Consider "light" use of sparse user-supplied MIP solution

[jajhall]

Taking a sparse user-supplied MIP solution, fixing the corresponding variables, and solving the resulting MIP to optimality may be very expensive, and not worth the investment.

Introduce a "light" options, where primal heuristics are used to (try to) use the user-supplied MIP solution to find an integer feasible solution

[mathgeekcoder]

FYI: @jajhall, I'm not sure if you remember, but HighsCallbackInput::repairSolution() will somewhat do this already. When the user provides a partial solution in a callback, this will call the solver to find a solution but terminates on the first feasible solution (rather than optimal).

Is this similar to what you were thinking, or am I misinterpreting?

[jajhall]

FYI: @jajhall, I'm not sure if you remember, but HighsCallbackInput::repairSolution() will somewhat do this already. When the user provides a partial solution in a callback, this will call the solver to find a solution but terminates on the first feasible solution (rather than optimal).

Is this similar to what you were thinking, or am I misinterpreting?

That was one thing in my mind, but a discussion last week with @odow revolved around a very much lighter touch. Apparently commercial solvers do little to exploit a partial MIP solution - probably reflecting its value in finding an optimal solution of the original MIP.