using', fvs1) <- rnLExpr using -- Rename the stmts and the 'by' expression -- Keep track of the variables mentioned in the 'by' expression ; ((stmts', (by', used_bndrs, thing)), fvs2) <- rnStmts (TransStmtCtxt ctxt) rnExpr stmts $ \ bndrs -> do { (by', fvs_by) <- mapMaybeFvRn rnLExpr by ; (thing, fvs_thing) <- thing_inside bndrs ; let fvs = fvs_by `plusFV` fvs_thing used_bndrs = filter (`elemNameSet` fvs) bndrs -- The paper (Fig 5) has a bug here; we must treat any free variable -- of the "thing inside", **or of the by-expression**, as used ; return ((by', used_bndrs, thing), fvs) } -- Lookup `return`, `(>>=)` and `liftM` for monad comprehensions ; (return_op, fvs3) <- lookupStmtName ctxt returnMName ; (bind_op, fvs4) <- lookupStmtName ctxt bindMName ; (fmap_op, fvs5) <- case form of ThenForm -> return (noExpr, emptyFVs) _ -> lookupStmtNamePoly ctxt fmapName ; let all_fvs = fvs1 `plusFV` fvs2 `plusFV` fvs3 `plusFV` fvs4 `plusFV` fvs5 bndr_map = used_bndrs `zip` used_bndrs -- See Note [TransStmt binder map] in GHC.Hs.Expr ; traceRn "rnStmt: implicitly rebound these used binders:" (ppr bndr_map) ; return (([(L loc (TransStmt { trS_ext = noExtField , trS_stmts = stmts', trS_bndrs = bndr_map , trS_by = by', trS_using = using', trS_form = form , trS_ret = return_op, trS_bind = bind_op , trS_fmap = fmap_op }), fvs2)], thing), all_fvs) } rnStmt _ _ (L _ ApplicativeStmt{}) _ = panic "rnStmt: ApplicativeStmt" rnParallelStmts :: forall thing. HsStmtContext GhcRn -> SyntaxExpr GhcRn -> [ParStmtBlock GhcPs GhcPs] -> ([Name] -> RnM (thing, FreeVars)) -> RnM (([ParStmtBlock GhcRn GhcRn], thing), FreeVars) -- Note [Renaming parallel Stmts] rnParallelStmts ctxt return_op segs thing_inside = do { orig_lcl_env <- getLocalRdrEnv ; rn_segs orig_lcl_env [] segs } where rn_segs :: LocalRdrEnv -> [Name] -> [ParStmtBlock GhcPs GhcPs] -> RnM (([ParStmtBlock GhcRn GhcRn], thing), FreeVars) rn_segs _ bndrs_so_far [] = do { let (bndrs', dups) = removeDups cmpByOcc bndrs_so_far ; mapM_ dupErr dups ; (thing, fvs) <- bindLocalNames bndrs' (thing_inside bndrs') ; return (([], thing), fvs) } rn_segs env bndrs_so_far (ParStmtBlock x stmts _ _ : segs) = do { ((stmts', (used_bndrs, segs', thing)), fvs) <- rnStmts ctxt rnExpr stmts $ \ bndrs -> setLocalRdrEnv env $ do { ((segs', thing), fvs) <- rn_segs env (bndrs ++ bndrs_so_far) segs ; let used_bndrs = filter (`elemNameSet` fvs) bndrs ; return ((used_bndrs, segs', thing), fvs) } ; let seg' = ParStmtBlock x stmts' used_bndrs return_op ; return ((seg':segs', thing), fvs) } cmpByOcc n1 n2 = nameOccName n1 `compare` nameOccName n2 dupErr vs = addErr (text "Duplicate binding in parallel list comprehension for:" <+> quotes (ppr (NE.head vs))) lookupQualifiedDoStmtName :: HsStmtContext GhcRn -> Name -> RnM (SyntaxExpr GhcRn, FreeVars) -- Like lookupStmtName, but respects QualifiedDo lookupQualifiedDoStmtName ctxt n = case qualifiedDoModuleName_maybe ctxt of Nothing -> lookupStmtName ctxt n Just modName -> first (mkSyntaxExpr . nl_HsVar) <$> lookupNameWithQualifier n modName lookupStmtName :: HsStmtContext GhcRn -> Name -> RnM (SyntaxExpr GhcRn, FreeVars) -- Like lookupSyntax, but respects contexts lookupStmtName ctxt n | rebindableContext ctxt = lookupSyntax n | otherwise = return (mkRnSyntaxExpr n, emptyFVs) lookupStmtNamePoly :: HsStmtContext GhcRn -> Name -> RnM (HsExpr GhcRn, FreeVars) lookupStmtNamePoly ctxt name | rebindableContext ctxt = do { rebindable_on <- xoptM LangExt.RebindableSyntax ; if rebindable_on then do { fm <- lookupOccRn (nameRdrName name) ; return (HsVar noExtField (noLocA fm), unitFV fm) } else not_rebindable } | otherwise = not_rebindable where not_rebindable = return (HsVar noExtField (noLocA name), emptyFVs) -- | Is this a context where we respect RebindableSyntax? -- but ListComp are never rebindable -- Neither is ArrowExpr, which has its own desugarer in GHC.HsToCore.Arrows rebindableContext :: HsStmtContext GhcRn -> Bool rebindableContext ctxt = case ctxt of ListComp -> False ArrowExpr -> False PatGuard {} -> False DoExpr m -> isNothing m MDoExpr m -> isNothing m MonadComp -> True GhciStmtCtxt -> True -- I suppose? ParStmtCtxt c -> rebindableContext c -- Look inside to TransStmtCtxt c -> rebindableContext c -- the parent context {- Note [Renaming parallel Stmts] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Renaming parallel statements is painful. Given, say [ a+c | a <- as, bs <- bss | c <- bs, a <- ds ] Note that (a) In order to report "Defined but not used" about 'bs', we must rename each group of Stmts with a thing_inside whose FreeVars include at least {a,c} (b) We want to report that 'a' is illegally bound in both branches (c) The 'bs' in the second group must obviously not be captured by the binding in the first group To satisfy (a) we nest the segements. To satisfy (b) we check for duplicates just before thing_inside. To satisfy (c) we reset the LocalRdrEnv each time. ************************************************************************ * * \subsubsection{mdo expressions} * * ************************************************************************ -} type FwdRefs = NameSet type Segment stmts = (Defs, Uses, -- May include defs FwdRefs, -- A subset of uses that are -- (a) used before they are bound in this segment, or -- (b) used here, and bound in subsequent segments stmts) -- Either Stmt or [Stmt] -- wrapper that does both the left- and right-hand sides rnRecStmtsAndThen :: AnnoBody body => HsStmtContext GhcRn -> (body GhcPs -> RnM (body GhcRn, FreeVars)) -> [LStmt GhcPs (LocatedA (body GhcPs))] -- assumes that the FreeVars returned includes -- the FreeVars of the Segments -> ([Segment (LStmt GhcRn (LocatedA (body GhcRn)))] -> RnM (a, FreeVars)) -> RnM (a, FreeVars) rnRecStmtsAndThen ctxt rnBody s cont = do { -- (A) Make the mini fixity env for all of the stmts fix_env <- makeMiniFixityEnv (collectRecStmtsFixities s) -- (B) Do the LHSes ; new_lhs_and_fv <- rn_rec_stmts_lhs fix_env s -- ...bring them and their fixities into scope ; let bound_names = collectLStmtsBinders CollNoDictBinders (map fst new_lhs_and_fv) -- Fake uses of variables introduced implicitly (warning suppression, see #4404) rec_uses = lStmtsImplicits (map fst new_lhs_and_fv) implicit_uses = mkNameSet $ concatMap snd $ rec_uses ; bindLocalNamesFV bound_names $ addLocalFixities fix_env bound_names $ do -- (C) do the right-hand-sides and thing-inside { segs <- rn_rec_stmts ctxt rnBody bound_names new_lhs_and_fv ; (res, fvs) <- cont segs ; mapM_ (\(loc, ns) -> checkUnusedRecordWildcard loc fvs (Just ns)) rec_uses ; warnUnusedLocalBinds bound_names (fvs `unionNameSet` implicit_uses) ; return (res, fvs) }} -- get all the fixity decls in any Let stmt collectRecStmtsFixities :: [LStmtLR GhcPs GhcPs body] -> [LFixitySig GhcPs] collectRecStmtsFixities l = foldr (\ s -> \acc -> case s of (L _ (LetStmt _ (HsValBinds _ (ValBinds _ _ sigs)))) -> foldr (\ sig -> \ acc -> case sig of (L loc (FixSig _ s)) -> (L loc s) : acc _ -> acc) acc sigs _ -> acc) [] l -- left-hand sides rn_rec_stmt_lhs :: AnnoBody body => MiniFixityEnv -> LStmt GhcPs (LocatedA (body GhcPs)) -- rename LHS, and return its FVs -- Warning: we will only need the FreeVars below in the case of a BindStmt, -- so we don't bother to compute it accurately in the other cases -> RnM [(LStmtLR GhcRn GhcPs (LocatedA (body GhcPs)), FreeVars)] rn_rec_stmt_lhs _ (L loc (BodyStmt _ body a b)) = return [(L loc (BodyStmt noExtField body a b), emptyFVs)] rn_rec_stmt_lhs _ (L loc (LastStmt _ body noret a)) = return [(L loc (LastStmt noExtField body noret a), emptyFVs)] rn_rec_stmt_lhs fix_env (L loc (BindStmt _ pat body)) = do -- should the ctxt be MDo instead? (pat', fv_pat) <- rnBindPat (localRecNameMaker fix_env) pat return [(L loc (BindStmt noAnn pat' body), fv_pat)] rn_rec_stmt_lhs _ (L _ (LetStmt _ binds@(HsIPBinds {}))) = failWith (badIpBinds (text "an mdo expression") binds) rn_rec_stmt_lhs fix_env (L loc (LetStmt _ (HsValBinds x binds))) = do (_bound_names, binds') <- rnLocalValBindsLHS fix_env binds return [(L loc (LetStmt noAnn (HsValBinds x binds')), -- Warning: this is bogus; see function invariant emptyFVs )] -- XXX Do we need to do something with the return and mfix names? rn_rec_stmt_lhs fix_env (L _ (RecStmt { recS_stmts = L _ stmts })) -- Flatten Rec inside Rec = rn_rec_stmts_lhs fix_env stmts rn_rec_stmt_lhs _ stmt@(L _ (ParStmt {})) -- Syntactically illegal in mdo = pprPanic "rn_rec_stmt" (ppr stmt) rn_rec_stmt_lhs _ stmt@(L _ (TransStmt {})) -- Syntactically illegal in mdo = pprPanic "rn_rec_stmt" (ppr stmt) rn_rec_stmt_lhs _ stmt@(L _ (ApplicativeStmt {})) -- Shouldn't appear yet = pprPanic "rn_rec_stmt" (ppr stmt) rn_rec_stmt_lhs _ (L _ (LetStmt _ (EmptyLocalBinds _))) = panic "rn_rec_stmt LetStmt EmptyLocalBinds" rn_rec_stmts_lhs :: AnnoBody body => MiniFixityEnv -> [LStmt GhcPs (LocatedA (body GhcPs))] -> RnM [(LStmtLR GhcRn GhcPs (LocatedA (body GhcPs)), FreeVars)] rn_rec_stmts_lhs fix_env stmts = do { ls <- concatMapM (rn_rec_stmt_lhs fix_env) stmts ; let boundNames = collectLStmtsBinders CollNoDictBinders (map fst ls) -- First do error checking: we need to check for dups here because we -- don't bind all of the variables from the Stmt at once -- with bindLocatedLocals. ; checkDupNames boundNames ; return ls } -- right-hand-sides rn_rec_stmt :: AnnoBody body => HsStmtContext GhcRn -> (body GhcPs -> RnM (body GhcRn, FreeVars)) -> [Name] -> (LStmtLR GhcRn GhcPs (LocatedA (body GhcPs)), FreeVars) -> RnM [Segment (LStmt GhcRn (LocatedA (body GhcRn)))] -- Rename a Stmt that is inside a RecStmt (or mdo) -- Assumes all binders are already in scope -- Turns each stmt into a singleton Stmt rn_rec_stmt ctxt rnBody _ (L loc (LastStmt _ (L lb body) noret _), _) = do { (body', fv_expr) <- rnBody body ; (ret_op, fvs1) <- lookupQualifiedDo ctxt returnMName ; return [(emptyNameSet, fv_expr `plusFV` fvs1, emptyNameSet, L loc (LastStmt noExtField (L lb body') noret ret_op))] } rn_rec_stmt ctxt rnBody _ (L loc (BodyStmt _ (L lb body) _ _), _) = do { (body', fvs) <- rnBody body ; (then_op, fvs1) <- lookupQualifiedDo ctxt thenMName ; return [(emptyNameSet, fvs `plusFV` fvs1, emptyNameSet, L loc (BodyStmt noExtField (L lb body') then_op noSyntaxExpr))] } rn_rec_stmt ctxt rnBody _ (L loc (BindStmt _ pat' (L lb body)), fv_pat) = do { (body', fv_expr) <- rnBody body ; (bind_op, fvs1) <- lookupQualifiedDo ctxt bindMName ; (fail_op, fvs2) <- getMonadFailOp ctxt ; let bndrs = mkNameSet (collectPatBinders CollNoDictBinders pat') fvs = fv_expr `plusFV` fv_pat `plusFV` fvs1 `plusFV` fvs2 ; let xbsrn = XBindStmtRn { xbsrn_bindOp = bind_op, xbsrn_failOp = fail_op } ; return [(bndrs, fvs, bndrs `intersectNameSet` fvs, L loc (BindStmt xbsrn pat' (L lb body')))] } rn_rec_stmt _ _ _ (L _ (LetStmt _ binds@(HsIPBinds {})), _) = failWith (badIpBinds (text "an mdo expression") binds) rn_rec_stmt _ _ all_bndrs (L loc (LetStmt _ (HsValBinds x binds')), _) = do { (binds', du_binds) <- rnLocalValBindsRHS (mkNameSet all_bndrs) binds' -- fixities and unused are handled above in rnRecStmtsAndThen ; let fvs = allUses du_binds ; return [(duDefs du_binds, fvs, emptyNameSet, L loc (LetStmt noAnn (HsValBinds x binds')))] } -- no RecStmt case because they get flattened above when doing the LHSes rn_rec_stmt _ _ _ stmt@(L _ (RecStmt {}), _) = pprPanic "rn_rec_stmt: RecStmt" (ppr stmt) rn_rec_stmt _ _ _ stmt@(L _ (ParStmt {}), _) -- Syntactically illegal in mdo = pprPanic "rn_rec_stmt: ParStmt" (ppr stmt) rn_rec_stmt _ _ _ stmt@(L _ (TransStmt {}), _) -- Syntactically illegal in mdo = pprPanic "rn_rec_stmt: TransStmt" (ppr stmt) rn_rec_stmt _ _ _ (L _ (LetStmt _ (EmptyLocalBinds _)), _) = panic "rn_rec_stmt: LetStmt EmptyLocalBinds" rn_rec_stmt _ _ _ stmt@(L _ (ApplicativeStmt {}), _) = pprPanic "rn_rec_stmt: ApplicativeStmt" (ppr stmt) rn_rec_stmts :: AnnoBody body => HsStmtContext GhcRn -> (body GhcPs -> RnM (body GhcRn, FreeVars)) -> [Name] -> [(LStmtLR GhcRn GhcPs (LocatedA (body GhcPs)), FreeVars)] -> RnM [Segment (LStmt GhcRn (LocatedA (body GhcRn)))] rn_rec_stmts ctxt rnBody bndrs stmts = do { segs_s <- mapM (rn_rec_stmt ctxt rnBody bndrs) stmts ; return (concat segs_s) } --------------------------------------------- segmentRecStmts :: AnnoBody body => SrcSpan -> HsStmtContext GhcRn -> Stmt GhcRn (LocatedA (body GhcRn)) -> [Segment (LStmt GhcRn (LocatedA (body GhcRn)))] -> FreeVars -> ([LStmt GhcRn (LocatedA (body GhcRn))], FreeVars) segmentRecStmts loc ctxt empty_rec_stmt segs fvs_later | null segs = ([], fvs_later) | MDoExpr _ <- ctxt = segsToStmts empty_rec_stmt grouped_segs fvs_later -- Step 4: Turn the segments into Stmts -- Use RecStmt when and only when there are fwd refs -- Also gather up the uses from the end towards the -- start, so we can tell the RecStmt which things are -- used 'after' the RecStmt | otherwise = ([ L (noAnnSrcSpan loc) $ empty_rec_stmt { recS_stmts = noLocA ss , recS_later_ids = nameSetElemsStable (defs `intersectNameSet` fvs_later) , recS_rec_ids = nameSetElemsStable (defs `intersectNameSet` uses) }] -- See Note [Deterministic ApplicativeDo and RecursiveDo desugaring] , uses `plusFV` fvs_later) where (defs_s, uses_s, _, ss) = unzip4 segs defs = plusFVs defs_s uses = plusFVs uses_s -- Step 2: Fill in the fwd refs. -- The segments are all singletons, but their fwd-ref -- field mentions all the things used by the segment -- that are bound after their use segs_w_fwd_refs = addFwdRefs segs -- Step 3: Group together the segments to make bigger segments -- Invariant: in the result, no segment uses a variable -- bound in a later segment grouped_segs = glomSegments ctxt segs_w_fwd_refs ---------------------------- addFwdRefs :: [Segment a] -> [Segment a] -- So far the segments only have forward refs *within* the Stmt -- (which happens for bind: x <- ...x...) -- This function adds the cross-seg fwd ref info addFwdRefs segs = fst (foldr mk_seg ([], emptyNameSet) segs) where mk_seg (defs, uses, fwds, stmts) (segs, later_defs) = (new_seg : segs, all_defs) where new_seg = (defs, uses, new_fwds, stmts) all_defs = later_defs `unionNameSet` defs new_fwds = fwds `unionNameSet` (uses `intersectNameSet` later_defs) -- Add the downstream fwd refs here {- Note [Segmenting mdo] ~~~~~~~~~~~~~~~~~~~~~ NB. June 7 2012: We only glom segments that appear in an explicit mdo; and leave those found in "do rec"'s intact. See https://gitlab.haskell.org/ghc/ghc/issues/4148 for the discussion leading to this design choice. Hence the test in segmentRecStmts. Note [Glomming segments] ~~~~~~~~~~~~~~~~~~~~~~~~ Glomming the singleton segments of an mdo into minimal recursive groups. At first I thought this was just strongly connected components, but there's an important constraint: the order of the stmts must not change. Consider mdo { x <- ...y... p <- z y <- ...x... q <- x z <- y r <- x } Here, the first stmt mention 'y', which is bound in the third. But that means that the innocent second stmt (p <- z) gets caught up in the recursion. And that in turn means that the binding for 'z' has to be included... and so on. Start at the tail { r <- x } Now add the next one { z <- y ; r <- x } Now add one more { q <- x ; z <- y ; r <- x } Now one more... but this time we have to group a bunch into rec { rec { y <- ...x... ; q <- x ; z <- y } ; r <- x } Now one more, which we can add on without a rec { p <- z ; rec { y <- ...x... ; q <- x ; z <- y } ; r <- x } Finally we add the last one; since it mentions y we have to glom it together with the first two groups { rec { x <- ...y...; p <- z ; y <- ...x... ; q <- x ; z <- y } ; r <- x } -} glomSegments :: HsStmtContext GhcRn -> [Segment (LStmt GhcRn body)] -> [Segment [LStmt GhcRn body]] -- Each segment has a non-empty list of Stmts -- See Note [Glomming segments] glomSegments _ [] = [] glomSegments ctxt ((defs,uses,fwds,stmt) : segs) -- Actually stmts will always be a singleton = (seg_defs, seg_uses, seg_fwds, seg_stmts) : others where segs' = glomSegments ctxt segs (extras, others) = grab uses segs' (ds, us, fs, ss) = unzip4 extras seg_defs = plusFVs ds `plusFV` defs seg_uses = plusFVs us `plusFV` uses seg_fwds = plusFVs fs `plusFV` fwds seg_stmts = stmt : concat ss grab :: NameSet -- The client -> [Segment a] -> ([Segment a], -- Needed by the 'client' [Segment a]) -- Not needed by the client -- The result is simply a split of the input grab uses dus = (reverse yeses, reverse noes) where (noes, yeses) = span not_needed (reverse dus) not_needed (defs,_,_,_) = disjointNameSet defs uses ---------------------------------------------------- segsToStmts :: Stmt GhcRn (LocatedA (body GhcRn)) -- A RecStmt with the SyntaxOps filled in -> [Segment [LStmt GhcRn (LocatedA (body GhcRn))]] -- Each Segment has a non-empty list of Stmts -> FreeVars -- Free vars used 'later' -> ([LStmt GhcRn (LocatedA (body GhcRn))], FreeVars) segsToStmts _ [] fvs_later = ([], fvs_later) segsToStmts empty_rec_stmt ((defs, uses, fwds, ss) : segs) fvs_later = ASSERT( not (null ss) ) (new_stmt : later_stmts, later_uses `plusFV` uses) where (later_stmts, later_uses) = segsToStmts empty_rec_stmt segs fvs_later new_stmt | non_rec = head ss | otherwise = L (getLoc (head ss)) rec_stmt rec_stmt = empty_rec_stmt { recS_stmts = noLocA ss , recS_later_ids = nameSetElemsStable used_later , recS_rec_ids = nameSetElemsStable fwds } -- See Note [Deterministic ApplicativeDo and RecursiveDo desugaring] non_rec = isSingleton ss && isEmptyNameSet fwds used_later = defs `intersectNameSet` later_uses -- The ones needed after the RecStmt {- ************************************************************************ * * ApplicativeDo * * ************************************************************************ Note [ApplicativeDo] = Example = For a sequence of statements do x <- A y <- B x z <- C return (f x y z) We want to transform this to (\(x,y) z -> f x y z) <$> (do x <- A; y <- B x; return (x,y)) <*> C It would be easy to notice that "y <- B x" and "z <- C" are independent and do something like this: do x <- A (y,z) <- (,) <$> B x <*> C return (f x y z) But this isn't enough! A and C were also independent, and this transformation loses the ability to do A and C in parallel. The algorithm works by first splitting the sequence of statements into independent "segments", and a separate "tail" (the final statement). In our example above, the segements would be [ x <- A , y <- B x ] [ z <- C ] and the tail is: return (f x y z) Then we take these segments and make an Applicative expression from them: (\(x,y) z -> return (f x y z)) <$> do { x <- A; y <- B x; return (x,y) } <*> C Finally, we recursively apply the transformation to each segment, to discover any nested parallelism. = Syntax & spec = expr ::= ... | do {stmt_1; ..; stmt_n} expr | ... stmt ::= pat <- expr | (arg_1 | ... | arg_n) -- applicative composition, n>=1 | ... -- other kinds of statement (e.g. let) arg ::= pat <- expr | {stmt_1; ..; stmt_n} {var_1..var_n} (note that in the actual implementation,the expr in a do statement is represented by a LastStmt as the final stmt, this is just a representational issue and may change later.) == Transformation to introduce applicative stmts == ado {} tail = tail ado {pat <- expr} {return expr'} = (mkArg(pat <- expr)); return expr' ado {one} tail = one : tail ado stmts tail | n == 1 = ado before (ado after tail) where (before,after) = split(stmts_1) | n > 1 = (mkArg(stmts_1) | ... | mkArg(stmts_n)); tail where {stmts_1 .. stmts_n} = segments(stmts) segments(stmts) = -- divide stmts into segments with no interdependencies mkArg({pat <- expr}) = (pat <- expr) mkArg({stmt_1; ...; stmt_n}) = {stmt_1; ...; stmt_n} {vars(stmt_1) u .. u vars(stmt_n)} split({stmt_1; ..; stmt_n) = ({stmt_1; ..; stmt_i}, {stmt_i+1; ..; stmt_n}) -- 1 <= i <= n -- i is a good place to insert a bind == Desugaring for do == dsDo {} expr = expr dsDo {pat <- rhs; stmts} expr = rhs >>= \pat -> dsDo stmts expr dsDo {(arg_1 | ... | arg_n)} (return expr) = (\argpat (arg_1) .. argpat(arg_n) -> expr) <$> argexpr(arg_1) <*> ... <*> argexpr(arg_n) dsDo {(arg_1 | ... | arg_n); stmts} expr = join (\argpat (arg_1) .. argpat(arg_n) -> dsDo stmts expr) <$> argexpr(arg_1) <*> ... <*> argexpr(arg_n) = Relevant modules in the rest of the compiler = ApplicativeDo touches a few phases in the compiler: * Renamer: The journey begins here in the renamer, where do-blocks are scheduled as outlined above and transformed into applicative combinators. However, the code is still represented as a do-block with special forms of applicative statements. This allows us to recover the original do-block when e.g. printing type errors, where we don't want to show any of the applicative combinators since they don't exist in the source code. See ApplicativeStmt and ApplicativeArg in HsExpr. * Typechecker: ApplicativeDo passes through the typechecker much like any other form of expression. The only crux is that the typechecker has to be aware of the special ApplicativeDo statements in the do-notation, and typecheck them appropriately. Relevant module: GHC.Tc.Gen.Match * Desugarer: Any do-block which contains applicative statements is desugared as outlined above, to use the Applicative combinators. Relevant module: GHC.HsToCore.Expr -} -- | The 'Name's of @return@ and @pure@. These may not be 'returnName' and -- 'pureName' due to @QualifiedDo@ or @RebindableSyntax@. data MonadNames = MonadNames { return_name, pure_name :: Name } instance Outputable MonadNames where ppr (MonadNames {return_name=return_name,pure_name=pure_name}) = hcat [text "MonadNames { return_name = " ,ppr return_name ,text ", pure_name = " ,ppr pure_name ,text "}" ] -- | rearrange a list of statements using ApplicativeDoStmt. See -- Note [ApplicativeDo]. rearrangeForApplicativeDo :: HsStmtContext GhcRn -> [(ExprLStmt GhcRn, FreeVars)] -> RnM ([ExprLStmt GhcRn], FreeVars) rearrangeForApplicativeDo _ [] = return ([], emptyNameSet) rearrangeForApplicativeDo _ [(one,_)] = return ([one], emptyNameSet) rearrangeForApplicativeDo ctxt stmts0 = do optimal_ado <- goptM Opt_OptimalApplicativeDo let stmt_tree | optimal_ado = mkStmtTreeOptimal stmts | otherwise = mkStmtTreeHeuristic stmts traceRn "rearrangeForADo" (ppr stmt_tree) (return_name, _) <- lookupQualifiedDoName ctxt returnMName (pure_name, _) <- lookupQualifiedDoName ctxt pureAName let monad_names = MonadNames { return_name = return_name , pure_name = pure_name } stmtTreeToStmts monad_names ctxt stmt_tree [last] last_fvs where (stmts,(last,last_fvs)) = findLast stmts0 findLast [] = error "findLast" findLast [last] = ([],last) findLast (x:xs) = (x:rest,last) where (rest,last) = findLast xs -- | A tree of statements using a mixture of applicative and bind constructs. data StmtTree a = StmtTreeOne a | StmtTreeBind (StmtTree a) (StmtTree a) | StmtTreeApplicative [StmtTree a] instance Outputable a => Outputable (StmtTree a) where ppr (StmtTreeOne x) = parens (text "StmtTreeOne" <+> ppr x) ppr (StmtTreeBind x y) = parens (hang (text "StmtTreeBind") 2 (sep [ppr x, ppr y])) ppr (StmtTreeApplicative xs) = parens (hang (text "StmtTreeApplicative") 2 (vcat (map ppr xs))) flattenStmtTree :: StmtTree a -> [a] flattenStmtTree t = go t [] where go (StmtTreeOne a) as = a : as go (StmtTreeBind l r) as = go l (go r as) go (StmtTreeApplicative ts) as = foldr go as ts type ExprStmtTree = StmtTree (ExprLStmt GhcRn, FreeVars) type Cost = Int -- | Turn a sequence of statements into an ExprStmtTree using a -- heuristic algorithm. /O(n^2)/ mkStmtTreeHeuristic :: [(ExprLStmt GhcRn, FreeVars)] -> ExprStmtTree mkStmtTreeHeuristic [one] = StmtTreeOne one mkStmtTreeHeuristic stmts = case segments stmts of [one] -> split one segs -> StmtTreeApplicative (map split segs) where split [one] = StmtTreeOne one split stmts = StmtTreeBind (mkStmtTreeHeuristic before) (mkStmtTreeHeuristic after) where (before, after) = splitSegment stmts -- | Turn a sequence of statements into an ExprStmtTree optimally, -- using dynamic programming. /O(n^3)/ mkStmtTreeOptimal :: [(ExprLStmt GhcRn, FreeVars)] -> ExprStmtTree mkStmtTreeOptimal stmts = ASSERT(not (null stmts)) -- the empty case is handled by the caller; -- we don't support empty StmtTrees. fst (arr ! (0,n)) where n = length stmts - 1 stmt_arr = listArray (0,n) stmts -- lazy cache of optimal trees for subsequences of the input arr :: Array (Int,Int) (ExprStmtTree, Cost) arr = array ((0,0),(n,n)) [ ((lo,hi), tree lo hi) | lo <- [0..n] , hi <- [lo..n] ] -- compute the optimal tree for the sequence [lo..hi] tree lo hi | hi == lo = (StmtTreeOne (stmt_arr ! lo), 1) | otherwise = case segments [ stmt_arr ! i | i <- [lo..hi] ] of [] -> panic "mkStmtTree" [_one] -> split lo hi segs -> (StmtTreeApplicative trees, maximum costs) where bounds = scanl (\(_,hi) a -> (hi+1, hi + length a)) (0,lo-1) segs (trees,costs) = unzip (map (uncurry split) (tail bounds)) -- find the best place to split the segment [lo..hi] split :: Int -> Int -> (ExprStmtTree, Cost) split lo hi | hi == lo = (StmtTreeOne (stmt_arr ! lo), 1) | otherwise = (StmtTreeBind before after, c1+c2) where -- As per the paper, for a sequence s1...sn, we want to find -- the split with the minimum cost, where the cost is the -- sum of the cost of the left and right subsequences. -- -- As an optimisation (also in the paper) if the cost of -- s1..s(n-1) is different from the cost of s2..sn, we know -- that the optimal solution is the lower of the two. Only -- in the case that these two have the same cost do we need -- to do the exhaustive search. -- ((before,c1),(after,c2)) | hi - lo == 1 = ((StmtTreeOne (stmt_arr ! lo), 1), (StmtTreeOne (stmt_arr ! hi), 1)) | left_cost < right_cost = ((left,left_cost), (StmtTreeOne (stmt_arr ! hi), 1)) | left_cost > right_cost = ((StmtTreeOne (stmt_arr ! lo), 1), (right,right_cost)) | otherwise = minimumBy (comparing cost) alternatives where (left, left_cost) = arr ! (lo,hi-1) (right, right_cost) = arr ! (lo+1,hi) cost ((_,c1),(_,c2)) = c1 + c2 alternatives = [ (arr ! (lo,k), arr ! (k+1,hi)) | k <- [lo .. hi-1] ] -- | Turn the ExprStmtTree back into a sequence of statements, using -- ApplicativeStmt where necessary. stmtTreeToStmts :: MonadNames -> HsStmtContext GhcRn -> ExprStmtTree -> [ExprLStmt GhcRn] -- ^ the "tail" -> FreeVars -- ^ free variables of the tail -> RnM ( [ExprLStmt GhcRn] -- ( output statements, , FreeVars ) -- , things we needed -- If we have a single bind, and we can do it without a join, transform -- to an ApplicativeStmt. This corresponds to the rule -- dsBlock [pat <- rhs] (return expr) = expr <$> rhs -- In the spec, but we do it here rather than in the desugarer, -- because we need the typechecker to typecheck the <$> form rather than -- the bind form, which would give rise to a Monad constraint. stmtTreeToStmts monad_names ctxt (StmtTreeOne (L _ (BindStmt xbs pat rhs), _)) tail _tail_fvs | not (isStrictPattern pat), (False,tail') <- needJoin monad_names tail -- See Note [ApplicativeDo and strict patterns] = mkApplicativeStmt ctxt [ApplicativeArgOne { xarg_app_arg_one = xbsrn_failOp xbs , app_arg_pattern = pat , arg_expr = rhs , is_body_stmt = False }] False tail' stmtTreeToStmts monad_names ctxt (StmtTreeOne (L _ (BodyStmt _ rhs _ _),_)) tail _tail_fvs | (False,tail') <- needJoin monad_names tail = mkApplicativeStmt ctxt [ApplicativeArgOne { xarg_app_arg_one = Nothing , app_arg_pattern = nlWildPatName , arg_expr = rhs , is_body_stmt = True }] False tail' stmtTreeToStmts _monad_names _ctxt (StmtTreeOne (s,_)) tail _tail_fvs = return (s : tail, emptyNameSet) stmtTreeToStmts monad_names ctxt (StmtTreeBind before after) tail tail_fvs = do (stmts1, fvs1) <- stmtTreeToStmts monad_names ctxt after tail tail_fvs let tail1_fvs = unionNameSets (tail_fvs : map snd (flattenStmtTree after)) (stmts2, fvs2) <- stmtTreeToStmts monad_names ctxt before stmts1 tail1_fvs return (stmts2, fvs1 `plusFV` fvs2) stmtTreeToStmts monad_names ctxt (StmtTreeApplicative trees) tail tail_fvs = do pairs <- mapM (stmtTreeArg ctxt tail_fvs) trees dflags <- getDynFlags let (stmts', fvss) = unzip pairs let (need_join, tail') = -- See Note [ApplicativeDo and refutable patterns] if any (hasRefutablePattern dflags) stmts' then (True, tail) else needJoin monad_names tail (stmts, fvs) <- mkApplicativeStmt ctxt stmts' need_join tail' return (stmts, unionNameSets (fvs:fvss)) where stmtTreeArg _ctxt _tail_fvs (StmtTreeOne (L _ (BindStmt xbs pat exp), _)) = return (ApplicativeArgOne { xarg_app_arg_one = xbsrn_failOp xbs , app_arg_pattern = pat , arg_expr = exp , is_body_stmt = False }, emptyFVs) stmtTreeArg _ctxt _tail_fvs (StmtTreeOne (L _ (BodyStmt _ exp _ _), _)) = return (ApplicativeArgOne { xarg_app_arg_one = Nothing , app_arg_pattern = nlWildPatName , arg_expr = exp , is_body_stmt = True }, emptyFVs) stmtTreeArg ctxt tail_fvs tree = do let stmts = flattenStmtTree tree pvarset = mkNameSet (concatMap (collectStmtBinders CollNoDictBinders . unLoc . fst) stmts) `intersectNameSet` tail_fvs pvars = nameSetElemsStable pvarset -- See Note [Deterministic ApplicativeDo and RecursiveDo desugaring] pat = mkBigLHsVarPatTup pvars tup = mkBigLHsVarTup pvars noExtField (stmts',fvs2) <- stmtTreeToStmts monad_names ctxt tree [] pvarset (mb_ret, fvs1) <- if | L _ ApplicativeStmt{} <- last stmts' -> return (unLoc tup, emptyNameSet) | otherwise -> do (ret, _) <- lookupQualifiedDoExpr ctxt returnMName let expr = HsApp noComments (noLocA ret) tup return (expr, emptyFVs) return ( ApplicativeArgMany { xarg_app_arg_many = noExtField , app_stmts = stmts' , final_expr = mb_ret , bv_pattern = pat , stmt_context = ctxt } , fvs1 `plusFV` fvs2) -- | Divide a sequence of statements into segments, where no segment -- depends on any variables defined by a statement in another segment. segments :: [(ExprLStmt GhcRn, FreeVars)] -> [[(ExprLStmt GhcRn, FreeVars)]] segments stmts = map fst $ merge $ reverse $ map reverse $ walk (reverse stmts) where allvars = mkNameSet (concatMap (collectStmtBinders CollNoDictBinders . unLoc . fst) stmts) -- We would rather not have a segment that just has LetStmts in -- it, so combine those with an adjacent segment where possible. merge [] = [] merge (seg : segs) = case rest of [] -> [(seg,all_lets)] ((s,s_lets):ss) | all_lets || s_lets -> (seg ++ s, all_lets && s_lets) : ss _otherwise -> (seg,all_lets) : rest where rest = merge segs all_lets = all (isLetStmt . fst) seg -- walk splits the statement sequence into segments, traversing -- the sequence from the back to the front, and keeping track of -- the set of free variables of the current segment. Whenever -- this set of free variables is empty, we have a complete segment. walk :: [(ExprLStmt GhcRn, FreeVars)] -> [[(ExprLStmt GhcRn, FreeVars)]] walk [] = [] walk ((stmt,fvs) : stmts) = ((stmt,fvs) : seg) : walk rest where (seg,rest) = chunter fvs' stmts (_, fvs') = stmtRefs stmt fvs chunter _ [] = ([], []) chunter vars ((stmt,fvs) : rest) | not (isEmptyNameSet vars) || isStrictPatternBind stmt -- See Note [ApplicativeDo and strict patterns] = ((stmt,fvs) : chunk, rest') where (chunk,rest') = chunter vars' rest (pvars, evars) = stmtRefs stmt fvs vars' = (vars `minusNameSet` pvars) `unionNameSet` evars chunter _ rest = ([], rest) stmtRefs stmt fvs | isLetStmt stmt = (pvars, fvs' `minusNameSet` pvars) | otherwise = (pvars, fvs') where fvs' = fvs `intersectNameSet` allvars pvars = mkNameSet (collectStmtBinders CollNoDictBinders (unLoc stmt)) isStrictPatternBind :: ExprLStmt GhcRn -> Bool isStrictPatternBind (L _ (BindStmt _ pat _)) = isStrictPattern pat isStrictPatternBind _ = False {- Note [ApplicativeDo and strict patterns] A strict pattern match is really a dependency. For example, do (x,y) <- A z <- B return C The pattern (_,_) must be matched strictly before we do B. If we allowed this to be transformed into (\(x,y) -> \z -> C) <$> A <*> B then it could be lazier than the standard desuraging using >>=. See #13875 for more examples. Thus, whenever we have a strict pattern match, we treat it as a dependency between that statement and the following one. The dependency prevents those two statements from being performed "in parallel" in an ApplicativeStmt, but doesn't otherwise affect what we can do with the rest of the statements in the same "do" expression. -} isStrictPattern :: LPat (GhcPass p) -> Bool isStrictPattern lpat = case unLoc lpat of WildPat{} -> False VarPat{} -> False LazyPat{} -> False AsPat _ _ p -> isStrictPattern p ParPat _ p -> isStrictPattern p ViewPat _ _ p -> isStrictPattern p SigPat _ p _ -> isStrictPattern p BangPat{} -> True ListPat{} -> True TuplePat{} -> True SumPat{} -> True ConPat{} -> True LitPat{} -> True NPat{} -> True NPlusKPat{} -> True SplicePat{} -> True XPat{} -> panic "isStrictPattern: XPat" {- Note [ApplicativeDo and refutable patterns] Refutable patterns in do blocks are desugared to use the monadic 'fail' operation. This means that sometimes an applicative block needs to be wrapped in 'join' simply because of a refutable pattern, in order for the types to work out. -} hasRefutablePattern :: DynFlags -> ApplicativeArg GhcRn -> Bool hasRefutablePattern dflags (ApplicativeArgOne { app_arg_pattern = pat , is_body_stmt = False}) = not (isIrrefutableHsPat dflags pat) hasRefutablePattern _ _ = False isLetStmt :: LStmt (GhcPass a) b -> Bool isLetStmt (L _ LetStmt{}) = True isLetStmt _ = False -- | Find a "good" place to insert a bind in an indivisible segment. -- This is the only place where we use heuristics. The current -- heuristic is to peel off the first group of independent statements -- and put the bind after those. splitSegment :: [(ExprLStmt GhcRn, FreeVars)] -> ( [(ExprLStmt GhcRn, FreeVars)] , [(ExprLStmt GhcRn, FreeVars)] ) splitSegment [one,two] = ([one],[two]) -- there is no choice when there are only two statements; this just saves -- some work in a common case. splitSegment stmts | Just (lets,binds,rest) <- slurpIndependentStmts stmts = if not (null lets) then (lets, binds++rest) else (lets++binds, rest) | otherwise = case stmts of (x:xs) -> ([x],xs) _other -> (stmts,[]) slurpIndependentStmts :: [(LStmt GhcRn (LocatedA (body GhcRn)), FreeVars)] -> Maybe ( [(LStmt GhcRn (LocatedA (body GhcRn)), FreeVars)] -- LetStmts , [(LStmt GhcRn (LocatedA (body GhcRn)), FreeVars)] -- BindStmts , [(LStmt GhcRn (LocatedA (body GhcRn)), FreeVars)] ) slurpIndependentStmts stmts = go [] [] emptyNameSet stmts where -- If we encounter a BindStmt that doesn't depend on a previous BindStmt -- in this group, then add it to the group. We have to be careful about -- strict patterns though; splitSegments expects that if we return Just -- then we have actually done some splitting. Otherwise it will go into -- an infinite loop (#14163). go lets indep bndrs ((L loc (BindStmt xbs pat body), fvs): rest) | disjointNameSet bndrs fvs && not (isStrictPattern pat) = go lets ((L loc (BindStmt xbs pat body), fvs) : indep) bndrs' rest where bndrs' = bndrs `unionNameSet` mkNameSet (collectPatBinders CollNoDictBinders pat) -- If we encounter a LetStmt that doesn't depend on a BindStmt in this -- group, then move it to the beginning, so that it doesn't interfere with -- grouping more BindStmts. -- TODO: perhaps we shouldn't do this if there are any strict bindings, -- because we might be moving evaluation earlier. go lets indep bndrs ((L loc (LetStmt noExtField binds), fvs) : rest) | disjointNameSet bndrs fvs = go ((L loc (LetStmt noExtField binds), fvs) : lets) indep bndrs rest go _ [] _ _ = Nothing go _ [_] _ _ = Nothing go lets indep _ stmts = Just (reverse lets, reverse indep, stmts) -- | Build an ApplicativeStmt, and strip the "return" from the tail -- if necessary. -- -- For example, if we start with -- do x <- E1; y <- E2; return (f x y) -- then we get -- do (E1[x] | E2[y]); f x y -- -- the LastStmt in this case has the return removed, but we set the -- flag on the LastStmt to indicate this, so that we can print out the -- original statement correctly in error messages. It is easier to do -- it this way rather than try to ignore the return later in both the -- typechecker and the desugarer (I tried it that way first!). mkApplicativeStmt :: HsStmtContext GhcRn -> [ApplicativeArg GhcRn] -- ^ The args -> Bool -- ^ True <=> need a join -> [ExprLStmt GhcRn] -- ^ The body statements -> RnM ([ExprLStmt GhcRn], FreeVars) mkApplicativeStmt ctxt args need_join body_stmts = do { (fmap_op, fvs1) <- lookupQualifiedDoStmtName ctxt fmapName ; (ap_op, fvs2) <- lookupQualifiedDoStmtName ctxt apAName ; (mb_join, fvs3) <- if need_join then do { (join_op, fvs) <- lookupQualifiedDoStmtName ctxt joinMName ; return (Just join_op, fvs) } else return (Nothing, emptyNameSet) ; let applicative_stmt = noLocA $ ApplicativeStmt noExtField (zip (fmap_op : repeat ap_op) args) mb_join ; return ( applicative_stmt : body_stmts , fvs1 `plusFV` fvs2 `plusFV` fvs3) } -- | Given the statements following an ApplicativeStmt, determine whether -- we need a @join@ or not, and remove the @return@ if necessary. needJoin :: MonadNames -> [ExprLStmt GhcRn] -> (Bool, [ExprLStmt GhcRn]) needJoin _monad_names [] = (False, []) -- we're in an ApplicativeArg needJoin monad_names [L loc (LastStmt _ e _ t)] | Just (arg, wasDollar) <- isReturnApp monad_names e = (False, [L loc (LastStmt noExtField arg (Just wasDollar) t)]) needJoin _monad_names stmts = (True, stmts) -- | @(Just e, False)@, if the expression is @return e@ -- @(Just e, True)@ if the expression is @return $ e@, -- otherwise @Nothing@. isReturnApp :: MonadNames -> LHsExpr GhcRn -> Maybe (LHsExpr GhcRn, Bool) isReturnApp monad_names (L _ (HsPar _ expr)) = isReturnApp monad_names expr isReturnApp monad_names (L _ e) = case e of OpApp _ l op r | is_return l, is_dollar op -> Just (r, True) HsApp _ f arg | is_return f -> Just (arg, False) _otherwise -> Nothing where is_var f (L _ (HsPar _ e)) = is_var f e is_var f (L _ (HsAppType _ e _)) = is_var f e is_var f (L _ (HsVar _ (L _ r))) = f r -- TODO: I don't know how to get this right for rebindable syntax is_var _ _ = False is_return = is_var (\n -> n == return_name monad_names || n == pure_name monad_names) is_dollar = is_var (`hasKey` dollarIdKey) {- ************************************************************************ * * \subsubsection{Errors} * * ************************************************************************ -} checkEmptyStmts :: HsStmtContext GhcRn -> RnM () -- We've seen an empty sequence of Stmts... is that ok? checkEmptyStmts ctxt = unless (okEmpty ctxt) (addErr (emptyErr ctxt)) okEmpty :: HsStmtContext a -> Bool okEmpty (PatGuard {}) = True okEmpty _ = False emptyErr :: HsStmtContext GhcRn -> SDoc emptyErr (ParStmtCtxt {}) = text "Empty statement group in parallel comprehension" emptyErr (TransStmtCtxt {}) = text "Empty statement group preceding 'group' or 'then'" emptyErr ctxt = text "Empty" <+> pprStmtContext ctxt ---------------------- checkLastStmt :: AnnoBody body => HsStmtContext GhcRn -> LStmt GhcPs (LocatedA (body GhcPs)) -> RnM (LStmt GhcPs (LocatedA (body GhcPs))) checkLastStmt ctxt lstmt@(L loc stmt) = case ctxt of ListComp -> check_comp MonadComp -> check_comp ArrowExpr -> check_do DoExpr{} -> check_do MDoExpr{} -> check_do _ -> check_other where check_do -- Expect BodyStmt, and change it to LastStmt = case stmt of BodyStmt _ e _ _ -> return (L loc (mkLastStmt e)) LastStmt {} -> return lstmt -- "Deriving" clauses may generate a -- LastStmt directly (unlike the parser) _ -> do { addErr (hang last_error 2 (ppr stmt)); return lstmt } last_error = (text "The last statement in" <+> pprAStmtContext ctxt <+> text "must be an expression") check_comp -- Expect LastStmt; this should be enforced by the parser! = case stmt of LastStmt {} -> return lstmt _ -> pprPanic "checkLastStmt" (ppr lstmt) check_other -- Behave just as if this wasn't the last stmt = do { checkStmt ctxt lstmt; return lstmt } -- Checking when a particular Stmt is ok checkStmt :: HsStmtContext GhcRn -> LStmt GhcPs (LocatedA (body GhcPs)) -> RnM () checkStmt ctxt (L _ stmt) = do { dflags <- getDynFlags ; case okStmt dflags ctxt stmt of IsValid -> return () NotValid extra -> addErr (msg $$ extra) } where msg = sep [ text "Unexpected" <+> pprStmtCat stmt <+> ptext (sLit "statement") , text "in" <+> pprAStmtContext ctxt ] pprStmtCat :: Stmt (GhcPass a) body -> SDoc pprStmtCat (TransStmt {}) = text "transform" pprStmtCat (LastStmt {}) = text "return expression" pprStmtCat (BodyStmt {}) = text "body" pprStmtCat (BindStmt {}) = text "binding" pprStmtCat (LetStmt {}) = text "let" pprStmtCat (RecStmt {}) = text "rec" pprStmtCat (ParStmt {}) = text "parallel" pprStmtCat (ApplicativeStmt {}) = panic "pprStmtCat: ApplicativeStmt" ------------ emptyInvalid :: Validity -- Payload is the empty document emptyInvalid = NotValid Outputable.empty okStmt, okDoStmt, okCompStmt, okParStmt :: DynFlags -> HsStmtContext GhcRn -> Stmt GhcPs (LocatedA (body GhcPs)) -> Validity -- Return Nothing if OK, (Just extra) if not ok -- The "extra" is an SDoc that is appended to a generic error message okStmt dflags ctxt stmt = case ctxt of PatGuard {} -> okPatGuardStmt stmt ParStmtCtxt ctxt -> okParStmt dflags ctxt stmt DoExpr{} -> okDoStmt dflags ctxt stmt MDoExpr{} -> okDoStmt dflags ctxt stmt ArrowExpr -> okDoStmt dflags ctxt stmt GhciStmtCtxt -> okDoStmt dflags ctxt stmt ListComp -> okCompStmt dflags ctxt stmt MonadComp -> okCompStmt dflags ctxt stmt TransStmtCtxt ctxt -> okStmt dflags ctxt stmt ------------- okPatGuardStmt :: Stmt GhcPs (LocatedA (body GhcPs)) -> Validity okPatGuardStmt stmt = case stmt of BodyStmt {} -> IsValid BindStmt {} -> IsValid LetStmt {} -> IsValid _ -> emptyInvalid ------------- okParStmt dflags ctxt stmt = case stmt of LetStmt _ (HsIPBinds {}) -> emptyInvalid _ -> okStmt dflags ctxt stmt ---------------- okDoStmt dflags ctxt stmt = case stmt of RecStmt {} | LangExt.RecursiveDo `xopt` dflags -> IsValid | ArrowExpr <- ctxt -> IsValid -- Arrows allows 'rec' | otherwise -> NotValid (text "Use RecursiveDo") BindStmt {} -> IsValid LetStmt {} -> IsValid BodyStmt {} -> IsValid _ -> emptyInvalid ---------------- okCompStmt dflags _ stmt = case stmt of BindStmt {} -> IsValid LetStmt {} -> IsValid BodyStmt {} -> IsValid ParStmt {} | LangExt.ParallelListComp `xopt` dflags -> IsValid | otherwise -> NotValid (text "Use ParallelListComp") TransStmt {} | LangExt.TransformListComp `xopt` dflags -> IsValid | otherwise -> NotValid (text "Use TransformListComp") RecStmt {} -> emptyInvalid LastStmt {} -> emptyInvalid -- Should not happen (dealt with by checkLastStmt) ApplicativeStmt {} -> emptyInvalid --------- checkTupleSection :: [HsTupArg GhcPs] -> RnM () checkTupleSection args = do { tuple_section <- xoptM LangExt.TupleSections ; checkErr (all tupArgPresent args || tuple_section) msg } where msg = text "Illegal tuple section: use TupleSections" --------- sectionErr :: HsExpr GhcPs -> SDoc sectionErr expr = hang (text "A section must be enclosed in parentheses") 2 (text "thus:" <+> (parens (ppr expr))) badIpBinds :: Outputable a => SDoc -> a -> SDoc badIpBinds what binds = hang (text "Implicit-parameter bindings illegal in" <+> what) 2 (ppr binds) --------- monadFailOp :: LPat GhcPs -> HsStmtContext GhcRn -> RnM (FailOperator GhcRn, FreeVars) monadFailOp pat ctxt = do dflags <- getDynFlags -- If the pattern is irrefutable (e.g.: wildcard, tuple, ~pat, etc.) -- we should not need to fail. if | isIrrefutableHsPat dflags pat -> return (Nothing, emptyFVs) -- For non-monadic contexts (e.g. guard patterns, list -- comprehensions, etc.) we should not need to fail, or failure is handled in -- a different way. See Note [Failing pattern matches in Stmts]. | not (isMonadStmtContext ctxt) -> return (Nothing, emptyFVs) | otherwise -> getMonadFailOp ctxt {- Note [Monad fail : Rebindable syntax, overloaded strings] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Given the code foo x = do { Just y <- x; return y } we expect it to desugar as foo x = x >>= \r -> case r of Just y -> return y Nothing -> fail "Pattern match error" But with RebindableSyntax and OverloadedStrings, we really want it to desugar thus: foo x = x >>= \r -> case r of Just y -> return y Nothing -> fail (fromString "Patterm match error") So, in this case, we synthesize the function \x -> fail (fromString x) (rather than plain 'fail') for the 'fail' operation. This is done in 'getMonadFailOp'. Similarly with QualifiedDo and OverloadedStrings, we also want to desugar using fromString: foo x = M.do { Just y <- x; return y } ===> foo x = x M.>>= \r -> case r of Just y -> return y Nothing -> M.fail (fromString "Pattern match error") -} getMonadFailOp :: HsStmtContext p -> RnM (FailOperator GhcRn, FreeVars) -- Syntax expr fail op getMonadFailOp ctxt = do { xOverloadedStrings <- fmap (xopt LangExt.OverloadedStrings) getDynFlags ; xRebindableSyntax <- fmap (xopt LangExt.RebindableSyntax) getDynFlags ; (fail, fvs) <- reallyGetMonadFailOp xRebindableSyntax xOverloadedStrings ; return (Just fail, fvs) } where isQualifiedDo = isJust (qualifiedDoModuleName_maybe ctxt) reallyGetMonadFailOp rebindableSyntax overloadedStrings | (isQualifiedDo || rebindableSyntax) && overloadedStrings = do (failExpr, failFvs) <- lookupQualifiedDoExpr ctxt failMName (fromStringExpr, fromStringFvs) <- lookupSyntaxExpr fromStringName let arg_lit = mkVarOcc "arg" arg_name <- newSysName arg_lit let arg_syn_expr = nlHsVar arg_name body :: LHsExpr GhcRn = nlHsApp (noLocA failExpr) (nlHsApp (noLocA $ fromStringExpr) arg_syn_expr) let failAfterFromStringExpr :: HsExpr GhcRn = unLoc $ mkHsLam [noLocA $ VarPat noExtField $ noLocA arg_name] body let failAfterFromStringSynExpr :: SyntaxExpr GhcRn = mkSyntaxExpr failAfterFromStringExpr return (failAfterFromStringSynExpr, failFvs `plusFV` fromStringFvs) | otherwise = lookupQualifiedDo ctxt failMName {- ********************************************************************* * * Generating code for HsExpanded See Note [Handling overloaded and rebindable constructs] * * ********************************************************************* -} genHsApps :: Name -> [LHsExpr GhcRn] -> HsExpr GhcRn genHsApps fun args = foldl genHsApp (genHsVar fun) args genHsApp :: HsExpr GhcRn -> LHsExpr GhcRn -> HsExpr GhcRn genHsApp fun arg = HsApp noAnn (wrapGenSpan fun) arg genLHsVar :: Name -> LHsExpr GhcRn genLHsVar nm = wrapGenSpan $ genHsVar nm genHsVar :: Name -> HsExpr GhcRn genHsVar nm = HsVar noExtField $ wrapGenSpan nm genAppType :: HsExpr GhcRn -> HsType (NoGhcTc GhcRn) -> HsExpr GhcRn genAppType expr = HsAppType noExtField (wrapGenSpan expr) . mkEmptyWildCardBndrs . wrapGenSpan genHsTyLit :: FastString -> HsType GhcRn genHsTyLit = HsTyLit noExtField . HsStrTy NoSourceText wrapGenSpan :: a -> LocatedAn an a -- Wrap something in a "generatedSrcSpan" -- See Note [Rebindable syntax and HsExpansion] wrapGenSpan x = L (noAnnSrcSpan generatedSrcSpan) x -- | Build a 'HsExpansion' out of an extension constructor, -- and the two components of the expansion: original and -- desugared expressions. mkExpandedExpr :: HsExpr GhcRn -- ^ source expression -> HsExpr GhcRn -- ^ expanded expression -> HsExpr GhcRn -- ^ suitably wrapped 'HsExpansion' mkExpandedExpr a b = XExpr (HsExpanded a b) ----------------------------------------- -- Bits and pieces for RecordDotSyntax. -- -- See Note [Overview of record dot syntax] in GHC.Hs.Expr. -- mkGetField arg field calcuates a get_field @field arg expression. -- e.g. z.x = mkGetField z x = get_field @x z mkGetField :: Name -> LHsExpr GhcRn -> Located FieldLabelString -> HsExpr GhcRn mkGetField get_field arg field = unLoc (head $ mkGet get_field [arg] field) -- mkSetField a field b calculates a set_field @field expression. -- e.g mkSetSetField a field b = set_field @"field" a b (read as "set field 'field' on a to b"). mkSetField :: Name -> LHsExpr GhcRn -> Located FieldLabelString -> LHsExpr GhcRn -> HsExpr GhcRn mkSetField set_field a (L _ field) b = genHsApp (genHsApp (genHsVar set_field `genAppType` genHsTyLit field) a) b mkGet :: Name -> [LHsExpr GhcRn] -> Located FieldLabelString -> [LHsExpr GhcRn] mkGet get_field l@(r : _) (L _ field) = wrapGenSpan (genHsApp (genHsVar get_field `genAppType` genHsTyLit field) r) : l mkGet _ [] _ = panic "mkGet : The impossible has happened!" mkSet :: Name -> LHsExpr GhcRn -> (Located FieldLabelString, LHsExpr GhcRn) -> LHsExpr GhcRn mkSet set_field acc (field, g) = wrapGenSpan (mkSetField set_field g field acc) -- mkProjection fields calculates a projection. -- e.g. .x = mkProjection [x] = getField @"x" -- .x.y = mkProjection [.x, .y] = (.y) . (.x) = getField @"y" . getField @"x" mkProjection :: Name -> Name -> NonEmpty (Located FieldLabelString) -> HsExpr GhcRn mkProjection getFieldName circName (field :| fields) = foldl' f (proj field) fields where f :: HsExpr GhcRn -> Located FieldLabelString -> HsExpr GhcRn f acc field = genHsApps circName $ map wrapGenSpan [proj field, acc] proj :: Located FieldLabelString -> HsExpr GhcRn proj (L _ f) = genHsVar getFieldName `genAppType` genHsTyLit f -- mkProjUpdateSetField calculates functions representing dot notation record updates. -- e.g. Suppose an update like foo.bar = 1. -- We calculate the function \a -> setField @"foo" a (setField @"bar" (getField @"foo" a) 1). mkProjUpdateSetField :: Name -> Name -> LHsRecProj GhcRn (LHsExpr GhcRn) -> (LHsExpr GhcRn -> LHsExpr GhcRn) mkProjUpdateSetField get_field set_field (L _ (HsRecField { hsRecFieldLbl = (L _ (FieldLabelStrings flds')), hsRecFieldArg = arg } )) = let { ; flds = map (fmap (unLoc . hflLabel)) flds' ; final = last flds -- quux ; fields = init flds -- [foo, bar, baz] ; getters = \a -> foldl' (mkGet get_field) [a] fields -- Ordered from deep to shallow. -- [getField@"baz"(getField@"bar"(getField@"foo" a), getField@"bar"(getField@"foo" a), getField@"foo" a, a] ; zips = \a -> (final, head (getters a)) : zip (reverse fields) (tail (getters a)) -- Ordered from deep to shallow. -- [("quux", getField@"baz"(getField@"bar"(getField@"foo" a)), ("baz", getField@"bar"(getField@"foo" a)), ("bar", getField@"foo" a), ("foo", a)] } in (\a -> foldl' (mkSet set_field) arg (zips a)) -- setField@"foo" (a) (setField@"bar" (getField @"foo" (a))(setField@"baz" (getField @"bar" (getField @"foo" (a)))(setField@"quux" (getField @"baz" (getField @"bar" (getField @"foo" (a))))(quux)))) mkRecordDotUpd :: Name -> Name -> LHsExpr GhcRn -> [LHsRecUpdProj GhcRn] -> HsExpr GhcRn mkRecordDotUpd get_field set_field exp updates = foldl' fieldUpdate (unLoc exp) updates where fieldUpdate :: HsExpr GhcRn -> LHsRecUpdProj GhcRn -> HsExpr GhcRn fieldUpdate acc lpu = unLoc $ (mkProjUpdateSetField get_field set_field lpu) (wrapGenSpan acc) rnHsUpdProjs :: [LHsRecUpdProj GhcPs] -> RnM ([LHsRecUpdProj GhcRn], FreeVars) rnHsUpdProjs us = do (u, fvs) <- unzip <$> mapM rnRecUpdProj us pure (u, plusFVs fvs) where rnRecUpdProj :: LHsRecUpdProj GhcPs -> RnM (LHsRecUpdProj GhcRn, FreeVars) rnRecUpdProj (L l (HsRecField _ fs arg pun)) = do { (arg, fv) <- rnLExpr arg ; return $ (L l (HsRecField { hsRecFieldAnn = noAnn , hsRecFieldLbl = fmap rnFieldLabelStrings fs , hsRecFieldArg = arg , hsRecPun = pun}), fv) }