multidimensional-array - 如何编写一个将 ndarray Array 或 ArrayView 作为输入的通用函数？

Question

我正在使用 ndarray 编写一组数学函数，我想在任何类型的 ArrayBase 上执行这些函数。但是，我无法指定所涉及的特征/类型。

此基本函数适用于 OwnedRepr 或 ViewRepr 数据:

use ndarray::{prelude::*, Data}; // 0.13.1

fn sum_owned(x: Array<f64, Ix1>) -> f64 {
    x.sum()
}

fn sum_view(x: ArrayView<f64, Ix1>) -> f64 {
    x.sum()
}

fn main() {
    let a = Array::from_shape_vec((4,), vec![1.0, 2.0, 3.0, 4.0]).unwrap();
    println!("{:?}", sum_owned(a.clone()));

    let b = a.slice(s![..]);
    println!("{:?}", sum_view(b));

    // Complains that OwnedRepr is not ViewRepr
    //println!("{:?}", sum_view(a.clone()));
}

我能理解为什么注释掉的部分无法编译，但我对泛型的理解还不够深入，无法编写更多...泛型。

这是我尝试过的:

use ndarray::prelude::*;
use ndarray::Data;

fn sum_general<S>(x: ArrayBase<S, Ix1>) -> f64
where
    S: Data,
{
    x.sum()
}

编译器错误表明 Data 不够具体，但我无法很好地解析它以找出解决方案:

error[E0277]: the trait bound `<S as ndarray::data_traits::RawData>::Elem: std::clone::Clone` is not satisfied
 --> src/lib.rs:8:7
  |
6 |     S: Data,
  |             - help: consider further restricting the associated type: `, <S as ndarray::data_traits::RawData>::Elem: std::clone::Clone`
7 | {
8 |     x.sum()
  |       ^^^ the trait `std::clone::Clone` is not implemented for `<S as ndarray::data_traits::RawData>::Elem`

error[E0277]: the trait bound `<S as ndarray::data_traits::RawData>::Elem: num_traits::identities::Zero` is not satisfied
 --> src/lib.rs:8:7
  |
6 |     S: Data,
  |             - help: consider further restricting the associated type: `, <S as ndarray::data_traits::RawData>::Elem: num_traits::identities::Zero`
7 | {
8 |     x.sum()
  |       ^^^ the trait `num_traits::identities::Zero` is not implemented for `<S as ndarray::data_traits::RawData>::Elem`

error[E0308]: mismatched types
 --> src/lib.rs:8:5
  |
4 | fn sum_general<S>(x: ArrayBase<S, Ix1>) -> f64
  |                                            --- expected `f64` because of return type
...
8 |     x.sum()
  |     ^^^^^^^ expected `f64`, found associated type
  |
  = note:         expected type `f64`
          found associated type `<S as ndarray::data_traits::RawData>::Elem`
  = note: consider constraining the associated type `<S as ndarray::data_traits::RawData>::Elem` to `f64`
  = note: for more information, visit https://doc.rust-lang.org/book/ch19-03-advanced-traits.html

Answer 1

如果您查看 ndarray::ArrayBase::sum 的定义您尝试调用的函数:

impl<A, S, D> ArrayBase<S, D>
where
    S: Data<Elem = A>,
    D: Dimension,
{
    pub fn sum(&self) -> A
    where
       A: Clone + Add<Output = A> + Zero
    {
         // etc.
    }
}

很明显，在你的情况下 A = f64和 D = Ix1 , 但您仍然需要指定约束 S: Data<Elem = f64> .因此:

use ndarray::prelude::*;
use ndarray::Data;

fn sum_general<S>(x: ArrayBase<S, Ix1>) -> f64
where
    S: Data<Elem = f64>,
{
    x.sum()
}

这正是编译器在提示时的意思:

  = note:         expected type `f64`          found associated type `<S as ndarray::data_traits::RawData>::Elem`  = note: consider constraining the associated type `<S as ndarray::data_traits::RawData>::Elem` to `f64`