diff options
| author | aarne <aarne@cs.chalmers.se> | 2008-06-25 16:54:35 +0000 |
|---|---|---|
| committer | aarne <aarne@cs.chalmers.se> | 2008-06-25 16:54:35 +0000 |
| commit | e9e80fc389365e24d4300d7d5390c7d833a96c50 (patch) | |
| tree | f0b58473adaa670bd8fc52ada419d8cad470ee03 /src/GF/Compile | |
| parent | b96b36f43de3e2f8b58d5f539daa6f6d47f25870 (diff) | |
changed names of resource-1.3; added a note on homepage on release
Diffstat (limited to 'src/GF/Compile')
| -rw-r--r-- | src/GF/Compile/BackOpt.hs | 105 | ||||
| -rw-r--r-- | src/GF/Compile/CheckGrammar.hs | 1105 | ||||
| -rw-r--r-- | src/GF/Compile/Compute.hs | 429 | ||||
| -rw-r--r-- | src/GF/Compile/Export.hs | 61 | ||||
| -rw-r--r-- | src/GF/Compile/Extend.hs | 138 | ||||
| -rw-r--r-- | src/GF/Compile/GFCCtoHaskell.hs | 213 | ||||
| -rw-r--r-- | src/GF/Compile/GFCCtoJS.hs | 117 | ||||
| -rw-r--r-- | src/GF/Compile/GenerateFCFG.hs | 526 | ||||
| -rw-r--r-- | src/GF/Compile/GeneratePMCFG.hs | 356 | ||||
| -rw-r--r-- | src/GF/Compile/GetGrammar.hs | 55 | ||||
| -rw-r--r-- | src/GF/Compile/GrammarToGFCC.hs | 561 | ||||
| -rw-r--r-- | src/GF/Compile/ModDeps.hs | 153 | ||||
| -rw-r--r-- | src/GF/Compile/Optimize.hs | 235 | ||||
| -rw-r--r-- | src/GF/Compile/OptimizeGF.hs | 277 | ||||
| -rw-r--r-- | src/GF/Compile/OptimizeGFCC.hs | 124 | ||||
| -rw-r--r-- | src/GF/Compile/ReadFiles.hs | 195 | ||||
| -rw-r--r-- | src/GF/Compile/Rebuild.hs | 104 | ||||
| -rw-r--r-- | src/GF/Compile/Refresh.hs | 133 | ||||
| -rw-r--r-- | src/GF/Compile/RemoveLiT.hs | 64 | ||||
| -rw-r--r-- | src/GF/Compile/Rename.hs | 338 | ||||
| -rw-r--r-- | src/GF/Compile/TC.hs | 292 | ||||
| -rw-r--r-- | src/GF/Compile/TypeCheck.hs | 118 | ||||
| -rw-r--r-- | src/GF/Compile/Update.hs | 135 |
23 files changed, 5834 insertions, 0 deletions
diff --git a/src/GF/Compile/BackOpt.hs b/src/GF/Compile/BackOpt.hs new file mode 100644 index 000000000..8667023c0 --- /dev/null +++ b/src/GF/Compile/BackOpt.hs @@ -0,0 +1,105 @@ +---------------------------------------------------------------------- +-- | +-- Module : BackOpt +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/04/21 16:21:33 $ +-- > CVS $Author: bringert $ +-- > CVS $Revision: 1.6 $ +-- +-- Optimizations on GF source code: sharing, parametrization, value sets. +-- +-- optimization: sharing branches in tables. AR 25\/4\/2003. +-- following advice of Josef Svenningsson +----------------------------------------------------------------------------- + +module GF.Compile.BackOpt (shareModule, OptSpec) where + +import GF.Grammar.Grammar +import GF.Infra.Ident +import GF.Infra.Option +import qualified GF.Grammar.Macros as C +import GF.Grammar.PrGrammar (prt) +import GF.Data.Operations +import Data.List +import qualified GF.Infra.Modules as M +import qualified Data.ByteString.Char8 as BS + +import Data.Set (Set) +import qualified Data.Set as Set + +type OptSpec = Set Optimization + +shareModule :: OptSpec -> (Ident, SourceModInfo) -> (Ident, SourceModInfo) +shareModule opt (i,m) = case m of + M.ModMod mo -> + (i,M.ModMod (M.replaceJudgements mo (mapTree (shareInfo opt) (M.jments mo)))) + _ -> (i,m) + +shareInfo opt (c, CncCat ty (Yes t) m) = (c,CncCat ty (Yes (shareOptim opt c t)) m) +shareInfo opt (c, CncFun kxs (Yes t) m) = (c,CncFun kxs (Yes (shareOptim opt c t)) m) +shareInfo opt (c, ResOper ty (Yes t)) = (c,ResOper ty (Yes (shareOptim opt c t))) +shareInfo _ i = i + +-- the function putting together optimizations +shareOptim :: OptSpec -> Ident -> Term -> Term +shareOptim opt c = (if OptValues `Set.member` opt then values else id) + . (if OptParametrize `Set.member` opt then factor c 0 else id) + +-- do even more: factor parametric branches + +factor :: Ident -> Int -> Term -> Term +factor c i t = case t of + T _ [_] -> t + T _ [] -> t + T (TComp ty) cs -> + T (TTyped ty) $ factors i [(p, factor c (i+1) v) | (p, v) <- cs] + _ -> C.composSafeOp (factor c i) t + where + + factors i psvs = -- we know psvs has at least 2 elements + let p = qqIdent c i + vs' = map (mkFun p) psvs + in if allEqs vs' + then mkCase p vs' + else psvs + + mkFun p (patt, val) = replace (C.patt2term patt) (Vr p) val + + allEqs (v:vs) = all (==v) vs + + mkCase p (v:_) = [(PV p, v)] + +--- we hope this will be fresh and don't check... in GFC would be safe + +qqIdent c i = identC (BS.pack ("q_" ++ prt c ++ "__" ++ show i)) + + +-- we need to replace subterms + +replace :: Term -> Term -> Term -> Term +replace old new trm = case trm of + + -- these are the important cases, since they can correspond to patterns + QC _ _ | trm == old -> new + App t ts | trm == old -> new + App t ts -> App (repl t) (repl ts) + R _ | isRec && trm == old -> new + _ -> C.composSafeOp repl trm + where + repl = replace old new + isRec = case trm of + R _ -> True + _ -> False + +-- It is very important that this is performed only after case +-- expansion since otherwise the order and number of values can +-- be incorrect. Guaranteed by the TComp flag. + +values :: Term -> Term +values t = case t of + T ty [(ps,t)] -> T ty [(ps,values t)] -- don't destroy parametrization + T (TComp ty) cs -> V ty [values t | (_, t) <- cs] + _ -> C.composSafeOp values t diff --git a/src/GF/Compile/CheckGrammar.hs b/src/GF/Compile/CheckGrammar.hs new file mode 100644 index 000000000..0a8361d36 --- /dev/null +++ b/src/GF/Compile/CheckGrammar.hs @@ -0,0 +1,1105 @@ +{-# LANGUAGE PatternGuards #-} +---------------------------------------------------------------------- +-- | +-- Module : CheckGrammar +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/11/11 23:24:33 $ +-- > CVS $Author: aarne $ +-- > CVS $Revision: 1.31 $ +-- +-- AR 4\/12\/1999 -- 1\/4\/2000 -- 8\/9\/2001 -- 15\/5\/2002 -- 27\/11\/2002 -- 18\/6\/2003 +-- +-- type checking also does the following modifications: +-- +-- - types of operations and local constants are inferred and put in place +-- +-- - both these types and linearization types are computed +-- +-- - tables are type-annotated +----------------------------------------------------------------------------- + +module GF.Compile.CheckGrammar ( + showCheckModule, justCheckLTerm, allOperDependencies, topoSortOpers) where + +import GF.Infra.Ident +import GF.Infra.Modules + +import GF.Compile.TypeCheck + +import GF.Compile.Refresh +import GF.Grammar.Grammar +import GF.Grammar.PrGrammar +import GF.Grammar.Lookup +import GF.Grammar.LookAbs +import GF.Grammar.Predef +import GF.Grammar.Macros +import GF.Grammar.ReservedWords +import GF.Grammar.PatternMatch +import GF.Grammar.AppPredefined +import GF.Grammar.Lockfield (isLockLabel) + +import GF.Data.Operations +import GF.Infra.CheckM + +import Data.List +import qualified Data.Set as Set +import qualified Data.Map as Map +import Control.Monad +import Debug.Trace --- + + +showCheckModule :: [SourceModule] -> SourceModule -> Err ([SourceModule],String) +showCheckModule mos m = do + (st,(_,msg)) <- checkStart $ checkModule mos m + return (st, unlines $ reverse msg) + +mapsCheckTree :: + (Ord a) => ((a,b) -> Check (a,c)) -> BinTree a b -> Check (BinTree a c) +mapsCheckTree f = checkErr . mapsErrTree (\t -> checkStart (f t) >>= return . fst) + + +-- | checking is performed in the dependency order of modules +checkModule :: [SourceModule] -> SourceModule -> Check [SourceModule] +checkModule ms (name,mod) = checkIn ("checking module" +++ prt name) $ case mod of + + ModMod mo -> do + let js = jments mo + checkRestrictedInheritance ms (name, mo) + js' <- case mtype mo of + MTAbstract -> mapsCheckTree (checkAbsInfo gr name mo) js + + MTTransfer a b -> mapsCheckTree (checkAbsInfo gr name mo) js + + MTResource -> mapsCheckTree (checkResInfo gr name mo) js + + MTConcrete a -> do + checkErr $ topoSortOpers $ allOperDependencies name js + ModMod abs <- checkErr $ lookupModule gr a + js1 <- checkCompleteGrammar abs mo + mapsCheckTree (checkCncInfo gr name mo (a,abs)) js1 + + MTInterface -> mapsCheckTree (checkResInfo gr name mo) js + + MTInstance a -> do + ModMod abs <- checkErr $ lookupModule gr a + -- checkCompleteInstance abs mo -- this is done in Rebuild + mapsCheckTree (checkResInfo gr name mo) js + + return $ (name, ModMod (replaceJudgements mo js')) : ms + + _ -> return $ (name,mod) : ms + where + gr = MGrammar $ (name,mod):ms + +-- check if restricted inheritance modules are still coherent +-- i.e. that the defs of remaining names don't depend on omitted names +---checkRestrictedInheritance :: [SourceModule] -> SourceModule -> Check () +checkRestrictedInheritance mos (name,mo) = do + let irs = [ii | ii@(_,mi) <- extend mo, mi /= MIAll] -- names with restr. inh. + let mrs = [((i,m),mi) | (i,ModMod m) <- mos, Just mi <- [lookup i irs]] + -- the restr. modules themself, with restr. infos + mapM_ checkRem mrs + where + checkRem ((i,m),mi) = do + let (incl,excl) = partition (isInherited mi) (map fst (tree2list (jments m))) + let incld c = Set.member c (Set.fromList incl) + let illegal c = Set.member c (Set.fromList excl) + let illegals = [(f,is) | + (f,cs) <- allDeps, incld f, let is = filter illegal cs, not (null is)] + case illegals of + [] -> return () + cs -> fail $ "In inherited module" +++ prt i ++ + ", dependence of excluded constants:" ++++ + unlines [" " ++ prt f +++ "on" +++ unwords (map prt is) | + (f,is) <- cs] + allDeps = ---- transClosure $ Map.fromList $ + concatMap (allDependencies (const True)) + [jments m | (_,ModMod m) <- mos] + transClosure ds = ds ---- TODO: check in deeper modules + +-- | check if a term is typable +justCheckLTerm :: SourceGrammar -> Term -> Err Term +justCheckLTerm src t = do + ((t',_),_) <- checkStart (inferLType src t) + return t' + +checkAbsInfo :: + SourceGrammar -> Ident -> Module Ident Info -> (Ident,Info) -> Check (Ident,Info) +checkAbsInfo st m mo (c,info) = do +---- checkReservedId c + case info of + AbsCat (Yes cont) _ -> mkCheck "category" $ + checkContext st cont ---- also cstrs + AbsFun (Yes typ0) md -> do + typ <- compAbsTyp [] typ0 -- to calculate let definitions + mkCheck "type of function" $ checkTyp st typ + md' <- case md of + Yes d -> do + let d' = elimTables d + mkCheckWarn "definition of function" $ checkEquation st (m,c) d' + return $ Yes d' + _ -> return md + return $ (c,AbsFun (Yes typ) md') + _ -> return (c,info) + where + mkCheck cat ss = case ss of + [] -> return (c,info) + ["[]"] -> return (c,info) ---- + _ -> checkErr $ Bad (unlines ss ++++ "in" +++ cat +++ prt c +++ pos c) + ---- temporary solution when tc of defs is incomplete + mkCheckWarn cat ss = case ss of + [] -> return (c,info) + ["[]"] -> return (c,info) ---- + _ -> do + checkWarn (unlines ss ++++ "in" +++ cat +++ prt c +++ pos c) + return (c,info) + + pos c = showPosition mo c + + compAbsTyp g t = case t of + Vr x -> maybe (fail ("no value given to variable" +++ prt x)) return $ lookup x g + Let (x,(_,a)) b -> do + a' <- compAbsTyp g a + compAbsTyp ((x, a'):g) b + Prod x a b -> do + a' <- compAbsTyp g a + b' <- compAbsTyp ((x,Vr x):g) b + return $ Prod x a' b' + Abs _ _ -> return t + _ -> composOp (compAbsTyp g) t + + elimTables e = case e of + S t a -> elimSel (elimTables t) (elimTables a) + T _ cs -> Eqs [(elimPatt p, elimTables t) | (p,t) <- cs] + _ -> composSafeOp elimTables e + elimPatt p = case p of + PR lps -> map snd lps + _ -> [p] + elimSel t a = case a of + R fs -> mkApp t (map (snd . snd) fs) + _ -> mkApp t [a] + +checkCompleteGrammar :: SourceAbs -> SourceCnc -> Check (BinTree Ident Info) +checkCompleteGrammar abs cnc = do + let js = jments cnc + let fs = tree2list $ jments abs + foldM checkOne js fs + where + checkOne js i@(c,info) = case info of + AbsFun (Yes _) _ -> case lookupIdent c js of + Ok _ -> return js + _ -> do + checkWarn $ "WARNING: no linearization of" +++ prt c + return js + AbsCat (Yes _) _ -> case lookupIdent c js of + Ok (AnyInd _ _) -> return js + Ok (CncCat (Yes _) _ _) -> return js + Ok (CncCat _ mt mp) -> do + checkWarn $ + "Warning: no linearization type for" +++ prt c ++ + ", inserting default {s : Str}" + return $ updateTree (c,CncCat (Yes defLinType) mt mp) js + _ -> do + checkWarn $ + "Warning: no linearization type for" +++ prt c ++ + ", inserting default {s : Str}" + return $ updateTree (c,CncCat (Yes defLinType) nope nope) js + _ -> return js + +-- | General Principle: only Yes-values are checked. +-- A May-value has always been checked in its origin module. +checkResInfo :: + SourceGrammar -> Ident -> Module Ident Info -> (Ident,Info) -> Check (Ident,Info) +checkResInfo gr mo mm (c,info) = do + checkReservedId c + case info of + ResOper pty pde -> chIn "operation" $ do + (pty', pde') <- case (pty,pde) of + (Yes ty, Yes de) -> do + ty' <- check ty typeType >>= comp . fst + (de',_) <- check de ty' + return (Yes ty', Yes de') + (_, Yes de) -> do + (de',ty') <- infer de + return (Yes ty', Yes de') + (_,Nope) -> do + checkWarn "No definition given to oper" + return (pty,pde) + _ -> return (pty, pde) --- other cases are uninteresting + return (c, ResOper pty' pde') + + ResOverload os tysts -> chIn "overloading" $ do + tysts' <- mapM (uncurry $ flip check) tysts -- return explicit ones + tysts0 <- checkErr $ lookupOverload gr mo c -- check against inherited ones too + tysts1 <- mapM (uncurry $ flip check) + [(mkFunType args val,tr) | (args,(val,tr)) <- tysts0] + let tysts2 = [(y,x) | (x,y) <- tysts1] + --- this can only be a partial guarantee, since matching + --- with value type is only possible if expected type is given + checkUniq $ + sort [t : map snd xs | (x,_) <- tysts2, Ok (xs,t) <- [typeFormCnc x]] + return (c,ResOverload os [(y,x) | (x,y) <- tysts']) + + ResParam (Yes (pcs,_)) -> chIn "parameter type" $ do +---- mapM ((mapM (computeLType gr . snd)) . snd) pcs + mapM_ ((mapM_ (checkIfParType gr . snd)) . snd) pcs + ts <- checkErr $ lookupParamValues gr mo c + return (c,ResParam (Yes (pcs, Just ts))) + + _ -> return (c,info) + where + infer = inferLType gr + check = checkLType gr + chIn cat = checkIn ("Happened in" +++ cat +++ prt c +++ pos c +++ ":") + comp = computeLType gr + pos c = showPosition mm c + + checkUniq xss = case xss of + x:y:xs + | x == y -> raise $ "ambiguous for type" +++ + prtType gr (mkFunType (tail x) (head x)) + | otherwise -> checkUniq $ y:xs + _ -> return () + + +checkCncInfo :: SourceGrammar -> Ident -> Module Ident Info -> + (Ident,SourceAbs) -> + (Ident,Info) -> Check (Ident,Info) +checkCncInfo gr m mo (a,abs) (c,info) = do + checkReservedId c + case info of + + CncFun _ (Yes trm) mpr -> chIn "linearization of" $ do + typ <- checkErr $ lookupFunType gr a c + cat0 <- checkErr $ valCat typ + (cont,val) <- linTypeOfType gr m typ -- creates arg vars + (trm',_) <- check trm (mkFunType (map snd cont) val) -- erases arg vars + checkPrintname gr mpr + cat <- return $ snd cat0 + return (c, CncFun (Just (cat,(cont,val))) (Yes trm') mpr) + -- cat for cf, typ for pe + + CncCat (Yes typ) mdef mpr -> chIn "linearization type of" $ do + checkErr $ lookupCatContext gr a c + typ' <- checkIfLinType gr typ + mdef' <- case mdef of + Yes def -> do + (def',_) <- checkLType gr def (mkFunType [typeStr] typ) + return $ Yes def' + _ -> return mdef + checkPrintname gr mpr + return (c,CncCat (Yes typ') mdef' mpr) + + _ -> checkResInfo gr m mo (c,info) + + where + env = gr + infer = inferLType gr + comp = computeLType gr + check = checkLType gr + chIn cat = checkIn ("Happened in" +++ cat +++ prt c +++ pos c +++ ":") + pos c = showPosition mo c + +checkIfParType :: SourceGrammar -> Type -> Check () +checkIfParType st typ = checkCond ("Not parameter type" +++ prt typ) (isParType typ) + where + isParType ty = True ---- +{- case ty of + Cn typ -> case lookupConcrete st typ of + Ok (CncParType _ _ _) -> True + Ok (CncOper _ ty' _) -> isParType ty' + _ -> False + Q p t -> case lookupInPackage st (p,t) of + Ok (CncParType _ _ _) -> True + _ -> False + RecType r -> all (isParType . snd) r + _ -> False +-} + +checkIfStrType :: SourceGrammar -> Type -> Check () +checkIfStrType st typ = case typ of + Table arg val -> do + checkIfParType st arg + checkIfStrType st val + _ | typ == typeStr -> return () + _ -> prtFail "not a string type" typ + + +checkIfLinType :: SourceGrammar -> Type -> Check Type +checkIfLinType st typ0 = do + typ <- computeLType st typ0 +{- ---- should check that not fun type + case typ of + RecType r -> do + let (lins,ihs) = partition (isLinLabel .fst) r + --- checkErr $ checkUnique $ map fst r + mapM_ checkInh ihs + mapM_ checkLin lins + _ -> prtFail "a linearization type cannot be" typ +-} + return typ + + where + checkInh (label,typ) = checkIfParType st typ + checkLin (label,typ) = return () ---- checkIfStrType st typ + + +computeLType :: SourceGrammar -> Type -> Check Type +computeLType gr t = do + g0 <- checkGetContext + let g = [(x, Vr x) | (x,_) <- g0] + checkInContext g $ comp t + where + comp ty = case ty of + _ | Just _ <- isTypeInts ty -> return ty ---- shouldn't be needed + | isPredefConstant ty -> return ty ---- shouldn't be needed + + Q m ident -> checkIn ("module" +++ prt m) $ do + ty' <- checkErr (lookupResDef gr m ident) + if ty' == ty then return ty else comp ty' --- is this necessary to test? + + Vr ident -> checkLookup ident -- never needed to compute! + + App f a -> do + f' <- comp f + a' <- comp a + case f' of + Abs x b -> checkInContext [(x,a')] $ comp b + _ -> return $ App f' a' + + Prod x a b -> do + a' <- comp a + b' <- checkInContext [(x,Vr x)] $ comp b + return $ Prod x a' b' + + Abs x b -> do + b' <- checkInContext [(x,Vr x)] $ comp b + return $ Abs x b' + + ExtR r s -> do + r' <- comp r + s' <- comp s + case (r',s') of + (RecType rs, RecType ss) -> checkErr (plusRecType r' s') >>= comp + _ -> return $ ExtR r' s' + + RecType fs -> do + let fs' = sortRec fs + liftM RecType $ mapPairsM comp fs' + + _ | ty == typeTok -> return typeStr + _ | isPredefConstant ty -> return ty + + _ -> composOp comp ty + +checkPrintname :: SourceGrammar -> Perh Term -> Check () +checkPrintname st (Yes t) = checkLType st t typeStr >> return () +checkPrintname _ _ = return () + +-- | for grammars obtained otherwise than by parsing ---- update!! +checkReservedId :: Ident -> Check () +checkReservedId x = let c = prt x in + if isResWord c + then checkWarn ("Warning: reserved word used as identifier:" +++ c) + else return () + +-- to normalize records and record types +labelIndex :: Type -> Label -> Int +labelIndex ty lab = case ty of + RecType ts -> maybe (error ("label index" +++ prt lab)) id $ lookup lab $ labs ts + _ -> error $ "label index" +++ prt ty + where + labs ts = zip (map fst (sortRec ts)) [0..] + +-- the underlying algorithms + +inferLType :: SourceGrammar -> Term -> Check (Term, Type) +inferLType gr trm = case trm of + + Q m ident | isPredef m -> termWith trm $ checkErr (typPredefined ident) + + Q m ident -> checks [ + termWith trm $ checkErr (lookupResType gr m ident) >>= comp + , + checkErr (lookupResDef gr m ident) >>= infer + , + prtFail "cannot infer type of constant" trm + ] + + QC m ident | isPredef m -> termWith trm $ checkErr (typPredefined ident) + + QC m ident -> checks [ + termWith trm $ checkErr (lookupResType gr m ident) >>= comp + , + checkErr (lookupResDef gr m ident) >>= infer + , + prtFail "cannot infer type of canonical constant" trm + ] + + Val ty i -> termWith trm $ return ty + + Vr ident -> termWith trm $ checkLookup ident + + Typed e t -> do + t' <- comp t + check e t' + return (e,t') + + App f a -> do + over <- getOverload gr Nothing trm + case over of + Just trty -> return trty + _ -> do + (f',fty) <- infer f + fty' <- comp fty + case fty' of + Prod z arg val -> do + a' <- justCheck a arg + ty <- if isWildIdent z + then return val + else substituteLType [(z,a')] val + return (App f' a',ty) + _ -> raise ("function type expected for"+++ + prt f +++"instead of" +++ prtType env fty) + + S f x -> do + (f', fty) <- infer f + case fty of + Table arg val -> do + x'<- justCheck x arg + return (S f' x', val) + _ -> prtFail "table lintype expected for the table in" trm + + P t i -> do + (t',ty) <- infer t --- ?? + ty' <- comp ty +----- let tr2 = PI t' i (labelIndex ty' i) + let tr2 = P t' i + termWith tr2 $ checkErr $ case ty' of + RecType ts -> maybeErr ("unknown label" +++ prt i +++ "in" +++ prt ty') $ + lookup i ts + _ -> prtBad ("record type expected for" +++ prt t +++ "instead of") ty' + PI t i _ -> infer $ P t i + + R r -> do + let (ls,fs) = unzip r + fsts <- mapM inferM fs + let ts = [ty | (Just ty,_) <- fsts] + checkCond ("cannot infer type of record"+++ prt trm) (length ts == length fsts) + return $ (R (zip ls fsts), RecType (zip ls ts)) + + T (TTyped arg) pts -> do + (_,val) <- checks $ map (inferCase (Just arg)) pts + check trm (Table arg val) + T (TComp arg) pts -> do + (_,val) <- checks $ map (inferCase (Just arg)) pts + check trm (Table arg val) + T ti pts -> do -- tries to guess: good in oper type inference + let pts' = [pt | pt@(p,_) <- pts, isConstPatt p] + case pts' of + [] -> prtFail "cannot infer table type of" trm +---- PInt k : _ -> return $ Ints $ max [i | PInt i <- pts'] + _ -> do + (arg,val) <- checks $ map (inferCase Nothing) pts' + check trm (Table arg val) + V arg pts -> do + (_,val) <- checks $ map infer pts + return (trm, Table arg val) + + K s -> do + if elem ' ' s + then do + let ss = foldr C Empty (map K (words s)) + ----- removed irritating warning AR 24/5/2008 + ----- checkWarn ("WARNING: token \"" ++ s ++ + ----- "\" converted to token list" ++ prt ss) + return (ss, typeStr) + else return (trm, typeStr) + + EInt i -> return (trm, typeInt) + + EFloat i -> return (trm, typeFloat) + + Empty -> return (trm, typeStr) + + C s1 s2 -> + check2 (flip justCheck typeStr) C s1 s2 typeStr + + Glue s1 s2 -> + check2 (flip justCheck typeStr) Glue s1 s2 typeStr ---- typeTok + +---- hack from Rename.identRenameTerm, to live with files with naming conflicts 18/6/2007 + Strs (Cn c : ts) | c == cConflict -> do + trace ("WARNING: unresolved constant, could be any of" +++ unwords (map prt ts)) (infer $ head ts) +-- checkWarn ("WARNING: unresolved constant, could be any of" +++ unwords (map prt ts)) +-- infer $ head ts + + Strs ts -> do + ts' <- mapM (\t -> justCheck t typeStr) ts + return (Strs ts', typeStrs) + + Alts (t,aa) -> do + t' <- justCheck t typeStr + aa' <- flip mapM aa (\ (c,v) -> do + c' <- justCheck c typeStr + v' <- justCheck v typeStrs + return (c',v')) + return (Alts (t',aa'), typeStr) + + RecType r -> do + let (ls,ts) = unzip r + ts' <- mapM (flip justCheck typeType) ts + return (RecType (zip ls ts'), typeType) + + ExtR r s -> do + (r',rT) <- infer r + rT' <- comp rT + (s',sT) <- infer s + sT' <- comp sT + + let trm' = ExtR r' s' + ---- trm' <- checkErr $ plusRecord r' s' + case (rT', sT') of + (RecType rs, RecType ss) -> do + rt <- checkErr $ plusRecType rT' sT' + check trm' rt ---- return (trm', rt) + _ | rT' == typeType && sT' == typeType -> return (trm', typeType) + _ -> prtFail "records or record types expected in" trm + + Sort _ -> + termWith trm $ return typeType + + Prod x a b -> do + a' <- justCheck a typeType + b' <- checkInContext [(x,a')] $ justCheck b typeType + return (Prod x a' b', typeType) + + Table p t -> do + p' <- justCheck p typeType --- check p partype! + t' <- justCheck t typeType + return $ (Table p' t', typeType) + + FV vs -> do + (_,ty) <- checks $ map infer vs +--- checkIfComplexVariantType trm ty + check trm ty + + EPattType ty -> do + ty' <- justCheck ty typeType + return (ty',typeType) + EPatt p -> do + ty <- inferPatt p + return (trm, EPattType ty) + + _ -> prtFail "cannot infer lintype of" trm + + where + env = gr + infer = inferLType env + comp = computeLType env + + check = checkLType env + + isPredef m = elem m [cPredef,cPredefAbs] + + justCheck ty te = check ty te >>= return . fst + + -- for record fields, which may be typed + inferM (mty, t) = do + (t', ty') <- case mty of + Just ty -> check ty t + _ -> infer t + return (Just ty',t') + + inferCase mty (patt,term) = do + arg <- maybe (inferPatt patt) return mty + cont <- pattContext env arg patt + i <- checkUpdates cont + (_,val) <- infer term + checkResets i + return (arg,val) + isConstPatt p = case p of + PC _ ps -> True --- all isConstPatt ps + PP _ _ ps -> True --- all isConstPatt ps + PR ps -> all (isConstPatt . snd) ps + PT _ p -> isConstPatt p + PString _ -> True + PInt _ -> True + PFloat _ -> True + PChar -> True + PChars _ -> True + PSeq p q -> isConstPatt p && isConstPatt q + PAlt p q -> isConstPatt p && isConstPatt q + PRep p -> isConstPatt p + PNeg p -> isConstPatt p + PAs _ p -> isConstPatt p + _ -> False + + inferPatt p = case p of + PP q c ps | q /= cPredef -> checkErr $ lookupResType gr q c >>= valTypeCnc + PAs _ p -> inferPatt p + PNeg p -> inferPatt p + PAlt p q -> checks [inferPatt p, inferPatt q] + PSeq _ _ -> return $ typeStr + PRep _ -> return $ typeStr + PChar -> return $ typeStr + PChars _ -> return $ typeStr + _ -> infer (patt2term p) >>= return . snd + + +-- type inference: Nothing, type checking: Just t +-- the latter permits matching with value type +getOverload :: SourceGrammar -> Maybe Type -> Term -> Check (Maybe (Term,Type)) +getOverload env@gr mt ot = case appForm ot of + (f@(Q m c), ts) -> case lookupOverload gr m c of + Ok typs -> do + ttys <- mapM infer ts + v <- matchOverload f typs ttys + return $ Just v + _ -> return Nothing + _ -> return Nothing + where + infer = inferLType env + matchOverload f typs ttys = do + let (tts,tys) = unzip ttys + let vfs = lookupOverloadInstance tys typs + let matches = [vf | vf@((v,_),_) <- vfs, matchVal mt v] + + case ([vf | (vf,True) <- matches],[vf | (vf,False) <- matches]) of + ([(val,fun)],_) -> return (mkApp fun tts, val) + ([],[(val,fun)]) -> do + checkWarn ("ignoring lock fields in resolving" +++ prt ot) + return (mkApp fun tts, val) + ([],[]) -> do + raise $ "no overload instance of" +++ prt f +++ + "for" +++ unwords (map (prtType env) tys) +++ "among" ++++ + unlines [" " ++ unwords (map (prtType env) ty) | (ty,_) <- typs] ++ + maybe [] (("with value type" +++) . prtType env) mt + + (vfs1,vfs2) -> case (noProds vfs1,noProds vfs2) of + ([(val,fun)],_) -> do + return (mkApp fun tts, val) + ([],[(val,fun)]) -> do + checkWarn ("ignoring lock fields in resolving" +++ prt ot) + return (mkApp fun tts, val) + +----- unsafely exclude irritating warning AR 24/5/2008 +----- checkWarn $ "WARNING: overloading of" +++ prt f +++ +----- "resolved by excluding partial applications:" ++++ +----- unlines [prtType env ty | (ty,_) <- vfs', not (noProd ty)] + + + _ -> raise $ "ambiguous overloading of" +++ prt f +++ + "for" +++ unwords (map (prtType env) tys) ++++ "with alternatives" ++++ + unlines [prtType env ty | (ty,_) <- if (null vfs1) then vfs2 else vfs2] + + matchVal mt v = elem mt [Nothing,Just v,Just (unlocked v)] + + unlocked v = case v of + RecType fs -> RecType $ filter (not . isLockLabel . fst) fs + _ -> v + ---- TODO: accept subtypes + ---- TODO: use a trie + lookupOverloadInstance tys typs = + [((mkFunType rest val, t),isExact) | + let lt = length tys, + (ty,(val,t)) <- typs, length ty >= lt, + let (pre,rest) = splitAt lt ty, + let isExact = pre == tys, + isExact || map unlocked pre == map unlocked tys + ] + + noProds vfs = [(v,f) | (v,f) <- vfs, noProd v] + + noProd ty = case ty of + Prod _ _ _ -> False + _ -> True + +checkLType :: SourceGrammar -> Term -> Type -> Check (Term, Type) +checkLType env trm typ0 = do + + typ <- comp typ0 + + case trm of + + Abs x c -> do + case typ of + Prod z a b -> do + checkUpdate (x,a) + (c',b') <- if isWildIdent z + then check c b + else do + b' <- checkIn "abs" $ substituteLType [(z,Vr x)] b + check c b' + checkReset + return $ (Abs x c', Prod x a b') + _ -> raise $ "product expected instead of" +++ prtType env typ + + App f a -> do + over <- getOverload env (Just typ) trm + case over of + Just trty -> return trty + _ -> do + (trm',ty') <- infer trm + termWith trm' $ checkEq typ ty' trm' + + Q _ _ -> do + over <- getOverload env (Just typ) trm + case over of + Just trty -> return trty + _ -> do + (trm',ty') <- infer trm + termWith trm' $ checkEq typ ty' trm' + + T _ [] -> + prtFail "found empty table in type" typ + T _ cs -> case typ of + Table arg val -> do + case allParamValues env arg of + Ok vs -> do + let ps0 = map fst cs + ps <- checkErr $ testOvershadow ps0 vs + if null ps + then return () + else checkWarn $ "WARNING: patterns never reached:" +++ + concat (intersperse ", " (map prt ps)) + + _ -> return () -- happens with variable types + cs' <- mapM (checkCase arg val) cs + return (T (TTyped arg) cs', typ) + _ -> raise $ "table type expected for table instead of" +++ prtType env typ + + R r -> case typ of --- why needed? because inference may be too difficult + RecType rr -> do + let (ls,_) = unzip rr -- labels of expected type + fsts <- mapM (checkM r) rr -- check that they are found in the record + return $ (R fsts, typ) -- normalize record + + _ -> prtFail "record type expected in type checking instead of" typ + + ExtR r s -> case typ of + _ | typ == typeType -> do + trm' <- comp trm + case trm' of + RecType _ -> termWith trm $ return typeType + ExtR (Vr _) (RecType _) -> termWith trm $ return typeType + -- ext t = t ** ... + _ -> prtFail "invalid record type extension" trm + RecType rr -> do + (r',ty,s') <- checks [ + do (r',ty) <- infer r + return (r',ty,s) + , + do (s',ty) <- infer s + return (s',ty,r) + ] + case ty of + RecType rr1 -> do + let (rr0,rr2) = recParts rr rr1 + r2 <- justCheck r' rr0 + s2 <- justCheck s' rr2 + return $ (ExtR r2 s2, typ) + _ -> raise ("record type expected in extension of" +++ prt r +++ + "but found" +++ prt ty) + + ExtR ty ex -> do + r' <- justCheck r ty + s' <- justCheck s ex + return $ (ExtR r' s', typ) --- is this all? + + _ -> prtFail "record extension not meaningful for" typ + + FV vs -> do + ttys <- mapM (flip check typ) vs +--- checkIfComplexVariantType trm typ + return (FV (map fst ttys), typ) --- typ' ? + + S tab arg -> checks [ do + (tab',ty) <- infer tab + ty' <- comp ty + case ty' of + Table p t -> do + (arg',val) <- check arg p + checkEq typ t trm + return (S tab' arg', t) + _ -> raise $ "table type expected for applied table instead of" +++ + prtType env ty' + , do + (arg',ty) <- infer arg + ty' <- comp ty + (tab',_) <- check tab (Table ty' typ) + return (S tab' arg', typ) + ] + Let (x,(mty,def)) body -> case mty of + Just ty -> do + (def',ty') <- check def ty + checkUpdate (x,ty') + body' <- justCheck body typ + checkReset + return (Let (x,(Just ty',def')) body', typ) + _ -> do + (def',ty) <- infer def -- tries to infer type of local constant + check (Let (x,(Just ty,def')) body) typ + + _ -> do + (trm',ty') <- infer trm + termWith trm' $ checkEq typ ty' trm' + where + cnc = env + infer = inferLType env + comp = computeLType env + + check = checkLType env + + justCheck ty te = check ty te >>= return . fst + + checkEq = checkEqLType env + + recParts rr t = (RecType rr1,RecType rr2) where + (rr1,rr2) = partition (flip elem (map fst t) . fst) rr + + checkM rms (l,ty) = case lookup l rms of + Just (Just ty0,t) -> do + checkEq ty ty0 t + (t',ty') <- check t ty + return (l,(Just ty',t')) + Just (_,t) -> do + (t',ty') <- check t ty + return (l,(Just ty',t')) + _ -> prtFail "cannot find value for label" l + + checkCase arg val (p,t) = do + cont <- pattContext env arg p + i <- checkUpdates cont + t' <- justCheck t val + checkResets i + return (p,t') + +pattContext :: LTEnv -> Type -> Patt -> Check Context +pattContext env typ p = case p of + PV x | not (isWildIdent x) -> return [(x,typ)] + PP q c ps | q /= cPredef -> do ---- why this /=? AR 6/1/2006 + t <- checkErr $ lookupResType cnc q c + (cont,v) <- checkErr $ typeFormCnc t + checkCond ("wrong number of arguments for constructor in" +++ prt p) + (length cont == length ps) + checkEqLType env typ v (patt2term p) + mapM (uncurry (pattContext env)) (zip (map snd cont) ps) >>= return . concat + PR r -> do + typ' <- computeLType env typ + case typ' of + RecType t -> do + let pts = [(ty,tr) | (l,tr) <- r, Just ty <- [lookup l t]] + ----- checkWarn $ prt p ++++ show pts ----- debug + mapM (uncurry (pattContext env)) pts >>= return . concat + _ -> prtFail "record type expected for pattern instead of" typ' + PT t p' -> do + checkEqLType env typ t (patt2term p') + pattContext env typ p' + + PAs x p -> do + g <- pattContext env typ p + return $ (x,typ):g + + PAlt p' q -> do + g1 <- pattContext env typ p' + g2 <- pattContext env typ q + let pts = [pt | pt <- g1, notElem pt g2] ++ [pt | pt <- g2, notElem pt g1] + checkCond + ("incompatible bindings of" +++ + unwords (nub (map (prt . fst) pts))+++ + "in pattern alterantives" +++ prt p) (null pts) + return g1 -- must be g1 == g2 + PSeq p q -> do + g1 <- pattContext env typ p + g2 <- pattContext env typ q + return $ g1 ++ g2 + PRep p' -> noBind typeStr p' + PNeg p' -> noBind typ p' + + _ -> return [] ---- check types! + where + cnc = env + noBind typ p' = do + co <- pattContext env typ p' + if not (null co) + then checkWarn ("no variable bound inside pattern" +++ prt p) + >> return [] + else return [] + +-- auxiliaries + +type LTEnv = SourceGrammar + +termWith :: Term -> Check Type -> Check (Term, Type) +termWith t ct = do + ty <- ct + return (t,ty) + +-- | light-weight substitution for dep. types +substituteLType :: Context -> Type -> Check Type +substituteLType g t = case t of + Vr x -> return $ maybe t id $ lookup x g + _ -> composOp (substituteLType g) t + +-- | compositional check\/infer of binary operations +check2 :: (Term -> Check Term) -> (Term -> Term -> Term) -> + Term -> Term -> Type -> Check (Term,Type) +check2 chk con a b t = do + a' <- chk a + b' <- chk b + return (con a' b', t) + +checkEqLType :: LTEnv -> Type -> Type -> Term -> Check Type +checkEqLType env t u trm = do + (b,t',u',s) <- checkIfEqLType env t u trm + case b of + True -> return t' + False -> raise $ s +++ "type of" +++ prt trm +++ + ": expected:" +++ prtType env t ++++ + "inferred:" +++ prtType env u + +checkIfEqLType :: LTEnv -> Type -> Type -> Term -> Check (Bool,Type,Type,String) +checkIfEqLType env t u trm = do + t' <- comp t + u' <- comp u + case t' == u' || alpha [] t' u' of + True -> return (True,t',u',[]) + -- forgive missing lock fields by only generating a warning. + --- better: use a flag to forgive? (AR 31/1/2006) + _ -> case missingLock [] t' u' of + Ok lo -> do + checkWarn $ "WARNING: missing lock field" +++ unwords (map prt lo) + return (True,t',u',[]) + Bad s -> return (False,t',u',s) + + where + + -- t is a subtype of u + --- quick hack version of TC.eqVal + alpha g t u = case (t,u) of + + -- error (the empty type!) is subtype of any other type + (_,u) | u == typeError -> True + + -- contravariance + (Prod x a b, Prod y c d) -> alpha g c a && alpha ((x,y):g) b d + + -- record subtyping + (RecType rs, RecType ts) -> all (\ (l,a) -> + any (\ (k,b) -> alpha g a b && l == k) ts) rs + (ExtR r s, ExtR r' s') -> alpha g r r' && alpha g s s' + (ExtR r s, t) -> alpha g r t || alpha g s t + + -- the following say that Ints n is a subset of Int and of Ints m >= n + (t,u) | Just m <- isTypeInts t, Just n <- isTypeInts t -> m >= n + | Just _ <- isTypeInts t, u == typeInt -> True ---- check size! + | t == typeInt, Just _ <- isTypeInts u -> True ---- why this ???? AR 11/12/2005 + + ---- this should be made in Rename + (Q m a, Q n b) | a == b -> elem m (allExtendsPlus env n) + || elem n (allExtendsPlus env m) + || m == n --- for Predef + (QC m a, QC n b) | a == b -> elem m (allExtendsPlus env n) + || elem n (allExtendsPlus env m) + (QC m a, Q n b) | a == b -> elem m (allExtendsPlus env n) + || elem n (allExtendsPlus env m) + (Q m a, QC n b) | a == b -> elem m (allExtendsPlus env n) + || elem n (allExtendsPlus env m) + + (Table a b, Table c d) -> alpha g a c && alpha g b d + (Vr x, Vr y) -> x == y || elem (x,y) g || elem (y,x) g + _ -> t == u + --- the following should be one-way coercions only. AR 4/1/2001 + || elem t sTypes && elem u sTypes + || (t == typeType && u == typePType) + || (u == typeType && t == typePType) + + missingLock g t u = case (t,u) of + (RecType rs, RecType ts) -> + let + ls = [l | (l,a) <- rs, + not (any (\ (k,b) -> alpha g a b && l == k) ts)] + (locks,others) = partition isLockLabel ls + in case others of + _:_ -> Bad $ "missing record fields" +++ unwords (map prt others) + _ -> return locks + -- contravariance + (Prod x a b, Prod y c d) -> do + ls1 <- missingLock g c a + ls2 <- missingLock g b d + return $ ls1 ++ ls2 + + _ -> Bad "" + + sTypes = [typeStr, typeTok, typeString] + comp = computeLType env + +-- printing a type with a lock field lock_C as C +prtType :: LTEnv -> Type -> String +prtType env ty = case ty of + RecType fs -> case filter isLockLabel $ map fst fs of + [lock] -> (drop 5 $ prt lock) --- ++++ "Full form" +++ prt ty + _ -> prtt ty + Prod x a b -> prtType env a +++ "->" +++ prtType env b + _ -> prtt ty + where + prtt t = prt t + ---- use computeLType gr to check if really equal to the cat with lock + + +-- | linearization types and defaults +linTypeOfType :: SourceGrammar -> Ident -> Type -> Check (Context,Type) +linTypeOfType cnc m typ = do + (cont,cat) <- checkErr $ typeSkeleton typ + val <- lookLin cat + args <- mapM mkLinArg (zip [0..] cont) + return (args, val) + where + mkLinArg (i,(n,mc@(m,cat))) = do + val <- lookLin mc + let vars = mkRecType varLabel $ replicate n typeStr + symb = argIdent n cat i + rec <- if n==0 then return val else + checkErr $ errIn ("extending" +++ prt vars +++ "with" +++ prt val) $ + plusRecType vars val + return (symb,rec) + lookLin (_,c) = checks [ --- rather: update with defLinType ? + checkErr (lookupLincat cnc m c) >>= computeLType cnc + ,return defLinType + ] + +-- | dependency check, detecting circularities and returning topo-sorted list + +allOperDependencies :: Ident -> BinTree Ident Info -> [(Ident,[Ident])] +allOperDependencies m = allDependencies (==m) + +allDependencies :: (Ident -> Bool) -> BinTree Ident Info -> [(Ident,[Ident])] +allDependencies ism b = + [(f, nub (concatMap opty (pts i))) | (f,i) <- tree2list b] + where + opersIn t = case t of + Q n c | ism n -> [c] + QC n c | ism n -> [c] + _ -> collectOp opersIn t + opty (Yes ty) = opersIn ty + opty _ = [] + pts i = case i of + ResOper pty pt -> [pty,pt] + ResParam (Yes (ps,_)) -> [Yes t | (_,cont) <- ps, (_,t) <- cont] + CncCat pty _ _ -> [pty] + CncFun _ pt _ -> [pt] ---- (Maybe (Ident,(Context,Type)) + AbsFun pty ptr -> [pty] --- ptr is def, which can be mutual + AbsCat (Yes co) _ -> [Yes ty | (_,ty) <- co] + _ -> [] + +topoSortOpers :: [(Ident,[Ident])] -> Err [Ident] +topoSortOpers st = do + let eops = topoTest st + either + return + (\ops -> Bad ("circular definitions:" +++ unwords (map prt (head ops)))) + eops diff --git a/src/GF/Compile/Compute.hs b/src/GF/Compile/Compute.hs new file mode 100644 index 000000000..f35e7c6a9 --- /dev/null +++ b/src/GF/Compile/Compute.hs @@ -0,0 +1,429 @@ +---------------------------------------------------------------------- +-- | +-- Module : Compute +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/11/01 15:39:12 $ +-- > CVS $Author: aarne $ +-- > CVS $Revision: 1.19 $ +-- +-- Computation of source terms. Used in compilation and in @cc@ command. +----------------------------------------------------------------------------- + +module GF.Compile.Compute (computeConcrete, computeTerm,computeConcreteRec) where + +import GF.Data.Operations +import GF.Grammar.Grammar +import GF.Infra.Ident +import GF.Infra.Option +import GF.Data.Str +import GF.Grammar.PrGrammar +import GF.Infra.Modules +import GF.Grammar.Predef +import GF.Grammar.Macros +import GF.Grammar.Lookup +import GF.Compile.Refresh +import GF.Grammar.PatternMatch +import GF.Grammar.Lockfield (isLockLabel) ---- + +import GF.Grammar.AppPredefined + +import Data.List (nub,intersperse) +import Control.Monad (liftM2, liftM) + +-- | computation of concrete syntax terms into normal form +-- used mainly for partial evaluation +computeConcrete :: SourceGrammar -> Term -> Err Term +computeConcrete g t = {- refreshTerm t >>= -} computeTerm g [] t +computeConcreteRec g t = {- refreshTerm t >>= -} computeTermOpt True g [] t + +computeTerm :: SourceGrammar -> Substitution -> Term -> Err Term +computeTerm = computeTermOpt False + +-- rec=True is used if it cannot be assumed that looked-up constants +-- have already been computed (mainly with -optimize=noexpand in .gfr) + +computeTermOpt :: Bool -> SourceGrammar -> Substitution -> Term -> Err Term +computeTermOpt rec gr = comput True where + + comput full g t = ---- errIn ("subterm" +++ prt t) $ --- for debugging + case t of + + Q p c | p == cPredef -> return t + | otherwise -> look p c + + -- if computed do nothing + Computed t' -> return $ unComputed t' + + Vr x -> do + t' <- maybe (prtBad ("no value given to variable") x) return $ lookup x g + case t' of + _ | t == t' -> return t + _ -> comp g t' + + -- Abs x@(IA _) b -> do + Abs x b | full -> do + let (xs,b1) = termFormCnc t + b' <- comp ([(x,Vr x) | x <- xs] ++ g) b1 + return $ mkAbs xs b' + -- b' <- comp (ext x (Vr x) g) b + -- return $ Abs x b' + Abs _ _ -> return t -- hnf + + Let (x,(_,a)) b -> do + a' <- comp g a + comp (ext x a' g) b + + Prod x a b -> do + a' <- comp g a + b' <- comp (ext x (Vr x) g) b + return $ Prod x a' b' + + -- beta-convert + App f a -> case appForm t of + (h,as) | length as > 1 -> do + h' <- hnf g h + as' <- mapM (comp g) as + case h' of + _ | not (null [() | FV _ <- as']) -> compApp g (mkApp h' as') + c@(QC _ _) -> do + return $ mkApp c as' + Q mod f | mod == cPredef -> do + (t',b) <- appPredefined (mkApp h' as') + if b then return t' else comp g t' + + Abs _ _ -> do + let (xs,b) = termFormCnc h' + let g' = (zip xs as') ++ g + let as2 = drop (length xs) as' + let xs2 = drop (length as') xs + b' <- comp g' (mkAbs xs2 b) + if null as2 then return b' else comp g (mkApp b' as2) + + _ -> compApp g (mkApp h' as') + _ -> compApp g t + + P t l | isLockLabel l -> return $ R [] + ---- a workaround 18/2/2005: take this away and find the reason + ---- why earlier compilation destroys the lock field + + + P t l -> do + t' <- comp g t + case t' of + FV rs -> mapM (\c -> comp g (P c l)) rs >>= returnC . variants + R r -> maybe (prtBad "no value for label" l) (comp g . snd) $ + lookup l $ reverse r + + ExtR a (R b) -> + case comp g (P (R b) l) of + Ok v -> return v + _ -> comp g (P a l) + +--- { - --- this is incorrect, since b can contain the proper value + ExtR (R a) b -> -- NOT POSSIBLE both a and b records! + case comp g (P (R a) l) of + Ok v -> return v + _ -> comp g (P b l) +--- - } --- + + S (T i cs) e -> prawitz g i (flip P l) cs e + S (V i cs) e -> prawitzV g i (flip P l) cs e + + _ -> returnC $ P t' l + + PI t l i -> comp g $ P t l ----- + + S t v -> do + t' <- compTable g t + v' <- comp g v + t1 <- case t' of +---- V (RecType fs) _ -> uncurrySelect g fs t' v' +---- T (TComp (RecType fs)) _ -> uncurrySelect g fs t' v' + _ -> return $ S t' v' + compSelect g t1 + + -- normalize away empty tokens + K "" -> return Empty + + -- glue if you can + Glue x0 y0 -> do + x <- comp g x0 + y <- comp g y0 + case (x,y) of + (FV ks,_) -> do + kys <- mapM (comp g . flip Glue y) ks + return $ variants kys + (_,FV ks) -> do + xks <- mapM (comp g . Glue x) ks + return $ variants xks + + (S (T i cs) e, s) -> prawitz g i (flip Glue s) cs e + (s, S (T i cs) e) -> prawitz g i (Glue s) cs e + (S (V i cs) e, s) -> prawitzV g i (flip Glue s) cs e + (s, S (V i cs) e) -> prawitzV g i (Glue s) cs e + (_,Empty) -> return x + (Empty,_) -> return y + (K a, K b) -> return $ K (a ++ b) + (_, Alts (d,vs)) -> do +---- (K a, Alts (d,vs)) -> do + let glx = Glue x + comp g $ Alts (glx d, [(glx v,c) | (v,c) <- vs]) + (Alts _, ka) -> checks [do + y' <- strsFromTerm ka +---- (Alts _, K a) -> checks [do + x' <- strsFromTerm x -- this may fail when compiling opers + return $ variants [ + foldr1 C (map K (str2strings (glueStr v u))) | v <- x', u <- y'] +---- foldr1 C (map K (str2strings (glueStr v (str a)))) | v <- x'] + ,return $ Glue x y + ] + (C u v,_) -> comp g $ C u (Glue v y) + + _ -> do + mapM_ checkNoArgVars [x,y] + r <- composOp (comp g) t + returnC r + + Alts _ -> do + r <- composOp (comp g) t + returnC r + + -- remove empty + C a b -> do + a' <- comp g a + b' <- comp g b + case (a',b') of + (Alts _, K a) -> checks [do + as <- strsFromTerm a' -- this may fail when compiling opers + return $ variants [ + foldr1 C (map K (str2strings (plusStr v (str a)))) | v <- as] + , + return $ C a' b' + ] + (Empty,_) -> returnC b' + (_,Empty) -> returnC a' + _ -> returnC $ C a' b' + + -- reduce free variation as much as you can + FV ts -> mapM (comp g) ts >>= returnC . variants + + -- merge record extensions if you can + ExtR r s -> do + r' <- comp g r + s' <- comp g s + case (r',s') of + (R rs, R ss) -> plusRecord r' s' + (RecType rs, RecType ss) -> plusRecType r' s' + _ -> return $ ExtR r' s' + + T _ _ -> compTable g t + V _ _ -> compTable g t + + -- otherwise go ahead + _ -> composOp (comp g) t >>= returnC + + where + + compApp g (App f a) = do + f' <- hnf g f + a' <- comp g a + case (f',a') of + (Abs x b, FV as) -> + mapM (\c -> comp (ext x c g) b) as >>= return . variants + (_, FV as) -> mapM (\c -> comp g (App f' c)) as >>= return . variants + (FV fs, _) -> mapM (\c -> comp g (App c a')) fs >>= return . variants + (Abs x b,_) -> comp (ext x a' g) b + + (QC _ _,_) -> returnC $ App f' a' + + (S (T i cs) e,_) -> prawitz g i (flip App a') cs e + (S (V i cs) e,_) -> prawitzV g i (flip App a') cs e + + _ -> do + (t',b) <- appPredefined (App f' a') + if b then return t' else comp g t' + + hnf = comput False + comp = comput True + + look p c + | rec = lookupResDef gr p c >>= comp [] + | otherwise = lookupResDef gr p c + + ext x a g = (x,a):g + + returnC = return --- . computed + + variants ts = case nub ts of + [t] -> t + ts -> FV ts + + isCan v = case v of + Con _ -> True + QC _ _ -> True + App f a -> isCan f && isCan a + R rs -> all (isCan . snd . snd) rs + _ -> False + + compPatternMacro p = case p of + PM m c -> case look m c of + Ok (EPatt p') -> compPatternMacro p' + _ -> prtBad "pattern expected as value of" p ---- should be in CheckGr + PAs x p -> do + p' <- compPatternMacro p + return $ PAs x p' + PAlt p q -> do + p' <- compPatternMacro p + q' <- compPatternMacro q + return $ PAlt p' q' + PSeq p q -> do + p' <- compPatternMacro p + q' <- compPatternMacro q + return $ PSeq p' q' + PRep p -> do + p' <- compPatternMacro p + return $ PRep p' + PNeg p -> do + p' <- compPatternMacro p + return $ PNeg p' + PR rs -> do + rs' <- mapPairsM compPatternMacro rs + return $ PR rs' + + _ -> return p + + compSelect g (S t' v') = case v' of + FV vs -> mapM (\c -> comp g (S t' c)) vs >>= returnC . variants + _ -> case t' of + FV ccs -> mapM (\c -> comp g (S c v')) ccs >>= returnC . variants + + T _ [(PV IW,c)] -> comp g c --- an optimization + T _ [(PT _ (PV IW),c)] -> comp g c + + T _ [(PV z,c)] -> comp (ext z v' g) c --- another optimization + T _ [(PT _ (PV z),c)] -> comp (ext z v' g) c + + -- course-of-values table: look up by index, no pattern matching needed + V ptyp ts -> do + vs <- allParamValues gr ptyp + case lookup v' (zip vs [0 .. length vs - 1]) of + Just i -> comp g $ ts !! i + _ -> return $ S t' v' -- if v' is not canonical + T _ cc -> case matchPattern cc v' of + Ok (c,g') -> comp (g' ++ g) c + _ | isCan v' -> prtBad ("missing case" +++ prt v' +++ "in") t + _ -> return $ S t' v' -- if v' is not canonical + + S (T i cs) e -> prawitz g i (flip S v') cs e + S (V i cs) e -> prawitzV g i (flip S v') cs e + _ -> returnC $ S t' v' + + -- case-expand tables + -- if already expanded, don't expand again + compTable g t = case t of + T i@(TComp ty) cs -> do + -- if there are no variables, don't even go inside + cs' <- if (null g) then return cs else mapPairsM (comp g) cs +---- return $ V ty (map snd cs') + return $ T i cs' + V ty cs -> do + ty' <- comp g ty + -- if there are no variables, don't even go inside + cs' <- if (null g) then return cs else mapM (comp g) cs + return $ V ty' cs' + + T i cs -> do + pty0 <- getTableType i + ptyp <- comp g pty0 + case allParamValues gr ptyp of + Ok vs -> do + + ps0 <- mapM (compPatternMacro . fst) cs + cs' <- mapM (compBranchOpt g) (zip ps0 (map snd cs)) + sts <- mapM (matchPattern cs') vs + ts <- mapM (\ (c,g') -> comp (g' ++ g) c) sts + ps <- mapM term2patt vs + let ps' = ps --- PT ptyp (head ps) : tail ps +---- return $ V ptyp ts -- to save space, just course of values + return $ T (TComp ptyp) (zip ps' ts) + _ -> do + cs' <- mapM (compBranch g) cs + return $ T i cs' -- happens with variable types + _ -> comp g t + + compBranch g (p,v) = do + let g' = contP p ++ g + v' <- comp g' v + return (p,v') + + compBranchOpt g c@(p,v) = case contP p of + [] -> return c + _ -> err (const (return c)) return $ compBranch g c + + contP p = case p of + PV x -> [(x,Vr x)] + PC _ ps -> concatMap contP ps + PP _ _ ps -> concatMap contP ps + PT _ p -> contP p + PR rs -> concatMap (contP . snd) rs + + PAs x p -> (x,Vr x) : contP p + + PSeq p q -> concatMap contP [p,q] + PAlt p q -> concatMap contP [p,q] + PRep p -> contP p + PNeg p -> contP p + + _ -> [] + + prawitz g i f cs e = do + cs' <- mapM (compBranch g) [(p, f v) | (p,v) <- cs] + return $ S (T i cs') e + prawitzV g i f cs e = do + cs' <- mapM (comp g) [(f v) | v <- cs] + return $ S (V i cs') e + +{- ---- + uncurrySelect g fs t v = do + ts <- mapM (allParamValues gr . snd) fs + vs <- mapM (comp g) [P v r | r <- map fst fs] + return $ reorderSelect t fs ts vs + + reorderSelect t fs pss vs = case (t,fs,pss,vs) of + (V _ ts, f:fs1, ps:pss1, v:vs1) -> + S (V (snd f) + [reorderSelect (V (RecType fs1) t) fs1 pss1 vs1 | + t <- segments (length ts `div` length ps) ts]) v + (T (TComp _) cs, f:fs1, ps:pss1, v:vs1) -> + S (T (TComp (snd f)) + [(p,reorderSelect (T (TComp (RecType fs1)) c) fs1 pss1 vs1) | + (ep,c) <- zip ps (segments (length cs `div` length ps) cs), + let Ok p = term2patt ep]) v + _ -> t + + segments i xs = + let (x0,xs1) = splitAt i xs in x0 : takeWhile (not . null) (segments i xs1) +-} + + +-- | argument variables cannot be glued +checkNoArgVars :: Term -> Err Term +checkNoArgVars t = case t of + Vr (IA _ _) -> Bad $ glueErrorMsg $ prt t + Vr (IAV _ _ _) -> Bad $ glueErrorMsg $ prt t + _ -> composOp checkNoArgVars t + +glueErrorMsg s = + "Cannot glue (+) term with run-time variable" +++ s ++ "." ++++ + "Use Prelude.bind instead." + +getArgType t = case t of + V ty _ -> return ty + T (TComp ty) _ -> return ty + _ -> prtBad "cannot get argument type of table" t + + + diff --git a/src/GF/Compile/Export.hs b/src/GF/Compile/Export.hs new file mode 100644 index 000000000..9e9a99e99 --- /dev/null +++ b/src/GF/Compile/Export.hs @@ -0,0 +1,61 @@ +module GF.Compile.Export where + +import PGF.CId +import PGF.Data (PGF(..)) +import PGF.Raw.Print (printTree) +import PGF.Raw.Convert (fromPGF) +import GF.Compile.GFCCtoHaskell +import GF.Compile.GFCCtoJS +import GF.Infra.Option +import GF.Speech.CFG +import GF.Speech.PGFToCFG +import GF.Speech.SRGS_XML +import GF.Speech.JSGF +import GF.Speech.GSL +import GF.Speech.VoiceXML +import GF.Speech.SLF +import GF.Speech.PrRegExp +import GF.Text.UTF8 + +import Data.Maybe +import System.FilePath + +-- top-level access to code generation + +exportPGF :: Options + -> OutputFormat + -> PGF + -> [(FilePath,String)] -- ^ List of recommended file names and contents. +exportPGF opts fmt pgf = + case fmt of + FmtPGF -> multi "pgf" printPGF + FmtJavaScript -> multi "js" pgf2js + FmtHaskell -> multi "hs" (grammar2haskell name) + FmtHaskell_GADT -> multi "hs" (grammar2haskellGADT name) + FmtBNF -> single "bnf" bnfPrinter + FmtSRGS_XML -> single "grxml" (srgsXmlPrinter sisr) + FmtJSGF -> single "jsgf" (jsgfPrinter sisr) + FmtGSL -> single "gsl" gslPrinter + FmtVoiceXML -> single "vxml" grammar2vxml + FmtSLF -> single ".slf" slfPrinter + FmtRegExp -> single ".rexp" regexpPrinter + FmtFA -> single ".dot" slfGraphvizPrinter + where + name = fromMaybe (prCId (absname pgf)) (moduleFlag optName opts) + sisr = flag optSISR opts + + multi :: String -> (PGF -> String) -> [(FilePath,String)] + multi ext pr = [(name <.> ext, pr pgf)] + + single :: String -> (PGF -> CId -> String) -> [(FilePath,String)] + single ext pr = [(prCId cnc <.> ext, pr pgf cnc) | cnc <- cncnames pgf] + +-- | Get the name of the concrete syntax to generate output from. +-- FIXME: there should be an option to change this. +outputConcr :: PGF -> CId +outputConcr pgf = case cncnames pgf of + [] -> error "No concrete syntax." + cnc:_ -> cnc + +printPGF :: PGF -> String +printPGF = encodeUTF8 . printTree . fromPGF diff --git a/src/GF/Compile/Extend.hs b/src/GF/Compile/Extend.hs new file mode 100644 index 000000000..8344a1696 --- /dev/null +++ b/src/GF/Compile/Extend.hs @@ -0,0 +1,138 @@ +---------------------------------------------------------------------- +-- | +-- Module : Extend +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/05/30 21:08:14 $ +-- > CVS $Author: aarne $ +-- > CVS $Revision: 1.18 $ +-- +-- AR 14\/5\/2003 -- 11\/11 +-- +-- The top-level function 'extendModule' +-- extends a module symbol table by indirections to the module it extends +----------------------------------------------------------------------------- + +module GF.Compile.Extend (extendModule, extendMod + ) where + +import GF.Grammar.Grammar +import GF.Infra.Ident +import GF.Grammar.PrGrammar +import GF.Infra.Modules +import GF.Compile.Update +import GF.Grammar.Macros +import GF.Data.Operations + +import Control.Monad + +extendModule :: [SourceModule] -> SourceModule -> Err SourceModule +extendModule ms (name,mod) = case mod of + + ---- Just to allow inheritance in incomplete concrete (which are not + ---- compiled anyway), extensions are not built for them. + ---- Should be replaced by real control. AR 4/2/2005 + ModMod m | mstatus m == MSIncomplete && isModCnc m -> return (name,mod) + + ModMod m -> do + mod' <- foldM extOne m (extend m) + return (name,ModMod mod') + where + extOne mo (n,cond) = do + (m0,isCompl) <- do + m <- lookupModMod (MGrammar ms) n + + -- test that the module types match, and find out if the old is complete + testErr (sameMType (mtype m) (mtype mo)) + ("illegal extension type to module" +++ prt name) + return (m, isCompleteModule m) + + -- build extension in a way depending on whether the old module is complete + js1 <- extendMod isCompl (n, isInherited cond) name (jments m0) (jments mo) + + -- if incomplete, throw away extension information + let es = extend mo + let es' = if isCompl then es else (filter ((/=n) . fst) es) + return $ mo {extend = es', jments = js1} + +-- | When extending a complete module: new information is inserted, +-- and the process is interrupted if unification fails. +-- If the extended module is incomplete, its judgements are just copied. +extendMod :: Bool -> (Ident,Ident -> Bool) -> Ident -> + BinTree Ident Info -> BinTree Ident Info -> + Err (BinTree Ident Info) +extendMod isCompl (name,cond) base old new = foldM try new $ tree2list old where + try t i@(c,_) | not (cond c) = return t + try t i@(c,_) = errIn ("constant" +++ prt c) $ + tryInsert (extendAnyInfo isCompl name base) indirIf t i + indirIf = if isCompl then indirInfo name else id + +indirInfo :: Ident -> Info -> Info +indirInfo n info = AnyInd b n' where + (b,n') = case info of + ResValue _ -> (True,n) + ResParam _ -> (True,n) + AbsFun _ (Yes EData) -> (True,n) + AnyInd b k -> (b,k) + _ -> (False,n) ---- canonical in Abs + +perhIndir :: Ident -> Perh a -> Perh a +perhIndir n p = case p of + Yes _ -> May n + _ -> p + +extendAnyInfo :: Bool -> Ident -> Ident -> Info -> Info -> Err Info +extendAnyInfo isc n o i j = + errIn ("building extension for" +++ prt n +++ "in" +++ prt o) $ case (i,j) of + (AbsCat mc1 mf1, AbsCat mc2 mf2) -> + liftM2 AbsCat (updn isc n mc1 mc2) (updn isc n mf1 mf2) --- add cstrs + (AbsFun mt1 md1, AbsFun mt2 md2) -> + liftM2 AbsFun (updn isc n mt1 mt2) (updn isc n md1 md2) --- add defs + (ResParam mt1, ResParam mt2) -> + liftM ResParam $ updn isc n mt1 mt2 + (ResValue mt1, ResValue mt2) -> + liftM ResValue $ updn isc n mt1 mt2 + (_, ResOverload ms t) | elem n ms -> + return $ ResOverload ms t + (ResOper mt1 m1, ResOper mt2 m2) -> ---- extendResOper n mt1 m1 mt2 m2 + liftM2 ResOper (updn isc n mt1 mt2) (updn isc n m1 m2) + (CncCat mc1 mf1 mp1, CncCat mc2 mf2 mp2) -> + liftM3 CncCat (updn isc n mc1 mc2) (updn isc n mf1 mf2) (updn isc n mp1 mp2) + (CncFun m mt1 md1, CncFun _ mt2 md2) -> + liftM2 (CncFun m) (updn isc n mt1 mt2) (updn isc n md1 md2) + +---- (AnyInd _ _, ResOper _ _) -> return j ---- + + (AnyInd b1 m1, AnyInd b2 m2) -> do + testErr (b1 == b2) "inconsistent indirection status" +---- commented out as work-around for a spurious problem in +---- TestResourceFre; should look at building of completion. 17/11/2004 + testErr (m1 == m2) $ + "different sources of indirection: " +++ show m1 +++ show m2 + return i + + _ -> Bad $ "cannot unify information in" ++++ show i ++++ "and" ++++ show j + +--- where + +updn isc n = if isc then (updatePerhaps n) else (updatePerhapsHard n) +updc isc n = if True then (updatePerhaps n) else (updatePerhapsHard n) + + + +{- ---- no more needed: this is done in Rebuild +-- opers declared in an interface and defined in an instance are a special case + +extendResOper n mt1 m1 mt2 m2 = case (m1,m2) of + (Nope,_) -> return $ ResOper (strip mt1) m2 + _ -> liftM2 ResOper (updatePerhaps n mt1 mt2) (updatePerhaps n m1 m2) + where + strip (Yes t) = Yes $ strp t + strip m = m + strp t = case t of + Q _ c -> Vr c + QC _ c -> Vr c + _ -> composSafeOp strp t +-} diff --git a/src/GF/Compile/GFCCtoHaskell.hs b/src/GF/Compile/GFCCtoHaskell.hs new file mode 100644 index 000000000..59db9c364 --- /dev/null +++ b/src/GF/Compile/GFCCtoHaskell.hs @@ -0,0 +1,213 @@ +---------------------------------------------------------------------- +-- | +-- Module : GFCCtoHaskell +-- Maintainer : Aarne Ranta +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/06/17 12:39:07 $ +-- > CVS $Author: bringert $ +-- > CVS $Revision: 1.8 $ +-- +-- to write a GF abstract grammar into a Haskell module with translations from +-- data objects into GF trees. Example: GSyntax for Agda. +-- AR 11/11/1999 -- 7/12/2000 -- 18/5/2004 +----------------------------------------------------------------------------- + +module GF.Compile.GFCCtoHaskell (grammar2haskell, grammar2haskellGADT) where + +import PGF.CId +import PGF.Data +import PGF.Macros + +import GF.Data.Operations +import GF.Text.UTF8 + +import Data.List --(isPrefixOf, find, intersperse) +import qualified Data.Map as Map + +-- | the main function +grammar2haskell :: String -- ^ Module name. + -> PGF + -> String +grammar2haskell name gr = encodeUTF8 $ foldr (++++) [] $ + haskPreamble name ++ [datatypes gr', gfinstances gr'] + where gr' = hSkeleton gr + +grammar2haskellGADT :: String -> PGF -> String +grammar2haskellGADT name gr = encodeUTF8 $ foldr (++++) [] $ + ["{-# OPTIONS_GHC -fglasgow-exts #-}"] ++ + haskPreamble name ++ [datatypesGADT gr', gfinstances gr'] + where gr' = hSkeleton gr + +-- | by this you can prefix all identifiers with stg; the default is 'G' +gId :: OIdent -> OIdent +gId i = 'G':i + +haskPreamble name = + [ + "module " ++ name ++ " where", + "", + "import PGF", + "----------------------------------------------------", + "-- automatic translation from GF to Haskell", + "----------------------------------------------------", + "", + "class Gf a where", + " gf :: a -> Tree", + " fg :: Tree -> a", + "", + predefInst "GString" "String" "Lit (LStr s)", + "", + predefInst "GInt" "Integer" "Lit (LInt s)", + "", + predefInst "GFloat" "Double" "Lit (LFlt s)", + "", + "----------------------------------------------------", + "-- below this line machine-generated", + "----------------------------------------------------", + "" + ] + +predefInst gtyp typ patt = + "newtype" +++ gtyp +++ "=" +++ gtyp +++ typ +++ " deriving Show" +++++ + "instance Gf" +++ gtyp +++ "where" ++++ + " gf (" ++ gtyp +++ "s) =" +++ patt ++++ + " fg t =" ++++ + " case t of" ++++ + " " +++ patt +++ " ->" +++ gtyp +++ "s" ++++ + " _ -> error (\"no" +++ gtyp +++ "\" ++ show t)" + +type OIdent = String + +type HSkeleton = [(OIdent, [(OIdent, [OIdent])])] + +datatypes, gfinstances :: (String,HSkeleton) -> String +datatypes = (foldr (+++++) "") . (filter (/="")) . (map hDatatype) . snd +gfinstances (m,g) = (foldr (+++++) "") $ (filter (/="")) $ (map (gfInstance m)) g + +hDatatype :: (OIdent, [(OIdent, [OIdent])]) -> String +gfInstance :: String -> (OIdent, [(OIdent, [OIdent])]) -> String + +hDatatype ("Cn",_) = "" --- +hDatatype (cat,[]) = "" +hDatatype (cat,rules) | isListCat (cat,rules) = + "newtype" +++ gId cat +++ "=" +++ gId cat +++ "[" ++ gId (elemCat cat) ++ "]" + +++ "deriving Show" +hDatatype (cat,rules) = + "data" +++ gId cat +++ "=" ++ + (if length rules == 1 then "" else "\n ") +++ + foldr1 (\x y -> x ++ "\n |" +++ y) + [gId f +++ foldr (+++) "" (map gId xx) | (f,xx) <- rules] ++++ + " deriving Show" + +-- GADT version of data types +datatypesGADT :: (String,HSkeleton) -> String +datatypesGADT (_,skel) = + unlines (concatMap hCatTypeGADT skel) + +++++ + "data Tree :: * -> * where" ++++ unlines (concatMap (map (" "++) . hDatatypeGADT) skel) + +hCatTypeGADT :: (OIdent, [(OIdent, [OIdent])]) -> [String] +hCatTypeGADT (cat,rules) + = ["type"+++gId cat+++"="+++"Tree"+++gId cat++"_", + "data"+++gId cat++"_"] + +hDatatypeGADT :: (OIdent, [(OIdent, [OIdent])]) -> [String] +hDatatypeGADT (cat, rules) + | isListCat (cat,rules) = [gId cat+++"::"+++"["++gId (elemCat cat)++"]" +++ "->" +++ t] + | otherwise = + [ gId f +++ "::" +++ concatMap (\a -> gId a +++ "-> ") args ++ t | (f,args) <- rules ] + where t = "Tree" +++ gId cat ++ "_" + +gfInstance m crs = hInstance m crs ++++ fInstance m crs + +----hInstance m ("Cn",_) = "" --- seems to belong to an old applic. AR 18/5/2004 +hInstance m (cat,[]) = "" +hInstance m (cat,rules) + | isListCat (cat,rules) = + "instance Gf" +++ gId cat +++ "where" ++++ + " gf (" ++ gId cat +++ "[" ++ concat (intersperse "," baseVars) ++ "])" + +++ "=" +++ mkRHS ("Base"++ec) baseVars ++++ + " gf (" ++ gId cat +++ "(x:xs)) = " + ++ mkRHS ("Cons"++ec) ["x",prParenth (gId cat+++"xs")] +-- no show for GADTs +-- ++++ " gf (" ++ gId cat +++ "xs) = error (\"Bad " ++ cat ++ " value: \" ++ show xs)" + | otherwise = + "instance Gf" +++ gId cat +++ "where\n" ++ + unlines [mkInst f xx | (f,xx) <- rules] + where + ec = elemCat cat + baseVars = mkVars (baseSize (cat,rules)) + mkInst f xx = let xx' = mkVars (length xx) in " gf " ++ + (if length xx == 0 then gId f else prParenth (gId f +++ foldr1 (+++) xx')) +++ + "=" +++ mkRHS f xx' + mkVars n = ["x" ++ show i | i <- [1..n]] + mkRHS f vars = "Fun (mkCId \"" ++ f ++ "\")" +++ + "[" ++ prTList ", " ["gf" +++ x | x <- vars] ++ "]" + + +----fInstance m ("Cn",_) = "" --- +fInstance m (cat,[]) = "" +fInstance m (cat,rules) = + " fg t =" ++++ + " case t of" ++++ + unlines [mkInst f xx | (f,xx) <- rules] ++++ + " _ -> error (\"no" +++ cat ++ " \" ++ show t)" + where + mkInst f xx = + " Fun i " ++ + "[" ++ prTList "," xx' ++ "]" +++ + "| i == mkCId \"" ++ f ++ "\" ->" +++ mkRHS f xx' + where xx' = ["x" ++ show i | (_,i) <- zip xx [1..]] + mkRHS f vars + | isListCat (cat,rules) = + if "Base" `isPrefixOf` f then + gId cat +++ "[" ++ prTList ", " [ "fg" +++ x | x <- vars ] ++ "]" + else + let (i,t) = (init vars,last vars) + in "let" +++ gId cat +++ "xs = fg " ++ t +++ "in" +++ + gId cat +++ prParenth (prTList ":" (["fg"+++v | v <- i] ++ ["xs"])) + | otherwise = + gId f +++ + prTList " " [prParenth ("fg" +++ x) | x <- vars] + + +--type HSkeleton = [(OIdent, [(OIdent, [OIdent])])] +hSkeleton :: PGF -> (String,HSkeleton) +hSkeleton gr = + (prCId (absname gr), + [(prCId c, [(prCId f, map prCId cs) | (f, (cs,_)) <- fs]) | + fs@((_, (_,c)):_) <- fns] + ) + where + fns = groupBy valtypg (sortBy valtyps (map jty (Map.assocs (funs (abstract gr))))) + valtyps (_, (_,x)) (_, (_,y)) = compare x y + valtypg (_, (_,x)) (_, (_,y)) = x == y + jty (f,(ty,_)) = (f,catSkeleton ty) + +updateSkeleton :: OIdent -> HSkeleton -> (OIdent, [OIdent]) -> HSkeleton +updateSkeleton cat skel rule = + case skel of + (cat0,rules):rr | cat0 == cat -> (cat0, rule:rules) : rr + (cat0,rules):rr -> (cat0, rules) : updateSkeleton cat rr rule + +isListCat :: (OIdent, [(OIdent, [OIdent])]) -> Bool +isListCat (cat,rules) = "List" `isPrefixOf` cat && length rules == 2 + && ("Base"++c) `elem` fs && ("Cons"++c) `elem` fs + where c = elemCat cat + fs = map fst rules + +-- | Gets the element category of a list category. +elemCat :: OIdent -> OIdent +elemCat = drop 4 + +isBaseFun :: OIdent -> Bool +isBaseFun f = "Base" `isPrefixOf` f + +isConsFun :: OIdent -> Bool +isConsFun f = "Cons" `isPrefixOf` f + +baseSize :: (OIdent, [(OIdent, [OIdent])]) -> Int +baseSize (_,rules) = length bs + where Just (_,bs) = find (("Base" `isPrefixOf`) . fst) rules diff --git a/src/GF/Compile/GFCCtoJS.hs b/src/GF/Compile/GFCCtoJS.hs new file mode 100644 index 000000000..8259e7385 --- /dev/null +++ b/src/GF/Compile/GFCCtoJS.hs @@ -0,0 +1,117 @@ +module GF.Compile.GFCCtoJS (pgf2js) where + +import PGF.CId +import PGF.Data +import qualified PGF.Macros as M +import qualified GF.JavaScript.AbsJS as JS +import qualified GF.JavaScript.PrintJS as JS + +import GF.Text.UTF8 +import GF.Data.ErrM +import GF.Infra.Option + +import Control.Monad (mplus) +import Data.Array (Array) +import qualified Data.Array as Array +import Data.Maybe (fromMaybe) +import qualified Data.Map as Map + +pgf2js :: PGF -> String +pgf2js pgf = + encodeUTF8 $ JS.printTree $ JS.Program [JS.ElStmt $ JS.SDeclOrExpr $ JS.Decl [JS.DInit (JS.Ident n) grammar]] + where + n = prCId $ absname pgf + as = abstract pgf + cs = Map.assocs (concretes pgf) + start = M.lookStartCat pgf + grammar = new "GFGrammar" [js_abstract, js_concrete] + js_abstract = abstract2js start as + js_concrete = JS.EObj $ map (concrete2js start n) cs + +abstract2js :: String -> Abstr -> JS.Expr +abstract2js start ds = new "GFAbstract" [JS.EStr start, JS.EObj $ map absdef2js (Map.assocs (funs ds))] + +absdef2js :: (CId,(Type,Expr)) -> JS.Property +absdef2js (f,(typ,_)) = + let (args,cat) = M.catSkeleton typ in + JS.Prop (JS.IdentPropName (JS.Ident (prCId f))) (new "Type" [JS.EArray [JS.EStr (prCId x) | x <- args], JS.EStr (prCId cat)]) + +concrete2js :: String -> String -> (CId,Concr) -> JS.Property +concrete2js start n (c, cnc) = + JS.Prop l (new "GFConcrete" ([(JS.EObj $ ((map (cncdef2js n (prCId c)) ds) ++ litslins))] ++ + maybe [] (parser2js start) (parser cnc))) + where + l = JS.IdentPropName (JS.Ident (prCId c)) + ds = concatMap Map.assocs [lins cnc, opers cnc, lindefs cnc] + litslins = [JS.Prop (JS.StringPropName "Int") (JS.EFun [children] [JS.SReturn $ new "Arr" [JS.EIndex (JS.EVar children) (JS.EInt 0)]]), + JS.Prop (JS.StringPropName "Float") (JS.EFun [children] [JS.SReturn $ new "Arr" [JS.EIndex (JS.EVar children) (JS.EInt 0)]]), + JS.Prop (JS.StringPropName "String") (JS.EFun [children] [JS.SReturn $ new "Arr" [JS.EIndex (JS.EVar children) (JS.EInt 0)]])] + + +cncdef2js :: String -> String -> (CId,Term) -> JS.Property +cncdef2js n l (f, t) = JS.Prop (JS.IdentPropName (JS.Ident (prCId f))) (JS.EFun [children] [JS.SReturn (term2js n l t)]) + +term2js :: String -> String -> Term -> JS.Expr +term2js n l t = f t + where + f t = + case t of + R xs -> new "Arr" (map f xs) + P x y -> JS.ECall (JS.EMember (f x) (JS.Ident "sel")) [f y] + S xs -> mkSeq (map f xs) + K t -> tokn2js t + V i -> JS.EIndex (JS.EVar children) (JS.EInt i) + C i -> new "Int" [JS.EInt i] + F f -> JS.ECall (JS.EMember (JS.EIndex (JS.EMember (JS.EVar $ JS.Ident n) (JS.Ident "concretes")) (JS.EStr l)) (JS.Ident "rule")) [JS.EStr (prCId f), JS.EVar children] + FV xs -> new "Variants" (map f xs) + W str x -> new "Suffix" [JS.EStr str, f x] + TM _ -> new "Meta" [] + +tokn2js :: Tokn -> JS.Expr +tokn2js (KS s) = mkStr s +tokn2js (KP ss vs) = mkSeq (map mkStr ss) -- FIXME + +mkStr :: String -> JS.Expr +mkStr s = new "Str" [JS.EStr s] + +mkSeq :: [JS.Expr] -> JS.Expr +mkSeq [x] = x +mkSeq xs = new "Seq" xs + +argIdent :: Integer -> JS.Ident +argIdent n = JS.Ident ("x" ++ show n) + +children :: JS.Ident +children = JS.Ident "cs" + +-- Parser +parser2js :: String -> ParserInfo -> [JS.Expr] +parser2js start p = [new "Parser" [JS.EStr start, + JS.EArray $ map frule2js (Array.elems (allRules p)), + JS.EObj $ map cats (Map.assocs (startupCats p))]] + where + cats (c,is) = JS.Prop (JS.IdentPropName (JS.Ident (prCId c))) (JS.EArray (map JS.EInt is)) + +frule2js :: FRule -> JS.Expr +frule2js (FRule f ps args res lins) = new "Rule" [JS.EInt res, name2js (f,ps), JS.EArray (map JS.EInt args), lins2js lins] + +name2js :: (CId,[Profile]) -> JS.Expr +name2js (f,ps) | f == wildCId = fromProfile (head ps) + | otherwise = new "FunApp" $ [JS.EStr $ prCId f, JS.EArray (map fromProfile ps)] + where + fromProfile :: Profile -> JS.Expr + fromProfile [] = new "MetaVar" [] + fromProfile [x] = daughter x + fromProfile args = new "Unify" [JS.EArray (map daughter args)] + + daughter i = new "Arg" [JS.EInt i] + +lins2js :: Array FIndex (Array FPointPos FSymbol) -> JS.Expr +lins2js ls = JS.EArray [ JS.EArray [ sym2js s | s <- Array.elems l] | l <- Array.elems ls] + +sym2js :: FSymbol -> JS.Expr +sym2js (FSymCat l n) = new "ArgProj" [JS.EInt n, JS.EInt l] +sym2js (FSymTok t) = new "Terminal" [JS.EStr t] + +new :: String -> [JS.Expr] -> JS.Expr +new f xs = JS.ENew (JS.Ident f) xs diff --git a/src/GF/Compile/GenerateFCFG.hs b/src/GF/Compile/GenerateFCFG.hs new file mode 100644 index 000000000..c2854ef3d --- /dev/null +++ b/src/GF/Compile/GenerateFCFG.hs @@ -0,0 +1,526 @@ +---------------------------------------------------------------------- +-- | +-- Maintainer : Krasimir Angelov +-- Stability : (stable) +-- Portability : (portable) +-- +-- Converting SimpleGFC grammars to fast nonerasing MCFG grammar. +-- +-- the resulting grammars might be /very large/ +-- +-- the conversion is only equivalent if the GFC grammar has a context-free backbone. +----------------------------------------------------------------------------- + + +module GF.Compile.GenerateFCFG + (convertConcrete) where + +import PGF.CId +import PGF.Data +import PGF.Macros --hiding (prt) +import PGF.Parsing.FCFG.Utilities + +import GF.Data.BacktrackM +import GF.Data.SortedList +import GF.Data.Utilities (updateNthM, sortNub) + +import qualified Data.Map as Map +import qualified Data.Set as Set +import qualified Data.List as List +import qualified Data.ByteString.Char8 as BS +import Data.Array +import Data.Maybe +import Control.Monad + +---------------------------------------------------------------------- +-- main conversion function + +convertConcrete :: Abstr -> Concr -> FGrammar +convertConcrete abs cnc = fixHoasFuns $ convert abs_defs' conc' cats' + where abs_defs = Map.assocs (funs abs) + conc = Map.union (opers cnc) (lins cnc) -- "union big+small most efficient" + cats = lincats cnc + (abs_defs',conc',cats') = expandHOAS abs_defs conc cats + +expandHOAS :: [(CId,(Type,Expr))] -> TermMap -> TermMap -> ([(CId,(Type,Expr))],TermMap,TermMap) +expandHOAS funs lins lincats = (funs' ++ hoFuns ++ varFuns, + Map.unions [lins, hoLins, varLins], + Map.unions [lincats, hoLincats, varLincat]) + where + -- replace higher-order fun argument types with new categories + funs' = [(f,(fixType ty,e)) | (f,(ty,e)) <- funs] + where + fixType :: Type -> Type + fixType ty = let (ats,rt) = typeSkeleton ty in cftype (map catName ats) rt + + hoTypes :: [(Int,CId)] + hoTypes = sortNub [(n,c) | (_,(ty,_)) <- funs, (n,c) <- fst (typeSkeleton ty), n > 0] + hoCats = sortNub (map snd hoTypes) + -- for each Cat with N bindings, we add a new category _NCat + -- each new category contains a single function __NCat : Cat -> _Var -> ... -> _Var -> _NCat + hoFuns = [(funName ty,(cftype (c : replicate n varCat) (catName ty),EEq [])) | ty@(n,c) <- hoTypes] + -- lincats for the new categories + hoLincats = Map.fromList [(catName ty, modifyRec (++ replicate n (S [])) (lincatOf c)) | ty@(n,c) <- hoTypes] + -- linearizations of the new functions, lin __NCat v_0 ... v_n-1 x = { s1 = x.s1; ...; sk = x.sk; $0 = v_0.s ... + hoLins = Map.fromList [ (funName ty, mkLin c n) | ty@(n,c) <- hoTypes] + where mkLin c n = modifyRec (\fs -> [P (V 0) (C j) | j <- [0..length fs-1]] ++ [P (V i) (C 0) | i <- [1..n]]) (lincatOf c) + -- for each Cat, we a add a fun _Var_Cat : _Var -> Cat + varFuns = [(varFunName cat, (cftype [varCat] cat,EEq [])) | cat <- hoCats] + -- linearizations of the _Var_Cat functions + varLins = Map.fromList [(varFunName cat, R [P (V 0) (C 0)]) | cat <- hoCats] + -- lincat for the _Var category + varLincat = Map.singleton varCat (R [S []]) + + lincatOf c = fromMaybe (error $ "No lincat for " ++ prCId c) $ Map.lookup c lincats + + modifyRec :: ([Term] -> [Term]) -> Term -> Term + modifyRec f (R xs) = R (f xs) + modifyRec _ t = error $ "Not a record: " ++ show t + + varCat = mkCId "_Var" + + catName :: (Int,CId) -> CId + catName (0,c) = c + catName (n,c) = mkCId ("_" ++ show n ++ prCId c) + + funName :: (Int,CId) -> CId + funName (n,c) = mkCId ("__" ++ show n ++ prCId c) + + varFunName :: CId -> CId + varFunName c = mkCId ("_Var_" ++ prCId c) + +-- replaces __NCat with _B and _Var_Cat with _. +-- the temporary names are just there to avoid name collisions. +fixHoasFuns :: FGrammar -> FGrammar +fixHoasFuns (rs, cs) = ([FRule (fixName n) ps args cat lins | FRule n ps args cat lins <- rs], cs) + where fixName (CId n) | BS.pack "__" `BS.isPrefixOf` n = (mkCId "_B") + | BS.pack "_Var_" `BS.isPrefixOf` n = wildCId + fixName n = n + +convert :: [(CId,(Type,Expr))] -> TermMap -> TermMap -> FGrammar +convert abs_defs cnc_defs cat_defs = getFGrammar (loop frulesEnv) + where + srules = [ + (XRule id args res (map findLinType args) (findLinType res) term) | + (id, (ty,_)) <- abs_defs, let (args,res) = catSkeleton ty, + term <- Map.lookup id cnc_defs] + + findLinType id = fromMaybe (error $ "No lincat for " ++ show id) (Map.lookup id cat_defs) + + (xrulesMap,frulesEnv) = List.foldl' helper (Map.empty,emptyFRulesEnv) srules + where + helper (xrulesMap,frulesEnv) rule@(XRule id abs_args abs_res cnc_args cnc_res term) = + let xrulesMap' = Map.insertWith (++) abs_res [rule] xrulesMap + frulesEnv' = List.foldl' (\env selector -> convertRule cnc_defs selector rule env) + frulesEnv + (mkSingletonSelectors cnc_defs cnc_res) + in xrulesMap' `seq` frulesEnv' `seq` (xrulesMap',frulesEnv') + + loop frulesEnv = + let (todo, frulesEnv') = takeToDoRules xrulesMap frulesEnv + in case todo of + [] -> frulesEnv' + _ -> loop $! List.foldl' (\env (srules,selector) -> + List.foldl' (\env srule -> convertRule cnc_defs selector srule env) env srules) frulesEnv' todo + +convertRule :: TermMap -> TermSelector -> XRule -> FRulesEnv -> FRulesEnv +convertRule cnc_defs selector (XRule fun args cat ctypes ctype term) frulesEnv = + foldBM addRule + frulesEnv + (convertTerm cnc_defs selector term [([],[])]) + (protoFCat cat, map (\scat -> (protoFCat scat,[])) args, ctype, ctypes) + where + addRule linRec (newCat', newArgs', _, _) env0 = + let (env1, newCat) = genFCatHead env0 newCat' + (env2, newArgs,idxArgs) = foldr (\((xcat@(PFCat cat rcs tcs),xpaths),ctype,idx) (env,args,all_args) -> + let xargs = xcat:[PFCat cat [path] tcs | path <- reverse xpaths] + (env1, xargs1) = List.mapAccumL (genFCatArg cnc_defs ctype) env xargs + in case xcat of + PFCat _ [] _ -> (env , args, all_args) + _ -> (env1,xargs1++args,(idx,zip xargs1 xargs):all_args)) (env1,[],[]) (zip3 newArgs' ctypes [0..]) + + newLinRec = listArray (0,length linRec-1) [translateLin idxArgs path linRec | path <- case newCat' of {PFCat _ rcs _ -> rcs}] + + (_,newProfile) = List.mapAccumL accumProf 0 newArgs' + where + accumProf nr (PFCat _ [] _,_ ) = (nr, [] ) + accumProf nr (_ ,xpaths) = (nr+cnt+1, [nr..nr+cnt]) + where cnt = length xpaths + + rule = FRule fun newProfile newArgs newCat newLinRec + in addFRule env2 rule + +translateLin idxArgs lbl' [] = array (0,-1) [] +translateLin idxArgs lbl' ((lbl,syms) : lins) + | lbl' == lbl = listArray (0,length syms-1) (map instSym syms) + | otherwise = translateLin idxArgs lbl' lins + where + instSym = either (\(lbl, nr, xnr) -> instCat lbl nr xnr 0 idxArgs) FSymTok + instCat lbl nr xnr nr' ((idx,xargs):idxArgs) + | nr == idx = let (fcat, PFCat _ rcs _) = xargs !! xnr + in FSymCat (index lbl rcs 0) (nr'+xnr) + | otherwise = instCat lbl nr xnr (nr'+length xargs) idxArgs + + index lbl' (lbl:lbls) idx + | lbl' == lbl = idx + | otherwise = index lbl' lbls $! (idx+1) + + +---------------------------------------------------------------------- +-- term conversion + +type CnvMonad a = BacktrackM Env a + +type FPath = [FIndex] +type Env = (ProtoFCat, [(ProtoFCat,[FPath])], Term, [Term]) +type LinRec = [(FPath, [Either (FPath, FIndex, Int) FToken])] + +type TermMap = Map.Map CId Term + +convertTerm :: TermMap -> TermSelector -> Term -> LinRec -> CnvMonad LinRec +convertTerm cnc_defs selector (V nr) ((lbl_path,lin) : lins) = convertArg selector nr [] lbl_path lin lins +convertTerm cnc_defs selector (C nr) ((lbl_path,lin) : lins) = convertCon selector nr lbl_path lin lins +convertTerm cnc_defs selector (R record) ((lbl_path,lin) : lins) = convertRec cnc_defs selector 0 record lbl_path lin lins + +convertTerm cnc_defs selector (P term sel) lins = do nr <- evalTerm cnc_defs [] sel + convertTerm cnc_defs (TuplePrj nr selector) term lins +convertTerm cnc_defs selector (FV vars) lins = do term <- member vars + convertTerm cnc_defs selector term lins +convertTerm cnc_defs selector (S ts) ((lbl_path,lin) : lins) = do projectHead lbl_path + foldM (\lins t -> convertTerm cnc_defs selector t lins) ((lbl_path,lin) : lins) (reverse ts) +convertTerm cnc_defs selector (K (KS str)) ((lbl_path,lin) : lins) = + do projectHead lbl_path + return ((lbl_path,Right str : lin) : lins) +convertTerm cnc_defs selector (K (KP strs vars))((lbl_path,lin) : lins) = + do projectHead lbl_path + toks <- member (strs:[strs' | Alt strs' _ <- vars]) + return ((lbl_path, map Right toks ++ lin) : lins) +convertTerm cnc_defs selector (F id) lins = do term <- Map.lookup id cnc_defs + convertTerm cnc_defs selector term lins +convertTerm cnc_defs selector (W s t) ((lbl_path,lin) : lins) = do + ss <- case t of + R ss -> return ss + F f -> do + t <- Map.lookup f cnc_defs + case t of + R ss -> return ss + convertRec cnc_defs selector 0 [K (KS (s ++ s1)) | K (KS s1) <- ss] lbl_path lin lins +convertTerm cnc_defs selector x lins = error ("convertTerm ("++show x++")") + + +convertArg (TupleSel record) nr path lbl_path lin lins = + foldM (\lins (lbl, selector) -> convertArg selector nr (lbl:path) (lbl:lbl_path) lin lins) lins record +convertArg (TuplePrj lbl selector) nr path lbl_path lin lins = + convertArg selector nr (lbl:path) lbl_path lin lins +convertArg (ConSel indices) nr path lbl_path lin lins = do + index <- member indices + restrictHead lbl_path index + restrictArg nr path index + return lins +convertArg StrSel nr path lbl_path lin lins = do + projectHead lbl_path + xnr <- projectArg nr path + return ((lbl_path, Left (path, nr, xnr) : lin) : lins) + +convertCon (ConSel indices) index lbl_path lin lins = do + guard (index `elem` indices) + restrictHead lbl_path index + return lins +convertCon x _ _ _ _ = error $ "SimpleToFCFG,convertCon: " ++ show x + +convertRec cnc_defs selector index [] lbl_path lin lins = return lins +convertRec cnc_defs selector@(TupleSel fields) index (val:record) lbl_path lin lins = select fields + where + select [] = convertRec cnc_defs selector (index+1) record lbl_path lin lins + select ((index',sub_sel) : fields) + | index == index' = do lins <- convertTerm cnc_defs sub_sel val ((index:lbl_path,lin) : lins) + convertRec cnc_defs selector (index+1) record lbl_path lin lins + | otherwise = select fields +convertRec cnc_defs (TuplePrj index' sub_sel) index record lbl_path lin lins = do + convertTerm cnc_defs sub_sel (record !! (index'-index)) ((lbl_path,lin) : lins) + + +------------------------------------------------------------ +-- eval a term to ground terms + +evalTerm :: TermMap -> FPath -> Term -> CnvMonad FIndex +evalTerm cnc_defs path (V nr) = do term <- readArgCType nr + unifyPType nr (reverse path) (selectTerm path term) +evalTerm cnc_defs path (C nr) = return nr +evalTerm cnc_defs path (R record) = case path of + (index:path) -> evalTerm cnc_defs path (record !! index) +evalTerm cnc_defs path (P term sel) = do index <- evalTerm cnc_defs [] sel + evalTerm cnc_defs (index:path) term +evalTerm cnc_defs path (FV terms) = member terms >>= evalTerm cnc_defs path +evalTerm cnc_defs path (F id) = do term <- Map.lookup id cnc_defs + evalTerm cnc_defs path term +evalTerm cnc_defs path x = error ("evalTerm ("++show x++")") + +unifyPType :: FIndex -> FPath -> Term -> CnvMonad FIndex +unifyPType nr path (C max_index) = + do (_, args, _, _) <- readState + let (PFCat _ _ tcs,_) = args !! nr + case lookup path tcs of + Just index -> return index + Nothing -> do index <- member [0..max_index] + restrictArg nr path index + return index +unifyPType nr path t = error $ "unifyPType " ++ show t ---- AR 2/10/2007 + +selectTerm :: FPath -> Term -> Term +selectTerm [] term = term +selectTerm (index:path) (R record) = selectTerm path (record !! index) + + +---------------------------------------------------------------------- +-- FRulesEnv + +data FRulesEnv = FRulesEnv {-# UNPACK #-} !Int FCatSet [FRule] +type FCatSet = Map.Map CId (Map.Map [FPath] (Map.Map [(FPath,FIndex)] (Either FCat FCat))) + +data ProtoFCat = PFCat CId [FPath] [(FPath,FIndex)] + +protoFCat :: CId -> ProtoFCat +protoFCat cat = PFCat cat [] [] + +emptyFRulesEnv = FRulesEnv 0 (ins fcatString (mkCId "String") [[0]] [] $ + ins fcatInt (mkCId "Int") [[0]] [] $ + ins fcatFloat (mkCId "Float") [[0]] [] $ + ins fcatVar (mkCId "_Var") [[0]] [] $ + Map.empty) [] + where + ins fcat cat rcs tcs fcatSet = + Map.insertWith (\_ -> Map.insertWith (\_ -> Map.insert tcs right_fcat) rcs tmap_s) cat rmap_s fcatSet + where + right_fcat = Right fcat + tmap_s = Map.singleton tcs right_fcat + rmap_s = Map.singleton rcs tmap_s + +addFRule :: FRulesEnv -> FRule -> FRulesEnv +addFRule (FRulesEnv last_id fcatSet rules) rule = FRulesEnv last_id fcatSet (rule:rules) + +getFGrammar :: FRulesEnv -> FGrammar +getFGrammar (FRulesEnv last_id fcatSet rules) = (rules, Map.map getFCatList fcatSet) + where + getFCatList rcs = Map.fold (\tcs lst -> Map.fold (\x lst -> either id id x : lst) lst tcs) [] rcs + +genFCatHead :: FRulesEnv -> ProtoFCat -> (FRulesEnv, FCat) +genFCatHead env@(FRulesEnv last_id fcatSet rules) (PFCat cat rcs tcs) = + case Map.lookup cat fcatSet >>= Map.lookup rcs >>= Map.lookup tcs of + Just (Left fcat) -> (FRulesEnv last_id (ins fcat) rules, fcat) + Just (Right fcat) -> (env, fcat) + Nothing -> let fcat = last_id+1 + in (FRulesEnv fcat (ins fcat) rules, fcat) + where + ins fcat = Map.insertWith (\_ -> Map.insertWith (\_ -> Map.insert tcs right_fcat) rcs tmap_s) cat rmap_s fcatSet + where + right_fcat = Right fcat + tmap_s = Map.singleton tcs right_fcat + rmap_s = Map.singleton rcs tmap_s + +genFCatArg :: TermMap -> Term -> FRulesEnv -> ProtoFCat -> (FRulesEnv, FCat) +genFCatArg cnc_defs ctype env@(FRulesEnv last_id fcatSet rules) (PFCat cat rcs tcs) = + case Map.lookup cat fcatSet >>= Map.lookup rcs of + Just tmap -> case Map.lookup tcs tmap of + Just (Left fcat) -> (env, fcat) + Just (Right fcat) -> (env, fcat) + Nothing -> ins tmap + Nothing -> ins Map.empty + where + ins tmap = + let fcat = last_id+1 + (either_fcat,last_id1,tmap1,rules1) + = foldBM (\tcs st (either_fcat,last_id,tmap,rules) -> + let (last_id1,tmap1,fcat_arg) = addArg tcs last_id tmap + rule = FRule wildCId [[0]] [fcat_arg] fcat + (listArray (0,length rcs-1) [listArray (0,0) [FSymCat lbl 0] | lbl <- [0..length rcs-1]]) + in if st + then (Right fcat, last_id1,tmap1,rule:rules) + else (either_fcat,last_id, tmap, rules)) + (Left fcat,fcat,Map.insert tcs either_fcat tmap,rules) + (gen_tcs ctype [] []) + False + rmap1 = Map.singleton rcs tmap1 + in (FRulesEnv last_id1 (Map.insertWith (\_ -> Map.insert rcs tmap1) cat rmap1 fcatSet) rules1, fcat) + where + addArg tcs last_id tmap = + case Map.lookup tcs tmap of + Just (Left fcat) -> (last_id, tmap, fcat) + Just (Right fcat) -> (last_id, tmap, fcat) + Nothing -> let fcat = last_id+1 + in (fcat, Map.insert tcs (Left fcat) tmap, fcat) + + gen_tcs :: Term -> FPath -> [(FPath,FIndex)] -> BacktrackM Bool [(FPath,FIndex)] + gen_tcs (R record) path acc = foldM (\acc (label,ctype) -> gen_tcs ctype (label:path) acc) acc (zip [0..] record) + gen_tcs (S _) path acc = return acc + gen_tcs (C max_index) path acc = + case List.lookup path tcs of + Just index -> return $! addConstraint path index acc + Nothing -> do writeState True + index <- member [0..max_index] + return $! addConstraint path index acc + where + addConstraint path0 index0 (c@(path,index) : cs) + | path0 > path = c:addConstraint path0 index0 cs + addConstraint path0 index0 cs = (path0,index0) : cs + gen_tcs (F id) path acc = case Map.lookup id cnc_defs of + Just term -> gen_tcs term path acc + Nothing -> error ("unknown identifier: "++prCId id) + + + +------------------------------------------------------------ +-- TODO queue organization + +type XRulesMap = Map.Map CId [XRule] +data XRule = XRule CId {- function -} + [CId] {- argument types -} + CId {- result type -} + [Term] {- argument lin-types representation -} + Term {- result lin-type representation -} + Term {- body -} + +takeToDoRules :: XRulesMap -> FRulesEnv -> ([([XRule], TermSelector)], FRulesEnv) +takeToDoRules xrulesMap (FRulesEnv last_id fcatSet rules) = (todo,FRulesEnv last_id fcatSet' rules) + where + (todo,fcatSet') = + Map.mapAccumWithKey (\todo cat rmap -> + let (todo1,rmap1) = Map.mapAccumWithKey (\todo rcs tmap -> + let (tcss,tmap') = Map.mapAccumWithKey (\tcss tcs either_xcat -> + case either_xcat of + Left xcat -> (tcs:tcss,Right xcat) + Right xcat -> ( tcss,either_xcat)) [] tmap + in case tcss of + [] -> ( todo,tmap ) + _ -> ((srules,mkSelector rcs tcss) : todo,tmap')) todo rmap + mb_srules = Map.lookup cat xrulesMap + Just srules = mb_srules + + in case mb_srules of + Just srules -> (todo1,rmap1) + Nothing -> (todo ,rmap1)) [] fcatSet + + +------------------------------------------------------------ +-- The TermSelector + +data TermSelector + = TupleSel [(FIndex, TermSelector)] + | TuplePrj FIndex TermSelector + | ConSel [FIndex] + | StrSel + deriving Show + +mkSingletonSelectors :: TermMap + -> Term -- ^ Type representation term + -> [TermSelector] -- ^ list of selectors containing just one string field +mkSingletonSelectors cnc_defs term = sels0 + where + (sels0,tcss0) = loop [] ([],[]) term + + loop path st (R record) = List.foldl' (\st (index,term) -> loop (index:path) st term) st (zip [0..] record) + loop path (sels,tcss) (C i) = ( sels,map ((,) path) [0..i] : tcss) + loop path (sels,tcss) (S _) = (mkSelector [path] tcss0 : sels, tcss) + loop path (sels,tcss) (F id) = case Map.lookup id cnc_defs of + Just term -> loop path (sels,tcss) term + Nothing -> error ("unknown identifier: "++prCId id) + +mkSelector :: [FPath] -> [[(FPath,FIndex)]] -> TermSelector +mkSelector rcs tcss = + List.foldl' addRestriction (case xs of + (path:xs) -> List.foldl' addProjection (path2selector StrSel path) xs) ys + where + xs = [ reverse path | path <- rcs] + ys = [(reverse path,term) | tcs <- tcss, (path,term) <- tcs] + + addRestriction :: TermSelector -> (FPath,FIndex) -> TermSelector + addRestriction (ConSel indices) ([] ,n_index) = ConSel (add indices) + where + add [] = [n_index] + add (index':indices) + | n_index == index' = index': indices + | otherwise = index':add indices + addRestriction (TupleSel fields) (index : path,n_index) = TupleSel (add fields) + where + add [] = [(index,path2selector (ConSel [n_index]) path)] + add (field@(index',sub_sel):fields) + | index == index' = (index',addRestriction sub_sel (path,n_index)):fields + | otherwise = field : add fields + + addProjection :: TermSelector -> FPath -> TermSelector + addProjection StrSel [] = StrSel + addProjection (TupleSel fields) (index : path) = TupleSel (add fields) + where + add [] = [(index,path2selector StrSel path)] + add (field@(index',sub_sel):fields) + | index == index' = (index',addProjection sub_sel path):fields + | otherwise = field : add fields + + path2selector base [] = base + path2selector base (index : path) = TupleSel [(index,path2selector base path)] + +------------------------------------------------------------ +-- updating the MCF rule + +readArgCType :: FIndex -> CnvMonad Term +readArgCType nr = do (_, _, _, ctypes) <- readState + return (ctypes !! nr) + +restrictArg :: FIndex -> FPath -> FIndex -> CnvMonad () +restrictArg nr path index = do + (head, args, ctype, ctypes) <- readState + args' <- updateNthM (\(xcat,xs) -> do xcat <- restrictProtoFCat path index xcat + return (xcat,xs) ) nr args + writeState (head, args', ctype, ctypes) + +projectArg :: FIndex -> FPath -> CnvMonad Int +projectArg nr path = do + (head, args, ctype, ctypes) <- readState + (xnr,args') <- updateArgs nr args + writeState (head, args', ctype, ctypes) + return xnr + where + updateArgs :: FIndex -> [(ProtoFCat,[FPath])] -> CnvMonad (Int,[(ProtoFCat,[FPath])]) + updateArgs 0 ((a@(PFCat _ rcs _),xpaths) : as) + | path `elem` rcs = return (length xpaths+1,(a,path:xpaths):as) + | otherwise = do a <- projectProtoFCat path a + return (0,(a,xpaths):as) + updateArgs n (a : as) = do + (xnr,as) <- updateArgs (n-1) as + return (xnr,a:as) + +readHeadCType :: CnvMonad Term +readHeadCType = do (_, _, ctype, _) <- readState + return ctype + +restrictHead :: FPath -> FIndex -> CnvMonad () +restrictHead path term + = do (head, args, ctype, ctypes) <- readState + head' <- restrictProtoFCat path term head + writeState (head', args, ctype, ctypes) + +projectHead :: FPath -> CnvMonad () +projectHead path + = do (head, args, ctype, ctypes) <- readState + head' <- projectProtoFCat path head + writeState (head', args, ctype, ctypes) + +restrictProtoFCat :: FPath -> FIndex -> ProtoFCat -> CnvMonad ProtoFCat +restrictProtoFCat path0 index0 (PFCat cat rcs tcs) = do + tcs <- addConstraint tcs + return (PFCat cat rcs tcs) + where + addConstraint (c@(path,index) : cs) + | path0 > path = liftM (c:) (addConstraint cs) + | path0 == path = guard (index0 == index) >> + return (c : cs) + addConstraint cs = return ((path0,index0) : cs) + +projectProtoFCat :: FPath -> ProtoFCat -> CnvMonad ProtoFCat +projectProtoFCat path0 (PFCat cat rcs tcs) = do + return (PFCat cat (addConstraint rcs) tcs) + where + addConstraint (path : rcs) + | path0 > path = path : addConstraint rcs + | path0 == path = path : rcs + addConstraint rcs = path0 : rcs diff --git a/src/GF/Compile/GeneratePMCFG.hs b/src/GF/Compile/GeneratePMCFG.hs new file mode 100644 index 000000000..e0343e8d6 --- /dev/null +++ b/src/GF/Compile/GeneratePMCFG.hs @@ -0,0 +1,356 @@ +{-# OPTIONS -fbang-patterns #-} +---------------------------------------------------------------------- +-- | +-- Maintainer : Krasimir Angelov +-- Stability : (stable) +-- Portability : (portable) +-- +-- Converting SimpleGFC grammars to fast nonerasing MCFG grammar. +-- +-- the resulting grammars might be /very large/ +-- +-- the conversion is only equivalent if the GFC grammar has a context-free backbone. +----------------------------------------------------------------------------- + + +module GF.Compile.GeneratePMCFG + (convertConcrete) where + +import PGF.CId +import PGF.Data +import PGF.Macros --hiding (prt) +import PGF.Parsing.FCFG.Utilities + +import GF.Data.BacktrackM +import GF.Data.SortedList +import GF.Data.Utilities (updateNthM, sortNub) + +import qualified Data.Map as Map +import qualified Data.Set as Set +import qualified Data.List as List +import qualified Data.ByteString.Char8 as BS +import Data.Array +import Data.Maybe +import Control.Monad +import Debug.Trace + +---------------------------------------------------------------------- +-- main conversion function + +convertConcrete :: Abstr -> Concr -> FGrammar +convertConcrete abs cnc = fixHoasFuns $ convert abs_defs' conc' cats' + where abs_defs = Map.assocs (funs abs) + conc = Map.union (opers cnc) (lins cnc) -- "union big+small most efficient" + cats = lincats cnc + (abs_defs',conc',cats') = expandHOAS abs_defs conc cats + +expandHOAS :: [(CId,(Type,Expr))] -> TermMap -> TermMap -> ([(CId,(Type,Expr))],TermMap,TermMap) +expandHOAS funs lins lincats = (funs' ++ hoFuns ++ varFuns, + Map.unions [lins, hoLins, varLins], + Map.unions [lincats, hoLincats, varLincat]) + where + -- replace higher-order fun argument types with new categories + funs' = [(f,(fixType ty,e)) | (f,(ty,e)) <- funs] + where + fixType :: Type -> Type + fixType ty = let (ats,rt) = typeSkeleton ty in cftype (map catName ats) rt + + hoTypes :: [(Int,CId)] + hoTypes = sortNub [(n,c) | (_,(ty,_)) <- funs, (n,c) <- fst (typeSkeleton ty), n > 0] + hoCats = sortNub (map snd hoTypes) + -- for each Cat with N bindings, we add a new category _NCat + -- each new category contains a single function __NCat : Cat -> _Var -> ... -> _Var -> _NCat + hoFuns = [(funName ty,(cftype (c : replicate n varCat) (catName ty),EEq [])) | ty@(n,c) <- hoTypes] + -- lincats for the new categories + hoLincats = Map.fromList [(catName ty, modifyRec (++ replicate n (S [])) (lincatOf c)) | ty@(n,c) <- hoTypes] + -- linearizations of the new functions, lin __NCat v_0 ... v_n-1 x = { s1 = x.s1; ...; sk = x.sk; $0 = v_0.s ... + hoLins = Map.fromList [ (funName ty, mkLin c n) | ty@(n,c) <- hoTypes] + where mkLin c n = modifyRec (\fs -> [P (V 0) (C j) | j <- [0..length fs-1]] ++ [P (V i) (C 0) | i <- [1..n]]) (lincatOf c) + -- for each Cat, we a add a fun _Var_Cat : _Var -> Cat + varFuns = [(varFunName cat, (cftype [varCat] cat,EEq [])) | cat <- hoCats] + -- linearizations of the _Var_Cat functions + varLins = Map.fromList [(varFunName cat, R [P (V 0) (C 0)]) | cat <- hoCats] + -- lincat for the _Var category + varLincat = Map.singleton varCat (R [S []]) + + lincatOf c = fromMaybe (error $ "No lincat for " ++ prCId c) $ Map.lookup c lincats + + modifyRec :: ([Term] -> [Term]) -> Term -> Term + modifyRec f (R xs) = R (f xs) + modifyRec _ t = error $ "Not a record: " ++ show t + + varCat = mkCId "_Var" + + catName :: (Int,CId) -> CId + catName (0,c) = c + catName (n,c) = mkCId ("_" ++ show n ++ prCId c) + + funName :: (Int,CId) -> CId + funName (n,c) = mkCId ("__" ++ show n ++ prCId c) + + varFunName :: CId -> CId + varFunName c = mkCId ("_Var_" ++ prCId c) + +-- replaces __NCat with _B and _Var_Cat with _. +-- the temporary names are just there to avoid name collisions. +fixHoasFuns :: FGrammar -> FGrammar +fixHoasFuns (!rs, !cs) = ([FRule (fixName n) ps args cat lins | FRule n ps args cat lins <- rs], cs) + where fixName (CId n) | BS.pack "__" `BS.isPrefixOf` n = (mkCId "_B") + | BS.pack "_Var_" `BS.isPrefixOf` n = wildCId + fixName n = n + +convert :: [(CId,(Type,Expr))] -> TermMap -> TermMap -> FGrammar +convert abs_defs cnc_defs cat_defs = getFGrammar (List.foldl' (convertRule cnc_defs) emptyFRulesEnv srules) + where + srules = [ + (XRule id args res (map findLinType args) (findLinType res) term) | + (id, (ty,_)) <- abs_defs, let (args,res) = catSkeleton ty, + term <- Map.lookup id cnc_defs] + + findLinType id = fromMaybe (error $ "No lincat for " ++ show id) (Map.lookup id cat_defs) + + +convertRule :: TermMap -> FRulesEnv -> XRule -> FRulesEnv +convertRule cnc_defs frulesEnv (XRule fun args cat ctypes ctype term) = + foldBM addRule + frulesEnv + (convertTerm cnc_defs [] ctype term [([],[])]) + (protoFCat cnc_defs cat ctype, zipWith (protoFCat cnc_defs) args ctypes) + where + addRule linRec (newCat', newArgs') env0 = + let (env1, newCat) = genFCatHead env0 newCat' + (env2, newArgs) = List.mapAccumL (genFCatArg cnc_defs) env1 newArgs' + + newLinRec = mkArray (map (mkArray . snd) linRec) + mkArray lst = listArray (0,length lst-1) lst + + rule = FRule fun [] newArgs newCat newLinRec + in addFRule env2 rule + +---------------------------------------------------------------------- +-- term conversion + +type CnvMonad a = BacktrackM Env a + +type FPath = [FIndex] +data ProtoFCat = PFCat CId [FPath] [(FPath,FIndex)] Term +type Env = (ProtoFCat, [ProtoFCat]) +type LinRec = [(FPath, [FSymbol])] +data XRule = XRule CId {- function -} + [CId] {- argument types -} + CId {- result type -} + [Term] {- argument lin-types representation -} + Term {- result lin-type representation -} + Term {- body -} + +protoFCat :: TermMap -> CId -> Term -> ProtoFCat +protoFCat cnc_defs cat ctype = PFCat cat (getRCS cnc_defs ctype) [] ctype + +type TermMap = Map.Map CId Term + +convertTerm :: TermMap -> FPath -> Term -> Term -> LinRec -> CnvMonad LinRec +convertTerm cnc_defs sel ctype (V nr) ((lbl_path,lin) : lins) = convertArg ctype nr (reverse sel) lbl_path lin lins +convertTerm cnc_defs sel ctype (C nr) ((lbl_path,lin) : lins) = convertCon ctype nr (reverse sel) lbl_path lin lins +convertTerm cnc_defs sel ctype (R record) ((lbl_path,lin) : lins) = convertRec cnc_defs sel ctype record lbl_path lin lins +convertTerm cnc_defs sel ctype (P term p) lins = do nr <- evalTerm cnc_defs [] p + convertTerm cnc_defs (nr:sel) ctype term lins +convertTerm cnc_defs sel ctype (FV vars) lins = do term <- member vars + convertTerm cnc_defs sel ctype term lins +convertTerm cnc_defs sel ctype (S ts) ((lbl_path,lin) : lins) = foldM (\lins t -> convertTerm cnc_defs sel ctype t lins) ((lbl_path,lin) : lins) (reverse ts) +convertTerm cnc_defs sel ctype (K (KS str)) ((lbl_path,lin) : lins) = return ((lbl_path,FSymTok str : lin) : lins) +convertTerm cnc_defs sel ctype (K (KP strs vars))((lbl_path,lin) : lins) = + do toks <- member (strs:[strs' | Alt strs' _ <- vars]) + return ((lbl_path, map FSymTok toks ++ lin) : lins) +convertTerm cnc_defs sel ctype (F id) lins = do term <- Map.lookup id cnc_defs + convertTerm cnc_defs sel ctype term lins +convertTerm cnc_defs sel ctype (W s t) ((lbl_path,lin) : lins) = do + ss <- case t of + R ss -> return ss + F f -> do + t <- Map.lookup f cnc_defs + case t of + R ss -> return ss + convertRec cnc_defs sel ctype [K (KS (s ++ s1)) | K (KS s1) <- ss] lbl_path lin lins +convertTerm cnc_defs sel ctype x lins = error ("convertTerm ("++show x++")") + + +convertArg (R record) nr path lbl_path lin lins = + foldM (\lins (lbl, ctype) -> convertArg ctype nr (lbl:path) (lbl:lbl_path) lin lins) lins (zip [0..] record) +convertArg (C max) nr path lbl_path lin lins = do + index <- member [0..max] + restrictHead lbl_path index + restrictArg nr path index + return lins +convertArg (S _) nr path lbl_path lin lins = do + (_, args) <- readState + let PFCat cat rcs tcs _ = args !! nr + return ((lbl_path, FSymCat (index path rcs 0) nr : lin) : lins) + where + index lbl' (lbl:lbls) idx + | lbl' == lbl = idx + | otherwise = index lbl' lbls $! (idx+1) + + +convertCon (C max) index [] lbl_path lin lins = do + guard (index <= max) + restrictHead lbl_path index + return lins +convertCon x _ _ _ _ _ = error $ "SimpleToFCFG,convertCon: " ++ show x + +convertRec cnc_defs [] (R ctypes) record lbl_path lin lins = + foldM (\lins (index,ctype,val) -> convertTerm cnc_defs [] ctype val ((index:lbl_path,lin) : lins)) + lins + (zip3 [0..] ctypes record) +convertRec cnc_defs (index:sub_sel) ctype record lbl_path lin lins = do + convertTerm cnc_defs sub_sel ctype (record !! index) ((lbl_path,lin) : lins) + + +------------------------------------------------------------ +-- eval a term to ground terms + +evalTerm :: TermMap -> FPath -> Term -> CnvMonad FIndex +evalTerm cnc_defs path (V nr) = do (_, args) <- readState + let PFCat _ _ _ ctype = args !! nr + unifyPType nr (reverse path) (selectTerm path ctype) +evalTerm cnc_defs path (C nr) = return nr +evalTerm cnc_defs path (R record) = case path of + (index:path) -> evalTerm cnc_defs path (record !! index) +evalTerm cnc_defs path (P term sel) = do index <- evalTerm cnc_defs [] sel + evalTerm cnc_defs (index:path) term +evalTerm cnc_defs path (FV terms) = member terms >>= evalTerm cnc_defs path +evalTerm cnc_defs path (F id) = do term <- Map.lookup id cnc_defs + evalTerm cnc_defs path term +evalTerm cnc_defs path x = error ("evalTerm ("++show x++")") + +unifyPType :: FIndex -> FPath -> Term -> CnvMonad FIndex +unifyPType nr path (C max_index) = + do (_, args) <- readState + let PFCat _ _ tcs _ = args !! nr + case lookup path tcs of + Just index -> return index + Nothing -> do index <- member [0..max_index] + restrictArg nr path index + return index +unifyPType nr path t = error $ "unifyPType " ++ show t ---- AR 2/10/2007 + +selectTerm :: FPath -> Term -> Term +selectTerm [] term = term +selectTerm (index:path) (R record) = selectTerm path (record !! index) + + +---------------------------------------------------------------------- +-- FRulesEnv + +data FRulesEnv = FRulesEnv {-# UNPACK #-} !Int FCatSet [FRule] +type FCatSet = Map.Map CId (Map.Map [(FPath,FIndex)] FCat) + +emptyFRulesEnv = FRulesEnv 0 (ins fcatString (mkCId "String") [] $ + ins fcatInt (mkCId "Int") [] $ + ins fcatFloat (mkCId "Float") [] $ + ins fcatVar (mkCId "_Var") [] $ + Map.empty) [] + where + ins fcat cat tcs fcatSet = + Map.insertWith (\_ -> Map.insert tcs fcat) cat tmap_s fcatSet + where + tmap_s = Map.singleton tcs fcat + +addFRule :: FRulesEnv -> FRule -> FRulesEnv +addFRule (FRulesEnv last_id fcatSet rules) rule = FRulesEnv last_id fcatSet (rule:rules) + +getFGrammar :: FRulesEnv -> FGrammar +getFGrammar (FRulesEnv last_id fcatSet rules) = (rules, Map.map Map.elems fcatSet) + +genFCatHead :: FRulesEnv -> ProtoFCat -> (FRulesEnv, FCat) +genFCatHead env@(FRulesEnv last_id fcatSet rules) (PFCat cat rcs tcs _) = + case Map.lookup cat fcatSet >>= Map.lookup tcs of + Just fcat -> (env, fcat) + Nothing -> let fcat = last_id+1 + in (FRulesEnv fcat (ins fcat) rules, fcat) + where + ins fcat = Map.insertWith (\_ -> Map.insert tcs fcat) cat tmap_s fcatSet + where + tmap_s = Map.singleton tcs fcat + +genFCatArg :: TermMap -> FRulesEnv -> ProtoFCat -> (FRulesEnv, FCat) +genFCatArg cnc_defs env@(FRulesEnv last_id fcatSet rules) (PFCat cat rcs tcs ctype) = + case Map.lookup cat fcatSet of + Just tmap -> case Map.lookup tcs tmap of + Just fcat -> (env, fcat) + Nothing -> ins tmap + Nothing -> ins Map.empty + where + ins tmap = + let fcat = last_id+1 + (last_id1,tmap1,rules1) + = foldBM (\tcs st (last_id,tmap,rules) -> + let (last_id1,tmap1,fcat_arg) = addArg tcs last_id tmap + rule = FRule wildCId [[0]] [fcat_arg] fcat + (listArray (0,length rcs-1) [listArray (0,0) [FSymCat lbl 0] | lbl <- [0..length rcs-1]]) + in if st + then (last_id1,tmap1,rule:rules) + else (last_id, tmap, rules)) + (fcat,Map.insert tcs fcat tmap,rules) + (gen_tcs ctype [] []) + False + in (FRulesEnv last_id1 (Map.insert cat tmap1 fcatSet) rules1, fcat) + where + addArg tcs last_id tmap = + case Map.lookup tcs tmap of + Just fcat -> (last_id, tmap, fcat) + Nothing -> let fcat = last_id+1 + in (fcat, Map.insert tcs fcat tmap, fcat) + + gen_tcs :: Term -> FPath -> [(FPath,FIndex)] -> BacktrackM Bool [(FPath,FIndex)] + gen_tcs (R record) path acc = foldM (\acc (label,ctype) -> gen_tcs ctype (label:path) acc) acc (zip [0..] record) + gen_tcs (S _) path acc = return acc + gen_tcs (C max_index) path acc = + case List.lookup path tcs of + Just index -> return $! addConstraint path index acc + Nothing -> do writeState True + index <- member [0..max_index] + return $! addConstraint path index acc + where + addConstraint path0 index0 (c@(path,index) : cs) + | path0 > path = c:addConstraint path0 index0 cs + addConstraint path0 index0 cs = (path0,index0) : cs + gen_tcs (F id) path acc = case Map.lookup id cnc_defs of + Just term -> gen_tcs term path acc + Nothing -> error ("unknown identifier: "++prCId id) + + +getRCS :: TermMap -> Term -> [FPath] +getRCS cnc_defs = loop [] [] + where + loop path rcs (R record) = List.foldl' (\rcs (index,term) -> loop (index:path) rcs term) rcs (zip [0..] record) + loop path rcs (C i) = rcs + loop path rcs (S _) = path:rcs + loop path rcs (F id) = case Map.lookup id cnc_defs of + Just term -> loop path rcs term + Nothing -> error ("unknown identifier: "++show id) + +------------------------------------------------------------ +-- updating the MCF rule + +restrictArg :: FIndex -> FPath -> FIndex -> CnvMonad () +restrictArg nr path index = do + (head, args) <- readState + args' <- updateNthM (restrictProtoFCat path index) nr args + writeState (head, args') + +restrictHead :: FPath -> FIndex -> CnvMonad () +restrictHead path term + = do (head, args) <- readState + head' <- restrictProtoFCat path term head + writeState (head', args) + +restrictProtoFCat :: FPath -> FIndex -> ProtoFCat -> CnvMonad ProtoFCat +restrictProtoFCat path0 index0 (PFCat cat rcs tcs ctype) = do + tcs <- addConstraint tcs + return (PFCat cat rcs tcs ctype) + where + addConstraint (c@(path,index) : cs) + | path0 > path = liftM (c:) (addConstraint cs) + | path0 == path = guard (index0 == index) >> + return (c : cs) + addConstraint cs = return ((path0,index0) : cs) diff --git a/src/GF/Compile/GetGrammar.hs b/src/GF/Compile/GetGrammar.hs new file mode 100644 index 000000000..a8eb8b749 --- /dev/null +++ b/src/GF/Compile/GetGrammar.hs @@ -0,0 +1,55 @@ +---------------------------------------------------------------------- +-- | +-- Module : GetGrammar +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/11/15 17:56:13 $ +-- > CVS $Author: aarne $ +-- > CVS $Revision: 1.16 $ +-- +-- this module builds the internal GF grammar that is sent to the type checker +----------------------------------------------------------------------------- + +module GF.Compile.GetGrammar where + +import GF.Data.Operations +import qualified GF.Source.ErrM as E + +import GF.Infra.UseIO +import GF.Infra.Modules +import GF.Grammar.Grammar +import qualified GF.Source.AbsGF as A +import GF.Source.SourceToGrammar +---- import Macros +---- import Rename +import GF.Infra.Option +--- import Custom +import GF.Source.ParGF +import qualified GF.Source.LexGF as L + +import GF.Compile.ReadFiles + +import Data.Char (toUpper) +import Data.List (nub) +import qualified Data.ByteString.Char8 as BS +import Control.Monad (foldM) +import System.Cmd (system) + +getSourceModule :: Options -> FilePath -> IOE SourceModule +getSourceModule opts file0 = do + file <- foldM runPreprocessor file0 (moduleFlag optPreprocessors opts) + string <- readFileIOE file + let tokens = myLexer string + mo1 <- ioeErr $ pModDef tokens + ioeErr $ transModDef mo1 + +-- FIXME: should use System.IO.openTempFile +runPreprocessor :: FilePath -> String -> IOE FilePath +runPreprocessor file0 p = + do let tmp = "_gf_preproc.tmp" + cmd = p +++ file0 ++ ">" ++ tmp + ioeIO $ system cmd + -- ioeIO $ putStrLn $ "preproc" +++ cmd + return tmp diff --git a/src/GF/Compile/GrammarToGFCC.hs b/src/GF/Compile/GrammarToGFCC.hs new file mode 100644 index 000000000..d14a914f1 --- /dev/null +++ b/src/GF/Compile/GrammarToGFCC.hs @@ -0,0 +1,561 @@ +{-# LANGUAGE PatternGuards #-} +module GF.Compile.GrammarToGFCC (prGrammar2gfcc,mkCanon2gfcc,addParsers) where + +import GF.Compile.Export +import GF.Compile.OptimizeGF (unshareModule) +import qualified GF.Compile.GenerateFCFG as FCFG +import qualified GF.Compile.GeneratePMCFG as PMCFG + +import PGF.CId +import PGF.BuildParser (buildParserInfo) +import qualified PGF.Macros as CM +import qualified PGF.Data as C +import qualified PGF.Data as D +import GF.Grammar.Predef +import GF.Grammar.PrGrammar +import GF.Grammar.Grammar +import qualified GF.Grammar.Lookup as Look +import qualified GF.Grammar.Abstract as A +import qualified GF.Grammar.Macros as GM +import qualified GF.Compile.Compute as Compute ---- +import qualified GF.Infra.Modules as M +import qualified GF.Infra.Option as O + +import GF.Infra.Ident +import GF.Infra.Option +import GF.Data.Operations +import GF.Text.UTF8 + +import Data.List +import Data.Char (isDigit,isSpace) +import qualified Data.Map as Map +import qualified Data.ByteString.Char8 as BS +import Debug.Trace ---- + +-- when developing, swap commenting + +--traceD s t = trace s t +traceD s t = t + + +-- the main function: generate PGF from GF. + +prGrammar2gfcc :: Options -> String -> SourceGrammar -> (String,String) +prGrammar2gfcc opts cnc gr = (abs,printPGF gc) where + (abs,gc) = mkCanon2gfcc opts cnc gr + +mkCanon2gfcc :: Options -> String -> SourceGrammar -> (String,D.PGF) +mkCanon2gfcc opts cnc gr = + (prIdent abs, (canon2gfcc opts pars . reorder abs . canon2canon abs) gr) + where + abs = err error id $ M.abstractOfConcrete gr (identC (BS.pack cnc)) + pars = mkParamLincat gr + +-- Adds parsers for all concretes +addParsers :: D.PGF -> D.PGF +addParsers pgf = pgf { D.concretes = Map.map conv (D.concretes pgf) } + where + conv cnc = cnc { D.parser = Just (buildParserInfo fcfg) } + where + fcfg + | Map.lookup (mkCId "erasing") (D.cflags cnc) == Just "on" = PMCFG.convertConcrete (D.abstract pgf) cnc + | otherwise = FCFG.convertConcrete (D.abstract pgf) cnc + + +-- Generate PGF from GFCM. +-- this assumes a grammar translated by canon2canon + +canon2gfcc :: Options -> (Ident -> Ident -> C.Term) -> SourceGrammar -> D.PGF +canon2gfcc opts pars cgr@(M.MGrammar ((a,M.ModMod abm):cms)) = + (if dump opts DumpCanon then trace (prGrammar cgr) else id) $ + D.PGF an cns gflags abs cncs + where + -- abstract + an = (i2i a) + cns = map (i2i . fst) cms + abs = D.Abstr aflags funs cats catfuns + gflags = Map.empty + aflags = Map.fromList [(mkCId f,x) | (f,x) <- moduleOptionsGFO (M.flags abm)] + mkDef pty = case pty of + Yes t -> mkExp t + _ -> CM.primNotion + + -- concretes + lfuns = [(f', (mkType ty, mkDef pty)) | + (f,AbsFun (Yes ty) pty) <- tree2list (M.jments abm), let f' = i2i f] + funs = Map.fromAscList lfuns + lcats = [(i2i c, mkContext cont) | + (c,AbsCat (Yes cont) _) <- tree2list (M.jments abm)] + cats = Map.fromAscList lcats + catfuns = Map.fromList + [(cat,[f | (f, (C.DTyp _ c _,_)) <- lfuns, c==cat]) | (cat,_) <- lcats] + + cncs = Map.fromList [mkConcr lang (i2i lang) mo | (lang,M.ModMod mo) <- cms] + mkConcr lang0 lang mo = + (lang,D.Concr flags lins opers lincats lindefs printnames params fcfg) + where + js = tree2list (M.jments mo) + flags = Map.fromList [(mkCId f,x) | (f,x) <- moduleOptionsGFO (M.flags mo)] + opers = Map.fromAscList [] -- opers will be created as optimization + utf = if moduleFlag optEncoding (moduleOptions (M.flags mo)) == UTF_8 + then D.convertStringsInTerm decodeUTF8 else id + lins = Map.fromAscList + [(i2i f, utf (mkTerm tr)) | (f,CncFun _ (Yes tr) _) <- js] + lincats = Map.fromAscList + [(i2i c, mkCType ty) | (c,CncCat (Yes ty) _ _) <- js] + lindefs = Map.fromAscList + [(i2i c, mkTerm tr) | (c,CncCat _ (Yes tr) _) <- js] + printnames = Map.union + (Map.fromAscList [(i2i f, mkTerm tr) | (f,CncFun _ _ (Yes tr)) <- js]) + (Map.fromAscList [(i2i f, mkTerm tr) | (f,CncCat _ _ (Yes tr)) <- js]) + params = Map.fromAscList + [(i2i c, pars lang0 c) | (c,CncCat (Yes ty) _ _) <- js] + fcfg = Nothing + +i2i :: Ident -> CId +i2i = CId . ident2bs + +mkType :: A.Type -> C.Type +mkType t = case GM.typeForm t of + Ok (hyps,(_,cat),args) -> C.DTyp (mkContext hyps) (i2i cat) (map mkExp args) + +mkExp :: A.Term -> C.Expr +mkExp t = case t of + A.Eqs eqs -> C.EEq [C.Equ (map mkPatt ps) (mkExp e) | (ps,e) <- eqs] + _ -> case GM.termForm t of + Ok (xs,c,args) -> mkAbs xs (mkApp c (map mkExp args)) + where + mkAbs xs t = foldr (C.EAbs . i2i) t xs + mkApp c args = case c of + Q _ c -> foldl C.EApp (C.EVar (i2i c)) args + QC _ c -> foldl C.EApp (C.EVar (i2i c)) args + Vr x -> C.EVar (i2i x) + EInt i -> C.ELit (C.LInt i) + EFloat f -> C.ELit (C.LFlt f) + K s -> C.ELit (C.LStr s) + Meta (MetaSymb i) -> C.EMeta i + _ -> C.EMeta 0 + mkPatt p = case p of + A.PP _ c ps -> foldl C.EApp (C.EVar (i2i c)) (map mkPatt ps) + A.PV x -> C.EVar (i2i x) + A.PW -> C.EVar wildCId + A.PInt i -> C.ELit (C.LInt i) + +mkContext :: A.Context -> [C.Hypo] +mkContext hyps = [C.Hyp (i2i x) (mkType ty) | (x,ty) <- hyps] + +mkTerm :: Term -> C.Term +mkTerm tr = case tr of + Vr (IA _ i) -> C.V i + Vr (IAV _ _ i) -> C.V i + Vr (IC s) | isDigit (BS.last s) -> + C.V ((read . BS.unpack . snd . BS.spanEnd isDigit) s) + ---- from gf parser of gfc + EInt i -> C.C $ fromInteger i + R rs -> C.R [mkTerm t | (_, (_,t)) <- rs] + P t l -> C.P (mkTerm t) (C.C (mkLab l)) + TSh _ _ -> error $ show tr + T _ cs -> C.R [mkTerm t | (_,t) <- cs] ------ + V _ cs -> C.R [mkTerm t | t <- cs] + S t p -> C.P (mkTerm t) (mkTerm p) + C s t -> C.S $ concatMap flats [mkTerm x | x <- [s,t]] + FV ts -> C.FV [mkTerm t | t <- ts] + K s -> C.K (C.KS s) +----- K (KP ss _) -> C.K (C.KP ss []) ---- TODO: prefix variants + Empty -> C.S [] + App _ _ -> prtTrace tr $ C.C 66661 ---- for debugging + Abs _ t -> mkTerm t ---- only on toplevel + Alts (td,tvs) -> + C.K (C.KP (strings td) [C.Alt (strings u) (strings v) | (u,v) <- tvs]) + _ -> prtTrace tr $ C.S [C.K (C.KS (A.prt tr +++ "66662"))] ---- for debugging + where + mkLab (LIdent l) = case BS.unpack l of + '_':ds -> (read ds) :: Int + _ -> prtTrace tr $ 66663 + strings t = case t of + K s -> [s] + C u v -> strings u ++ strings v + Strs ss -> concatMap strings ss + _ -> prtTrace tr $ ["66660"] + flats t = case t of + C.S ts -> concatMap flats ts + _ -> [t] + +-- encoding PGF-internal lincats as terms +mkCType :: Type -> C.Term +mkCType t = case t of + EInt i -> C.C $ fromInteger i + RecType rs -> C.R [mkCType t | (_, t) <- rs] + Table pt vt -> case pt of + EInt i -> C.R $ replicate (1 + fromInteger i) $ mkCType vt + RecType rs -> mkCType $ foldr Table vt (map snd rs) + Sort s | s == cStr -> C.S [] --- Str only + _ | Just i <- GM.isTypeInts t -> C.C $ fromInteger i + _ -> error $ "mkCType " ++ show t + +-- encoding showable lincats (as in source gf) as terms +mkParamLincat :: SourceGrammar -> Ident -> Ident -> C.Term +mkParamLincat sgr lang cat = errVal (C.R [C.S []]) $ do + typ <- Look.lookupLincat sgr lang cat + mkPType typ + where + mkPType typ = case typ of + RecType lts -> do + ts <- mapM (mkPType . snd) lts + return $ C.R [ C.P (kks $ prt_ l) t | ((l,_),t) <- zip lts ts] + Table (RecType lts) v -> do + ps <- mapM (mkPType . snd) lts + v' <- mkPType v + return $ foldr (\p v -> C.S [p,v]) v' ps + Table p v -> do + p' <- mkPType p + v' <- mkPType v + return $ C.S [p',v'] + Sort s | s == cStr -> return $ C.S [] + _ -> return $ + C.FV $ map (kks . filter showable . prt_) $ + errVal [] $ Look.allParamValues sgr typ + showable c = not (isSpace c) ---- || (c == ' ') -- to eliminate \n in records + kks = C.K . C.KS + +-- return just one module per language + +reorder :: Ident -> SourceGrammar -> SourceGrammar +reorder abs cg = M.MGrammar $ + (abs, M.ModMod $ + M.Module M.MTAbstract M.MSComplete aflags [] [] adefs poss): + [(c, M.ModMod $ + M.Module (M.MTConcrete abs) M.MSComplete fs [] [] (sorted2tree js) poss) + | (c,(fs,js)) <- cncs] + where + poss = emptyBinTree -- positions no longer needed + mos = M.allModMod cg + adefs = sorted2tree $ sortIds $ + predefADefs ++ Look.allOrigInfos cg abs + predefADefs = + [(c, AbsCat (Yes []) Nope) | c <- [cFloat,cInt,cString]] + aflags = + concatModuleOptions [M.flags mo | (_,mo) <- M.allModMod cg, M.isModAbs mo] + + cncs = sortIds [(lang, concr lang) | lang <- M.allConcretes cg abs] + concr la = (flags, + sortIds (predefCDefs ++ jments)) where + jments = Look.allOrigInfos cg la + flags = concatModuleOptions + [M.flags mo | + (i,mo) <- mos, M.isModCnc mo, + Just r <- [lookup i (M.allExtendSpecs cg la)]] + + predefCDefs = + [(c, CncCat (Yes GM.defLinType) Nope Nope) | c <- [cInt,cFloat,cString]] + + sortIds = sortBy (\ (f,_) (g,_) -> compare f g) + + +-- one grammar per language - needed for symtab generation +repartition :: Ident -> SourceGrammar -> [SourceGrammar] +repartition abs cg = [M.partOfGrammar cg (lang,mo) | + let mos = M.allModMod cg, + lang <- M.allConcretes cg abs, + let mo = errVal + (error ("no module found for " ++ A.prt lang)) $ M.lookupModule cg lang + ] + + +-- translate tables and records to arrays, parameters and labels to indices + +canon2canon :: Ident -> SourceGrammar -> SourceGrammar +canon2canon abs = + recollect . map cl2cl . repartition abs . purgeGrammar abs + where + recollect = M.MGrammar . nubBy (\ (i,_) (j,_) -> i==j) . concatMap M.modules + cl2cl = M.MGrammar . js2js . map (c2c p2p) . M.modules + + js2js ms = map (c2c (j2j (M.MGrammar ms))) ms + + c2c f2 (c,m) = case m of + M.ModMod mo -> + (c, M.ModMod $ M.replaceJudgements mo $ mapTree f2 (M.jments mo)) + _ -> (c,m) + j2j cg (f,j) = case j of + CncFun x (Yes tr) z -> (f,CncFun x (Yes ({-trace ("+ " ++ prt f)-} (t2t tr))) z) + CncCat (Yes ty) (Yes x) y -> (f,CncCat (Yes (ty2ty ty)) (Yes (t2t x)) y) + _ -> (f,j) + where + t2t = term2term cg pv + ty2ty = type2type cg pv + pv@(labels,untyps,typs) = trs $ paramValues cg + + -- flatten record arguments of param constructors + p2p (f,j) = case j of + ResParam (Yes (ps,v)) -> + (f,ResParam (Yes ([(c,concatMap unRec cont) | (c,cont) <- ps],Nothing))) + _ -> (f,j) + unRec (x,ty) = case ty of + RecType fs -> [ity | (_,typ) <- fs, ity <- unRec (identW,typ)] + _ -> [(x,ty)] + +---- + trs v = traceD (tr v) v + + tr (labels,untyps,typs) = + ("LABELS:" ++++ + unlines [A.prt c ++ "." ++ unwords (map A.prt l) +++ "=" +++ show i | + ((c,l),i) <- Map.toList labels]) ++++ + ("UNTYPS:" ++++ unlines [A.prt t +++ "=" +++ show i | + (t,i) <- Map.toList untyps]) ++++ + ("TYPS:" ++++ unlines [A.prt t +++ "=" +++ show (Map.assocs i) | + (t,i) <- Map.toList typs]) +---- + +purgeGrammar :: Ident -> SourceGrammar -> SourceGrammar +purgeGrammar abstr gr = + (M.MGrammar . list . map unopt . filter complete . purge . M.modules) gr + where + list ms = traceD ("MODULES" +++ unwords (map (prt . fst) ms)) ms + purge = nubBy (\x y -> fst x == fst y) . filter (flip elem needed . fst) + needed = nub $ concatMap (requiredCanModules isSingle gr) acncs + acncs = abstr : M.allConcretes gr abstr + isSingle = True + complete (i,M.ModMod m) = M.isCompleteModule m --- not . isIncompleteCanon + unopt = unshareModule gr -- subexp elim undone when compiled + +type ParamEnv = + (Map.Map (Ident,[Label]) (Type,Integer), -- numbered labels + Map.Map Term Integer, -- untyped terms to values + Map.Map Type (Map.Map Term Integer)) -- types to their terms to values + +--- gathers those param types that are actually used in lincats and lin terms +paramValues :: SourceGrammar -> ParamEnv +paramValues cgr = (labels,untyps,typs) where + partyps = nub $ + --- [App (Q (IC "Predef") (IC "Ints")) (EInt i) | i <- [1,9]] ---linTypeInt + [ty | + (_,(_,CncCat (Yes ty0) _ _)) <- jments, + ty <- typsFrom ty0 + ] ++ [ + Q m ty | + (m,(ty,ResParam _)) <- jments + ] ++ [ty | + (_,(_,CncFun _ (Yes tr) _)) <- jments, + ty <- err (const []) snd $ appSTM (typsFromTrm tr) [] + ] + params = [(ty, errVal (traceD ("UNKNOWN PARAM TYPE" +++ show ty) []) $ + Look.allParamValues cgr ty) | ty <- partyps] + typsFrom ty = unlockTy ty : case ty of + Table p t -> typsFrom p ++ typsFrom t + RecType ls -> concat [typsFrom t | (_, t) <- ls] + _ -> [] + + typsFromTrm :: Term -> STM [Type] Term + typsFromTrm tr = case tr of + R fs -> mapM_ (typsFromField . snd) fs >> return tr + where + typsFromField (mty, t) = case mty of + Just x -> updateSTM (x:) >> typsFromTrm t + _ -> typsFromTrm t + V ty ts -> updateSTM (ty:) >> mapM_ typsFromTrm ts >> return tr + T (TTyped ty) cs -> + updateSTM (ty:) >> mapM_ typsFromTrm [t | (_, t) <- cs] >> return tr + T (TComp ty) cs -> + updateSTM (ty:) >> mapM_ typsFromTrm [t | (_, t) <- cs] >> return tr + _ -> GM.composOp typsFromTrm tr + + jments = + [(m,j) | (m,mo) <- M.allModMod cgr, j <- tree2list $ M.jments mo] + typs = + Map.fromList [(ci,Map.fromList (zip vs [0..])) | (ci,vs) <- params] + untyps = + Map.fromList $ concatMap Map.toList [typ | (_,typ) <- Map.toList typs] + lincats = + [(cat,[f | let RecType fs = GM.defLinType, f <- fs]) | cat <- [cInt,cFloat, cString]] ++ + reverse ---- TODO: really those lincats that are reached + ---- reverse is enough to expel overshadowed ones... + [(cat,ls) | (_,(cat,CncCat (Yes ty) _ _)) <- jments, + RecType ls <- [unlockTy ty]] + labels = Map.fromList $ concat + [((cat,[lab]),(typ,i)): + [((cat,[LVar v]),(typ,toInteger (mx + v))) | v <- [0,1]] ++ ---- 1 or 2 vars + [((cat,[lab,lab2]),(ty,j)) | + rs <- getRec typ, ((lab2, ty),j) <- zip rs [0..]] + | + (cat,ls) <- lincats, ((lab, typ),i) <- zip ls [0..], let mx = length ls] + -- go to tables recursively + ---- TODO: even go to deeper records + where + getRec typ = case typ of + RecType rs -> [rs] ---- [unlockTyp rs] -- (sort (unlockTyp ls)) + Table _ t -> getRec t + _ -> [] + +type2type :: SourceGrammar -> ParamEnv -> Type -> Type +type2type cgr env@(labels,untyps,typs) ty = case ty of + RecType rs -> + RecType [(mkLab i, t2t t) | (i,(l, t)) <- zip [0..] (unlockTyp rs)] + Table pt vt -> Table (t2t pt) (t2t vt) + QC _ _ -> look ty + _ -> ty + where + t2t = type2type cgr env + look ty = EInt $ (+ (-1)) $ toInteger $ case Map.lookup ty typs of + Just vs -> length $ Map.assocs vs + _ -> trace ("unknown partype " ++ show ty) 66669 + +term2term :: SourceGrammar -> ParamEnv -> Term -> Term +term2term cgr env@(labels,untyps,typs) tr = case tr of + App _ _ -> mkValCase (unrec tr) + QC _ _ -> mkValCase tr + R rs -> R [(mkLab i, (Nothing, t2t t)) | + (i,(l,(_,t))) <- zip [0..] (GM.sortRec (unlock rs))] + P t l -> r2r tr + PI t l i -> EInt $ toInteger i + + T (TWild _) _ -> error $ "wild" +++ prt tr + T (TComp ty) cs -> t2t $ V ty $ map snd cs ---- should be elim'ed in tc + T (TTyped ty) cs -> t2t $ V ty $ map snd cs ---- should be elim'ed in tc + V ty ts -> mkCurry $ V ty [t2t t | t <- ts] + S t p -> mkCurrySel (t2t t) (t2t p) + + _ -> GM.composSafeOp t2t tr + where + t2t = term2term cgr env + + unrec t = case t of + App f (R fs) -> GM.mkApp (unrec f) [unrec u | (_,(_,u)) <- fs] + _ -> GM.composSafeOp unrec t + + mkValCase tr = case appSTM (doVar tr) [] of + Ok (tr', st@(_:_)) -> t2t $ comp $ foldr mkCase tr' st + _ -> valNum $ comp tr + + --- this is mainly needed for parameter record projections + ---- was: + comp t = errVal t $ Compute.computeConcreteRec cgr t + compt t = case t of + T (TComp typ) ts -> comp $ V typ (map (comp . snd) ts) ---- should... + T (TTyped typ) ts -> comp $ V typ (map (comp . snd) ts) ---- should + V typ ts -> V typ (map comp ts) + S tb (FV ts) -> FV $ map (comp . S tb) ts + S tb@(V typ ts) v0 -> err error id $ do + let v = comp v0 + let mv1 = Map.lookup v untyps + case mv1 of + Just v1 -> return $ (comp . (ts !!) . fromInteger) v1 + _ -> return (S (comp tb) v) + + R r -> R [(l,(ty,comp t)) | (l,(ty,t)) <- r] + P (R r) l -> maybe t (comp . snd) $ lookup l r + _ -> GM.composSafeOp comp t + + doVar :: Term -> STM [((Type,[Term]),(Term,Term))] Term + doVar tr = case getLab tr of + Ok (cat, lab) -> do + k <- readSTM >>= return . length + let tr' = Vr $ identC $ (BS.pack (show k)) ----- + + let tyvs = case Map.lookup (cat,lab) labels of + Just (ty,_) -> case Map.lookup ty typs of + Just vs -> (ty,[t | + (t,_) <- sortBy (\x y -> compare (snd x) (snd y)) + (Map.assocs vs)]) + _ -> error $ "doVar1" +++ A.prt ty + _ -> error $ "doVar2" +++ A.prt tr +++ show (cat,lab) ---- debug + updateSTM ((tyvs, (tr', tr)):) + return tr' + _ -> GM.composOp doVar tr + + r2r tr@(P (S (V ty ts) v) l) = t2t $ S (V ty [comp (P t l) | t <- ts]) v + + r2r tr@(P p _) = case getLab tr of + Ok (cat,labs) -> P (t2t p) . mkLab $ + maybe (prtTrace tr $ 66664) snd $ + Map.lookup (cat,labs) labels + _ -> K ((A.prt tr +++ prtTrace tr "66665")) + + -- this goes recursively into tables (ignored) and records (accumulated) + getLab tr = case tr of + Vr (IA cat _) -> return (identC cat,[]) + Vr (IAV cat _ _) -> return (identC cat,[]) + Vr (IC s) -> return (identC cat,[]) where + cat = BS.takeWhile (/='_') s ---- also to match IAVs; no _ in a cat tolerated + ---- init (reverse (dropWhile (/='_') (reverse s))) ---- from gf parser +---- Vr _ -> error $ "getLab " ++ show tr + P p lab2 -> do + (cat,labs) <- getLab p + return (cat,labs++[lab2]) + S p _ -> getLab p + _ -> Bad "getLab" + + + mkCase ((ty,vs),(x,p)) tr = + S (V ty [mkBranch x v tr | v <- vs]) p + mkBranch x t tr = case tr of + _ | tr == x -> t + _ -> GM.composSafeOp (mkBranch x t) tr + + valNum tr = maybe (valNumFV $ tryFV tr) EInt $ Map.lookup tr untyps + where + tryFV tr = case GM.appForm tr of + (c@(QC _ _), ts) -> [GM.mkApp c ts' | ts' <- combinations (map tryFV ts)] + (FV ts,_) -> ts + _ -> [tr] + valNumFV ts = case ts of + [tr] -> error ("valNum" +++ prt tr) ----- prtTrace tr $ K "66667" + _ -> FV $ map valNum ts + + mkCurry trm = case trm of + V (RecType [(_,ty)]) ts -> V ty ts + V (RecType ((_,ty):ltys)) ts -> + V ty [mkCurry (V (RecType ltys) cs) | + cs <- chop (product (map (lengthtyp . snd) ltys)) ts] + _ -> trm + lengthtyp ty = case Map.lookup ty typs of + Just m -> length (Map.assocs m) + _ -> error $ "length of type " ++ show ty + chop i xs = case splitAt i xs of + (xs1,[]) -> [xs1] + (xs1,xs2) -> xs1:chop i xs2 + + + mkCurrySel t p = S t p -- done properly in CheckGFCC + + +mkLab k = LIdent (BS.pack ("_" ++ show k)) + +-- remove lock fields; in fact, any empty records and record types +unlock = filter notlock where + notlock (l,(_, t)) = case t of --- need not look at l + R [] -> False + RecType [] -> False + _ -> True + +unlockTyp = filter notlock + +notlock (l, t) = case t of --- need not look at l + RecType [] -> False + _ -> True + +unlockTy ty = case ty of + RecType ls -> RecType $ GM.sortRec [(l, unlockTy t) | (l,t) <- ls, notlock (l,t)] + _ -> GM.composSafeOp unlockTy ty + + +prtTrace tr n = + trace ("-- INTERNAL COMPILER ERROR" +++ A.prt tr ++++ show n) n +prTrace tr n = trace ("-- OBSERVE" +++ A.prt tr +++ show n +++ show tr) n + + +-- | this function finds out what modules are really needed in the canonical gr. +-- its argument is typically a concrete module name +requiredCanModules :: (Ord i, Show i) => Bool -> M.MGrammar i a -> i -> [i] +requiredCanModules isSingle gr c = nub $ filter notReuse ops ++ exts where + exts = M.allExtends gr c + ops = if isSingle + then map fst (M.modules gr) + else iterFix (concatMap more) $ exts + more i = errVal [] $ do + m <- M.lookupModMod gr i + return $ M.extends m ++ [o | o <- map M.openedModule (M.opens m)] + notReuse i = errVal True $ do + m <- M.lookupModMod gr i + return $ M.isModRes m -- to exclude reused Cnc and Abs from required diff --git a/src/GF/Compile/ModDeps.hs b/src/GF/Compile/ModDeps.hs new file mode 100644 index 000000000..b5b1b798c --- /dev/null +++ b/src/GF/Compile/ModDeps.hs @@ -0,0 +1,153 @@ +---------------------------------------------------------------------- +-- | +-- Module : ModDeps +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/11/11 23:24:34 $ +-- > CVS $Author: aarne $ +-- > CVS $Revision: 1.14 $ +-- +-- Check correctness of module dependencies. Incomplete. +-- +-- AR 13\/5\/2003 +----------------------------------------------------------------------------- + +module GF.Compile.ModDeps (mkSourceGrammar, + moduleDeps, + openInterfaces, + requiredCanModules + ) where + +import GF.Grammar.Grammar +import GF.Infra.Ident +import GF.Infra.Option +import GF.Grammar.PrGrammar +import GF.Compile.Update +import GF.Grammar.Lookup +import GF.Infra.Modules + +import GF.Data.Operations + +import Control.Monad +import Data.List + +-- | to check uniqueness of module names and import names, the +-- appropriateness of import and extend types, +-- to build a dependency graph of modules, and to sort them topologically +mkSourceGrammar :: [(Ident,SourceModInfo)] -> Err SourceGrammar +mkSourceGrammar ms = do + let ns = map fst ms + checkUniqueErr ns + mapM (checkUniqueImportNames ns . snd) ms + deps <- moduleDeps ms + deplist <- either + return + (\ms -> Bad $ "circular modules" +++ unwords (map show ms)) $ + topoTest deps + return $ MGrammar [(m, maybe undefined id $ lookup m ms) | IdentM m _ <- deplist] + +checkUniqueErr :: (Show i, Eq i) => [i] -> Err () +checkUniqueErr ms = do + let msg = checkUnique ms + if null msg then return () else Bad $ unlines msg + +-- | check that import names don't clash with module names +checkUniqueImportNames :: [Ident] -> SourceModInfo -> Err () +checkUniqueImportNames ns mo = case mo of + ModMod m -> test [n | OQualif _ n v <- opens m, n /= v] + _ -> return () --- Bad $ "bug: ModDeps does not treat" +++ show mo + where + + test ms = testErr (all (`notElem` ns) ms) + ("import names clashing with module names among" +++ + unwords (map prt ms)) + +type Dependencies = [(IdentM Ident,[IdentM Ident])] + +-- | to decide what modules immediately depend on what, and check if the +-- dependencies are appropriate +moduleDeps :: [(Ident,SourceModInfo)] -> Err Dependencies +moduleDeps ms = mapM deps ms where + deps (c,mi) = errIn ("checking dependencies of module" +++ prt c) $ case mi of + ModMod m -> case mtype m of + MTConcrete a -> do + aty <- lookupModuleType gr a + testErr (aty == MTAbstract) "the of-module is not an abstract syntax" + chDep (IdentM c (MTConcrete a)) + (extends m) (MTConcrete a) (opens m) MTResource + t -> chDep (IdentM c t) (extends m) t (opens m) t + + chDep it es ety os oty = do + ests <- mapM (lookupModuleType gr) es + testErr (all (compatMType ety) ests) "inappropriate extension module type" +---- osts <- mapM (lookupModuleType gr . openedModule) os +---- testErr (all (compatOType oty) osts) "inappropriate open module type" + let ab = case it of + IdentM _ (MTConcrete a) -> [IdentM a MTAbstract] + _ -> [] ---- + return (it, ab ++ + [IdentM e ety | e <- es] ++ + [IdentM (openedModule o) oty | o <- os]) + + -- check for superficial compatibility, not submodule relation etc: what can be extended + compatMType mt0 mt = case (mt0,mt) of + (MTResource, MTConcrete _) -> True + (MTInstance _, MTConcrete _) -> True + (MTInterface, MTAbstract) -> True + (MTConcrete _, MTConcrete _) -> True + (MTInstance _, MTInstance _) -> True + (MTReuse _, MTReuse _) -> True + (MTInstance _, MTResource) -> True + (MTResource, MTInstance _) -> True + ---- some more? + _ -> mt0 == mt + -- in the same way; this defines what can be opened + compatOType mt0 mt = case mt0 of + MTAbstract -> mt == MTAbstract + MTTransfer _ _ -> mt == MTAbstract + _ -> case mt of + MTResource -> True + MTReuse _ -> True + MTInterface -> True + MTInstance _ -> True + _ -> False + + gr = MGrammar ms --- hack + +openInterfaces :: Dependencies -> Ident -> Err [Ident] +openInterfaces ds m = do + let deps = [(i,ds) | (IdentM i _,ds) <- ds] + let more (c,_) = [(i,mt) | Just is <- [lookup c deps], IdentM i mt <- is] + let mods = iterFix (concatMap more) (more (m,undefined)) + return $ [i | (i,MTInterface) <- mods] + +-- | this function finds out what modules are really needed in the canonical gr. +-- its argument is typically a concrete module name +requiredCanModules :: (Ord i, Show i) => Bool -> MGrammar i a -> i -> [i] +requiredCanModules isSingle gr c = nub $ filter notReuse ops ++ exts where + exts = allExtends gr c + ops = if isSingle + then map fst (modules gr) + else iterFix (concatMap more) $ exts + more i = errVal [] $ do + m <- lookupModMod gr i + return $ extends m ++ [o | o <- map openedModule (opens m)] + notReuse i = errVal True $ do + m <- lookupModMod gr i + return $ isModRes m -- to exclude reused Cnc and Abs from required + + +{- +-- to test +exampleDeps = [ + (ir "Nat",[ii "Gen", ir "Adj"]), + (ir "Adj",[ii "Num", ii "Gen", ir "Nou"]), + (ir "Nou",[ii "Cas"]) + ] + +ii s = IdentM (IC s) MTInterface +ir s = IdentM (IC s) MTResource +-} + diff --git a/src/GF/Compile/Optimize.hs b/src/GF/Compile/Optimize.hs new file mode 100644 index 000000000..83cbeb57a --- /dev/null +++ b/src/GF/Compile/Optimize.hs @@ -0,0 +1,235 @@ +{-# LANGUAGE PatternGuards #-} +---------------------------------------------------------------------- +-- | +-- Module : Optimize +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/09/16 13:56:13 $ +-- > CVS $Author: aarne $ +-- > CVS $Revision: 1.18 $ +-- +-- Top-level partial evaluation for GF source modules. +----------------------------------------------------------------------------- + +module GF.Compile.Optimize (optimizeModule) where + +import GF.Grammar.Grammar +import GF.Infra.Ident +import GF.Infra.Modules +import GF.Grammar.PrGrammar +import GF.Grammar.Macros +import GF.Grammar.Lookup +import GF.Grammar.Predef +import GF.Compile.Refresh +import GF.Compile.Compute +import GF.Compile.BackOpt +import GF.Compile.CheckGrammar +import GF.Compile.Update + +import GF.Data.Operations +import GF.Infra.CheckM +import GF.Infra.Option + +import Control.Monad +import Data.List +import qualified Data.Set as Set + +import Debug.Trace + + +-- conditional trace + +prtIf :: (Print a) => Bool -> a -> a +prtIf b t = if b then trace (" " ++ prt t) t else t + +-- | partial evaluation of concrete syntax. AR 6\/2001 -- 16\/5\/2003 -- 5\/2\/2005. + +type EEnv = () --- not used + +-- only do this for resource: concrete is optimized in gfc form +optimizeModule :: Options -> ([(Ident,SourceModInfo)],EEnv) -> + (Ident,SourceModInfo) -> Err ((Ident,SourceModInfo),EEnv) +optimizeModule opts mse@(ms,eenv) mo@(_,mi) = case mi of + ModMod m0 | mstatus m0 == MSComplete && isModRes m0 -> do + (mo1,_) <- evalModule oopts mse mo + let mo2 = shareModule optim mo1 + return (mo2,eenv) + _ -> evalModule oopts mse mo + where + oopts = addOptions opts (moduleOptions (flagsModule mo)) + optim = moduleFlag optOptimizations oopts + +evalModule :: Options -> ([(Ident,SourceModInfo)],EEnv) -> (Ident,SourceModInfo) -> + Err ((Ident,SourceModInfo),EEnv) +evalModule oopts (ms,eenv) mo@(name,mod) = case mod of + + ModMod m0 | mstatus m0 == MSComplete -> case mtype m0 of + _ | isModRes m0 -> do + let deps = allOperDependencies name (jments m0) + ids <- topoSortOpers deps + MGrammar (mod' : _) <- foldM evalOp gr ids + return $ (mod',eenv) + + MTConcrete a -> do + js' <- mapMTree (evalCncInfo oopts gr name a) (jments m0) + return $ ((name, ModMod (replaceJudgements m0 js')),eenv) + + _ -> return $ ((name,mod),eenv) + _ -> return $ ((name,mod),eenv) + where + gr0 = MGrammar $ ms + gr = MGrammar $ (name,mod) : ms + + evalOp g@(MGrammar ((_, ModMod m) : _)) i = do + info <- lookupTree prt i $ jments m + info' <- evalResInfo oopts gr (i,info) + return $ updateRes g name i info' + +-- | only operations need be compiled in a resource, and this is local to each +-- definition since the module is traversed in topological order +evalResInfo :: Options -> SourceGrammar -> (Ident,Info) -> Err Info +evalResInfo oopts gr (c,info) = case info of + + ResOper pty pde -> eIn "operation" $ do + pde' <- case pde of + Yes de | optres -> liftM yes $ comp de + _ -> return pde + return $ ResOper pty pde' + + _ -> return info + where + comp = if optres then computeConcrete gr else computeConcreteRec gr + eIn cat = errIn ("Error optimizing" +++ cat +++ prt c +++ ":") + optim = moduleFlag optOptimizations oopts + optres = OptExpand `Set.member` optim + + +evalCncInfo :: + Options -> SourceGrammar -> Ident -> Ident -> (Ident,Info) -> Err (Ident,Info) +evalCncInfo opts gr cnc abs (c,info) = do + + seq (prtIf (verbAtLeast opts Verbose) c) $ return () + + errIn ("optimizing" +++ prt c) $ case info of + + CncCat ptyp pde ppr -> do + pde' <- case (ptyp,pde) of + (Yes typ, Yes de) -> + liftM yes $ pEval ([(varStr, typeStr)], typ) de + (Yes typ, Nope) -> + liftM yes $ mkLinDefault gr typ >>= partEval noOptions gr ([(varStr, typeStr)],typ) + (May b, Nope) -> + return $ May b + _ -> return pde -- indirection + + ppr' <- liftM yes $ evalPrintname gr c ppr (yes $ K $ prt c) + + return (c, CncCat ptyp pde' ppr') + + CncFun (mt@(Just (_,ty@(cont,val)))) pde ppr -> --trace (prt c) $ + eIn ("linearization in type" +++ prt (mkProd (cont,val,[])) ++++ "of function") $ do + pde' <- case pde of + Yes de -> do + liftM yes $ pEval ty de + + _ -> return pde + ppr' <- liftM yes $ evalPrintname gr c ppr pde' + return $ (c, CncFun mt pde' ppr') -- only cat in type actually needed + + _ -> return (c,info) + where + pEval = partEval opts gr + eIn cat = errIn ("Error optimizing" +++ cat +++ prt c +++ ":") + +-- | the main function for compiling linearizations +partEval :: Options -> SourceGrammar -> (Context,Type) -> Term -> Err Term +partEval opts gr (context, val) trm = errIn ("parteval" +++ prt_ trm) $ do + let vars = map fst context + args = map Vr vars + subst = [(v, Vr v) | v <- vars] + trm1 = mkApp trm args + trm2 <- computeTerm gr subst trm1 + trm3 <- if rightType trm2 + then computeTerm gr subst trm2 + else recordExpand val trm2 >>= computeTerm gr subst + return $ mkAbs vars trm3 + where + -- don't eta expand records of right length (correct by type checking) + rightType (R rs) = case val of + RecType ts -> length rs == length ts + _ -> False + rightType _ = False + + + + +-- here we must be careful not to reduce +-- variants {{s = "Auto" ; g = N} ; {s = "Wagen" ; g = M}} +-- {s = variants {"Auto" ; "Wagen"} ; g = variants {N ; M}} ; + +recordExpand :: Type -> Term -> Err Term +recordExpand typ trm = case unComputed typ of + RecType tys -> case trm of + FV rs -> return $ FV [R [assign lab (P r lab) | (lab,_) <- tys] | r <- rs] + _ -> return $ R [assign lab (P trm lab) | (lab,_) <- tys] + _ -> return trm + + +-- | auxiliaries for compiling the resource + +mkLinDefault :: SourceGrammar -> Type -> Err Term +mkLinDefault gr typ = do + case unComputed typ of + RecType lts -> mapPairsM mkDefField lts >>= (return . Abs varStr . R . mkAssign) + _ -> liftM (Abs varStr) $ mkDefField typ +---- _ -> prtBad "linearization type must be a record type, not" typ + where + mkDefField typ = case unComputed typ of + Table p t -> do + t' <- mkDefField t + let T _ cs = mkWildCases t' + return $ T (TWild p) cs + Sort s | s == cStr -> return $ Vr varStr + QC q p -> lookupFirstTag gr q p + RecType r -> do + let (ls,ts) = unzip r + ts' <- mapM mkDefField ts + return $ R $ [assign l t | (l,t) <- zip ls ts'] + _ | Just _ <- isTypeInts typ -> return $ EInt 0 -- exists in all as first val + _ -> prtBad "linearization type field cannot be" typ + +-- | Form the printname: if given, compute. If not, use the computed +-- lin for functions, cat name for cats (dispatch made in evalCncDef above). +--- We cannot use linearization at this stage, since we do not know the +--- defaults we would need for question marks - and we're not yet in canon. +evalPrintname :: SourceGrammar -> Ident -> MPr -> Perh Term -> Err Term +evalPrintname gr c ppr lin = + case ppr of + Yes pr -> comp pr + _ -> case lin of + Yes t -> return $ K $ clean $ prt $ oneBranch t ---- stringFromTerm + _ -> return $ K $ prt c ---- + where + comp = computeConcrete gr + + oneBranch t = case t of + Abs _ b -> oneBranch b + R (r:_) -> oneBranch $ snd $ snd r + T _ (c:_) -> oneBranch $ snd c + V _ (c:_) -> oneBranch c + FV (t:_) -> oneBranch t + C x y -> C (oneBranch x) (oneBranch y) + S x _ -> oneBranch x + P x _ -> oneBranch x + Alts (d,_) -> oneBranch d + _ -> t + + --- very unclean cleaner + clean s = case s of + '+':'+':' ':cs -> clean cs + '"':cs -> clean cs + c:cs -> c: clean cs + _ -> s + diff --git a/src/GF/Compile/OptimizeGF.hs b/src/GF/Compile/OptimizeGF.hs new file mode 100644 index 000000000..41b828aa3 --- /dev/null +++ b/src/GF/Compile/OptimizeGF.hs @@ -0,0 +1,277 @@ +---------------------------------------------------------------------- +-- | +-- Module : OptimizeGF +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/04/21 16:21:33 $ +-- > CVS $Author: bringert $ +-- > CVS $Revision: 1.6 $ +-- +-- Optimizations on GF source code: sharing, parametrization, value sets. +-- +-- optimization: sharing branches in tables. AR 25\/4\/2003. +-- following advice of Josef Svenningsson +----------------------------------------------------------------------------- + +module GF.Compile.OptimizeGF ( + optModule,unshareModule,unsubexpModule,unoptModule,subexpModule,shareModule + ) where + +import GF.Grammar.Grammar +import GF.Grammar.Lookup +import GF.Infra.Ident +import qualified GF.Grammar.Macros as C +import GF.Grammar.PrGrammar (prt) +import qualified GF.Infra.Modules as M +import GF.Data.Operations + +import Control.Monad +import Data.Map (Map) +import qualified Data.Map as Map +import qualified Data.ByteString.Char8 as BS +import Data.List + +optModule :: (Ident, SourceModInfo) -> (Ident, SourceModInfo) +optModule = subexpModule . shareModule + +shareModule = processModule optim + +unoptModule :: SourceGrammar -> (Ident, SourceModInfo) -> (Ident, SourceModInfo) +unoptModule gr = unshareModule gr . unsubexpModule + +unshareModule :: SourceGrammar -> (Ident, SourceModInfo) -> (Ident, SourceModInfo) +unshareModule gr = processModule (const (unoptim gr)) + +processModule :: + (Ident -> Term -> Term) -> (Ident, SourceModInfo) -> (Ident, SourceModInfo) +processModule opt (i,m) = case m of + M.ModMod mo -> + (i,M.ModMod (M.replaceJudgements mo (mapTree (shareInfo opt) (M.jments mo)))) + _ -> (i,m) + +shareInfo opt (c, CncCat ty (Yes t) m) = (c,CncCat ty (Yes (opt c t)) m) +shareInfo opt (c, CncFun kxs (Yes t) m) = (c,CncFun kxs (Yes (opt c t)) m) +shareInfo opt (c, ResOper ty (Yes t)) = (c,ResOper ty (Yes (opt c t))) +shareInfo _ i = i + +-- the function putting together optimizations +optim :: Ident -> Term -> Term +optim c = values . factor c 0 + +-- we need no counter to create new variable names, since variables are +-- local to tables (only true in GFC) --- + +-- factor parametric branches + +factor :: Ident -> Int -> Term -> Term +factor c i t = case t of + T _ [_] -> t + T _ [] -> t + T (TComp ty) cs -> + T (TTyped ty) $ factors i [(p, factor c (i+1) v) | (p, v) <- cs] + _ -> C.composSafeOp (factor c i) t + where + + factors i psvs = -- we know psvs has at least 2 elements + let p = qqIdent c i + vs' = map (mkFun p) psvs + in if allEqs vs' + then mkCase p vs' + else psvs + + mkFun p (patt, val) = replace (C.patt2term patt) (Vr p) val + + allEqs (v:vs) = all (==v) vs + + mkCase p (v:_) = [(PV p, v)] + +--- we hope this will be fresh and don't check... in GFC would be safe + +qqIdent c i = identC (BS.pack ("q_" ++ prt c ++ "__" ++ show i)) + + +-- we need to replace subterms + +replace :: Term -> Term -> Term -> Term +replace old new trm = case trm of + + -- these are the important cases, since they can correspond to patterns + QC _ _ | trm == old -> new + App t ts | trm == old -> new + App t ts -> App (repl t) (repl ts) + R _ | isRec && trm == old -> new + _ -> C.composSafeOp repl trm + where + repl = replace old new + isRec = case trm of + R _ -> True + _ -> False + +-- It is very important that this is performed only after case +-- expansion since otherwise the order and number of values can +-- be incorrect. Guaranteed by the TComp flag. + +values :: Term -> Term +values t = case t of + T ty [(ps,t)] -> T ty [(ps,values t)] -- don't destroy parametrization + T (TComp ty) cs -> V ty [values t | (_, t) <- cs] + T (TTyped ty) cs -> V ty [values t | (_, t) <- cs] + ---- why are these left? + ---- printing with GrammarToSource does not preserve the distinction + _ -> C.composSafeOp values t + + +-- to undo the effect of factorization + +unoptim :: SourceGrammar -> Term -> Term +unoptim gr = unfactor gr + +unfactor :: SourceGrammar -> Term -> Term +unfactor gr t = case t of + T (TTyped ty) [(PV x,u)] -> V ty [restore x v (unfac u) | v <- vals ty] + _ -> C.composSafeOp unfac t + where + unfac = unfactor gr + vals = err error id . allParamValues gr + restore x u t = case t of + Vr y | y == x -> u + _ -> C.composSafeOp (restore x u) t + + +---------------------------------------------------------------------- + +{- +This module implements a simple common subexpression elimination + for gfc grammars, to factor out shared subterms in lin rules. +It works in three phases: + + (1) collectSubterms collects recursively all subterms of forms table and (P x..y) + from lin definitions (experience shows that only these forms + tend to get shared) and counts how many times they occur + (2) addSubexpConsts takes those subterms t that occur more than once + and creates definitions of form "oper A''n = t" where n is a + fresh number; notice that we assume no ids of this form are in + scope otherwise + (3) elimSubtermsMod goes through lins and the created opers by replacing largest + possible subterms by the newly created identifiers + +The optimization is invoked in gf by the flag i -subs. + +If an application does not support GFC opers, the effect of this +optimization can be undone by the function unSubelimCanon. + +The function unSubelimCanon can be used to diagnostisize how much +cse is possible in the grammar. It is used by the flag pg -printer=subs. + +-} + +subexpModule :: SourceModule -> SourceModule +subexpModule (n,m) = errVal (n,m) $ case m of + M.ModMod mo -> do + let ljs = tree2list (M.jments mo) + (tree,_) <- appSTM (getSubtermsMod n ljs) (Map.empty,0) + js2 <- liftM buildTree $ addSubexpConsts n tree $ ljs + return (n,M.ModMod (M.replaceJudgements mo js2)) + _ -> return (n,m) + +unsubexpModule :: SourceModule -> SourceModule +unsubexpModule sm@(i,m) = case m of + M.ModMod mo | hasSub ljs -> + (i, M.ModMod (M.replaceJudgements mo + (rebuild (map unparInfo ljs)))) + where ljs = tree2list (M.jments mo) + _ -> (i,m) + where + -- perform this iff the module has opers + hasSub ljs = not $ null [c | (c,ResOper _ _) <- ljs] + unparInfo (c,info) = case info of + CncFun xs (Yes t) m -> [(c, CncFun xs (Yes (unparTerm t)) m)] + ResOper (Yes (EInt 8)) _ -> [] -- subexp-generated opers + ResOper pty (Yes t) -> [(c, ResOper pty (Yes (unparTerm t)))] + _ -> [(c,info)] + unparTerm t = case t of + Q m c | isOperIdent c -> --- name convention of subexp opers + errVal t $ liftM unparTerm $ lookupResDef gr m c + _ -> C.composSafeOp unparTerm t + gr = M.MGrammar [sm] + rebuild = buildTree . concat + +-- implementation + +type TermList = Map Term (Int,Int) -- number of occs, id +type TermM a = STM (TermList,Int) a + +addSubexpConsts :: + Ident -> Map Term (Int,Int) -> [(Ident,Info)] -> Err [(Ident,Info)] +addSubexpConsts mo tree lins = do + let opers = [oper id trm | (trm,(_,id)) <- list] + mapM mkOne $ opers ++ lins + where + + mkOne (f,def) = case def of + CncFun xs (Yes trm) pn -> do + trm' <- recomp f trm + return (f,CncFun xs (Yes trm') pn) + ResOper ty (Yes trm) -> do + trm' <- recomp f trm + return (f,ResOper ty (Yes trm')) + _ -> return (f,def) + recomp f t = case Map.lookup t tree of + Just (_,id) | operIdent id /= f -> return $ Q mo (operIdent id) + _ -> C.composOp (recomp f) t + + list = Map.toList tree + + oper id trm = (operIdent id, ResOper (Yes (EInt 8)) (Yes trm)) + --- impossible type encoding generated opers + +getSubtermsMod :: Ident -> [(Ident,Info)] -> TermM (Map Term (Int,Int)) +getSubtermsMod mo js = do + mapM (getInfo (collectSubterms mo)) js + (tree0,_) <- readSTM + return $ Map.filter (\ (nu,_) -> nu > 1) tree0 + where + getInfo get fi@(f,i) = case i of + CncFun xs (Yes trm) pn -> do + get trm + return $ fi + ResOper ty (Yes trm) -> do + get trm + return $ fi + _ -> return fi + +collectSubterms :: Ident -> Term -> TermM Term +collectSubterms mo t = case t of + App f a -> do + collect f + collect a + add t + T ty cs -> do + let (_,ts) = unzip cs + mapM collect ts + add t + V ty ts -> do + mapM collect ts + add t +---- K (KP _ _) -> add t + _ -> C.composOp (collectSubterms mo) t + where + collect = collectSubterms mo + add t = do + (ts,i) <- readSTM + let + ((count,id),next) = case Map.lookup t ts of + Just (nu,id) -> ((nu+1,id), i) + _ -> ((1, i ), i+1) + writeSTM (Map.insert t (count,id) ts, next) + return t --- only because of composOp + +operIdent :: Int -> Ident +operIdent i = identC (operPrefix `BS.append` (BS.pack (show i))) --- + +isOperIdent :: Ident -> Bool +isOperIdent id = BS.isPrefixOf operPrefix (ident2bs id) + +operPrefix = BS.pack ("A''") diff --git a/src/GF/Compile/OptimizeGFCC.hs b/src/GF/Compile/OptimizeGFCC.hs new file mode 100644 index 000000000..c73d5bbcb --- /dev/null +++ b/src/GF/Compile/OptimizeGFCC.hs @@ -0,0 +1,124 @@ +module GF.Compile.OptimizeGFCC where + +import PGF.CId +import PGF.Data + +import GF.Data.Operations + +import Data.List +import qualified Data.Map as Map + + +-- back-end optimization: +-- suffix analysis followed by common subexpression elimination + +optPGF :: PGF -> PGF +optPGF = cseOptimize . suffixOptimize + +suffixOptimize :: PGF -> PGF +suffixOptimize pgf = pgf { + concretes = Map.map opt (concretes pgf) + } + where + opt cnc = cnc { + lins = Map.map optTerm (lins cnc), + lindefs = Map.map optTerm (lindefs cnc), + printnames = Map.map optTerm (printnames cnc) + } + +cseOptimize :: PGF -> PGF +cseOptimize pgf = pgf { + concretes = Map.map subex (concretes pgf) + } + +-- analyse word form lists into prefix + suffixes +-- suffix sets can later be shared by subex elim + +optTerm :: Term -> Term +optTerm tr = case tr of + R ts@(_:_:_) | all isK ts -> mkSuff $ optToks [s | K (KS s) <- ts] + R ts -> R $ map optTerm ts + P t v -> P (optTerm t) v + _ -> tr + where + optToks ss = prf : suffs where + prf = pref (head ss) (tail ss) + suffs = map (drop (length prf)) ss + pref cand ss = case ss of + s1:ss2 -> if isPrefixOf cand s1 then pref cand ss2 else pref (init cand) ss + _ -> cand + isK t = case t of + K (KS _) -> True + _ -> False + mkSuff ("":ws) = R (map (K . KS) ws) + mkSuff (p:ws) = W p (R (map (K . KS) ws)) + + +-- common subexpression elimination + +---subex :: [(CId,Term)] -> [(CId,Term)] +subex :: Concr -> Concr +subex cnc = err error id $ do + (tree,_) <- appSTM (getSubtermsMod cnc) (Map.empty,0) + return $ addSubexpConsts tree cnc + +type TermList = Map.Map Term (Int,Int) -- number of occs, id +type TermM a = STM (TermList,Int) a + +addSubexpConsts :: TermList -> Concr -> Concr +addSubexpConsts tree cnc = cnc { + opers = Map.fromList [(f,recomp f trm) | (f,trm) <- ops], + lins = rec lins, + lindefs = rec lindefs, + printnames = rec printnames + } + where + ops = [(fid id, trm) | (trm,(_,id)) <- Map.assocs tree] + mkOne (f,trm) = (f, recomp f trm) + recomp f t = case Map.lookup t tree of + Just (_,id) | fid id /= f -> F $ fid id -- not to replace oper itself + _ -> case t of + R ts -> R $ map (recomp f) ts + S ts -> S $ map (recomp f) ts + W s t -> W s (recomp f t) + P t p -> P (recomp f t) (recomp f p) + _ -> t + fid n = mkCId $ "_" ++ show n + rec field = Map.fromAscList [(f,recomp f trm) | (f,trm) <- Map.assocs (field cnc)] + + +getSubtermsMod :: Concr -> TermM TermList +getSubtermsMod cnc = do + mapM getSubterms (Map.assocs (lins cnc)) + mapM getSubterms (Map.assocs (lindefs cnc)) + mapM getSubterms (Map.assocs (printnames cnc)) + (tree0,_) <- readSTM + return $ Map.filter (\ (nu,_) -> nu > 1) tree0 + where + getSubterms (f,trm) = collectSubterms trm >> return () + +collectSubterms :: Term -> TermM () +collectSubterms t = case t of + R ts -> do + mapM collectSubterms ts + add t + S ts -> do + mapM collectSubterms ts + add t + W s u -> do + collectSubterms u + add t + P p u -> do + collectSubterms p + collectSubterms u + add t + _ -> return () + where + add t = do + (ts,i) <- readSTM + let + ((count,id),next) = case Map.lookup t ts of + Just (nu,id) -> ((nu+1,id), i) + _ -> ((1, i ), i+1) + writeSTM (Map.insert t (count,id) ts, next) + diff --git a/src/GF/Compile/ReadFiles.hs b/src/GF/Compile/ReadFiles.hs new file mode 100644 index 000000000..cd2faec15 --- /dev/null +++ b/src/GF/Compile/ReadFiles.hs @@ -0,0 +1,195 @@ +---------------------------------------------------------------------- +-- | +-- Module : ReadFiles +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/11/11 23:24:34 $ +-- > CVS $Author: aarne $ +-- > CVS $Revision: 1.26 $ +-- +-- Decide what files to read as function of dependencies and time stamps. +-- +-- make analysis for GF grammar modules. AR 11\/6\/2003--24\/2\/2004 +-- +-- to find all files that have to be read, put them in dependency order, and +-- decide which files need recompilation. Name @file.gf@ is returned for them, +-- and @file.gfo@ otherwise. +----------------------------------------------------------------------------- + +module GF.Compile.ReadFiles + ( getAllFiles,ModName,ModEnv,importsOfModule, + gfoFile,gfFile,isGFO, + getOptionsFromFile) where + +import GF.Infra.UseIO +import GF.Infra.Option +import GF.Data.Operations +import GF.Source.AbsGF hiding (FileName) +import GF.Source.LexGF +import GF.Source.ParGF + +import Control.Monad +import Data.Char +import Data.List +import qualified Data.ByteString.Char8 as BS +import qualified Data.Map as Map +import System.Time +import System.Directory +import System.FilePath + +type ModName = String +type ModEnv = Map.Map ModName (ClockTime,[ModName]) + + +-- | Returns a list of all files to be compiled in topological order i.e. +-- the low level (leaf) modules are first. +getAllFiles :: Options -> [InitPath] -> ModEnv -> FileName -> IOE [FullPath] +getAllFiles opts ps env file = do + -- read module headers from all files recursively + ds <- liftM reverse $ get [] [] (justModuleName file) + ioeIO $ putIfVerb opts $ "all modules:" +++ show [name | (name,_,_,_,_) <- ds] + return $ paths ds + where + -- construct list of paths to read + paths cs = [mk (p </> f) | (f,st,_,_,p) <- cs, mk <- mkFile st] + where + mkFile CSComp = [gfFile ] + mkFile CSRead = [gfoFile] + mkFile _ = [] + + -- | traverses the dependency graph and returns a topologicaly sorted + -- list of ModuleInfo. An error is raised if there is circular dependency + get :: [ModName] -- ^ keeps the current path in the dependency graph to avoid cycles + -> [ModuleInfo] -- ^ a list of already traversed modules + -> ModName -- ^ the current module + -> IOE [ModuleInfo] -- ^ the final + get trc ds name + | name `elem` trc = ioeErr $ Bad $ "circular modules" +++ unwords trc + | (not . null) [n | (n,_,_,_,_) <- ds, name == n] --- file already read + = return ds + | otherwise = do + (name,st0,t0,imps,p) <- findModule name + ds <- foldM (get (name:trc)) ds imps + let (st,t) | (not . null) [f | (f,CSComp,_,_,_) <- ds, elem f imps] + = (CSComp,Nothing) + | otherwise = (st0,t0) + return ((name,st,t,imps,p):ds) + + -- searches for module in the search path and if it is found + -- returns 'ModuleInfo'. It fails if there is no such module + findModule :: ModName -> IOE ModuleInfo + findModule name = do + (file,gfTime,gfoTime) <- do + mb_gfFile <- ioeIO $ getFilePathMsg "" ps (gfFile name) + case mb_gfFile of + Just gfFile -> do gfTime <- ioeIO $ getModificationTime gfFile + mb_gfoTime <- ioeIO $ catch (liftM Just $ getModificationTime (replaceExtension gfFile "gfo")) + (\_->return Nothing) + return (gfFile, Just gfTime, mb_gfoTime) + Nothing -> do mb_gfoFile <- ioeIO $ getFilePathMsg "" ps (gfoFile name) + case mb_gfoFile of + Just gfoFile -> do gfoTime <- ioeIO $ getModificationTime gfoFile + return (gfoFile, Nothing, Just gfoTime) + Nothing -> ioeErr $ Bad ("File " ++ gfFile name ++ " does not exist.") + + + let mb_envmod = Map.lookup name env + (st,t) = selectFormat opts (fmap fst mb_envmod) gfTime gfoTime + + imps <- if st == CSEnv + then return (maybe [] snd mb_envmod) + else do s <- ioeIO $ BS.readFile file + (mname,imps) <- ioeErr ((liftM importsOfModule . pModHeader . myLexer) s) + ioeErr $ testErr (mname == name) + ("module name" +++ mname +++ "differs from file name" +++ name) + return imps + + return (name,st,t,imps,dropFileName file) + + +isGFO :: FilePath -> Bool +isGFO = (== ".gfo") . takeExtensions + +gfoFile :: FilePath -> FilePath +gfoFile f = addExtension f "gfo" + +gfFile :: FilePath -> FilePath +gfFile f = addExtension f "gf" + + +-- From the given Options and the time stamps computes +-- whether the module have to be computed, read from .gfo or +-- the environment version have to be used +selectFormat :: Options -> Maybe ClockTime -> Maybe ClockTime -> Maybe ClockTime -> (CompStatus,Maybe ClockTime) +selectFormat opts mtenv mtgf mtgfo = + case (mtenv,mtgfo,mtgf) of + (_,_,Just tgf) | fromSrc -> (CSComp,Nothing) + (Just tenv,_,_) | fromComp -> (CSEnv, Just tenv) + (_,Just tgfo,_) | fromComp -> (CSRead,Just tgfo) + (Just tenv,_,Just tgf) | tenv > tgf -> (CSEnv, Just tenv) + (_,Just tgfo,Just tgf) | tgfo > tgf -> (CSRead,Just tgfo) + (Just tenv,_,Nothing) -> (CSEnv,Just tenv) -- source does not exist + (_,_, Nothing) -> (CSRead,Nothing) -- source does not exist + _ -> (CSComp,Nothing) + where + fromComp = flag optRecomp opts == NeverRecomp + fromSrc = flag optRecomp opts == AlwaysRecomp + + +-- internal module dep information + + +data CompStatus = + CSComp -- compile: read gf + | CSRead -- read gfo + | CSEnv -- gfo is in env + deriving Eq + +type ModuleInfo = (ModName,CompStatus,Maybe ClockTime,[ModName],InitPath) + + +importsOfModule :: ModDef -> (ModName,[ModName]) +importsOfModule (MModule _ typ body) = modType typ (modBody body []) + where + modType (MTAbstract m) xs = (modName m,xs) + modType (MTResource m) xs = (modName m,xs) + modType (MTInterface m) xs = (modName m,xs) + modType (MTConcrete m m2) xs = (modName m,modName m2:xs) + modType (MTInstance m m2) xs = (modName m,modName m2:xs) + modType (MTTransfer m o1 o2) xs = (modName m,open o1 (open o2 xs)) + + modBody (MBody e o _) xs = extend e (opens o xs) + modBody (MNoBody is) xs = foldr include xs is + modBody (MWith i os) xs = include i (foldr open xs os) + modBody (MWithBody i os o _) xs = include i (foldr open (opens o xs) os) + modBody (MWithE is i os) xs = foldr include (include i (foldr open xs os)) is + modBody (MWithEBody is i os o _) xs = foldr include (include i (foldr open (opens o xs) os)) is + modBody (MReuse m) xs = modName m:xs + modBody (MUnion is) xs = foldr include xs is + + include (IAll m) xs = modName m:xs + include (ISome m _) xs = modName m:xs + include (IMinus m _) xs = modName m:xs + + open (OName n) xs = modName n:xs + open (OQualQO _ n) xs = modName n:xs + open (OQual _ _ n) xs = modName n:xs + + extend NoExt xs = xs + extend (Ext is) xs = foldr include xs is + + opens NoOpens xs = xs + opens (OpenIn os) xs = foldr open xs os + + modName (PIdent (_,s)) = BS.unpack s + + +-- | options can be passed to the compiler by comments in @--#@, in the main file +getOptionsFromFile :: FilePath -> IOE Options +getOptionsFromFile file = do + s <- ioeIO $ readFileIfStrict file + let ls = filter (BS.isPrefixOf (BS.pack "--#")) $ BS.lines s + fs = map (BS.unpack . BS.unwords . BS.words . BS.drop 3) ls + ioeErr $ liftM moduleOptions $ parseModuleOptions fs diff --git a/src/GF/Compile/Rebuild.hs b/src/GF/Compile/Rebuild.hs new file mode 100644 index 000000000..ec9076e1c --- /dev/null +++ b/src/GF/Compile/Rebuild.hs @@ -0,0 +1,104 @@ +---------------------------------------------------------------------- +-- | +-- Module : Rebuild +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/05/30 21:08:14 $ +-- > CVS $Author: aarne $ +-- > CVS $Revision: 1.14 $ +-- +-- Rebuild a source module from incomplete and its with-instance. +----------------------------------------------------------------------------- + +module GF.Compile.Rebuild (rebuildModule) where + +import GF.Grammar.Grammar +import GF.Compile.ModDeps +import GF.Grammar.PrGrammar +import GF.Grammar.Lookup +import GF.Compile.Extend +import GF.Grammar.Macros + +import GF.Infra.Ident +import GF.Infra.Modules +import GF.Infra.Option +import GF.Data.Operations + +import Data.List (nub) + +-- | rebuilding instance + interface, and "with" modules, prior to renaming. +-- AR 24/10/2003 +rebuildModule :: [SourceModule] -> SourceModule -> Err SourceModule +rebuildModule ms mo@(i,mi) = do + let gr = MGrammar ms +---- deps <- moduleDeps ms +---- is <- openInterfaces deps i + let is = [] ---- the method above is buggy: try "i -src" for two grs. AR 8/3/2005 + mi' <- case mi of + + -- add the information given in interface into an instance module + ModMod m -> do + testErr (null is || mstatus m == MSIncomplete) + ("module" +++ prt i +++ + "has open interfaces and must therefore be declared incomplete") + case mtype m of + MTInstance i0 -> do + m1 <- lookupModMod gr i0 + testErr (isModRes m1) ("interface expected instead of" +++ prt i0) + m' <- do + js' <- extendMod False (i0,const True) i (jments m1) (jments m) + --- to avoid double inclusions, in instance I of I0 = J0 ** ... + case extends m of + [] -> return $ replaceJudgements m js' + j0s -> do + m0s <- mapM (lookupModMod gr) j0s + let notInM0 c _ = all (not . isInBinTree c . jments) m0s + let js2 = filterBinTree notInM0 js' + return $ (replaceJudgements m js2) + {positions = + buildTree (tree2list (positions m1) ++ + tree2list (positions m))} + return $ ModMod m' + _ -> return mi + + -- add the instance opens to an incomplete module "with" instances + -- ModWith mt stat ext me ops -> do + ModWith (Module mt stat fs_ me ops_ js_ ps_) (ext,incl) ops -> do + let insts = [(inf,inst) | OQualif _ inf inst <- ops] + let infs = map fst insts + let stat' = ifNull MSComplete (const MSIncomplete) + [i | i <- is, notElem i infs] + testErr (stat' == MSComplete || stat == MSIncomplete) + ("module" +++ prt i +++ "remains incomplete") + Module mt0 _ fs me' ops0 js ps0 <- lookupModMod gr ext + let ops1 = nub $ + ops_ ++ -- N.B. js has been name-resolved already + ops ++ [o | o <- ops0, notElem (openedModule o) infs] + ++ [oQualif i i | i <- map snd insts] ---- + ++ [oSimple i | i <- map snd insts] ---- + + --- check if me is incomplete + let fs1 = addModuleOptions fs fs_ -- new flags have priority + let js0 = [ci | ci@(c,_) <- tree2list js, isInherited incl c] + let js1 = buildTree (tree2list js_ ++ js0) + let ps1 = buildTree (tree2list ps_ ++ tree2list ps0) + return $ ModMod $ Module mt0 stat' fs1 me ops1 js1 ps1 + ---- (mapTree (qualifInstanceInfo insts) js) -- not needed + + _ -> return mi + return (i,mi') + +checkCompleteInstance :: SourceRes -> SourceRes -> Err () +checkCompleteInstance abs cnc = ifNull (return ()) (Bad . unlines) $ + checkComplete [f | (f, ResOper (Yes _) _) <- abs'] cnc' + where + abs' = tree2list $ jments abs + cnc' = jments cnc + checkComplete sought given = foldr ckOne [] sought + where + ckOne f = if isInBinTree f given + then id + else (("Error: no definition given to" +++ prt f):) + diff --git a/src/GF/Compile/Refresh.hs b/src/GF/Compile/Refresh.hs new file mode 100644 index 000000000..39fb57db0 --- /dev/null +++ b/src/GF/Compile/Refresh.hs @@ -0,0 +1,133 @@ +---------------------------------------------------------------------- +-- | +-- Module : Refresh +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/04/21 16:22:27 $ +-- > CVS $Author: bringert $ +-- > CVS $Revision: 1.6 $ +-- +-- (Description of the module) +----------------------------------------------------------------------------- + +module GF.Compile.Refresh (refreshTerm, refreshTermN, + refreshModule + ) where + +import GF.Data.Operations +import GF.Grammar.Grammar +import GF.Infra.Ident +import GF.Infra.Modules +import GF.Grammar.Macros +import Control.Monad + +refreshTerm :: Term -> Err Term +refreshTerm = refreshTermN 0 + +refreshTermN :: Int -> Term -> Err Term +refreshTermN i e = liftM snd $ refreshTermKN i e + +refreshTermKN :: Int -> Term -> Err (Int,Term) +refreshTermKN i e = liftM (\ (t,(_,i)) -> (i,t)) $ + appSTM (refresh e) (initIdStateN i) + +refresh :: Term -> STM IdState Term +refresh e = case e of + + Vr x -> liftM Vr (lookVar x) + Abs x b -> liftM2 Abs (refVarPlus x) (refresh b) + + Prod x a b -> do + a' <- refresh a + x' <- refVar x + b' <- refresh b + return $ Prod x' a' b' + + Let (x,(mt,a)) b -> do + a' <- refresh a + mt' <- case mt of + Just t -> refresh t >>= (return . Just) + _ -> return mt + x' <- refVar x + b' <- refresh b + return (Let (x',(mt',a')) b') + + R r -> liftM R $ refreshRecord r + + ExtR r s -> liftM2 ExtR (refresh r) (refresh s) + + T i cc -> liftM2 T (refreshTInfo i) (mapM refreshCase cc) + + _ -> composOp refresh e + +refreshCase :: (Patt,Term) -> STM IdState (Patt,Term) +refreshCase (p,t) = liftM2 (,) (refreshPatt p) (refresh t) + +refreshPatt p = case p of + PV x -> liftM PV (refVar x) + PC c ps -> liftM (PC c) (mapM refreshPatt ps) + PP q c ps -> liftM (PP q c) (mapM refreshPatt ps) + PR r -> liftM PR (mapPairsM refreshPatt r) + PT t p' -> liftM2 PT (refresh t) (refreshPatt p') + + PAs x p' -> liftM2 PAs (refVar x) (refreshPatt p') + + PSeq p' q' -> liftM2 PSeq (refreshPatt p') (refreshPatt q') + PAlt p' q' -> liftM2 PAlt (refreshPatt p') (refreshPatt q') + PRep p' -> liftM PRep (refreshPatt p') + PNeg p' -> liftM PNeg (refreshPatt p') + + _ -> return p + +refreshRecord r = case r of + [] -> return r + (x,(mt,a)):b -> do + a' <- refresh a + mt' <- case mt of + Just t -> refresh t >>= (return . Just) + _ -> return mt + b' <- refreshRecord b + return $ (x,(mt',a')) : b' + +refreshTInfo i = case i of + TTyped t -> liftM TTyped $ refresh t + TComp t -> liftM TComp $ refresh t + TWild t -> liftM TWild $ refresh t + _ -> return i + +-- for abstract syntax + +refreshEquation :: Equation -> Err ([Patt],Term) +refreshEquation pst = err Bad (return . fst) (appSTM (refr pst) initIdState) where + refr (ps,t) = liftM2 (,) (mapM refreshPatt ps) (refresh t) + +-- for concrete and resource in grammar, before optimizing + +refreshGrammar :: SourceGrammar -> Err SourceGrammar +refreshGrammar = liftM (MGrammar . snd) . foldM refreshModule (0,[]) . modules + +refreshModule :: (Int,[SourceModule]) -> SourceModule -> Err (Int,[SourceModule]) +refreshModule (k,ms) mi@(i,m) = case m of + ModMod mo | (isModCnc mo || isModRes mo) -> do + (k',js') <- foldM refreshRes (k,[]) $ tree2list $ jments mo + return (k', (i, ModMod(replaceJudgements mo (buildTree js'))) : ms) + _ -> return (k, mi:ms) + where + refreshRes (k,cs) ci@(c,info) = case info of + ResOper ptyp (Yes trm) -> do ---- refresh ptyp + (k',trm') <- refreshTermKN k trm + return $ (k', (c, ResOper ptyp (Yes trm')):cs) + ResOverload os tyts -> do + (k',tyts') <- liftM (\ (t,(_,i)) -> (i,t)) $ + appSTM (mapPairsM refresh tyts) (initIdStateN k) + return $ (k', (c, ResOverload os tyts'):cs) + CncCat mt (Yes trm) pn -> do ---- refresh mt, pn + (k',trm') <- refreshTermKN k trm + return $ (k', (c, CncCat mt (Yes trm') pn):cs) + CncFun mt (Yes trm) pn -> do ---- refresh pn + (k',trm') <- refreshTermKN k trm + return $ (k', (c, CncFun mt (Yes trm') pn):cs) + _ -> return (k, ci:cs) + diff --git a/src/GF/Compile/RemoveLiT.hs b/src/GF/Compile/RemoveLiT.hs new file mode 100644 index 000000000..d06b80400 --- /dev/null +++ b/src/GF/Compile/RemoveLiT.hs @@ -0,0 +1,64 @@ +---------------------------------------------------------------------- +-- | +-- Module : RemoveLiT +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/04/21 16:21:45 $ +-- > CVS $Author: bringert $ +-- > CVS $Revision: 1.6 $ +-- +-- remove obsolete (Lin C) expressions before doing anything else. AR 21/6/2003 +-- +-- What the program does is replace the occurrences of Lin C with the actual +-- definition T given in lincat C = T ; with {s : Str} if no lincat is found. +-- The procedure is uncertain, if T contains another Lin. +----------------------------------------------------------------------------- + +module GF.Compile.RemoveLiT (removeLiT) where + +import GF.Grammar.Grammar +import GF.Infra.Ident +import GF.Infra.Modules +import GF.Grammar.Macros +import GF.Grammar.Lookup +import GF.Grammar.Predef + +import GF.Data.Operations + +import Control.Monad + +removeLiT :: SourceGrammar -> Err SourceGrammar +removeLiT gr = liftM MGrammar $ mapM (remlModule gr) (modules gr) + +remlModule :: SourceGrammar -> (Ident,SourceModInfo) -> Err (Ident,SourceModInfo) +remlModule gr mi@(name,mod) = case mod of + ModMod mo -> do + js1 <- mapMTree (remlResInfo gr) (jments mo) + let mod2 = ModMod $ mo {jments = js1} + return $ (name,mod2) + _ -> return mi + +remlResInfo :: SourceGrammar -> (Ident,Info) -> Err (Ident,Info) +remlResInfo gr mi@(i,info) = case info of + ResOper pty ptr -> liftM ((,) i) $ liftM2 ResOper (ren pty) (ren ptr) + CncCat pty ptr ppr -> liftM ((,) i) $ liftM3 CncCat (ren pty) (ren ptr) (ren ppr) + CncFun mt ptr ppr -> liftM ((,) i) $ liftM2 (CncFun mt) (ren ptr) (ren ppr) + _ -> return mi + where + ren = remlPerh gr + +remlPerh gr pt = case pt of + Yes t -> liftM Yes $ remlTerm gr t + _ -> return pt + +remlTerm :: SourceGrammar -> Term -> Err Term +remlTerm gr trm = case trm of + LiT c -> look c >>= remlTerm gr + _ -> composOp (remlTerm gr) trm + where + look c = err (const $ return defLinType) return $ lookupLincat gr m c + m = case [cnc | (cnc,ModMod m) <- modules gr, isModCnc m] of + cnc:_ -> cnc -- actually there is always exactly one + _ -> cCNC diff --git a/src/GF/Compile/Rename.hs b/src/GF/Compile/Rename.hs new file mode 100644 index 000000000..7b4d09277 --- /dev/null +++ b/src/GF/Compile/Rename.hs @@ -0,0 +1,338 @@ +---------------------------------------------------------------------- +-- | +-- Module : Rename +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/05/30 18:39:44 $ +-- > CVS $Author: aarne $ +-- > CVS $Revision: 1.19 $ +-- +-- AR 14\/5\/2003 +-- The top-level function 'renameGrammar' does several things: +-- +-- - extends each module symbol table by indirections to extended module +-- +-- - changes unqualified and as-qualified imports to absolutely qualified +-- +-- - goes through the definitions and resolves names +-- +-- Dependency analysis between modules has been performed before this pass. +-- Hence we can proceed by @fold@ing "from left to right". +----------------------------------------------------------------------------- + +module GF.Compile.Rename (renameGrammar, + renameSourceTerm, + renameModule + ) where + +import GF.Grammar.Grammar +import GF.Grammar.Values +import GF.Grammar.Predef +import GF.Infra.Modules +import GF.Infra.Ident +import GF.Grammar.Macros +import GF.Grammar.PrGrammar +import GF.Grammar.AppPredefined +import GF.Grammar.Lookup +import GF.Compile.Extend +import GF.Data.Operations + +import Control.Monad +import Data.List (nub) +import Debug.Trace (trace) + +renameGrammar :: SourceGrammar -> Err SourceGrammar +renameGrammar g = liftM (MGrammar . reverse) $ foldM renameModule [] (modules g) + +-- | this gives top-level access to renaming term input in the cc command +renameSourceTerm :: SourceGrammar -> Ident -> Term -> Err Term +renameSourceTerm g m t = do + mo <- lookupErr m (modules g) + status <- buildStatus g m mo + renameTerm status [] t + +renameModule :: [SourceModule] -> SourceModule -> Err [SourceModule] +renameModule ms (name,mod) = errIn ("renaming module" +++ prt name) $ case mod of + ModMod mo -> do + let js1 = jments mo + status <- buildStatus (MGrammar ms) name mod + js2 <- mapsErrTree (renameInfo mo status) js1 + let mod2 = ModMod $ mo {opens = map forceQualif (opens mo), jments = js2} + return $ (name,mod2) : ms + +type Status = (StatusTree, [(OpenSpec Ident, StatusTree)]) + +type StatusTree = BinTree Ident StatusInfo + +type StatusInfo = Ident -> Term + +renameIdentTerm :: Status -> Term -> Err Term +renameIdentTerm env@(act,imps) t = + errIn ("atomic term" +++ prt t +++ "given" +++ unwords (map (prt . fst) qualifs)) $ + case t of + Vr c -> ident predefAbs c + Cn c -> ident (\_ s -> Bad s) c + Q m' c | m' == cPredef {- && isInPredefined c -} -> return t + Q m' c -> do + m <- lookupErr m' qualifs + f <- lookupTree prt c m + return $ f c + QC m' c | m' == cPredef {- && isInPredefined c -} -> return t + QC m' c -> do + m <- lookupErr m' qualifs + f <- lookupTree prt c m + return $ f c + _ -> return t + where + opens = [st | (OSimple _ _,st) <- imps] + qualifs = [(m, st) | (OQualif _ m _, st) <- imps] ++ + [(m, st) | (OSimple _ m, st) <- imps] -- qualif is always possible + + -- this facility is mainly for BWC with GF1: you need not import PredefAbs + predefAbs c s + | isPredefCat c = return $ Q cPredefAbs c + | otherwise = Bad s + + ident alt c = case lookupTree prt c act of + Ok f -> return $ f c + _ -> case lookupTreeManyAll prt opens c of + [f] -> return $ f c + [] -> alt c ("constant not found:" +++ prt c) + fs -> case nub [f c | f <- fs] of + [tr] -> return tr + ts@(t:_) -> trace ("WARNING: conflict" +++ unwords (map prt ts)) (return t) + -- a warning will be generated in CheckGrammar, and the head returned + -- in next V: + -- Bad $ "conflicting imports:" +++ unwords (map prt ts) + + +--- | would it make sense to optimize this by inlining? +renameIdentPatt :: Status -> Patt -> Err Patt +renameIdentPatt env p = do + let t = patt2term p + t' <- renameIdentTerm env t + term2patt t' + +info2status :: Maybe Ident -> (Ident,Info) -> (Ident,StatusInfo) +info2status mq (c,i) = (c, case i of + AbsFun _ (Yes EData) -> maybe Con QC mq + ResValue _ -> maybe Con QC mq + ResParam _ -> maybe Con QC mq + AnyInd True m -> maybe Con (const (QC m)) mq + AnyInd False m -> maybe Cn (const (Q m)) mq + _ -> maybe Cn Q mq + ) + +tree2status :: OpenSpec Ident -> BinTree Ident Info -> BinTree Ident StatusInfo +tree2status o = case o of + OSimple _ i -> mapTree (info2status (Just i)) + OQualif _ i j -> mapTree (info2status (Just j)) + +buildStatus :: SourceGrammar -> Ident -> SourceModInfo -> Err Status +buildStatus gr c mo = let mo' = self2status c mo in case mo of + ModMod m -> do + let gr1 = MGrammar $ (c,mo) : modules gr + ops = [OSimple OQNormal e | e <- allExtends gr1 c] ++ allOpens m + mods <- mapM (lookupModule gr1 . openedModule) ops + let sts = map modInfo2status $ zip ops mods + return $ if isModCnc m + then (emptyBinTree, reverse sts) -- the module itself does not define any names + else (mo',reverse sts) -- so the empty ident is not needed + +modInfo2status :: (OpenSpec Ident,SourceModInfo) -> (OpenSpec Ident, StatusTree) +modInfo2status (o,i) = (o,case i of + ModMod m -> tree2status o (jments m) + ) + +self2status :: Ident -> SourceModInfo -> StatusTree +self2status c i = mapTree (info2status (Just c)) js where -- qualify internal + js = case i of + ModMod m + | isModTrans m -> sorted2tree $ filter noTrans $ tree2list $ jments m + | otherwise -> jments m + noTrans (_,d) = case d of -- to enable other than transfer js in transfer module + AbsTrans _ -> False + _ -> True + +forceQualif o = case o of + OSimple q i -> OQualif q i i + OQualif q _ i -> OQualif q i i + +renameInfo :: Module Ident Info -> Status -> (Ident,Info) -> Err (Ident,Info) +renameInfo mo status (i,info) = errIn + ("renaming definition of" +++ prt i +++ showPosition mo i) $ + liftM ((,) i) $ case info of + AbsCat pco pfs -> liftM2 AbsCat (renPerh (renameContext status) pco) + (renPerh (mapM rent) pfs) + AbsFun pty ptr -> liftM2 AbsFun (ren pty) (ren ptr) + AbsTrans f -> liftM AbsTrans (rent f) + + ResOper pty ptr -> liftM2 ResOper (ren pty) (ren ptr) + ResOverload os tysts -> + liftM (ResOverload os) (mapM (pairM rent) tysts) + + ResParam (Yes (pp,m)) -> do + pp' <- mapM (renameParam status) pp + return $ ResParam $ Yes (pp',m) + ResValue (Yes (t,m)) -> do + t' <- rent t + return $ ResValue $ Yes (t',m) + CncCat pty ptr ppr -> liftM3 CncCat (ren pty) (ren ptr) (ren ppr) + CncFun mt ptr ppr -> liftM2 (CncFun mt) (ren ptr) (ren ppr) + _ -> return info + where + ren = renPerh rent + rent = renameTerm status [] + +renPerh ren pt = case pt of + Yes t -> liftM Yes $ ren t + _ -> return pt + +renameTerm :: Status -> [Ident] -> Term -> Err Term +renameTerm env vars = ren vars where + ren vs trm = case trm of + Abs x b -> liftM (Abs x) (ren (x:vs) b) + Prod x a b -> liftM2 (Prod x) (ren vs a) (ren (x:vs) b) + Typed a b -> liftM2 Typed (ren vs a) (ren vs b) + Vr x + | elem x vs -> return trm + | otherwise -> renid trm + Cn _ -> renid trm + Con _ -> renid trm + Q _ _ -> renid trm + QC _ _ -> renid trm + Eqs eqs -> liftM Eqs $ mapM (renameEquation env vars) eqs + T i cs -> do + i' <- case i of + TTyped ty -> liftM TTyped $ ren vs ty -- the only annotation in source + _ -> return i + liftM (T i') $ mapM (renCase vs) cs + + Let (x,(m,a)) b -> do + m' <- case m of + Just ty -> liftM Just $ ren vs ty + _ -> return m + a' <- ren vs a + b' <- ren (x:vs) b + return $ Let (x,(m',a')) b' + + P t@(Vr r) l -- for constant t we know it is projection + | elem r vs -> return trm -- var proj first + | otherwise -> case renid (Q r (label2ident l)) of -- qualif second + Ok t -> return t + _ -> case liftM (flip P l) $ renid t of + Ok t -> return t -- const proj last + _ -> prtBad "unknown qualified constant" trm + + EPatt p -> do + (p',_) <- renpatt p + return $ EPatt p' + + _ -> composOp (ren vs) trm + + renid = renameIdentTerm env + renCase vs (p,t) = do + (p',vs') <- renpatt p + t' <- ren (vs' ++ vs) t + return (p',t') + renpatt = renamePattern env + +-- | vars not needed in env, since patterns always overshadow old vars +renamePattern :: Status -> Patt -> Err (Patt,[Ident]) +renamePattern env patt = case patt of + + PMacro c -> do + c' <- renid $ Vr c + case c' of + Q p d -> renp $ PM p d + _ -> prtBad "unresolved pattern" patt + + PC c ps -> do + c' <- renameIdentTerm env $ Cn c + case c' of + QC p d -> renp $ PP p d ps +-- Q p d -> renp $ PP p d ps --- why this? AR 15/3/2008 + _ -> prtBad "unresolved pattern" c' ---- (PC c ps', concat vs) + + PP p c ps -> do + + (p', c') <- case renameIdentTerm env (QC p c) of + Ok (QC p' c') -> return (p',c') + _ -> return (p,c) --- temporarily, for bw compat + psvss <- mapM renp ps + let (ps',vs) = unzip psvss + return (PP p' c' ps', concat vs) + + PM p c -> do + (p', c') <- case renameIdentTerm env (Q p c) of + Ok (Q p' c') -> return (p',c') + _ -> prtBad "not a pattern macro" patt + return (PM p' c', []) + + PV x -> case renid (Vr x) of + Ok (QC m c) -> return (PP m c [],[]) + _ -> return (patt, [x]) + + PR r -> do + let (ls,ps) = unzip r + psvss <- mapM renp ps + let (ps',vs') = unzip psvss + return (PR (zip ls ps'), concat vs') + + PAlt p q -> do + (p',vs) <- renp p + (q',ws) <- renp q + return (PAlt p' q', vs ++ ws) + + PSeq p q -> do + (p',vs) <- renp p + (q',ws) <- renp q + return (PSeq p' q', vs ++ ws) + + PRep p -> do + (p',vs) <- renp p + return (PRep p', vs) + + PNeg p -> do + (p',vs) <- renp p + return (PNeg p', vs) + + PAs x p -> do + (p',vs) <- renp p + return (PAs x p', x:vs) + + _ -> return (patt,[]) + + where + renp = renamePattern env + renid = renameIdentTerm env + +renameParam :: Status -> (Ident, Context) -> Err (Ident, Context) +renameParam env (c,co) = do + co' <- renameContext env co + return (c,co') + +renameContext :: Status -> Context -> Err Context +renameContext b = renc [] where + renc vs cont = case cont of + (x,t) : xts + | isWildIdent x -> do + t' <- ren vs t + xts' <- renc vs xts + return $ (x,t') : xts' + | otherwise -> do + t' <- ren vs t + let vs' = x:vs + xts' <- renc vs' xts + return $ (x,t') : xts' + _ -> return cont + ren = renameTerm b + +-- | vars not needed in env, since patterns always overshadow old vars +renameEquation :: Status -> [Ident] -> Equation -> Err Equation +renameEquation b vs (ps,t) = do + (ps',vs') <- liftM unzip $ mapM (renamePattern b) ps + t' <- renameTerm b (concat vs' ++ vs) t + return (ps',t') diff --git a/src/GF/Compile/TC.hs b/src/GF/Compile/TC.hs new file mode 100644 index 000000000..c0c8a83ae --- /dev/null +++ b/src/GF/Compile/TC.hs @@ -0,0 +1,292 @@ +---------------------------------------------------------------------- +-- | +-- Module : TC +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/10/02 20:50:19 $ +-- > CVS $Author: aarne $ +-- > CVS $Revision: 1.11 $ +-- +-- Thierry Coquand's type checking algorithm that creates a trace +----------------------------------------------------------------------------- + +module GF.Compile.TC (AExp(..), + Theory, + checkExp, + inferExp, + checkEqs, + eqVal, + whnf + ) where + +import GF.Data.Operations +import GF.Grammar.Predef +import GF.Grammar.Abstract + +import Control.Monad +import Data.List (sortBy) + +data AExp = + AVr Ident Val + | ACn QIdent Val + | AType + | AInt Integer + | AFloat Double + | AStr String + | AMeta MetaSymb Val + | AApp AExp AExp Val + | AAbs Ident Val AExp + | AProd Ident AExp AExp + | AEqs [([Exp],AExp)] --- not used + | AData Val + deriving (Eq,Show) + +type Theory = QIdent -> Err Val + +lookupConst :: Theory -> QIdent -> Err Val +lookupConst th f = th f + +lookupVar :: Env -> Ident -> Err Val +lookupVar g x = maybe (prtBad "unknown variable" x) return $ lookup x ((IW,uVal):g) +-- wild card IW: no error produced, ?0 instead. + +type TCEnv = (Int,Env,Env) + +emptyTCEnv :: TCEnv +emptyTCEnv = (0,[],[]) + +whnf :: Val -> Err Val +whnf v = ---- errIn ("whnf" +++ prt v) $ ---- debug + case v of + VApp u w -> do + u' <- whnf u + w' <- whnf w + app u' w' + VClos env e -> eval env e + _ -> return v + +app :: Val -> Val -> Err Val +app u v = case u of + VClos env (Abs x e) -> eval ((x,v):env) e + _ -> return $ VApp u v + +eval :: Env -> Exp -> Err Val +eval env e = ---- errIn ("eval" +++ prt e +++ "in" +++ prEnv env) $ + case e of + Vr x -> lookupVar env x + Q m c -> return $ VCn (m,c) + QC m c -> return $ VCn (m,c) ---- == Q ? + Sort c -> return $ VType --- the only sort is Type + App f a -> join $ liftM2 app (eval env f) (eval env a) + _ -> return $ VClos env e + +eqVal :: Int -> Val -> Val -> Err [(Val,Val)] +eqVal k u1 u2 = ---- errIn (prt u1 +++ "<>" +++ prBracket (show k) +++ prt u2) $ + do + w1 <- whnf u1 + w2 <- whnf u2 + let v = VGen k + case (w1,w2) of + (VApp f1 a1, VApp f2 a2) -> liftM2 (++) (eqVal k f1 f2) (eqVal k a1 a2) + (VClos env1 (Abs x1 e1), VClos env2 (Abs x2 e2)) -> + eqVal (k+1) (VClos ((x1,v x1):env1) e1) (VClos ((x2,v x1):env2) e2) + (VClos env1 (Prod x1 a1 e1), VClos env2 (Prod x2 a2 e2)) -> + liftM2 (++) + (eqVal k (VClos env1 a1) (VClos env2 a2)) + (eqVal (k+1) (VClos ((x1,v x1):env1) e1) (VClos ((x2,v x1):env2) e2)) + (VGen i _, VGen j _) -> return [(w1,w2) | i /= j] + (VCn (_, i), VCn (_,j)) -> return [(w1,w2) | i /= j] + --- thus ignore qualifications; valid because inheritance cannot + --- be qualified. Simplifies annotation. AR 17/3/2005 + _ -> return [(w1,w2) | w1 /= w2] +-- invariant: constraints are in whnf + +checkType :: Theory -> TCEnv -> Exp -> Err (AExp,[(Val,Val)]) +checkType th tenv e = checkExp th tenv e vType + +checkExp :: Theory -> TCEnv -> Exp -> Val -> Err (AExp, [(Val,Val)]) +checkExp th tenv@(k,rho,gamma) e ty = do + typ <- whnf ty + let v = VGen k + case e of + Meta m -> return $ (AMeta m typ,[]) + EData -> return $ (AData typ,[]) + + Abs x t -> case typ of + VClos env (Prod y a b) -> do + a' <- whnf $ VClos env a --- + (t',cs) <- checkExp th + (k+1,(x,v x):rho, (x,a'):gamma) t (VClos ((y,v x):env) b) + return (AAbs x a' t', cs) + _ -> prtBad ("function type expected for" +++ prt e +++ "instead of") typ + +-- {- --- to get deprec when checkEqs works (15/9/2005) + Eqs es -> do + bcs <- mapM (\b -> checkBranch th tenv b typ) es + let (bs,css) = unzip bcs + return (AEqs bs, concat css) +-- - } + Prod x a b -> do + testErr (typ == vType) "expected Type" + (a',csa) <- checkType th tenv a + (b',csb) <- checkType th (k+1, (x,v x):rho, (x,VClos rho a):gamma) b + return (AProd x a' b', csa ++ csb) + + _ -> checkInferExp th tenv e typ + +checkInferExp :: Theory -> TCEnv -> Exp -> Val -> Err (AExp, [(Val,Val)]) +checkInferExp th tenv@(k,_,_) e typ = do + (e',w,cs1) <- inferExp th tenv e + cs2 <- eqVal k w typ + return (e',cs1 ++ cs2) + +inferExp :: Theory -> TCEnv -> Exp -> Err (AExp, Val, [(Val,Val)]) +inferExp th tenv@(k,rho,gamma) e = case e of + Vr x -> mkAnnot (AVr x) $ noConstr $ lookupVar gamma x + Q m c | m == cPredefAbs && isPredefCat c + -> return (ACn (m,c) vType, vType, []) + | otherwise -> mkAnnot (ACn (m,c)) $ noConstr $ lookupConst th (m,c) + QC m c -> mkAnnot (ACn (m,c)) $ noConstr $ lookupConst th (m,c) ---- + EInt i -> return (AInt i, valAbsInt, []) + EFloat i -> return (AFloat i, valAbsFloat, []) + K i -> return (AStr i, valAbsString, []) + Sort _ -> return (AType, vType, []) + App f t -> do + (f',w,csf) <- inferExp th tenv f + typ <- whnf w + case typ of + VClos env (Prod x a b) -> do + (a',csa) <- checkExp th tenv t (VClos env a) + b' <- whnf $ VClos ((x,VClos rho t):env) b + return $ (AApp f' a' b', b', csf ++ csa) + _ -> prtBad ("Prod expected for function" +++ prt f +++ "instead of") typ + _ -> prtBad "cannot infer type of expression" e + +checkEqs :: Theory -> TCEnv -> (Fun,Trm) -> Val -> Err [(Val,Val)] +checkEqs th tenv@(k,rho,gamma) (fun@(m,f),def) val = case def of + Eqs es -> liftM concat $ mapM checkBranch es + _ -> liftM snd $ checkExp th tenv def val + where + checkBranch (ps,df) = + let + (ps',_,vars) = foldr p2t ([],0,[]) ps + fps = mkApp (Q m f) ps' + in errIn ("branch" +++ prt fps) $ do + (aexp, typ, cs1) <- inferExp th tenv fps + let + bds = binds vars aexp + tenv' = (k, rho, bds ++ gamma) + (_,cs2) <- errIn (show bds) $ checkExp th tenv' df typ + return $ (cs1 ++ cs2) + p2t p (ps,i,g) = case p of + PW -> (Meta (MetaSymb i) : ps, i+1, g) + PV IW -> (Meta (MetaSymb i) : ps, i+1, g) + PV x -> (Meta (MetaSymb i) : ps, i+1,upd x i g) + PString s -> ( K s : ps, i, g) + PInt n -> (EInt n : ps, i, g) + PFloat n -> (EFloat n : ps, i, g) + PP m c xs -> (mkApp (qq (m,c)) xss : ps, i', g') + where (xss,i',g') = foldr p2t ([],i,g) xs + _ -> error $ "undefined p2t case" +++ prt p +++ "in checkBranch" + upd x i g = (x,i) : g --- to annotate pattern variables: treat as metas + + -- notice: in vars, the sequence 0.. is sorted. In subst aexp, all + -- this occurs and nothing else. + binds vars aexp = [(x,v) | ((x,_),v) <- zip vars metas] where + metas = map snd $ sortBy (\ (x,_) (y,_) -> compare x y) $ subst aexp + subst aexp = case aexp of + AMeta (MetaSymb i) v -> [(i,v)] + AApp c a _ -> subst c ++ subst a + _ -> [] -- never matter in patterns + +checkBranch :: Theory -> TCEnv -> Equation -> Val -> Err (([Exp],AExp),[(Val,Val)]) +checkBranch th tenv b@(ps,t) ty = errIn ("branch" +++ show b) $ + chB tenv' ps' ty + where + + (ps',_,rho2,k') = ps2ts k ps + tenv' = (k, rho2++rho, gamma) ---- k' ? + (k,rho,gamma) = tenv + + chB tenv@(k,rho,gamma) ps ty = case ps of + p:ps2 -> do + typ <- whnf ty + case typ of + VClos env (Prod y a b) -> do + a' <- whnf $ VClos env a + (p', sigma, binds, cs1) <- checkP tenv p y a' + let tenv' = (length binds, sigma ++ rho, binds ++ gamma) + ((ps',exp),cs2) <- chB tenv' ps2 (VClos ((y,p'):env) b) + return ((p:ps',exp), cs1 ++ cs2) -- don't change the patt + _ -> prtBad ("Product expected for definiens" +++prt t +++ "instead of") typ + [] -> do + (e,cs) <- checkExp th tenv t ty + return (([],e),cs) + checkP env@(k,rho,gamma) t x a = do + (delta,cs) <- checkPatt th env t a + let sigma = [(x, VGen i x) | ((x,_),i) <- zip delta [k..]] + return (VClos sigma t, sigma, delta, cs) + + ps2ts k = foldr p2t ([],0,[],k) + p2t p (ps,i,g,k) = case p of + PW -> (Meta (MetaSymb i) : ps, i+1,g,k) + PV IW -> (Meta (MetaSymb i) : ps, i+1,g,k) + PV x -> (Vr x : ps, i, upd x k g,k+1) + PString s -> (K s : ps, i, g, k) + PInt n -> (EInt n : ps, i, g, k) + PFloat n -> (EFloat n : ps, i, g, k) + PP m c xs -> (mkApp (qq (m,c)) xss : ps, j, g',k') + where (xss,j,g',k') = foldr p2t ([],i,g,k) xs + _ -> error $ "undefined p2t case" +++ prt p +++ "in checkBranch" + + upd x k g = (x, VGen k x) : g --- hack to recognize pattern variables + + +checkPatt :: Theory -> TCEnv -> Exp -> Val -> Err (Binds,[(Val,Val)]) +checkPatt th tenv exp val = do + (aexp,_,cs) <- checkExpP tenv exp val + let binds = extrBinds aexp + return (binds,cs) + where + extrBinds aexp = case aexp of + AVr i v -> [(i,v)] + AApp f a _ -> extrBinds f ++ extrBinds a + _ -> [] -- no other cases are possible + +--- ad hoc, to find types of variables + checkExpP tenv@(k,rho,gamma) exp val = case exp of + Meta m -> return $ (AMeta m val, val, []) + Vr x -> return $ (AVr x val, val, []) + EInt i -> return (AInt i, valAbsInt, []) + EFloat i -> return (AFloat i, valAbsFloat, []) + K s -> return (AStr s, valAbsString, []) + + Q m c -> do + typ <- lookupConst th (m,c) + return $ (ACn (m,c) typ, typ, []) + QC m c -> do + typ <- lookupConst th (m,c) + return $ (ACn (m,c) typ, typ, []) ---- + App f t -> do + (f',w,csf) <- checkExpP tenv f val + typ <- whnf w + case typ of + VClos env (Prod x a b) -> do + (a',_,csa) <- checkExpP tenv t (VClos env a) + b' <- whnf $ VClos ((x,VClos rho t):env) b + return $ (AApp f' a' b', b', csf ++ csa) + _ -> prtBad ("Prod expected for function" +++ prt f +++ "instead of") typ + _ -> prtBad "cannot typecheck pattern" exp + +-- auxiliaries + +noConstr :: Err Val -> Err (Val,[(Val,Val)]) +noConstr er = er >>= (\v -> return (v,[])) + +mkAnnot :: (Val -> AExp) -> Err (Val,[(Val,Val)]) -> Err (AExp,Val,[(Val,Val)]) +mkAnnot a ti = do + (v,cs) <- ti + return (a v, v, cs) + diff --git a/src/GF/Compile/TypeCheck.hs b/src/GF/Compile/TypeCheck.hs new file mode 100644 index 000000000..2d58a33ee --- /dev/null +++ b/src/GF/Compile/TypeCheck.hs @@ -0,0 +1,118 @@ +---------------------------------------------------------------------- +-- | +-- Module : TypeCheck +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/09/15 16:22:02 $ +-- > CVS $Author: aarne $ +-- > CVS $Revision: 1.16 $ +-- +-- (Description of the module) +----------------------------------------------------------------------------- + +module GF.Compile.TypeCheck (-- * top-level type checking functions; TC should not be called directly. + checkContext, + checkTyp, + checkEquation, + checkConstrs, + ) where + +import GF.Data.Operations +import GF.Data.Zipper + +import GF.Grammar.Abstract +import GF.Compile.Refresh +import GF.Grammar.LookAbs +import qualified GF.Grammar.Lookup as Lookup --- +import GF.Grammar.Unify --- + +import GF.Compile.TC + +import Control.Monad (foldM, liftM, liftM2) +import Data.List (nub) --- + +-- | invariant way of creating TCEnv from context +initTCEnv gamma = + (length gamma,[(x,VGen i x) | ((x,_),i) <- zip gamma [0..]], gamma) + +-- interface to TC type checker + +type2val :: Type -> Val +type2val = VClos [] + +aexp2tree :: (AExp,[(Val,Val)]) -> Err Tree +aexp2tree (aexp,cs) = do + (bi,at,vt,ts) <- treeForm aexp + ts' <- mapM aexp2tree [(t,[]) | t <- ts] + return $ Tr (N (bi,at,vt,(cs,[]),False),ts') + where + treeForm a = case a of + AAbs x v b -> do + (bi, at, vt, args) <- treeForm b + v' <- whnf v ---- should not be needed... + return ((x,v') : bi, at, vt, args) + AApp c a v -> do + (_,at,_,args) <- treeForm c + v' <- whnf v ---- + return ([],at,v',args ++ [a]) + AVr x v -> do + v' <- whnf v ---- + return ([],AtV x,v',[]) + ACn c v -> do + v' <- whnf v ---- + return ([],AtC c,v',[]) + AInt i -> do + return ([],AtI i,valAbsInt,[]) + AFloat i -> do + return ([],AtF i,valAbsFloat,[]) + AStr s -> do + return ([],AtL s,valAbsString,[]) + AMeta m v -> do + v' <- whnf v ---- + return ([],AtM m,v',[]) + _ -> Bad "illegal tree" -- AProd + +cont2exp :: Context -> Exp +cont2exp c = mkProd (c, eType, []) -- to check a context + +cont2val :: Context -> Val +cont2val = type2val . cont2exp + +-- some top-level batch-mode checkers for the compiler + +justTypeCheck :: Grammar -> Exp -> Val -> Err Constraints +justTypeCheck gr e v = do + (_,constrs0) <- checkExp (grammar2theory gr) (initTCEnv []) e v + return $ filter notJustMeta constrs0 +---- return $ fst $ splitConstraintsSrc gr constrs0 +---- this change was to force proper tc of abstract modules. +---- May not be quite right. AR 13/9/2005 + +notJustMeta (c,k) = case (c,k) of + (VClos g1 (Meta m1), VClos g2 (Meta m2)) -> False + _ -> True + +grammar2theory :: Grammar -> Theory +grammar2theory gr (m,f) = case lookupFunType gr m f of + Ok t -> return $ type2val t + Bad s -> case lookupCatContext gr m f of + Ok cont -> return $ cont2val cont + _ -> Bad s + +checkContext :: Grammar -> Context -> [String] +checkContext st = checkTyp st . cont2exp + +checkTyp :: Grammar -> Type -> [String] +checkTyp gr typ = err singleton prConstrs $ justTypeCheck gr typ vType + +checkEquation :: Grammar -> Fun -> Trm -> [String] +checkEquation gr (m,fun) def = err singleton id $ do + typ <- lookupFunType gr m fun + cs <- justTypeCheck gr def (vClos typ) + let cs1 = filter notJustMeta cs + return $ ifNull [] (singleton . prConstraints) cs1 + +checkConstrs :: Grammar -> Cat -> [Ident] -> [String] +checkConstrs gr cat _ = [] ---- check constructors! diff --git a/src/GF/Compile/Update.hs b/src/GF/Compile/Update.hs new file mode 100644 index 000000000..82d7a609e --- /dev/null +++ b/src/GF/Compile/Update.hs @@ -0,0 +1,135 @@ +---------------------------------------------------------------------- +-- | +-- Module : Update +-- Maintainer : AR +-- Stability : (stable) +-- Portability : (portable) +-- +-- > CVS $Date: 2005/05/30 18:39:44 $ +-- > CVS $Author: aarne $ +-- > CVS $Revision: 1.8 $ +-- +-- (Description of the module) +----------------------------------------------------------------------------- + +module GF.Compile.Update (updateRes, buildAnyTree, combineAnyInfos, unifyAnyInfo, + -- * these auxiliaries should be somewhere else + -- since they don't use the info types + groupInfos, sortInfos, combineInfos, unifyInfos, + tryInsert, unifAbsDefs, unifConstrs + ) where + +import GF.Infra.Ident +import GF.Grammar.Grammar +import GF.Grammar.PrGrammar +import GF.Infra.Modules + +import GF.Data.Operations + +import Data.List +import Control.Monad + +-- | update a resource module by adding a new or changing an old definition +updateRes :: SourceGrammar -> Ident -> Ident -> Info -> SourceGrammar +updateRes gr@(MGrammar ms) m i info = MGrammar $ map upd ms where + upd (n,mod) + | n /= m = (n,mod) + | n == m = case mod of + ModMod r -> (m,ModMod $ updateModule r i info) + _ -> (n,mod) --- no error msg + +-- | combine a list of definitions into a balanced binary search tree +buildAnyTree :: [(Ident,Info)] -> Err (BinTree Ident Info) +buildAnyTree ias = do + ias' <- combineAnyInfos ias + return $ buildTree ias' + + +-- | unifying information for abstract, resource, and concrete +combineAnyInfos :: [(Ident,Info)] -> Err [(Ident,Info)] +combineAnyInfos = combineInfos unifyAnyInfo + +unifyAnyInfo :: Ident -> Info -> Info -> Err Info +unifyAnyInfo c i j = errIn ("combining information for" +++ prt c) $ case (i,j) of + (AbsCat mc1 mf1, AbsCat mc2 mf2) -> + liftM2 AbsCat (unifPerhaps mc1 mc2) (unifConstrs mf1 mf2) -- adding constrs + (AbsFun mt1 md1, AbsFun mt2 md2) -> + liftM2 AbsFun (unifPerhaps mt1 mt2) (unifAbsDefs md1 md2) -- adding defs + + (ResParam mt1, ResParam mt2) -> liftM ResParam $ unifPerhaps mt1 mt2 + (ResOper mt1 m1, ResOper mt2 m2) -> + liftM2 ResOper (unifPerhaps mt1 mt2) (unifPerhaps m1 m2) + + (CncCat mc1 mf1 mp1, CncCat mc2 mf2 mp2) -> + liftM3 CncCat (unifPerhaps mc1 mc2) (unifPerhaps mf1 mf2) (unifPerhaps mp1 mp2) + (CncFun m mt1 md1, CncFun _ mt2 md2) -> + liftM2 (CncFun m) (unifPerhaps mt1 mt2) (unifPerhaps md1 md2) ---- adding defs +-- for bw compatibility with unspecified printnames in old GF + (CncFun Nothing Nope (Yes pr),_) -> + unifyAnyInfo c (CncCat Nope Nope (Yes pr)) j + (_,CncFun Nothing Nope (Yes pr)) -> + unifyAnyInfo c i (CncCat Nope Nope (Yes pr)) + + _ -> Bad $ "cannot unify informations in" ++++ show i ++++ "and" ++++ show j + +--- these auxiliaries should be somewhere else since they don't use the info types + +groupInfos :: Eq a => [(a,b)] -> [[(a,b)]] +groupInfos = groupBy (\i j -> fst i == fst j) + +sortInfos :: Ord a => [(a,b)] -> [(a,b)] +sortInfos = sortBy (\i j -> compare (fst i) (fst j)) + +combineInfos :: Ord a => (a -> b -> b -> Err b) -> [(a,b)] -> Err [(a,b)] +combineInfos f ris = do + let riss = groupInfos $ sortInfos ris + mapM (unifyInfos f) riss + +unifyInfos :: (a -> b -> b -> Err b) -> [(a,b)] -> Err (a,b) +unifyInfos _ [] = Bad "empty info list" +unifyInfos unif ris = do + let c = fst $ head ris + let infos = map snd ris + let ([i],is) = splitAt 1 infos + info <- foldM (unif c) i is + return (c,info) + + +tryInsert :: Ord a => (b -> b -> Err b) -> (b -> b) -> + BinTree a b -> (a,b) -> Err (BinTree a b) +tryInsert unif indir tree z@(x, info) = case justLookupTree x tree of + Ok info0 -> do + info1 <- unif info info0 + return $ updateTree (x,info1) tree + _ -> return $ updateTree (x,indir info) tree + +{- ---- +case tree of + NT -> return $ BT (x, indir info) NT NT + BT c@(a,info0) left right + | x < a -> do + left' <- tryInsert unif indir left z + return $ BT c left' right + | x > a -> do + right' <- tryInsert unif indir right z + return $ BT c left right' + | x == a -> do + info' <- unif info info0 + return $ BT (x,info') left right +-} + +--- addToMaybeList m c = maybe (return c) (\old -> return (c ++ old)) m + +unifAbsDefs :: Perh Term -> Perh Term -> Err (Perh Term) +unifAbsDefs p1 p2 = case (p1,p2) of + (Nope, _) -> return p2 + (_, Nope) -> return p1 + (Yes (Eqs bs), Yes (Eqs ds)) -> return $ yes $ Eqs $ bs ++ ds --- order! + _ -> Bad "update conflict for definitions" + +unifConstrs :: Perh [Term] -> Perh [Term] -> Err (Perh [Term]) +unifConstrs p1 p2 = case (p1,p2) of + (Nope, _) -> return p2 + (_, Nope) -> return p1 + (Yes bs, Yes ds) -> return $ yes $ bs ++ ds + _ -> Bad "update conflict for constructors" |
