函數(shù)中說到,函數(shù)是從參數(shù)多元組映射到返回值的對象,若沒有合適返回值則拋出異常。實際中常需要對不同類型的參數(shù)做同樣的運算,例如對整數(shù)做加法、對浮點數(shù)做加法、對整數(shù)與浮點數(shù)做加法,它們都是加法。在 Julia 中,它們都屬于同一對象: +
函數(shù)。
對同一概念做一系列實現(xiàn)時,可以逐個定義特定參數(shù)類型、個數(shù)所對應(yīng)的特定函數(shù)行為。方法是對函數(shù)中某一特定的行為定義。函數(shù)中可以定義多個方法。對一個特定的參數(shù)多元組調(diào)用函數(shù)時,最匹配此參數(shù)多元組的方法被調(diào)用。
函數(shù)調(diào)用時,選取調(diào)用哪個方法,被稱為重載。 Julia 依據(jù)參數(shù)個數(shù)、類型來進行重載。
Julia 的所有標準函數(shù)和運算符,如前面提到的 +
函數(shù),都有許多針對各種參數(shù)類型組合和不同參數(shù)個數(shù)而定義的方法。
定義函數(shù)時,可以像復(fù)合類型中介紹的那樣,使用 ::
類型斷言運算符,選擇性地對參數(shù)類型進行限制:
julia> f(x::Float64, y::Float64) = 2x + y;
此函數(shù)中參數(shù) x
和 y
只能是 Float64
類型:
julia> f(2.0, 3.0)
7.0
如果參數(shù)是其它類型,會引發(fā) “no method” 錯誤:
julia> f(2.0, 3)
ERROR: `f` has no method matching f(::Float64, ::Int64)
julia> f(float32(2.0), 3.0)
ERROR: `f` has no method matching f(::Float32, ::Float64)
julia> f(2.0, "3.0")
ERROR: `f` has no method matching f(::Float64, ::ASCIIString)
julia> f("2.0", "3.0")
ERROR: `f` has no method matching f(::ASCIIString, ::ASCIIString)
有時需要寫一些通用方法,這時應(yīng)聲明參數(shù)為抽象類型:
julia> f(x::Number, y::Number) = 2x - y;
julia> f(2.0, 3)
1.0
要想給一個函數(shù)定義多個方法,只需要多次定義這個函數(shù),每次定義的參數(shù)個數(shù)和類型需不同。函數(shù)調(diào)用時,最匹配的方法被重載:
julia> f(2.0, 3.0)
7.0
julia> f(2, 3.0)
1.0
julia> f(2.0, 3)
1.0
julia> f(2, 3)
1
對非數(shù)值的值,或參數(shù)個數(shù)少于 2,f
是未定義的,調(diào)用它會返回 “no method” 錯誤:
julia> f("foo", 3)
ERROR: `f` has no method matching f(::ASCIIString, ::Int64)
julia> f()
ERROR: `f` has no method matching f()
在交互式會話中輸入函數(shù)對象本身,可以看到函數(shù)所存在的方法:
julia> f
f (generic function with 2 methods)
這個輸出告訴我們,f
是一個含有兩個方法的函數(shù)對象。要找出這些方法的簽名,可以通過使用 methods
函數(shù)來實現(xiàn):
julia> methods(f)
# 2 methods for generic function "f":
f(x::Float64,y::Float64) at none:1
f(x::Number,y::Number) at none:1
這表明,f
有兩個方法,一個以兩個 Float64
類型作為參數(shù)和另一個則以一個 Number
類型作為參數(shù)。它也指示了定義方法的文件和行數(shù):因為這些方法在 REPL 中定義,我們得到明顯行數(shù)值:none:1
。
定義類型時如果沒使用 ::
,則方法參數(shù)的類型默認為 Any
。對 f
定義一個總括匹配的方法:
julia> f(x,y) = println("Whoa there, Nelly.");
julia> f("foo", 1)
Whoa there, Nelly.
總括匹配的方法,是重載時的最后選擇。
重載是 Julia 最強大最核心的特性。核心運算一般都有好幾十種方法:
julia> methods(+)
# 125 methods for generic function "+":
+(x::Bool) at bool.jl:36
+(x::Bool,y::Bool) at bool.jl:39
+(y::FloatingPoint,x::Bool) at bool.jl:49
+(A::BitArray{N},B::BitArray{N}) at bitarray.jl:848
+(A::Union(DenseArray{Bool,N},SubArray{Bool,N,A<:DenseArray{T,N},I<:(Union(Range{Int64},Int64)...,)}),B::Union(DenseArray{Bool,N},SubArray{Bool,N,A<:DenseArray{T,N},I<:(Union(Range{Int64},Int64)...,)})) at array.jl:797
+{S,T}(A::Union(SubArray{S,N,A<:DenseArray{T,N},I<:(Union(Range{Int64},Int64)...,)},DenseArray{S,N}),B::Union(SubArray{T,N,A<:DenseArray{T,N},I<:(Union(Range{Int64},Int64)...,)},DenseArray{T,N})) at array.jl:719
+{T<:Union(Int16,Int32,Int8)}(x::T<:Union(Int16,Int32,Int8),y::T<:Union(Int16,Int32,Int8)) at int.jl:16
+{T<:Union(Uint32,Uint16,Uint8)}(x::T<:Union(Uint32,Uint16,Uint8),y::T<:Union(Uint32,Uint16,Uint8)) at int.jl:20
+(x::Int64,y::Int64) at int.jl:33
+(x::Uint64,y::Uint64) at int.jl:34
+(x::Int128,y::Int128) at int.jl:35
+(x::Uint128,y::Uint128) at int.jl:36
+(x::Float32,y::Float32) at float.jl:119
+(x::Float64,y::Float64) at float.jl:120
+(z::Complex{T<:Real},w::Complex{T<:Real}) at complex.jl:110
+(x::Real,z::Complex{T<:Real}) at complex.jl:120
+(z::Complex{T<:Real},x::Real) at complex.jl:121
+(x::Rational{T<:Integer},y::Rational{T<:Integer}) at rational.jl:113
+(x::Char,y::Char) at char.jl:23
+(x::Char,y::Integer) at char.jl:26
+(x::Integer,y::Char) at char.jl:27
+(a::Float16,b::Float16) at float16.jl:132
+(x::BigInt,y::BigInt) at gmp.jl:194
+(a::BigInt,b::BigInt,c::BigInt) at gmp.jl:217
+(a::BigInt,b::BigInt,c::BigInt,d::BigInt) at gmp.jl:223
+(a::BigInt,b::BigInt,c::BigInt,d::BigInt,e::BigInt) at gmp.jl:230
+(x::BigInt,c::Uint64) at gmp.jl:242
+(c::Uint64,x::BigInt) at gmp.jl:246
+(c::Union(Uint32,Uint16,Uint8,Uint64),x::BigInt) at gmp.jl:247
+(x::BigInt,c::Union(Uint32,Uint16,Uint8,Uint64)) at gmp.jl:248
+(x::BigInt,c::Union(Int64,Int16,Int32,Int8)) at gmp.jl:249
+(c::Union(Int64,Int16,Int32,Int8),x::BigInt) at gmp.jl:250
+(x::BigFloat,c::Uint64) at mpfr.jl:147
+(c::Uint64,x::BigFloat) at mpfr.jl:151
+(c::Union(Uint32,Uint16,Uint8,Uint64),x::BigFloat) at mpfr.jl:152
+(x::BigFloat,c::Union(Uint32,Uint16,Uint8,Uint64)) at mpfr.jl:153
+(x::BigFloat,c::Int64) at mpfr.jl:157
+(c::Int64,x::BigFloat) at mpfr.jl:161
+(x::BigFloat,c::Union(Int64,Int16,Int32,Int8)) at mpfr.jl:162
+(c::Union(Int64,Int16,Int32,Int8),x::BigFloat) at mpfr.jl:163
+(x::BigFloat,c::Float64) at mpfr.jl:167
+(c::Float64,x::BigFloat) at mpfr.jl:171
+(c::Float32,x::BigFloat) at mpfr.jl:172
+(x::BigFloat,c::Float32) at mpfr.jl:173
+(x::BigFloat,c::BigInt) at mpfr.jl:177
+(c::BigInt,x::BigFloat) at mpfr.jl:181
+(x::BigFloat,y::BigFloat) at mpfr.jl:328
+(a::BigFloat,b::BigFloat,c::BigFloat) at mpfr.jl:339
+(a::BigFloat,b::BigFloat,c::BigFloat,d::BigFloat) at mpfr.jl:345
+(a::BigFloat,b::BigFloat,c::BigFloat,d::BigFloat,e::BigFloat) at mpfr.jl:352
+(x::MathConst{sym},y::MathConst{sym}) at constants.jl:23
+{T<:Number}(x::T<:Number,y::T<:Number) at promotion.jl:188
+{T<:FloatingPoint}(x::Bool,y::T<:FloatingPoint) at bool.jl:46
+(x::Number,y::Number) at promotion.jl:158
+(x::Integer,y::Ptr{T}) at pointer.jl:68
+(x::Bool,A::AbstractArray{Bool,N}) at array.jl:767
+(x::Number) at operators.jl:71
+(r1::OrdinalRange{T,S},r2::OrdinalRange{T,S}) at operators.jl:325
+{T<:FloatingPoint}(r1::FloatRange{T<:FloatingPoint},r2::FloatRange{T<:FloatingPoint}) at operators.jl:331
+(r1::FloatRange{T<:FloatingPoint},r2::FloatRange{T<:FloatingPoint}) at operators.jl:348
+(r1::FloatRange{T<:FloatingPoint},r2::OrdinalRange{T,S}) at operators.jl:349
+(r1::OrdinalRange{T,S},r2::FloatRange{T<:FloatingPoint}) at operators.jl:350
+(x::Ptr{T},y::Integer) at pointer.jl:66
+{S,T<:Real}(A::Union(SubArray{S,N,A<:DenseArray{T,N},I<:(Union(Range{Int64},Int64)...,)},DenseArray{S,N}),B::Range{T<:Real}) at array.jl:727
+{S<:Real,T}(A::Range{S<:Real},B::Union(SubArray{T,N,A<:DenseArray{T,N},I<:(Union(Range{Int64},Int64)...,)},DenseArray{T,N})) at array.jl:736
+(A::AbstractArray{Bool,N},x::Bool) at array.jl:766
+{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti},B::SparseMatrixCSC{Tv,Ti}) at sparse/sparsematrix.jl:530
+{TvA,TiA,TvB,TiB}(A::SparseMatrixCSC{TvA,TiA},B::SparseMatrixCSC{TvB,TiB}) at sparse/sparsematrix.jl:522
+(A::SparseMatrixCSC{Tv,Ti<:Integer},B::Array{T,N}) at sparse/sparsematrix.jl:621
+(A::Array{T,N},B::SparseMatrixCSC{Tv,Ti<:Integer}) at sparse/sparsematrix.jl:623
+(A::SymTridiagonal{T},B::SymTridiagonal{T}) at linalg/tridiag.jl:45
+(A::Tridiagonal{T},B::Tridiagonal{T}) at linalg/tridiag.jl:207
+(A::Tridiagonal{T},B::SymTridiagonal{T}) at linalg/special.jl:99
+(A::SymTridiagonal{T},B::Tridiagonal{T}) at linalg/special.jl:98
+{T,MT,uplo}(A::Triangular{T,MT,uplo,IsUnit},B::Triangular{T,MT,uplo,IsUnit}) at linalg/triangular.jl:10
+{T,MT,uplo1,uplo2}(A::Triangular{T,MT,uplo1,IsUnit},B::Triangular{T,MT,uplo2,IsUnit}) at linalg/triangular.jl:11
+(Da::Diagonal{T},Db::Diagonal{T}) at linalg/diagonal.jl:44
+(A::Bidiagonal{T},B::Bidiagonal{T}) at linalg/bidiag.jl:92
+{T}(B::BitArray{2},J::UniformScaling{T}) at linalg/uniformscaling.jl:26
+(A::Diagonal{T},B::Bidiagonal{T}) at linalg/special.jl:89
+(A::Bidiagonal{T},B::Diagonal{T}) at linalg/special.jl:90
+(A::Diagonal{T},B::Tridiagonal{T}) at linalg/special.jl:89
+(A::Tridiagonal{T},B::Diagonal{T}) at linalg/special.jl:90
+(A::Diagonal{T},B::Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}) at linalg/special.jl:89
+(A::Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit},B::Diagonal{T}) at linalg/special.jl:90
+(A::Diagonal{T},B::Array{T,2}) at linalg/special.jl:89
+(A::Array{T,2},B::Diagonal{T}) at linalg/special.jl:90
+(A::Bidiagonal{T},B::Tridiagonal{T}) at linalg/special.jl:89
+(A::Tridiagonal{T},B::Bidiagonal{T}) at linalg/special.jl:90
+(A::Bidiagonal{T},B::Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}) at linalg/special.jl:89
+(A::Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit},B::Bidiagonal{T}) at linalg/special.jl:90
+(A::Bidiagonal{T},B::Array{T,2}) at linalg/special.jl:89
+(A::Array{T,2},B::Bidiagonal{T}) at linalg/special.jl:90
+(A::Tridiagonal{T},B::Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}) at linalg/special.jl:89
+(A::Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit},B::Tridiagonal{T}) at linalg/special.jl:90
+(A::Tridiagonal{T},B::Array{T,2}) at linalg/special.jl:89
+(A::Array{T,2},B::Tridiagonal{T}) at linalg/special.jl:90
+(A::Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit},B::Array{T,2}) at linalg/special.jl:89
+(A::Array{T,2},B::Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}) at linalg/special.jl:90
+(A::SymTridiagonal{T},B::Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit}) at linalg/special.jl:98
+(A::Triangular{T,S<:AbstractArray{T,2},UpLo,IsUnit},B::SymTridiagonal{T}) at linalg/special.jl:99
+(A::SymTridiagonal{T},B::Array{T,2}) at linalg/special.jl:98
+(A::Array{T,2},B::SymTridiagonal{T}) at linalg/special.jl:99
+(A::Diagonal{T},B::SymTridiagonal{T}) at linalg/special.jl:107
+(A::SymTridiagonal{T},B::Diagonal{T}) at linalg/special.jl:108
+(A::Bidiagonal{T},B::SymTridiagonal{T}) at linalg/special.jl:107
+(A::SymTridiagonal{T},B::Bidiagonal{T}) at linalg/special.jl:108
+{T<:Number}(x::AbstractArray{T<:Number,N}) at abstractarray.jl:358
+(A::AbstractArray{T,N},x::Number) at array.jl:770
+(x::Number,A::AbstractArray{T,N}) at array.jl:771
+(J1::UniformScaling{T<:Number},J2::UniformScaling{T<:Number}) at linalg/uniformscaling.jl:25
+(J::UniformScaling{T<:Number},B::BitArray{2}) at linalg/uniformscaling.jl:27
+(J::UniformScaling{T<:Number},A::AbstractArray{T,2}) at linalg/uniformscaling.jl:28
+(J::UniformScaling{T<:Number},x::Number) at linalg/uniformscaling.jl:29
+(x::Number,J::UniformScaling{T<:Number}) at linalg/uniformscaling.jl:30
+{TA,TJ}(A::AbstractArray{TA,2},J::UniformScaling{TJ}) at linalg/uniformscaling.jl:33
+{T}(a::HierarchicalValue{T},b::HierarchicalValue{T}) at pkg/resolve/versionweight.jl:19
+(a::VWPreBuildItem,b::VWPreBuildItem) at pkg/resolve/versionweight.jl:82
+(a::VWPreBuild,b::VWPreBuild) at pkg/resolve/versionweight.jl:120
+(a::VersionWeight,b::VersionWeight) at pkg/resolve/versionweight.jl:164
+(a::FieldValue,b::FieldValue) at pkg/resolve/fieldvalue.jl:41
+(a::Vec2,b::Vec2) at graphics.jl:60
+(bb1::BoundingBox,bb2::BoundingBox) at graphics.jl:123
+(a,b,c) at operators.jl:82
+(a,b,c,xs...) at operators.jl:83
重載和靈活的參數(shù)化類型系統(tǒng)一起,使得 Julia 可以抽象表達高級算法,不需關(guān)注實現(xiàn)的具體細節(jié),生成有效率、運行時專用的代碼。
函數(shù)方法的適用范圍可能會重疊:
julia> g(x::Float64, y) = 2x + y;
julia> g(x, y::Float64) = x + 2y;
Warning: New definition
g(Any,Float64) at none:1
is ambiguous with:
g(Float64,Any) at none:1.
To fix, define
g(Float64,Float64)
before the new definition.
julia> g(2.0, 3)
7.0
julia> g(2, 3.0)
8.0
julia> g(2.0, 3.0)
7.0
此處 g(2.0, 3.0)
既可以調(diào)用 g(Float64, Any)
,也可以調(diào)用 g(Any, Float64)
,兩種方法沒有優(yōu)先級。遇到這種情況,Julia 會警告定義含糊,但仍會任選一個方法來繼續(xù)執(zhí)行。應(yīng)避免含糊的方法:
julia> g(x::Float64, y::Float64) = 2x + 2y;
julia> g(x::Float64, y) = 2x + y;
julia> g(x, y::Float64) = x + 2y;
julia> g(2.0, 3)
7.0
julia> g(2, 3.0)
8.0
julia> g(2.0, 3.0)
10.0
要消除 Julia 的警告,應(yīng)先定義清晰的方法。
構(gòu)造參數(shù)化方法,應(yīng)在方法名與參數(shù)多元組之間,添加類型參數(shù):
julia> same_type(x::T, y::T) where {T} = true
same_type (generic function with 1 method)
julia> same_type(x,y) = false
same_type (generic function with 2 methods)
這兩個方法定義了一個布爾函數(shù),它檢查兩個參數(shù)是否為同一類型:
julia> same_type(1, 2)
true
julia> same_type(1, 2.0)
false
julia> same_type(1.0, 2.0)
true
julia> same_type("foo", 2.0)
false
julia> same_type("foo", "bar")
true
julia> same_type(int32(1), int64(2))
false
類型參數(shù)可用于函數(shù)定義或函數(shù)體的任何地方:
julia> myappend{T}(v::Vector{T}, x::T) = [v..., x]
myappend (generic function with 1 method)
julia> myappend([1,2,3],4)
4-element Array{Int64,1}:
1
2
3
4
julia> myappend([1,2,3],2.5)
ERROR: `myappend` has no method matching myappend(::Array{Int64,1}, ::Float64)
julia> myappend([1.0,2.0,3.0],4.0)
4-element Array{Float64,1}:
1.0
2.0
3.0
4.0
julia> myappend([1.0,2.0,3.0],4)
ERROR: `myappend` has no method matching myappend(::Array{Float64,1}, ::Int64)
下例中,方法類型參數(shù) T
被用作返回值:
julia> mytypeof{T}(x::T) = T
mytypeof (generic function with 1 method)
julia> mytypeof(1)
Int64
julia> mytypeof(1.0)
Float64
方法的類型參數(shù)也可以被限制范圍:
same_type_numeric{T<:Number}(x::T, y::T) = true
same_type_numeric(x::Number, y::Number) = false
julia> same_type_numeric(1, 2)
true
julia> same_type_numeric(1, 2.0)
false
julia> same_type_numeric(1.0, 2.0)
true
julia> same_type_numeric("foo", 2.0)
no method same_type_numeric(ASCIIString,Float64)
julia> same_type_numeric("foo", "bar")
no method same_type_numeric(ASCIIString,ASCIIString)
julia> same_type_numeric(int32(1), int64(2))
false
same_type_numeric
函數(shù)與 same_type
大致相同,但只應(yīng)用于數(shù)對兒。
函數(shù)中曾簡略提到,可選參數(shù)是可由多方法定義語法的實現(xiàn)。例如:
f(a=1,b=2) = a+2b
可以翻譯為下面三個方法:
f(a,b) = a+2b
f(a) = f(a,2)
f() = f(1,2)
關(guān)鍵字參數(shù)則與普通的與位置有關(guān)的參數(shù)不同。它們不用于方法重載。方法重載僅基于位置參數(shù),選取了匹配的方法后,才處理關(guān)鍵字參數(shù)。
更多建議: