Bug with computed indexes of variables in set_bounds

[dirkschumacher]

set_bounds(u[e(k)], ub = 0, k = 1:m)

[dirkschumacher]

Hello past me, is this really a bug? Isn't that equal to set_bounds(u[k], ub = 0, k = e(1:m))?

[dirkschumacher]

Though this works with add_constraint. I guess that could be harmonized

[dirkschumacher]

  a <- identity
  m <- MILPModel() %>%
    add_variable(x[i], i = 1:3) %>%
    set_bounds(x[a(i)], i = 1:3, lb = 1)
  expect_equal(c(1, 1, 1), m@variables[[1]]@lb)

[rickyars]

how would one use this technique to set bounds if the computed indexes belong to a matrix?

[dirkschumacher]

Can you give me an example?

[rickyars]

Let's use your student assignment example and extend it. I want to create a matrix that says "student i cannot take course j". So for these (i,j) pairs, I want to set the upper-bound to zero. So how do I do this? I think this can be done by creating a matrix and adding a constraint.

Build the matrix:

blocked <- data.frame(
  crossing(i = 1:n, j = 1:m) %>%
    mutate(blocked = 1) %>%
    spread(j, blocked) %>%
    select(-i)
)
# student 10 cannot take course 4
blocked[10, 4] <- 0

# student 25 cannot take course 1 or 3
blocked[25, 1] <- 0
blocked[25, 3] <- 0

Build the model with additional constraint:

model <- MIPModel() %>%
  
  # 1 iff student i is assigned to course m
  add_variable(x[i, j], i = 1:n, j = 1:m, type = "binary") %>%
  
  # maximize the preferences
  set_objective(sum_expr(weight(i, j) * x[i, j], i = 1:n, j = 1:m)) %>%
  
  # we cannot exceed the capacity of a course
  add_constraint(sum_expr(x[i, j], i = 1:n) <= capacity[j], j = 1:m) %>% 
  
  # each student needs to be assigned to one course
  add_constraint(sum_expr(x[i, j], j = 1:m) == 1, i = 1:n) %>%

  # NEW: some students are blocked from certain courses
  add_constraint(x[i, j] <= blocked[i, j], i = 1:n, j = 1)

But it might be nice (and maybe in some instances preferable) to do this by setting the bounds on the decision variables. I'm not sure what this would look like though. E.g.,:

  # NEW: some students are blocked from certain courses
  set_bounds(x[i, j], ub = blocked[i, j])

Or maybe (with the 0/1 flipped in blocked[i,j]):

  # some students are blocked from certain courses
  set_bounds(x[blocked[i, j]], ub = 0, i = 1:n, j = 1:m)

[ghost]

Hey great work on this package! Looking forward to its release on CRAN.

Here's a related bug (I believe so) that I found today -

MIPModel() %>%
  add_variable(x[a, b], a=1:2, b=1:5, type = "integer", lb = 0, ub = 5) %>%
  set_objective(sum_expr(x[a, b], a=1:2, b=1:5), "max") %>%
  # add_constraint(x[i, i] == 0, i = 1:2) %>%
  set_bounds(x[i, i], i=1:2, ub = 0) %>%
  solve_model(with_ROI(solver = "glpk")) %>%
  get_solution(x[a, b])

   variable a b value
1         x 1 1     0
2         x 1 2     5
3         x 1 3     5
4         x 1 4     0
5         x 1 5     5
6         x 2 1     5
7         x 2 2     5
8         x 2 3     5
9         x 2 4     5
10        x 2 5     5

The expected output is x[1, 1] = 0 and x[2, 2] = 0. However I get x[1, 1] = 0 and x[1, 4] = 0.

I can get the correct results with add_constraint i.e. below code works fine

MIPModel() %>%
  add_variable(x[a, b], a=1:2, b=1:5, type = "integer", lb = 0, ub = 5) %>%
  set_objective(sum_expr(x[a, b], a=1:2, b=1:5), "max") %>%
  add_constraint(x[i, i] == 0, i = 1:2) %>%
  # set_bounds(x[i, i], i=1:2, ub = 0) %>%
  solve_model(with_ROI(solver = "glpk")) %>%
  get_solution(x[a, b])

   variable a b value
1         x 1 1     0
2         x 1 2     5
3         x 1 3     5
4         x 1 4     5
5         x 1 5     5
6         x 2 1     5
7         x 2 2     0
8         x 2 3     5
9         x 2 4     5
10        x 2 5     5

Any thoughts?

[dirkschumacher]

Thanks @shrinidhee indeed a bug. Will post a fix soon

[ShuchitaShukla]

hi @rickyars @dirkschumacher @shrinidhee

Does the below workf for you?

we cannot exceed the capacity of a course

add_constraint(sum_expr(x[i, j], i = 1:n) <= capacity[j], j = 1:m) %>%
Im not able to add the below line in my code:
model <- MIPModel() %>%
add_variable(Qtysold[i],i=1:n,type = "integer" ,lb=0 ) %>%
add_variable(Total_mat[i,j],i=1:n,j=1:68, type = "continuous", lb = 0) %>%
set_objective(sum_expr(Qtysold[i]*Revenueperunit[i],i=1:n), "max")%>%
add_constraint(Qtysold[i] <= qty_req[i], i = 1:n)%>%
add_constraint(Qtysold[i]*Raw_mat[i,j]==Total_mat[i,j],i=1:n,j=1:68)%>%
add_constraint(sum_expr(Total_mat[i,j],i=1:n)<=Inv_Stock[j],j=1:68)

The last line gives an error:Error in check_for_unknown_vars_impl(model, the_ast) :
The expression contains a variable that is not part of the model.

[dirkschumacher]

This seems to be fixed now with #130.