如何从多个嵌套函数中丢弃分隔的延续?

How to discard a delimited continuation from within multiple nested functions?

我研究定界延续,目前正在尝试丢弃它们以获得类似于引发异常的效果。

这是给我带来麻烦的原因:

const structure = type => cons => {
  const f = (f, args) =>
   ({["run" + type]: f, [Symbol.toStringTag]: type, [Symbol("args")]: args});

  return cons(f);
};

const Cont = structure("Cont")
  (Cont => f => Cont(f));

const runCont = tf => k =>
  tf.runCont(k);

const reset = tf =>
  of(tf.runCont(id));
  
const shift = f =>
  Cont(k => f(k).runCont(id));

const of = x =>
  Cont(k => k(x));
  
const liftM2 = f => tf => tg =>
  of(runCont(tf) (x => runCont(tg) (y => f(x) (y))));

const id = x => x;

const mulM = liftM2(x => y => x * y);
const addM = liftM2(x => y => x + y);
const subM = liftM2(x => y => x - y);

const z1 = mulM(of(5))
  (reset
    (addM
      (shift(k => of(3)))
        (of(3)))
  ).runCont(id); // 5 * 3 = 15 (as expected)

const z2 = mulM(of(5))
  (reset // A
    (mulM // B
      (addM
        (shift(k => of(3))) // C should unwind up to A instead of B
          (of(3)))
            (of(4)))
  ).runCont(id); // 5 * 3 * 4 = 60 (but 15 expected)

console.log(z1);
console.log(z2);

看来我只能将堆栈展开一帧。这是 shift/reset 的设计还是我的实现中的缺陷造成的?

[编辑]

我在 Haskell 中使用它,即它是一个实施问题:

reset :: ((a -> a) -> a) -> (a -> r) -> r
reset k f = f $ k id

shift :: ((a -> r) -> (r -> r) -> r) -> (a -> r) -> r
shift f k = f k id

return :: a -> (a -> r) -> r
return a k = k a

liftM2 :: (a -> b -> c) -> ((a -> r) -> r) -> ((b -> r) -> r) -> (c -> r) -> r
liftM2 f ma mb k = ma $ \a -> mb $ \b -> k (f a b)

example :: Num a => (a -> r) -> r
example = liftM2 (*) (return 5) (reset (liftM2 (*) (return 3) (liftM2 (+) (return 2) (shift (\k -> return 3)))))

我觉得你的liftM2坏了,因为它不懒惰。而不是使用 of,它应该构造一个新的延续:

const liftM2 = f => tf => tg =>
  Cont(k => runCont(tf) (x => runCont(tg) (y => k(f(x)(y)))));