Oscar Dowson

YALMIP has a great collection of debugging tips: * https://yalmip.github.io/debugginginfeasible * https://yalmip.github.io/debuggingunbounded * https://yalmip.github.io/inside/debuggingnumerics/ * https://yalmip.github.io/inside/debuggingnonsymmetricsquare/ * https://yalmip.github.io/inside/debug/ * https://yalmip.github.io/naninmodel * https://yalmip.github.io/slowmilp * https://yalmip.github.io/fixedconstraints

I've merged a tutorial: https://jump.dev/JuMP.jl/dev/tutorials/getting_started/debugging/ From discussion with @ccoffrin, one thing we could add is a modeling suggestion to structure models so that they have complete recourse (that is, they...

I can reproduce. An elusive bug in JuMP! This is my fault in #2978. I'll take a look.

Annoyingly, it works if the indices are `AbstractUnitRange`: ```Julia julia> x = Containers.@container([i=1:2], 0) 2-element Vector{Int64}: 0 0 julia> view(x, :) 2-element view(::Vector{Int64}, :) with eltype Int64: 0 0 julia>...

Seems like over-riding `view` for all instances would break existing code that currently works. But I don't know how to intercept a method where the method signature doesn't match. I...

For example, this is clearly wrong: ```Julia julia> c = Containers.@container([i=1:4, j=2:3], i + j) 2-dimensional DenseAxisArray{Int64,2,...} with index sets: Dimension 1, Base.OneTo(4) Dimension 2, 2:3 And data, a 4×2...

> Maybe we should use AbstractArray to allow SubArray as data storage? I'd prefer not too, because we'd need a new type parameter in `DenseAxisArray`. But I'm not really sure...

@metab0t what is the motivation for using a `view`? When is it useful?

Sure. But where does this show up as a problem when modeling? On Tue, 2 Aug 2022, 7:18 pm Y. Yang, ***@***.***> wrote: > For performance reasons: > > julia>...

Less convenient, but you could go something like ```julia for (i, key) in enumerate(axes(variables, 1)) @constraint( model, LinearAlgebra.dot(f(key), view(variables.data, i, :))