In r-opt/GLPKoptimizer: GLPK Bindings for MOI

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

GLPKoptimizer

The goal of GLPKoptimizer is to provide MOI GLPK bindings. Still very experimental and incomplete.

GLPKoptimizer interacts directly with the solver by creating a pointer to a GLPK instance and modifying that pointer directly. Most modelling operations are directly passed through to the solver without making copies of the data in R.

Installation

You can install the development version of GLPKoptimizer like so:

remotes::install_github("r-opt/GLPKoptimizer")

Example

You can use set_irowgen_callback to register a callback with the solver, while still being able to use rmpk's modelling features.

library(rmpk)
library(GLPKoptimizer)
solver <- GLPK_optimizer(presolve = TRUE)
model <- optimization_model(solver)
x <- model$add_variable("x", type = "binary", i = 1:10)
model$set_objective(sum_expr(x[i], i = 1:10), sense = "max")
model$add_constraint(sum_expr(x[i], i = 1:10) <= 7.5)

# When an integer solution is found, we dynamically add a constraint
# further restricting the search space.
# In this case, we add another constraint of the sum of x <= 4
set_irowgen_callback(solver, function() {
  values <- vapply(1:10, function(i) {
    # in GLPK you can only access the values of the relaxation
    glpk_get_col_prim(solver, x[i])
  }, numeric(1L))
  all_integral <- all(values %% 1 == 0)
  if (all_integral && sum(values) > 4) {
    model$add_constraint(sum_expr(x[i], i = 1:10) <= 4)
  }
})

model$optimize()
model$get_variable_value(x[i])