Rounding solution output in JuMP

Specific Domains Optimization (Mathematical)
,
1

When I solve a problem in JuMP, my code ends with

optimize!(model);
    @show value(m);
end

and then it prints the solution m to the problem–as directed by the @show value(m) line. But it prints many decimal places of m. I only want 2 or 3 decimal places. How do I get it to give me only the desired number of decimal places?

I know I can use @show value(round(m,digits=3)) but that is annoying to write when I want to print out many variables in the solution since then I have to write the round function many times. Is there a way to just impose the rounding automatically for all variables in the solution?

Thanks

2

Not that I know of.

I believe the usual is to save all of them rounded in the same or other data structure. But you need to apply round at least one time.

2 Likes
3

You may want to save all your values in a data structure as @Henrique_Becker suggested, but to answer your question directly, you can write your own macro:


julia> macro show2(x)
                  quote
                      print($(QuoteNode(x)))
                      print(" = ")
                      println(round($x; digits=2))
                      $x
                  end
                  end
@show2 (macro with 1 method)

julia> @show sqrt(3);
sqrt(3) = 1.7320508075688772

julia> @show2 sqrt(3);
sqrt(3) = 1.73
1 Like
4

That’s great thanks! But how do out save all values to a data structure? I see in the docs it talks about containers
so I could do that, but then I think I would have to rewrite the model functions in terms of a vector (i.e. a container) for the variables.

Is there something where I could do like

data_structure_for_solution_vars =  optimize!(model).solution # ???
@show2 data_structure_for_solution_vars
5

I’m not sure what’s idiomatic in JuMP, but you could probably do something like

round2(x) = round(x; digits=2);
map(round2∘value, (;x,y,z))

if your parameters are x,y,z.

6

You can query a vector of all variables with all_variables(model), but this doesn’t preserve the container structure. You’ll need to write something custom.