Julia:预期的类型是Vector{Array{Float32}},但得到的值是Vector{Vector{Float32}}。

huangapple go评论91阅读模式
英文:

Julia: Expected Vector{Array{Float32}}, got a value of type Vector{Vector{Float32}}

问题

Sure, here's the translated code:

我有一个函数 `g`,它以 Float32 数组的 Vector 作为输入。但是,当我像下面这样调用该函数时,出现了一个错误,说找不到匹配 `g(::Vector{Vector{Float32}})` 的方法:

代码:
```julia
function g(x::Array{Array{Float32}, 1})
    return x
end

function f()
    a = [[1f0], [2f0]]
    println( g(a) )
end

错误:

MethodError: no method matching g(::Vector{Vector{Float32}})
  Closest candidates are:
    g(::Vector{Array{Float32}}) at ~/.../runtests.jl:25

如何让 g 接受任意大小的 Float32 数组的向量作为输入,包括 1D 数组(向量)?我是 Julia 的新手,所以提前感谢您的帮助!


<details>
<summary>英文:</summary>

I have a function `g` that takes in a Vector of Float32 Arrays as input. However, when I call the function as follows, I get an error saying that no method matching `g(::Vector{Vector{Float32}})` was found:

Code:

function g(x::Array{Array{Float32}, 1})
return x
end

function f()
a = [[1f0], [2f0]]
println( g(a) )
end


Error:

MethodError: no method matching g(::Vector{Vector{Float32}})
Closest candidates are:
g(::Vector{Array{Float32}}) at ~/.../runtests.jl:25

How do I let `g` take in as input a vector of arbitrarily sized Float32 arrays, including 1D arrays (vectors)? I&#39;m new to Julia, so thanks in advance for your help!

</details>


# 答案1
**得分**: 1

`x::Array{Array{Float32}, 1}` 不同于 `x::Array{Array{Float32, 1}, 1}` — 也就是说,您在声明参数类型时表示您的参数是一个包含*任意维度*数组的数组,这是一种具体类型,与包含*1维*数组的数组不同。

也许您想要的是:

```julia
function g(x::Array{<:Array{Float32}, 1})

或者等效地:

function g(x::Vector{<:Array{Float32}})

或者更一般地(例如):

function g(x::AbstractVector{<:AbstractArray{<:Real}})

以接受任何类型的1维数组(AbstractVector 的任何子类型),其元素是任何类型的任意维度的实数数组。

请参阅Julia FAQ 中的不变类型

另请参阅Julia 手册中的参数类型声明。通常情况下,使用像 g(x::Array{Array{Float32, 1}, 1}) 这样具体的类型签名来声明函数几乎没有性能优势,并且几乎肯定比必要的要不够通用。

英文:

x::Array{Array{Float32}, 1} is different from x::Array{Array{Float32, 1}, 1} — that is, you are saying that your argument is an array of arbitrary-dimension arrays, which is a specific type and is different from an array of 1d arrays.

You maybe want

function g(x::Array{&lt;:Array{Float32}, 1})

or equivalently

function g(x::Vector{&lt;:Array{Float32}})

or more generally (for example):

function g(x::AbstractVector{&lt;:AbstractArray{&lt;:Real}})

to accept any type of 1d array (any subtype of AbstractVector) whose elements are any type of any-dimensional array of any type of real number.

See the Julia FAQ on invariant types.

See also the Julia manual on argument-type declarations. There is generally zero (nada, none, zip) performance advantage to declaring a function with a type signature as specific as g(x::Array{Array{Float32, 1}, 1}), and it's almost certainly less general than necessary.

huangapple
  • 本文由 发表于 2023年7月18日 10:35:11
  • 转载请务必保留本文链接:https://go.coder-hub.com/76709195.html
匿名

发表评论

匿名网友

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen:

确定