F# 引号:变量可能会转义作用域

发布于 2024-11-16 13:25:58 字数 1193 浏览 1 评论 0原文

我有这段代码:

let rec h n z = if n = 0 then z
                else <@ (fun x -> %(h (n - 1) <@ x + %z @>)) n @>

http://www.cs 中的 MetaOcaml 示例转换而来。 Rice.edu/~taha/publications/journal/dspg04a.pdf

在论文中解释了上面的例子将产生以下参数 3.<1>.(采用 MetaOcaml 表示法):

.<(fun x_1 -> (fun x_2 -> (fun x_3 -> x_3 + (x_2 + (x_1 + 1))) 1) 2) 3>.

如您所见,xx_1x_2 替换等等,因为x否则只会引用最里面的fun中的x

但在 F# 中这是不允许的。我收到编译时错误:“变量‘x’绑定在引号中,但用作拼接表达式的一部分。这是不允许的,因为它可能会逃脱其范围。”所以问题是:如何更改它才能编译并具有与 MetaOcaml 输出相同的语义?

更新评论:我使用 PowerPack 来实际评估报价。但我认为这与它没有任何关系,因为错误是在编译时发生的。到目前为止,QuotationEvaluation 有效。但是,我确实知道这可能不是最有效的实现。

更新 Tomas 的答案: 我真的不希望 x 是全局的,或者逃避范围。但我想要的是相当于

let rec h n z = if n = 0 then z
                else (fun x -> (h (n - 1) (x + z))) n

带引号的。您的答案给出 (h 3 <@ 1 @>).Eval() = 4,其中上述结果为 h 3 1 = 7。在这里,我希望 7 作为答案。

I have this bit of code:

let rec h n z = if n = 0 then z
                else <@ (fun x -> %(h (n - 1) <@ x + %z @>)) n @>

converted from a MetaOcaml example in http://www.cs.rice.edu/~taha/publications/journal/dspg04a.pdf

In the paper there is explained that the above example will yield the following with the parameters 3 and .<1>. (in MetaOcaml notation):

.<(fun x_1 -> (fun x_2 -> (fun x_3 -> x_3 + (x_2 + (x_1 + 1))) 1) 2) 3>.

As you can see the x´s gets replaced by x_1, x_2 etc. because the x would otherwise only refer to the x in the innermost fun.

But in F# this isn't allowed. I get the compile-time error: "The variable 'x' is bound in a quotation but is used as part of a spliced expression. This is not permitted since it may escape its scope." So the question is: how can this be changed so it will compile and have the same semantic as the MetaOcaml output?

Update to comment: I use the PowerPack to actually evaluating the quotation. But I don't think this have anything to do with it because the error is at compile-time. So far QuotationEvaluation works. However, I do know it may not be the most efficient implementation.

Update to Tomas´ answer:
I really don't want the x to be global, or to escape scope. But I want is the equivalent to

let rec h n z = if n = 0 then z
                else (fun x -> (h (n - 1) (x + z))) n

with quotations. Your answer gives (h 3 <@ 1 @>).Eval() = 4 where the above yields h 3 1 = 7. And here, I want 7 to be the answer.

如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。

扫码二维码加入Web技术交流群

发布评论

需要 登录 才能够评论, 你可以免费 注册 一个本站的账号。

评论(1

寂寞陪衬 2024-11-23 13:25:58

F# 引号语法不支持可能逃脱作用域的变量,因此您需要使用 Expr 操作显式构造树。像这样的东西应该可以解决问题:

open Microsoft.FSharp.Quotations

let rec h n (z:Expr<int>) = 
  if n = 0 then z                
  else 
    let v = new Var("x", typeof<int>)
    let ve = Expr.Var(v)
    Expr.Cast<int>
        (Expr.Application( Expr.Lambda(v, h (n - 1) <@ %%ve + %z @>), 
                           Expr.Value(n)))

但是,这是一个非常人为的示例(演示 MetaOCaml 中的变量捕获,这在 F# 中不可用)。它只是生成类似 (2 + (1 + ...)) 的表达式。您可以通过编写如下内容获得相同的结果:

let rec h n (z:Expr<int>) = 
  if n = 0 then z                
  else h (n - 1) <@ n + %z @>

或者甚至更好:

[ 1 .. 4 ] |> List.fold (fun st n -> <@ n + %st @>) <@ 0 @>

我也在 F# 引用中遇到了此限制,如果支持的话那就太好了。不过,我认为这在实践中并不是一个大问题,因为 F# 引用不用于分阶段元编程。它们对于分析现有 F# 代码比生成代码更有用。

F# quotation syntax doesn't support variables that could potentially escape the scope, so you'll need to construct the tree explicitly using the Expr operations. Something like this should do the trick:

open Microsoft.FSharp.Quotations

let rec h n (z:Expr<int>) = 
  if n = 0 then z                
  else 
    let v = new Var("x", typeof<int>)
    let ve = Expr.Var(v)
    Expr.Cast<int>
        (Expr.Application( Expr.Lambda(v, h (n - 1) <@ %%ve + %z @>), 
                           Expr.Value(n)))

However, this is quite artificial example (to demonstrate variable capturing in MetaOCaml, which isn't available in F#). It just generates expression like (2 + (1 + ...)). You can get the same result by writing something like this:

let rec h n (z:Expr<int>) = 
  if n = 0 then z                
  else h (n - 1) <@ n + %z @>

Or even better:

[ 1 .. 4 ] |> List.fold (fun st n -> <@ n + %st @>) <@ 0 @>

I also came accross this limitation in F# quotations and it would be nice if this was supported. However, I don't think it is such a big problem in practice, because F# quotations are not used for staged meta-programming. They are more useful for analyzing existing F# code than for generating code.

~没有更多了~
我们使用 Cookies 和其他技术来定制您的体验包括您的登录状态等。通过阅读我们的 隐私政策 了解更多相关信息。 单击 接受 或继续使用网站,即表示您同意使用 Cookies 和您的相关数据。
原文