🚀 go-pugleaf

In article <AKAMA.97Jan9005151@siena.is.s.u-tokyo.ac.jp> akama@is.s.u-tokyo.ac.jp (AKAMA Youji) writes:

> $B$5$F!"(B $B%W%m%0%i%`8@8l(B $B#M#L$K$*$$$F!"(B exception, raise, handle $B$J$I$G!"(B
> call/cc $B$_$?$$$J$3$H$,$I$l$/$i$$$G$-$k$N$G$7$g$&$+!)(B

> $B!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a(B
> Standard ML of New Jersey, Version 0.93, February 15, 1993
> val it = () : unit
> - val Peirce = fn w => let exception y of '1a in  w(fn z => raise(y z)) handle (y x) => x end;
> val Peirce = fn : (('1a -> 'b) -> '1a) -> '1a

> $B$3$l$G$O(B call/cc $B$h$j<e$$$G$9$+$M!#!#!#!#(B

$BNc30=hM}$H7QB3$H!"$I$A$i$b(Bgoto$B$_$?$$$J$3$H$,<BAu$G$-$k$+$i;w$F$$$k$H$$(B
$B$($P!";w$F$$$^$9$1$l$I$b!"0c$$$NJ}$,Bg$-$$$H;W$$$^$9!#(B

$BNc$($P!"(B($B0lEy(B)$B7QB3$O2?2s$G$b<B9T$G$-$k$,!">e$N$h$&$J$d$jJ}$G$O0l2s$7$+(B
$B$G$-$^$;$s!#Fs2sL\$O(Btoplevel$B$N%(%i!<$K$J$j$^$9!#A0$K$7$+%8%c%s%W$G$-$J(B
$B$$$h$&$J(Bgoto$B$G$9!#7?$,(Bcallcc$B$H;w$F$$$F$b!"I=8=NO$,0c$$$^$9!#(B

$B7QB3$K$D$$$FJ88%$O;3$[$I$"$j$^$9!#I=8=NO$KFC$K6=L#$,$"$k$J$i$P!"(BPOPL94
$B$N(BAndreszj Filinzki$B$NO@J8(B($B%b%J%I$r7QB3$G%(%s%3!<%I$9$k$b$N(B)$B$,LLGr$$$G(B
$B$7$g$&!#(B

	$B<f6j(B
-- 
---------------------------------------------------------------------------
Jacques Garrigue	Kyoto University      garrigue@kurims.kyoto-u.ac.jp
     <A HREF="http://wwwfun.kurims.kyoto-u.ac.jp/~garrigue/">JG</A>

$B<f6j(B $BMM!"(B

$B%U%)%m!<$"$j$,$H$&$4$6$$$^$7$?!#(B

$B<f6j(B> In article <AKAMA.97Jan9005151@siena.is.s.u-tokyo.ac.jp> akama@is.s.u-tokyo.ac.jp (AKAMA Youji) writes:
>> $B$5$F!"(B $B%W%m%0%i%`8@8l(B $B#M#L$K$*$$$F!"(B exception, raise, handle $B$J$I$G!"(B
>> call/cc $B$_$?$$$J$3$H$,$I$l$/$i$$$G$-$k$N$G$7$g$&$+!)(B

>> $B!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a!a(B
>> Standard ML of New Jersey, Version 0.93, February 15, 1993
>> val it = () : unit
>> - val Peirce = fn w => let exception y of '1a in  w(fn z => raise(y z)) handle (y x) => x end;
>> val Peirce = fn : (('1a -> 'b) -> '1a) -> '1a

>> $B$3$l$G$O(B call/cc $B$h$j<e$$$G$9$+$M!#!#!#!#(B

$B<f6j(B> $BNc30=hM}$H7QB3$H!"$I$A$i$b(Bgoto$B$_$?$$$J$3$H$,<BAu$G$-$k$+$i;w$F$$$k$H$$(B
$B<f6j(B> $B$($P!";w$F$$$^$9$1$l$I$b!"0c$$$NJ}$,Bg$-$$$H;W$$$^$9!#(B

$B<f6j(B> $BNc$($P!"(B($B0lEy(B)$B7QB3$O2?2s$G$b<B9T$G$-$k$,!">e$N$h$&$J$d$jJ}$G$O0l2s$7$+(B
$B<f6j(B> $B$G$-$^$;$s!#Fs2sL\$O(Btoplevel$B$N%(%i!<$K$J$j$^$9!#A0$K$7$+%8%c%s%W$G$-$J(B
$B<f6j(B> $B$$$h$&$J(Bgoto$B$G$9!#7?$,(Bcallcc$B$H;w$F$$$F$b!"I=8=NO$,0c$$$^$9!#(B

$B$O$$!"$=$&$G$9$M!#(B

$B<f6j(B> $B7QB3$K$D$$$FJ88%$O;3$[$I$"$j$^$9!#I=8=NO$KFC$K6=L#$,$"$k$J$i$P!"(BPOPL94
$B<f6j(B> $B$N(BAndreszj Filinzki$B$NO@J8(B($B%b%J%I$r7QB3$G%(%s%3!<%I$9$k$b$N(B)$B$,LLGr$$$G(B
$B<f6j(B> $B$7$g$&!#(B

$B>pJs$r$"$j$,$H$&$4$6$$$^$9!#$H$j$"$($:!"%j%W%i%$$r$7$^$9!#<ALd$NGX7J$O!"(B
$B0J2<$NO@J8(B

Carl Gunter, Didier Remy, and Jon Riecke. A generalization of
exceptions and control in ML-like languages. In: 1995 Conference on
Functional Programming and Computer Architecture, pages 12-23, La
Jolla, CA, June 1995.

$B$K(B

	It is folklore (the authors know of no published proof) that
	neither exceptions nor continuations can be expressed as a
	macro in terms of the other (at least if no references are
	present), even though they are closely relate.

$B$H$"$k$3$H$rCN$C$F!"!V$3$l$ONI$/CN$i$l$?(B folklore $B$J$N$+!)(B $B$9$0M}2r$G(B
$B$-$k$b$N$+!)!W$H;d$,<aA3$H$7$J$+$C$?$3$H$G$9!#;d$O!";DG0$J$,$i(B call/cc 
$B$H$+(B exception $B$H$+$,?H$K$7$_$F$^$;$s!#(B

$B!V(BCall/cc $B$,(B exception $B$r;H$C$F%^%/%m$H$7$FDj5A$G$-$J$$!W$O!"%U%)!<%^(B
$B%k$J>ZL@$O$o$+$j$^$;$s!#$G$b!"<f6j$5$s$N@bL@$G>/$746$8$,$o$+$j$^$7$?!#(B
$B$"$j$,$H$&$4$6$$$^$7$?!#(B

$B$=$N5U!V(BException $B$,(B Call/cc $B$r;H$C$F%^%/%m$H$7$FDj5A$G$-$J$$!W$K$D$$(B
$B$F$O!"OC$7$,HyL/$G$9!#(B

$B!z(B Mark Lillibridge, Exceptions Are Strictly More Powerful Than
Call/CC, CMU-CS-95-178 $B!J$A$g$C$HFI$s$@!K!'(B

	$B$=$N$3$H$r%U%)!<%^%k$K>ZL@$7$F$$$k$3$H$K$J$C$F$$$k!#(B
	
$B!z(B $B<f6j$5$s$,>R2p$7$?(B Filinski $B$NO@J8!J$A$g$C$H8+$?!K!'(B

	Exception (Monad) $B$r7QB3$K4X$9$k1i;;(B(shift, reset; call/cc $B$h(B
	$B$j4qNo$@$H$$$o$l$F$$$k(B)$B$J$I$G(B represent $B$7$F$$$k!#(B

$B$&!<$`!#>pJs!&0l2H8@$r$*;}$A$NJ}$O$h$m$7$/65$($F2<$5$$!#(B

						$B@V4V(B
-----------------------------------------------------------------------------
 $BEl5~Bg3X(B $BBg3X1!M}3X7O8&5f2J(B $B>pJs2J3X@l96(B $B@V4VM[Fs(B
$B")#1#1#3(B $BEl5~ET(B $BJ85~6h(B $BK\6?#7!]#3!]#1(B $BM}3XIt>pJs2J3X2J(B
$B!J#T#E#L!K#0#3(B($B#3#8#1#2(B)$B#4#1#7#7!JD>DL!K!"#0#3(B($B#3#8#1#2(B)$B#2#1#1#1!JFb@~#4#0#9#7!K!"(B
$B!J#F#A#X!K#0#3(B($B#3#8#1#8(B)$B#1#0#7#3(B
--

-----------------------------------------------------------------------------
 $BEl5~Bg3X(B $BBg3X1!M}3X7O8&5f2J(B $B>pJs2J3X@l96(B $B@V4VM[Fs(B
$B")#1#1#3(B $BEl5~ET(B $BJ85~6h(B $BK\6?#7!]#3!]#1(B $BM}3XIt>pJs2J3X2J(B
$B!J#T#E#L!K#0#3(B($B#3#8#1#2(B)$B#4#1#7#7!JD>DL!K!"#0#3(B($B#3#8#1#2(B)$B#2#1#1#1!JFb@~#4#0#9#7!K!"(B
$B!J#F#A#X!K#0#3(B($B#3#8#1#8(B)$B#1#0#7#3(B

$BN6C+$NCfLn$H?=$7$^$9!#(B

In article <AKAMA.97Jan23162641@amsterdam.is.s.u-tokyo.ac.jp>
akama@is.s.u-tokyo.ac.jp (AKAMA Youji) writes:
 |	  It is folklore (the authors know of no published proof) that
 |	  neither exceptions nor continuations can be expressed as a
 |	  macro in terms of the other (at least if no references are
 |	  present), even though they are closely relate.
 |
 |$B$H(B$B$"$k$3$H$rCN$C$F!"!V$3$l$ONI$/CN$i$l$?(B folklore $B$J$N$+!)(B $B$9$0M}2r$G(B
 |$B$-$k$b$N$+!)!W$H;d$,<aA3$H$7$J$+(B$B$C$?$3$H$G$9!#;d$O!";DG0$J$,$i(B call/cc 
 |$B$H$+(B exception $B$H$+$,?H$K$7$_$F$^$;$s!#(B

$B@N$N(B LISP $B$G$h$/;H(B$B$o$l$F$$$^$7$?$,!"JQ?t$N(B dynamic binding $B$H$$$&(B
semantics $B$,$"$j$^$9$h$M!#(B Common Lisp $B$@$HIaDL$O(B lexical binding $B$G$9(B
$B$,!"JQ?t$r(B special $B$H@k8@$9$k$H!"$?$7$+$3$N(B dynamic binding $B$b$G$-$k$s(B
$B$8$c$J$+$C$?(B$B$+$H;W$$$^$9!#(B

call/cc $B$H(B ML $B$N(B raise/handle $B$,$I$&$b$&$^$/3z$_9g$o$J$$$N$O!"(Bcall/cc
$B$,(B lexical binding $B$J$N$KBP$7!"0lJ}$N(B raise/handle $B$,(B dynamic binding
$B$@$+$i$G$9!#(B

$B$3$NFs$D$N0c$$$O(B scope $B$H(B extent $B$H$$$&8@MU$GI=8=$5$l$k$3$H$,B?$$$h$&(B
$B$G$9$,!"(Bcall/cc $B$H(B raise/handle $B$N4X78$rI=$K$9$k$H!"(B

                               \         $B7QB3(B $B$N(B scope         |
                                \---------------+--------------+
        $B7QB3$N(B extent            \    lexical   |  indefinite  |
                                  \   (static)  |   (dynamic)  |
 ----------------------------------+------------+--------------+
      indefinite (static)          |            |              |
  ($B$=$N7QB3$X$N;2>H$r;D(B$B$7$F$*$1$P(B  |  call/cc   |     ???      |
   $B$$$D$G$b$=$l$r;HMQ$G$-$k(B)       |            |              |
 ----------------------------------+------------+--------------+
             dynamic               |            |              |
  ($B7QB3$r@8@.$7$?9=J8MWAG$N<B9T$,(B  |    ???     | raise/handle |
   $B=*N;$7$?$i$=$N<BBN$O<:$o$l$k(B)   |            |              |
 ----------------------------------+------------+--------------+

$B$_$?$$$J46$8$8$c$J$$$+$H;W$$$^$9!#(B $BEvA3Fs$D$N(B ??? $B$KBP1~$9$k$b$N$bM-$j(B
$BF@(B$B$k$o$1$G$9$,!"$I$&$b(B($BH>(B)$B<BMQE*$J%W%m%0%i%_%s%08@8l$G$O$"$^$j8+$+$1$^(B
$B$;$s$M!#(B Common Lisp $B$N(B catch/throw $B$b(B dynamic binding $B$G$9$7!#(B

# C $B$N(B setjmp/longjmp $B$J$s$+$O!"(Blexical scope + dynamic extent $B$H9M$((B
# $B$k$3$H$b$G$-$^$9$,(B ... $B$D$$$G$K(B break $B$J$s$+$b(B ... :-)

 |$B$=$N5U!V(BException $B$,(B Call/cc $B$r;H$C$F%^%/(B$B%m$H$7$FDj5A$G$-$J$$!W$K$D$$(B
 |$B$F$O!"OC$7$,HyL/$G$9!#(B
 |
 |$B!z(B Mark Lillibridge, Exceptions Are Strictly More Powerful Than
 |Call/CC, CMU-CS-95-178 $B!J$A$g$C$HFI$s$@!K!'(B
 |
 |	  $B$=$N$3$H$r%U%)!<%^%k$K>ZL@(B$B$7$F$$$k$3$H$K$J$C$F$$$k!#(B

$B$3$NO@J8$G$OFC$K?($l$F$$$J$+$C$?$+$bCN$l$^$;$s$,!"!V(BException $B$,(B
Call/cc $B$r;H$C$F%^%/%m$H$7$FDj5A$G$-$J$$!W$N$O!"7QB3$N(B extent $B$N0c$$$O(B
$B$I$&$G$b$h$/$F!"(Bscoping $B$,(B static $B$+(B dynamic $B$+$N0c$$$NJ}$,K\<AE*$G$9!#(B
scoping $B$,F1$8$J$i!"EvA3(B call/cc $B$G(B raise/handle $B$,=q$1$^$9!#(B

# $B$=(B$B$&$$$($P!"(Bcall/cc $B$,(B raise/handle $B$r;H$C$?%^%/%m$G=q$1$J$$$H$$$&$N(B
# $B$O!"$I$&<($7$?$i$$$$$s$G$7$g$&$+(B??? $B$b$C$H$b!"$=$N$?$a$K$O!"%^%/%m$H(B
# $B$O2?$+!"$H$$$&$3$H$r$O$C$-$j$5$;$H$+$J$$$H$$$1$J$$$s$G$7$g$&$1(B$B$l$I(B ...
--
			Hiroshi Nakano <nakano@rins.ryukoku.ac.jp>
			Ryukoku Univ., Seta, Otsu, Japan

$B;{ED$C$F$$$$$^$9(B. 

$B$^$:$O$8$a$K(B. 
call/cc $B$C$F(B scheme $B$K$"$k%"%l$N$3$H$G$9$h$M(B?

In article <NAKANO.97Jan27151745@dugong37.math.ryukoku.ac.jp>
	nakano@math.ryukoku.ac.jp (Hiroshi NAKANO) writes:
> 
> call/cc $B$H(B ML $B$N(B raise/handle $B$,$I$&$b$&$^$/3z$_9g$o$J$$$N$O!"(Bcall/cc
> $B$,(B lexical binding $B$J$N$KBP$7!"0lJ}$N(B raise/handle $B$,(B dynamic binding
> $B$@$+$i$G$9!#(B
> 
> $B$3$NFs$D$N0c$$$O(B scope $B$H(B extent $B$H$$$&8@MU$GI=8=$5$l$k$3$H$,B?$$$h$&(B
> $B$G$9$,!"(Bcall/cc $B$H(B raise/handle $B$N4X78$rI=$K$9$k$H!"(B

$B$9$$$^$;$s(B, $B!V7QB3$N(B scope$B!W$H$$$&35G0$,(B
$B$h$/$o$+$i$J$$$b$s$G(B, $B$I$J$?$+$A$g$C$H@bL@$7$F$$$?$@$1$^$;$s$G$7$g$&$+(B. 
$B%^%5%+(B call/cc $B$KM?$($?4X?t$N0z?t$H$7$F(B continuation $B$,<j$KF~$k$+$i(B, 
$B$H$$$&$N$8$c$J$$$G$9$h$M(B?
$B$?$7$+$K$=$N0z?t$H(B continuation $B$N(B binding $B$O@EE*$G$9$1$l$I(B. 

object $B$H$7$F$N(B continuation $B$K$O!VL>A0!W$,$J$$$+$i(B, 
scope $B$C$F$N$,9M$($i$l$J$$(B...

In article <TERADA.97Feb4175227@kirin.sanpo.t.u-tokyo.ac.jp>
terada@sanpo.t.u-tokyo.ac.jp (Minoru TERADA) writes:
 |$B$^$:$O$8$a$K(B. 
 |call/cc $B$C$F(B scheme $B$K$"$k%"%l$N$3$H$G$9$h$M(B?

$B$O(B$B$$!#(B

 |$B$9$$$^$;$s(B, $B!V7QB3$N(B scope$B!W$H$$$&35G0$,(B
 |$B$h$/$o$+$i$J$$$b$s$G(B, $B$I$J$?$+$A$g$C$H@bL@$7$F$$$?(B$B$@$1$^$;$s$G$7$g$&$+(B. 

$B!V7QB3$N(B scope$B!W$d!V7QB3$N(B extent $B!W$H$$$&8@$$J}$,NI$/$J$+$C$?$G$9$M!#(B
$B@53N$K(B$B$O!V7QB3$r;2>H$9$kJQ?t(B($BL>A0$N(B) scope$B!W$d!V(Bextent$B!W$G$9!#(B

 |$B%^%5%+(B call/cc $B$KM?$($?4X?t$N0z?t$H$7$F(B continuation $B$,<j$KF~$k$+$i(B, 
 |$B$H$$$&$N$8$c$J$$$G$9$h$M(B?

$B$(!<$H!"7k6I$O$=$N$H$*$j$G$9!#(B Scheme $B$@$H7Q(B$BB3$r;2>H$9$kJQ?t$bB>$NJQ?t(B
$B$HF1$8$h$&$K07$o$l$^$9$+$i(B ...

$BCfLn$,!V(Bscope$B!W$H8@$C$?$H$-$K$O!"(B Scheme $B$N(B

	(call/cc (lambda (f) (... (f x) ...)))

$B$r(B SML $B$N(B

	(... (raise (f x)) ...) handle (f x) => x

$B$HBP(B$BHf$7$F!"$3$N(B f $B$N(B scope (raise $B$N(B f $B$H(B handle $B$N(B f $B$H$NBP1~(B)$B$r9M$((B
$B$F$$$^$7$?!#(B Scheme $B$G$O(B($BIaDL$NJQ?t(B$B$G$9$+$i(B) lexical scope $B$r;}$C$F$$(B
$B$^$9$,!"0lJ}(B SML $B$G$O(B indefinite (dynamic) scope $B$K$J$j$^$9!#(B $B$?$H(B$B$($P!"(B

  (call/cc (lambda (f)
              (+ 1 ((lambda (g) (call/cc (lambda (f) (g 1))))
                      (lambda (y) (f y))))))

$B$r(B Scheme $B$GI>2A$9$k$H(B 1 $B$K$J$j$^$9$,!"$3$l$KBP1~$9$k(B

  let exception f of int in
      (1 + (fn g => g(1) handle (f x) => x)(fn y => raise(f y)))
                                                handle (f x) => x end

$B$r(B SML $B$GI>2A$9$k$H(B 2 $B$K$J$C$F$7$^$$$^$9!#(B $B$b$A(B$B$m$s$3$N;~!"(B extent $B$N(B
$BJ}$N0c$$$O4X78$7$F$$$^$;$s!#(B

# $B$A$c$s$H@bL@$G$-$?$+$I$&$+$A$g$C$HIT0B$G$9(B ... ;-<
--
			Hiroshi Nakano <nakano@rins.ryukoku.ac.jp>
			Ryukoku Univ., Seta, Otsu, Japan

In article <NAKANO.97Feb5104254@orca20.math.ryukoku.ac.jp> nakano@math.ryukoku.ac.jp (Hiroshi NAKANO) writes:

 | 中野が「scope」と言ったときには、 Scheme の
 | 
 | 	(call/cc (lambda (f) (... (f x) ...)))
 | 
 | を SML の
 | 
 | 	(... (raise (f x)) ...) handle (f x) => x
 | 
 | と対比して、この f の scope (raise の f と handle の f との対応)を考え
 | ていました。 Scheme では(普通の変数ですから) lexical scope を持ってい
 | ますが、一方 SML では indefinite (dynamic) scope になります。 たとえば、
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^
indefiniteがSchemeでdynamicがSMLなのではありませんか?

 |   (call/cc (lambda (f)
 |               (+ 1 ((lambda (g) (call/cc (lambda (f) (g 1))))
 |                       (lambda (y) (f y))))))
 | 
 | を Scheme で評価すると 1 になりますが、これに対応する
 | 
 |   let exception f of int in
 |       (1 + (fn g => g(1) handle (f x) => x)(fn y => raise(f y)))
 |                                                 handle (f x) => x end
 | 
 | を SML で評価すると 2 になってしまいます。 もちろんこの時、 extent の
 | 方の違いは関係していません。
 | 
 | # ちゃんと説明できたかどうかちょっと不安です ... ;-<

Schemeの継続がindefinite extentを持つことを示したいのであれば、

> (define ccc #f)

> (+ 1 (call/cc (lambda (cc)
                  (set! ccc cc)
                  (+ 2 (cc 3)))))
4

> (ccc 100)
101

で十分でないでしょうか?一方、Common Lispですと(SMLで書けないものですみ
ません)、

>(+ 1 (catch 'tag (progn (+ 2 (throw 'tag 3)))))
4

>(throw 'tag 100)

Error: TAG is an undefined tag.
Fast links are on: do (si::use-fast-links nil) for debugging
Error signalled by EVAL.
Broken at THROW.  Type :H for Help.
>>

のようなかんじでdynamicである、というのが言えると思います。

;;; $Id: .signature,v 1.3 1996/09/20 07:06:26 kuroda Exp $
(define (くろだ ひさお)
  (email kuroda@msi.co.jp))

$B;{ED$C$F$$$$$^$9(B. 

$B$4@bL@$"$j$,$H$&$4$6$$$^$7$?(B. 
$BB4O@$GK;$7$/$C$F$*JV;v$,CY$l$^$7$?(B. 

In article <NAKANO.97Feb5104254@orca20.math.ryukoku.ac.jp>
	nakano@math.ryukoku.ac.jp (Hiroshi NAKANO) writes:
>  |$B$9$$$^$;$s(B, $B!V7QB3$N(B scope$B!W$H$$$&35G0$,(B
>  |$B$h$/$o$+$i$J$$$b$s$G(B, $B$I$J$?$+$A$g$C$H@bL@$7$F$$$?$@$1$^$;$s$G$7$g$&$+(B. 
> 
> $B!V7QB3$N(B scope$B!W$d!V7QB3$N(B extent $B!W$H$$$&8@$$J}$,NI$/$J$+$C$?$G$9$M!#(B
> $B@53N$K$O!V7QB3$r;2>H$9$kJQ?t(B($BL>A0$N(B) scope$B!W$d!V(Bextent$B!W$G$9!#(B
> 
>  |$B%^%5%+(B call/cc $B$KM?$($?4X?t$N0z?t$H$7$F(B continuation $B$,<j$KF~$k$+$i(B, 
>  |$B$H$$$&$N$8$c$J$$$G$9$h$M(B?
> 
> $B$(!<$H!"7k6I$O$=$N$H$*$j$G$9!#(B Scheme $B$@$H7QB3$r;2>H$9$kJQ?t$bB>$NJQ?t(B
> $B$HF1$8$h$&$K07$o$l$^$9$+$i(B ...

$B$J$k$[$I(B, $B$=$&$$$&$3$H$G$7$?$+(B. 

> $BCfLn$,!V(Bscope$B!W$H8@$C$?$H$-$K$O!"(B Scheme $B$N(B
> 
> 	(call/cc (lambda (f) (... (f x) ...)))
> 
> $B$r(B SML $B$N(B
> 
> 	(... (raise (f x)) ...) handle (f x) => x
> 
> $B$HBPHf$7$F!"$3$N(B f $B$N(B scope (raise $B$N(B f $B$H(B handle $B$N(B f $B$H$NBP1~(B)$B$r9M$((B
> $B$F$$$^$7$?!#(B Scheme $B$G$O(B($BIaDL$NJQ?t$G$9$+$i(B) lexical scope $B$r;}$C$F$$(B
> $B$^$9$,!"0lJ}(B SML $B$G$O(B indefinite (dynamic) scope $B$K$J$j$^$9!#(B $B$?$H$($P!"(B
> 
>   (call/cc (lambda (f)
>               (+ 1 ((lambda (g) (call/cc (lambda (f) (g 1))))
>                       (lambda (y) (f y))))))
> 
> $B$r(B Scheme $B$GI>2A$9$k$H(B 1 $B$K$J$j$^$9$,(B

$B$3$l$O(B, 2$B9TL\$N(B (g 1) -- $B$3$N$H$-(B g $B$O(B 3$B9TL\$N(B (lambda(y)(f y))
$B$H$$$&4X?t(B($B%/%m!<%8%c(B)$B$K$J$C$F$k(B -- $B$NI>2A$K$h$C$F;H$o$l$k(B f $B$,(B, 
2$B9TL\$N(B f $B$G$J$/(B 1$B9TL\$N(B f $B$K$J$C$F$$$k(B, $B$H$$$&$3$H$G$9$M(B. 

$B$3$l$OMW$9$k$K(B scheme $B$O(B static scope $B$G$"$k(B, $B$H$$$&$3$H$G$9$M(B. 

> $B$3$l$KBP1~$9$k(B
> 
>   let exception f of int in
>       (1 + (fn g => g(1) handle (f x) => x)(fn y => raise(f y)))
>                                                 handle (f x) => x end
> 
> $B$r(B SML $B$GI>2A$9$k$H(B 2 $B$K$J$C$F$7$^$$$^$9!#(B

$B$3$A$i$O(B, raise(f y) $B$K$h$k(B f $B$,(B, 2$B9TL\$N(B($B$9$J$o$AFbB&$N(B) handle $B$G(B
$BDj5A(B($B$C$F$$$&$N$+$J(B?)$B$5$l$?(B f $B$H$J$k(B, $B$H$$$&$3$H$J$o$1$G$9$M(B. 

$B$?$7$+$K(B dynamic bind $B$HF1$8?6Iq$$$r$7$^$9$M(B. 
($B$^$"$=$b$=$b(B exception handling $B$,@EE*$K$7$+$G$-$J$$$H$$$&$N$O(B
 $B$[$H$s$I0UL#$J$$$G$9$+$i$M(B. $B$=$l$J$i(B goto $B$G==J,(B)

$B$G(B, $B0c$$$O$o$+$C$?$N$G$9$,(B, $BKM$K$O$=$l$,%9%3!<%W$N0c$$$+(B, 
$B$H$$$&$N$,%$%^%$%A$O$C$-$j$7$J$$$N$G$9(B. 

$B$`$7$m(B, continuation $B$O$=$l<+?H$OL>A0$r;}$?$J$$(B object $B$G$"$k$N$K(B
$BBP$7$F(B, (SML $B$O<B$ONI$/CN$i$J$$$N$@$1$l$I(B) exception $B$H$$$&$N$O(B
$BFHN)$7$?L>A06u4V$r$b$C$F(B, $B$=$l$K$h$C$F$@$1;2>H$5$l$k$h$&$J(B
$B$b$N$J$N$@(B, $B$H$$$&$N$,0c$$$J$N$8$c$J$$$G$7$g$&$+(B. 

 continuation $B$H$$$&5!9=$=$N$b$N$K$O(B scope $B$H$$$&(B
$B35G0$O$D$$$F$J$/$F(B, $B$?$^$?$^(B scheme $B$NJQ?t$N(B binding $B$,(B
$B@EE*$G$"$C$?$@$1$G$O$J$$$G$9$+$M$((B. 

dynamic bind $B$N(B lisp $B$N>e$K$@$C$F(B continuation $B$O9M$($i$l$k$N$G$O(B
$B$J$$$G$7$g$+(B. deep bind $B$N8EE5E*$J(B evaluator $B$@$C$?$i(B, $BJQ?t4D6-$N(B
a-list $B$H(B $B%3%s%H%m!<%k%9%?%C%/$N%Z%"$r$b$F$P==J,$+$J(B?

$B$^$?5U$K(B, scheme $B$G$b(B, $BFHN)$7$?L>A06u4V(B($B$C$?$C$FBg0hE*$J(B a-list $B$G(B
$B==J,$G$9$1$l$I(B)$B$r:n$C$F$d$l$P(B, dynamic $B$J?6Iq$$$r$5$;$i$l$k$o$1$G(B. 

...

$B$H9M$($k$H(B, continuation $B$NJ}$,%(%i%$(B, $B$H$$$&(B, $B0JA0$N5-;v(B
(CMU $B$N(B Technical Report $B$G$7$?$C$1(B?)$B$H$O5U$N7kO@$,=P$F$7(B
$B$^$&$N$,$I$&$b(B...

SMLのraise/handleって，型付であることを除けばCommon Lispのcatch/throw
とおんなじ能力ではないでしょうか．

(defmacro handle (tag value handler)
  `(block #1=#:not-thrown
     (catch ,tag
       (return-from #1# ,value))
     ,handler))

(handle 'f
  ;; value
  (+ 1 ((lambda (g) (handle 'f
                            (funcall g 1)
                            1))
        #'(lambda (y) (throw 'f y))))
  ;; handler
  1)
=> 2

つまり，SMLのhandlerはCLのcatcherと同じくindefinite scope/dynamic extent
ですよね．

continuationはindefinite extentですから，SMLのraise/handleを用いて
call/ccを書くことはできません．

逆に，call/ccを使ってraise/handleやcatch/throwのようなものを書くことが
できるかというと，indefinite scope/dynamic extentの名前空間を持つ変数
というのが標準Schemeにはないので，簡単には書けない，という話では．

(実際にはfluid-letとか，call-with-dynamic-bindingなどという構文を持っ
た処理系が多いし，こういう構文がライブラリとして配布されているので，割
と簡単に書けてしまうでしょうけど．)

>>>>> "terada" == Minoru TERADA <terada@sanpo.t.u-tokyo.ac.jp> writes:
In article <TERADA.97Feb7215216@kirin.sanpo.t.u-tokyo.ac.jp> terada@sanpo.t.u-tokyo.ac.jp (Minoru TERADA) writes:

    terada> dynamic bind の lisp の上にだって continuation は考えられるのでは
    terada> ないでしょか. deep bind の古典的な evaluator だったら, 変数環境の
    terada> a-list と コントロールスタックのペアをもてば十分かな?

a-listとスタックポインタで，catcherのような制限付の(dynamic extentを持
つ)continuationになりますね．first class continuationを作るには，スタッ
ク全体をコピーするとか，スタックをもともとヒープ上にとっておくとか，結
局Schemeと同じようなことをすることになりますが．

    terada> また逆に, scheme でも, 独立した名前空間(ったって大域的な a-list で
    terada> 十分ですけれど)を作ってやれば, dynamic な振舞いをさせられるわけで. 

ですね．

    terada> と考えると, continuation の方がエライ, という, 以前の記事
    terada> (CMU の Technical Report でしたっけ?)とは逆の結論が出てし
    terada> まうのがどうも...

でも，MLやLispでfirst class continuationを実現するには，言語の意味論自
体を拡張する必要があって，MLやLisp自身では書けないと思いますが．

				前田敦司

自己フォローです．

    mad> (defmacro handle (tag value handler)
    mad>   `(block #1=#:not-thrown
    mad>      (catch ,tag
    mad>        (return-from #1# ,value))
    mad>      ,handler))

    mad> (handle 'f
    mad>   ;; value
    mad>   (+ 1 ((lambda (g) (handle 'f
    mad>                             (funcall g 1)
    mad>                             1))
    mad>         #'(lambda (y) (throw 'f y))))
    mad>   ;; handler
    mad>   1)
    mad> => 2

これ，

(defmacro handle (tag value handler)
  `(block #1=#:not-thrown
     (multiple-value-call ,handler
      (catch ,tag
       (return-from #1# ,value)))))

(handle 'f
  ;; value
  (+ 1 ((lambda (g) (handle 'f
                            (funcall g 1)
                            #'(lambda (x) x)))
        #'(lambda (y) (throw 'f y))))
  ;; handler
  #'(lambda (x) x))

とした方が感じ出てますね．

				前田敦司