I don’t understand this: julia> isa([1,2,3], Vector{Number}) false Each 1, 2, or 3 is certainly a Number: julia> isa(1, Number) true It is illogical to say [1, 2, 3] is not Vector{Number} when each element is a Number.

[image] Bryan_So: Will a Vulcan agree it is the logical thing to define? Probably not. Generally it is not very useful to consider programming language design decisions from the perspective of being “logical” — you can’t really derive them from basic principles of logic, so these criteria usu…

[image] Bryan_So: Your example is destructive. An auto promotion of some kind. Not what I’m expecting. Good example. There is no type-promotion in my example, as far as I understand. It just shows a function that works for Vector{Number} but fails for Vector{Int}. This means one cannot substi…

[image] Bryan_So: All these explanations fall into “it is what it is”. Mr. Nardelli explained in the video “It is a language design choice”. It’s not just arbitrary choice though. An object can be covariant only on types it “produces” and contravariant only on types it “consumes”. E.g., a f…

Julia’s subtyping is based on Set logic . In set logic it’s definitely possible that isa([white horse], Vector{<:Horse}) == true but isa([white horse], Vector{Horse}) == false.

The concept of subtyping is often defined via the Liskov substitution principle . Quoting from Wikipedia: Let \phi (x) be a property provable about objects x of type T. Then \phi (y) should be true for objects y of type S where S is a subtype of T. Speaking in looser terms, we can phrase the subs…

[image] CameronBieganek: that means that the main reason that parametric types are invariant in Julia is due to performance considerations Yes, it’s critical for performance. See my explanation here: Problem with Complex{Rationals} - #7 by stevengj

[image] stevengj: Yes, it’s critical for performance. See my explanation here: Problem with Complex{Rationals} - #7 by stevengj Whilst I understand the performance differences, I don’t fully get this: If you have a compiled function that operates on an Complex{Real} or an array of Complex{…

[image] greg_plowman: if Julia allowed subtyping from concrete types I am not sure what that would look like.

[image] greg_plowman: But Julia doesn’t allow subtyping from concrete types. Presumably there are good reasons for this. Imagine what this would mean for working with an array a of type Vector{Float64}, for example: Currently, this can be stored as a consecutive sequence of 8-byte chunks i…

I think the boolean question you are trying to ask is actually this: julia> typeof([1,2,3]) <: Vector{T} where T <: Number true or maybe simpler: julia> isa([1,2,3], Vector{<:Number}) true Also see this: julia> isa(Number[1,2,3], Vector{Number}) true

Why [1, 2, 3] is not a Vector{Number}?

New to Julia

stevengj January 2, 2021, 10:55pm 16

Yes, it’s critical for performance. See my explanation here: Problem with Complex{Rationals} - #7 by stevengj

Topic		Replies	Views
Why? isa([(x,1),(y,1)], Array{Tuple{Stuff,Number},1}) = false General Usage	25	1822	February 23, 2021
Need for explicitly defining subtypes in function arguments? New to Julia	30	1546	October 14, 2023
Simple question about Julia types New to Julia type	11	838	June 10, 2023
Why use subtypes instead of traits and duck typing? General Usage inheritance , traits	32	4639	April 15, 2021
Ambiguity on Vector{<:Abstract*}? General Usage	11	1338	April 28, 2018

Why [1, 2, 3] is not a Vector{Number}?

Related topics