(* Title: Pure/General/lazy_scan.ML ID: $Id: lazy_scan.ML,v 1.3 2005/04/07 07:26:18 wenzelm Exp $ Author: Sebastian Skalberg, TU Muenchen Scanner combinators for lazy sequences. *) signature LAZY_SCAN = sig exception SyntaxError type ('a,'b) scanner = 'a LazySeq.seq -> 'b * 'a LazySeq.seq val :-- : ('a,'b) scanner * ('b -> ('a,'c) scanner) -> ('a,'b*'c) scanner val -- : ('a,'b) scanner * ('a,'c) scanner -> ('a,'b*'c) scanner val >> : ('a,'b) scanner * ('b -> 'c) -> ('a,'c) scanner val --| : ('a,'b) scanner * ('a,'c) scanner -> ('a,'b) scanner val |-- : ('a,'b) scanner * ('a,'c) scanner -> ('a,'c) scanner val ^^ : ('a,string) scanner * ('a,string) scanner -> ('a,string) scanner val || : ('a,'b) scanner * ('a,'b) scanner -> ('a,'b) scanner val one : ('a -> bool) -> ('a,'a) scanner val succeed : 'b -> ('a,'b) scanner val any : ('a -> bool) -> ('a,'a list) scanner val any1 : ('a -> bool) -> ('a,'a list) scanner val optional : ('a,'b) scanner -> 'b -> ('a,'b) scanner val option : ('a,'b) scanner -> ('a,'b option) scanner val repeat : ('a,'b) scanner -> ('a,'b list) scanner val repeat1 : ('a,'b) scanner -> ('a,'b list) scanner val ahead : ('a,'b) scanner -> ('a,'b) scanner val unless : ('a LazySeq.seq -> bool) -> ('a,'b) scanner -> ('a,'b) scanner val $$ : ''a -> (''a,''a) scanner val !! : ('a LazySeq.seq -> string) -> ('a,'b) scanner -> ('a,'b) scanner val scan_full: ('a,'b) scanner -> 'a LazySeq.seq -> 'b LazySeq.seq end structure LazyScan :> LAZY_SCAN = struct infix 7 |-- --| infix 5 :-- -- ^^ infix 3 >> infix 0 || exception SyntaxError exception Fail of string type ('a,'b) scanner = 'a LazySeq.seq -> 'b * 'a LazySeq.seq fun (sc1 :-- sc2) toks = let val (x,toks2) = sc1 toks val (y,toks3) = sc2 x toks2 in ((x,y),toks3) end fun (sc1 -- sc2) toks = let val (x,toks2) = sc1 toks val (y,toks3) = sc2 toks2 in ((x,y),toks3) end fun (sc >> f) toks = let val (x,toks2) = sc toks in (f x,toks2) end fun (sc1 --| sc2) toks = let val (x,toks2) = sc1 toks val (_,toks3) = sc2 toks2 in (x,toks3) end fun (sc1 |-- sc2) toks = let val (_,toks2) = sc1 toks in sc2 toks2 end fun (sc1 ^^ sc2) toks = let val (x,toks2) = sc1 toks val (y,toks3) = sc2 toks2 in (x^y,toks3) end fun (sc1 || sc2) toks = (sc1 toks) handle SyntaxError => sc2 toks fun one p toks = case LazySeq.getItem toks of NONE => raise SyntaxError | SOME(t,toks) => if p t then (t,toks) else raise SyntaxError fun succeed e toks = (e,toks) fun any p toks = case LazySeq.getItem toks of NONE => ([],toks) | SOME(x,toks2) => if p x then let val (xs,toks3) = any p toks2 in (x::xs,toks3) end else ([],toks) fun any1 p toks = let val (x,toks2) = one p toks val (xs,toks3) = any p toks2 in (x::xs,toks3) end fun optional sc def = sc || succeed def fun option sc = (sc >> SOME) || succeed NONE (* fun repeat sc = let fun R toks = let val (x,toks2) = sc toks val (xs,toks3) = R toks2 in (x::xs,toks3) end handle SyntaxError => ([],toks) in R end *) (* A tail-recursive version of repeat. It is (ever so) slightly slower * than the above, non-tail-recursive version (due to the garbage generation * associated with the reversal of the list). However, this version will be * able to process input where the former version must give up (due to stack * overflow). The slowdown seems to be around the one percent mark. *) fun repeat sc = let fun R xs toks = case SOME (sc toks) handle SyntaxError => NONE of SOME (x,toks2) => R (x::xs) toks2 | NONE => (List.rev xs,toks) in R [] end fun repeat1 sc toks = let val (x,toks2) = sc toks val (xs,toks3) = repeat sc toks2 in (x::xs,toks3) end fun ahead (sc:'a->'b*'a) toks = (#1 (sc toks),toks) fun unless test sc toks = let val test_failed = (test toks;false) handle SyntaxError => true in if test_failed then sc toks else raise SyntaxError end fun $$ arg = one (fn x => x = arg) fun !! f sc toks = (sc toks handle SyntaxError => raise Fail (f toks)) fun scan_full sc toks = let fun F toks = if LazySeq.null toks then NONE else let val (x,toks') = sc toks in SOME(x,LazySeq.make (fn () => F toks')) end in LazySeq.make (fn () => F toks) end end