[英]Tail Call Optimization in F#
我真的可以在F#中的尾部调用优化方面提供一些帮助。 我试图解析一个像树的结构,并在每片叶子上执行计算。
我遇到问题的函数是calcLength
type Location = float * float
type Radius = float
type Width = float
type Angle = float
type Primitive =
| Circle of Location * Radius
| Ellipse of Location * Radius * Radius
| Square of Location * Width * Angle
| MultiPrimitive of Primitive List
type Primitive with
member x.Length =
let rec calcLength x =
match x with
| Circle (_,r) -> System.Math.PI * r * 2.
| Ellipse (_,r1,r2) -> System.Math.PI * 2. * sqrt( (r1 * r1 ) + (r2 * r2 ) / 2.)
| Square (_, w,_) -> w * 4.
| MultiPrimitive [] -> 0.
| MultiPrimitive (head::tail) -> calcLength (MultiPrimitive tail) + (calcLength head)
[<Fact>]
let ``test discriminated unions``() =
let pattern = MultiPrimitive(
[
MultiPrimitive(
[
MultiPrimitive(
[
Square( (10.,10.), 10., 45. );
Circle( (3.,7.), 3. );
Circle( (7.,7.), 3. );
Square( (5.,2.), 3., 45. );
] );
Square( (10.,10.), 10., 45. );
Circle( (3.,7.), 3. );
Circle( (7.,7.), 3. );
Square( (5.,2.), 3., 45. );
] );
Square( (10.,10.), 10., 45. );
Circle( (3.,7.), 3. );
Circle( (7.,7.), 3. );
Square( (5.,2.), 3., 45. );
] )
let r = pattern.Length
我尝试将延续方法与以下方法结合使用:
let rec calcLength x f =
match x with
| Circle (_,r) -> f() + System.Math.PI * r * 2.
| Ellipse (_,r1,r2) -> f() + System.Math.PI * 2. * sqrt( (r1 * r1 ) + (r2 * r2 ) / 2.)
| Square (_, w,_) -> f() + w * 4.
| MultiPrimitive [] -> f()
| MultiPrimitive (head::tail) -> calcLength head (fun () -> calcLength(MultiPrimitive tail) f )
calcLength x (fun () -> 0.)
但步进通过与调试器显示堆栈越来越大,任何帮助将非常感激。
使用CPS的通常方法是将结果传递给给定的延续:
let rec calcLength x k =
match x with
| Circle (_,r) -> k (System.Math.PI * r * 2.)
| Ellipse (_,r1,r2) -> k (System.Math.PI * 2. * sqrt( (r1 * r1 ) + (r2 * r2 ) / 2.))
| Square (_, w,_) -> k (w * 4.)
| MultiPrimitive [] -> k 0.
| MultiPrimitive (head::tail) -> (calcLength head (fun h -> calcLength(MultiPrimitive tail) (fun t -> k (h + t))))
因此在MultiPrimitive
情况下,您需要传递另一个延续来处理计算头部的结果。
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.