我有这段代码:
let rec collect ( t : BCFile list ) ( acc : Set<BCFile> ) : Set<BCFile> =
match t with
| [] -> acc
| hr::tl -> collect ( tl ) ( Set.union acc ( FindSourceFilesForTarget ( hr ) ) )
let s = collect (Set.toList targets) Set.empty
它看起来应该是尾递归,但它不是(看IL)。知道为什么不编译使用尾递归吗?
答案 0 :(得分:10)
据我所知,collect
函数实际上是尾递归的。第一种情况显然只返回acc
。第二种情况首先调用FindSourceFilesForTarget
,然后调用Set.union
然后返回。您可以按如下方式重写它(更清楚地显示尾递归):
| hr::tl ->
let sources = FindSourceFilesForTarget hr
let acc = Set.union acc sources
collect tl
因为这只是一个调用自身的函数,所以编译器会将其优化为循环。这是编译代码的外观(当您使用反射器将其转换为C#时):
public static FSharpSet<int> collect(FSharpList<int> t, FSharpSet<int> acc) {
while (true) {
FSharpList<int> fSharpList = t;
if (fSharpList.TailOrNull == null) break;
// The following corresponds to the second case
FSharpList<int> tl = fSharpList.TailOrNull;
int hr = fSharpList.HeadOrDefault;
// Variables 'acc' and 't' are mutated (instead of calling the function)
acc = SetModule.Union<int>(acc, Program.FindSourceFilesForTarget<int>(hr));
t = tl;
}
return acc;
}
稍微不相关的说明,您也可以使用标准库函数表达:
t |> Seq.map FindSourceFilesForTarget |> Set.unionMany