Merge pull request #20414 from stevengj/covariant

syntactic sugar Foo{<:Bar} for Foo{T} where T<:Bar

Author: JeffBezanson

NEWS.md

@@ -14,6 +14,8 @@ New language features
       `function inv(M::Matrix{T}) where T<:AbstractFloat`.
       Anonymous functions can have type parameters via the syntax
       `((x::Array{T}) where T<:Real) -> 2x`.
+    * Implicit type parameters, e.g. `Vector{<:Real}` is equivalent to
+      `Vector{T} where T<:Real`, and similarly for `Vector{>:Int}` ([#20414]).
     * Much more accurate subtype and type intersection algorithms. Method sorting and
       identification of equivalent and ambiguous methods are improved as a result.
 

doc/src/manual/types.md

@@ -577,7 +577,7 @@ false
     This last point is *very* important: even though `Float64 <: Real` we **DO NOT** have `Point{Float64} <: Point{Real}`.
 
 In other words, in the parlance of type theory, Julia's type parameters are *invariant*, rather
-than being covariant (or even contravariant). This is for practical reasons: while any instance
+than being [covariant (or even contravariant)](https://en.wikipedia.org/wiki/Covariance_and_contravariance_%28computer_science%29). This is for practical reasons: while any instance
 of `Point{Float64}` may conceptually be like an instance of `Point{Real}` as well, the two types
 have different representations in memory:
 
@@ -603,15 +603,18 @@ function norm(p::Point{Real})
 end
 ```
 
-The correct way to define a method that accepts all arguments of type `Point{T}` where `T` is
+A correct way to define a method that accepts all arguments of type `Point{T}` where `T` is
 a subtype of `Real` is:
 
 ```julia
-function norm{T<:Real}(p::Point{T})
+function norm(p::Point{<:Real})
     sqrt(p.x^2 + p.y^2)
 end
 ```
 
+(Equivalently, one could define `function norm{T<:Real}(p::Point{T})` or
+`function norm(p::Point{T} where T<:Real)`; see [UnionAll Types](@ref).)
+
 More examples will be discussed later in [Methods](@ref).
 
 How does one construct a `Point` object? It is possible to define custom constructors for composite
@@ -711,6 +714,17 @@ julia> Pointy{Real} <: Pointy{Float64}
 false
 ```
 
+The notation `Pointy{<:Real}` can be used to express the Julia analogue of a
+*covariant* type, while `Pointy{>:Int}` the analogue of a *contravariant* type,
+but technically these represent *sets* of types (see [UnionAll Types](@ref)).
+```jldoctest pointytype
+julia> Pointy{Float64} <: Pointy{<:Real}
+true
+
+julia> Pointy{Real} <: Pointy{>:Int}
+true
+```
+
 Much as plain old abstract types serve to create a useful hierarchy of types over concrete types,
 parametric abstract types serve the same purpose with respect to parametric composite types. We
 could, for example, have declared `Point{T}` to be a subtype of `Pointy{T}` as follows:
@@ -740,6 +754,9 @@ This relationship is also invariant:
 ```jldoctest pointytype
 julia> Point{Float64} <: Pointy{Real}
 false
+
+julia> Point{Float64} <: Pointy{<:Real}
+true
 ```
 
 What purpose do parametric abstract types like `Pointy` serve? Consider if we create a point-like
@@ -990,10 +1007,12 @@ Using explicit `where` syntax, any subset of parameters can be fixed. For exampl
 
 Type variables can be restricted with subtype relations.
 `Array{T} where T<:Integer` refers to all arrays whose element type is some kind of `Integer`.
+The syntax `Array{<:Integer}` is a convenient shorthand for `Array{T} where T<:Integer`.
 Type variables can have both lower and upper bounds.
 `Array{T} where Int<:T<:Number` refers to all arrays of `Number`s that are able to contain `Int`s
 (since `T` must be at least as big as `Int`).
-The syntax `where T>:Int` also works to specify only the lower bound of a type variable.
+The syntax `where T>:Int` also works to specify only the lower bound of a type variable,
+and `Array{>:Int}` is equivalent to `Array{T} where T>:Int`.
 
 Since `where` expressions nest, type variable bounds can refer to outer type variables.
 For example `Tuple{T,Array{S}} where S<:AbstractArray{T} where T<:Real` refers to 2-tuples whose first

src/julia-syntax.scm

@@ -1763,6 +1763,20 @@
       body
       (expand-where (expand-wheres body (cdr vars)) (car vars))))
 
+; given e = (curly T params...), return (newparams . whereparams) where any <:X expression
+; in params is converted to T and T<:X is added to whereparams; similarly for >:X.
+; (This implements the syntactic sugar Foo{<:Bar} --> Foo{T} where T<:Bar.)
+(define (extract-implicit-whereparams e)
+  (define (extract params newparams whereparams)
+    (if (null? params)
+        (cons (reverse newparams) (reverse whereparams))
+        (let ((p (car params)))
+          (if (and (list? p) (= (length p) 3) (eq? (car p) 'call) (or (eq? (cadr p) '|<:|) (eq? (cadr p) '|>:|)))
+              (let ((T (gensy)))
+                (extract (cdr params) (cons T newparams) (cons (list (cadr p) T (caddr p)) whereparams)))
+              (extract (cdr params) (cons p newparams) whereparams)))))
+  (extract (cddr e) '() '()))
+
 ;; table mapping expression head to a function expanding that form
 (define expand-table
   (table
@@ -1980,7 +1994,13 @@
            (expand-forms (partially-expand-ref e)))))
 
    'curly
-   (lambda (e) (expand-forms `(call (core apply_type) ,@(cdr e))))
+   (lambda (e)
+     (let* ((p (extract-implicit-whereparams e))
+            (curlyparams (car p))
+            (whereparams (cdr p)))
+       (if (null? whereparams)
+           (expand-forms `(call (core apply_type) ,@(cdr e)))
+           (expand-forms `(where (curly ,(cadr e) ,@curlyparams) ,@whereparams)))))
 
    'call
    (lambda (e)

test/subtype.jl

@@ -861,3 +861,30 @@ f18348{T<:Any}(::Type{T}, x::T) = 2
 # Issue #12721
 f12721{T<:Type{Int}}(::T) = true
 @test_throws MethodError f12721(Float64)
+
+# implicit "covariant" type parameters:
+type TwoParams{S,T}; x::S; y::T; end
+@test TwoParams{<:Real,<:Number} == (TwoParams{S,T} where S<:Real where T<:Number) ==
+      (TwoParams{S,<:Number} where S<:Real) == (TwoParams{<:Real,T} where T<:Number)
+@test TwoParams(3,0im) isa TwoParams{<:Real,<:Number}
+@test TwoParams(3,"foo") isa TwoParams{<:Real}
+@test !(TwoParams(3im,0im) isa TwoParams{<:Real,<:Number})
+@test !(TwoParams(3,"foo") isa TwoParams{<:Real,<:Number})
+ftwoparams(::TwoParams) = 1
+ftwoparams(::TwoParams{<:Real}) = 2
+ftwoparams(::TwoParams{<:Real,<:Real}) = 3
+@test ftwoparams(TwoParams('x',3)) == 1
+@test ftwoparams(TwoParams(3,'x')) == 2
+@test ftwoparams(TwoParams(3,4)) == 3
+@test !([TwoParams(3,4)] isa Vector{TwoParams{<:Real,<:Real}})
+@test TwoParams{<:Real,<:Real}[TwoParams(3,4)] isa Vector{TwoParams{<:Real,<:Real}}
+@test [TwoParams(3,4)] isa Vector{<:TwoParams{<:Real,<:Real}}
+@test [TwoParams(3,4)] isa (Vector{TwoParams{T,T}} where T<:Real)
+
+# implicit "contravariant" type parameters:
+@test TwoParams{>:Int,<:Number} == (TwoParams{S,T} where S>:Int where T<:Number) ==
+      (TwoParams{S,<:Number} where S>:Int) == (TwoParams{>:Int,T} where T<:Number)
+@test TwoParams(3,0im) isa TwoParams{>:Int,<:Number}
+@test TwoParams{Real,Complex}(3,0im) isa TwoParams{>:Int,<:Number}
+@test !(TwoParams(3.0,0im) isa TwoParams{>:Int,<:Number})
+@test !(TwoParams(3,'x') isa TwoParams{>:Int,<:Number})