Mateus Araújo
Mateus Araújo
Indeed, put that into `.JuliaFormatter.toml`, call `format`, and see it fail.
My syntax highlighting really doesn't like "nothing" without the quotation marks in a TOML file, so maybe this violates the format specification?
That's what the PR is doing.
I know, I'm not changing the code implementation, I'm only changing the documentation (and the style file `pretty.jl`).
I don't know, I don't understand the architecture of MOI. I guessed that one needed to define the complex cone in MOI because it has both `PositiveSemidefiniteConeTriangle` and `HermitianPositiveSemidefiniteConeTriangle`. To...
I see, so the only way to get a nice interface is to define the complex cones in JuMP. For the 2-norm cone it doesn't seem that a MOI version...
Well it's what I wrote in the first comment; if `x` is a complex vector, there's no way of rewriting `t >= norm(x,Inf)` in terms of the real infinity norm...
Yep, that's it, although I think the most convenient definition would be `[t, real(x[1]), imag(x[1]), real(x[2]), imag(x[2]), ...]` instead of `[t; real.(x); imag.(x)]`.
Because that's how Hypatia does it, so if you define it the same way no reshuffling will be necessary. I'm not a fan of the indexing in `HermitianPositiveSemidefiniteConeTriangle`, it drove...
I was referring to [this comment](https://github.com/jump-dev/Hypatia.jl/pull/822#issue-1648714960) and [this bug](https://github.com/jump-dev/Hypatia.jl/pull/826). In any case I wasn't suggesting that the indexing should be changed, I was just saying it is not an example...