Seq.cache 缓存 IEnumerable< T> 实例，以便序列中的每个项目仅计算一次。不过，就您而言，您要缓存一个函数返回的序列，并且每次调用该函数时，都会得到一个 new 缓存序列，这对您没有任何好处。正如您所概述的那样，我认为缓存并不是解决问题的正确方法；

Seq.cache caches an IEnumerable<T> instance so that each item in the sequence is only calculated once. In your case, though, you're caching the sequence returned by a function, and each time you call the function you get a new cached sequence, which doesn't do you any good. I don't think caching is really the right approach to your problem as you've outlined it; instead you should probably look into memoization.

如果要定义质数小于 n 的函数，而不是定义该函数来定义素数的无限可数序列，那么缓存就更有意义了。看起来更像这样：

If instead of defining a function giving the primes less than n you want to define an infinite enumerable sequence of primes, then caching makes more sense. That would look more like this:

let rec upFrom i =
  seq { 
    yield i
    yield! upFrom (i+1)
  }

let rec primes =
  seq { 
    yield 2
    yield!
      upFrom 3 |>
      Seq.filter (fun p -> primes |> Seq.takeWhile (fun j -> j*j <= p) |> Seq.forall (fun j -> p % j <> 0))
  }
  |> Seq.cache

我没有将这种方法的性能与您的方法进行比较。

I haven't compared the performance of this method compared to yours.