Closed matbesancon closed 10 months ago
One thing, I never returned nothing for the gap, since it starts with a default value that is not an obvious flag value (0.001 and 1e-6 for relative and absolute resp.), comparing against it makes the MOI wrapper potentially dependent on the HiGHS version
more broadly than this PR, this is a consequence of treating values that have been set differently than the default of the solver (I'm not a big fan of e.g. MOI.ListOfModelAttributesSet
because ultimately, it pushes down on most solvers to keep track of what has been set or not if they have a non-flag default value set by default
The original motivation for nothing
was that some solvers don't have a way of querying the default.
For MOI v2, I think we could make get
return nothing
OR the solver's default. set(..., nothing)
should probably restore the default though.
adds the absolute and relative gap limit attributes to HiGHS.