Júlio Hoffimann (Jan 17 2025 at 21:45): Júlio Hoffimann (Jan 17 2025 at 21:45):I am writing n-dimensional algorithms and it would be convenient to call this function with either numbers or matrices.
Júlio Hoffimann (Jan 17 2025 at 21:47):Could this method be added to Base? Any counter-argument?
jar (Jan 17 2025 at 21:50):If you're working on n-dimensional arrays and you need to support the n=0 case, Julia has a 0-dimensional array type Array{T, 0}:
julia> fill(4)
0-dimensional Array{Int64, 0}:
4
that might work for your application.
Júlio Hoffimann (Jan 17 2025 at 22:07):Would it have the same performance benefits of a Number though? Stack allocation?
Júlio Hoffimann (Jan 17 2025 at 22:07):Ideally the algorithm would retain its performance with Number inputs
Fredrik Bagge Carlson (Jan 20 2025 at 19:26):mydiag(x::AbstractArray) = diag(x)
mydiag(x::Number) = x
Yes. My question is if there is a problem in defining this in Base?
Mason Protter (Jan 20 2025 at 19:44):I don't really see why we'd put this in base. I think it'd be better to just define your own mydiag like Fredrik showed
Júlio Hoffimann (Jan 20 2025 at 19:53):Mathematically speaking, doesn't it make sense to extend the method?
Júlio Hoffimann (Jan 20 2025 at 19:55):I've seen Number used in places that expect matrices or vectors as inputs.
Mason Protter (Jan 20 2025 at 20:19):Numbers are not matrices or vectors. They're zero dimensional objects. diag is specifically defined on matrices
Júlio Hoffimann (Jan 20 2025 at 20:22):I thought that numbers could be seen as linear algebras over themselves.
Júlio Hoffimann (Jan 20 2025 at 20:23):In that case one can introduce a coord system and identify the number with itself. Consequently defining diag too
Júlio Hoffimann (Jan 20 2025 at 20:23):But if people already see a problem with this on Zulip. The chances it gets accepted in Base are low
Last updated: Jan 29 2025 at 04:38 UTC