R Interface to the Ipopt Nonlinear Optimizer • ipopt

Lightweight R interface to the Ipopt nonlinear optimizer through Ipopt’s C interface.

See the Solving nonlinear programs with ipopt vignette for a guided tour of the objective, gradient, constraint Jacobian, and Hessian-of-the-Lagrangian callbacks, with worked examples (HS071, Rosenbrock, limited-memory, and a maximum-entropy problem).

The package links against an Ipopt library located at build time. On Linux and macOS this is the system Ipopt found with pkg-config; on Windows it is one of Ipopt’s official prebuilt release binaries (see below). The package is always installed from source against that locally available Ipopt.

Installation

A system Ipopt (and pkg-config) must be available; the configure script requires

pkg-config --cflags --libs ipopt

to succeed.

Linux

Install the system library, then build from source:

# Debian / Ubuntu
sudo apt-get install -y coinor-libipopt-dev pkg-config
R CMD INSTALL .

macOS (build from source)

Homebrew provides Ipopt, so macOS users build from source. The configure script locates the Homebrew Ipopt automatically:

brew install ipopt pkg-config

install.packages("ipopt", repos = "https://bnaras.github.io/drat", type = "source")

Equivalently, clone the repository and run R CMD INSTALL ..

Windows (prebuilt Ipopt binary)

Ipopt does not ship in the Windows R toolchain, but the Ipopt project publishes prebuilt Windows binaries that this package can link against. Although those binaries are built with MSVC, they export the Ipopt C interface as plain C symbols, so Rtools’ (MinGW) linker can use them directly.

Install Rtools (matching your R version).
Download the latest win64-msvs2022-md zip from Ipopt releases and extract it, for example to C:\Ipopt, giving you C:\Ipopt\include\coin-or, C:\Ipopt\lib, and C:\Ipopt\bin.
Build from source, pointing the package at that folder (use forward slashes):
```
R CMD INSTALL ipopt ^
  --configure-vars="INCLUDE_DIR=C:/Ipopt/include/coin-or LIB_DIR=C:/Ipopt/lib"
```
Equivalently, set IPOPT_HOME=C:/Ipopt in the environment before installing and the script will find the headers, library, and DLL itself.
At run time, the package loads ipopt-3.dll and its dependencies (coinmumps-3.dll and the Intel runtime DLLs, all in C:\Ipopt\bin). Make sure that folder is on the PATH for the R session, e.g.
```
Sys.setenv(PATH = paste("C:/Ipopt/bin", Sys.getenv("PATH"), sep = ";"))
library(ipopt)
```
(Add C:\Ipopt\bin to the system PATH to avoid doing this each session.)

Smoke Test

library(ipopt)

ipopt_solve(
  x0 = 0,
  lower = -Inf,
  upper = Inf,
  eval_f = function(x) (x[1] - 2)^2,
  eval_grad_f = function(x) 2 * (x[1] - 2),
  hessian_structure = matrix(c(1L, 1L), ncol = 2),
  eval_h = function(x, obj_factor, lambda) 2 * obj_factor,
  options = list(print_level = 0L)
)