Check for Feasibility of a Model about powermodels.jl

[jac0320]

How about:
Reconstruct model -> Fix vars by go through varDict with solution value -> objective = min 0.0 -> solve -> return status
?

[ccoffrin]

Given the current state of JuMP this might be the best solution.

In an ideal world, checking feasibility should not require sending the problem to a solver. If you know the constraints and have a value for every variable, then you should be able to easily evaluate the constraints in Julia to test for feasibility up to some numerical tolerance.

@mlubin, @chriscoey suggestions on this? It seems like you do similar checks in Pajarito.

[mlubin]

Well, you can extract the constraint data and check feasibility. The easiest way is likely to make a dummy solver that does this through MathProgBase.

[ccoffrin]

Sounds like a good solution to me.

[ccoffrin]

@mlubin is this an expected feature in the next version of JuMP/MOI?

[mlubin]

We will have the infrastructure to implement this much more easily in JuMP/MOI, but this isn't a blocker or high priority for the JuMP 0.19 release.