Generating a simple problem that reliably triggers hits :UserLimit

[vtjeng]

I'm on Julia 1.4 but using an older version of the JuMP and Gurobi packages (I'm planning to upgrade soon, but one of my other packages constrains me to using JuMP v0.18.6.):

  [2e9cd046] Gurobi v0.8.1
  [4076af6c] JuMP v0.18.6

When solving an MILP via JuMP using the GurobiSolver, I've found that getobjbound appears to ignore any constant term (perhaps related to this issue) if the objective is a JuMP.AffExpr when my solve hits a :UserLimit. I haven't been able to confirm whether this is an issue if a) my solve does not hit a :UserLimit or b) for getobjectivevalue. (I can get around this issue by adding an additional variable constrained to be equal to the objective and optimizing over the variable instead.)

I wanted to add a test (although I'm still not sure which package that test should go into) to ensure that this is no longer the case, but I haven't been able to reproduce this with a simple example as I can't figure out how to reliably trigger :UserLimit with a small problem. Does anyone have a suggestion for a problem that is simple to understand but for which I can get the :UserLimit status?

[odow]

You could do something like:

Gurobi.jl/test/MOI/MOI_callbacks.jl

Lines 27 to 53 in 183a1c5

	function callback_knapsack_model()
	model = Gurobi.Optimizer(
	GUROBI_ENV,
	OutputFlag = 0,
	Cuts = 0,
	Presolve = 0,
	PreCrush = 1,
	Heuristics = 0
	)
	N = 30
	x = MOI.add_variables(model, N)
	MOI.add_constraints(model, MOI.SingleVariable.(x), MOI.ZeroOne())
	MOI.set.(model, MOI.VariablePrimalStart(), x, 0.0)
	Random.seed!(1)
	item_weights, item_values = rand(N), rand(N)
	MOI.add_constraint(
	model,
	MOI.ScalarAffineFunction(MOI.ScalarAffineTerm.(item_weights, x), 0.0),
	MOI.LessThan(10.0)
	)
	MOI.set(
	model,
	MOI.ObjectiveFunction{MOI.ScalarAffineFunction{Float64}}(),
	MOI.ScalarAffineFunction(MOI.ScalarAffineTerm.(item_values, x), 0.0)
	)
	MOI.set(model, MOI.ObjectiveSense(), MOI.MAX_SENSE)
	return model, x, item_weights

but set a time limit of 0 seconds or a solution limit of 1 or something.

[vtjeng]

For users who simply want to trigger a particular termination status from the optimizer for testing purposes (but are not interested in it being a Gurobi.Optimizer in particular), you can use the suggestion by @mlubin here.