@casadi/casadi-wasm

CasADi compiled to WebAssembly. Symbolic framework for algorithmic differentiation and numerical optimization.

Install

npm install @casadi/casadi-wasm

Use (Node.js)

const create = require('@casadi/casadi-wasm');

(async () => {
  const ca = await create();                 // async: instantiates the wasm module
  const x = ca.SX.sym('x', 2);
  const [x0, x1] = ca.vertsplit(x);
  const nlp = { x, f: ca.plus(ca.times(x0, x0), ca.times(x1, x1)),
                   g: ca.minus(ca.plus(x0, x1), ca.SX(10)) };

  await ca.load_nlpsol('ipopt');             // on-demand plugin load (async)
  const solver = ca.nlpsol('solver', 'ipopt', nlp);
  const sol = solver.call({ lbg: ca.DM(0) });
  console.log(sol['x'].nonzeros());         // [5, 5]
})();

JavaScript has no operator overloading — use ca.plus, ca.minus, ca.times, etc. instead of + - *. Solver plugins (ipopt, fatrop, sundials integrators, ...) ship as sibling .so side modules and load on demand via await ca.load_<type>('<name>').

See examples/ for ports of the CasADi documentation examples.

Plugin availability

plugins.json lists the plugins required by this build, including explicit exclusions and their reasons. A configured plugin without a Wasm recipe or exclusion is a CMake error. Solver dependencies are linked transitively from the ordinary CMake plugin targets; the CasADi core stays in the main module.

In browsers, await load_nlpsol, load_conic, load_linsol, load_blas, load_filesystem, or the other load_* functions before constructing a solver. Class entry points such as Linsol.load_plugin and XmlFile.load_plugin also return promises. Concurrent loads of the same plugin share one request; failed loads can be retried.