jlbun - Using Julia in Bun

• jlbun - Using Julia in Bun
  • Installation
  • Usage
    • Evaluate Julia code
    • Pass a Bun array to Julia
    • Pass a Julia Array to Bun
    • Do some linear algebra
    • Install and use new packages
    • Function calls with keyword arguments
    • Call JS functions from Julia
    • Run Julia with multiple threads
  • TODO

Installation

• You need to have Bun, CMake and Julia installed to use this library.
• bun install does not run the packages's install scripts, so npm install is used instead.

npm install jlbun

Usage

Evaluate Julia code

jlbun provides Julia.eval() and Julia.tagEval(). The former accepts a string, while the latter accepts a tagged template literal.

import { Julia } from "jlbun";

Julia.init();

Julia.eval('println("Hello from Julia")'); // "Hello from Julia"

const arr = [1, 2, "hello"];
Julia.tagEval`println(${arr})`; // "Any[1, 2, "hello"]"

const arr2 = new Uint8Array([1, 2, 3]);
Julia.tagEval`println(${arr2})`; // "UInt8[0x01, 0x02, 0x03]"

const obj = { foo: 0, bar: false, bal: [1, 2, 3] }
Julia.tagEval`println(${obj})`; // "(foo = 0, bar = false, bal = Any[1, 2, 3])"

Julia.close();

Pass a Bun array to Julia

import { Julia, JuliaArray } from "jlbun";

// This initializes Julia and loads prelude symbols.
Julia.init();

// Create a `TypedArray` at the Bun side.
const bunArray = new Float64Array([1, 2, 3, 4, 5]);

// Create a `JuliaArray` from the `TypedArray`.
const juliaArray = JuliaArray.from(bunArray);

// These two arrays now share the same memory.
Julia.Base.println(juliaArray); // [1.0, 2.0, 3.0, 4.0, 5.0]

// We can modify the array at the Bun side (0-indexed).
bunArray[1] = 100.0;
Julia.Base.println(juliaArray); // [1.0, 100.0, 3.0, 4.0, 5.0]

// Or we can modify the array at the Julia side (also 0-indexed).
juliaArray.set(0, -10.0);
Julia.Base.println(juliaArray); // [-10.0, 100.0, 3.0, 4.0, 5.0]

// This cleans up Julia-related stuff.
Julia.close();

Pass a Julia Array to Bun

import { Julia } from "jlbun";

Julia.init();

const juliaArray = Julia.Base.rand(10, 10);
const bunArray = juliaArray.rawValue;
console.log(bunArray);

Julia.close();

Do some linear algebra

import { Julia } from "jlbun";

Julia.init();

const juliaArray = Julia.Base.rand(2, 2);
const inv = Julia.Base.inv(juliaArray);
console.log(inv.value);

const anotherJuliaArray = Julia.Base.rand(2, 2);
const product = Julia.Base["*"](juliaArray, anotherJuliaArray);
console.log(product.value);

// We can also import Julia modules.
const LA = Julia.import("LinearAlgebra");
console.log(LA.norm(product).value);

Julia.close();

Install and use new packages

import { Julia } from "jlbun";
import { join } from "path";

Julia.init();

// Install `CairoMakie`
Julia.Pkg.add("CairoMakie");

// Import `CairoMakie`
const Cairo = Julia.import("CairoMakie");

// Plot and save
const plt = Cairo.plot(Julia.Base.rand(10), Julia.Base.rand(10));
Cairo.save(join(process.cwd(), "plot.png"), plt);

Julia.close();

Function calls with keyword arguments

import { Julia, JuliaArray } from "jlbun";

Julia.init();

const rawArray = new Int32Array([1, 10, 20, 30, 100]);
const arr = JuliaArray.from(rawArray);
Julia.Base["sort!"].callWithKwargs({ by: Julia.Base.string, rev: true }, arr);
console.log(rawArray); // Int32Array(5) [ 30, 20, 100, 10, 1 ]

Julia.close();

Call JS functions from Julia

import { Julia, JuliaArray, JuliaFunction } from "jlbun";

Julia.init();

const jsFunc = (x: number) => -x;

const cb = JuliaFunction.from(jsFunc, {
  returns: "i32",
  args: ["i32"],
});

const arr = JuliaArray.from(new Int32Array([1, 10, 20, 30, 100]));
const neg = arr.map(cb);
Julia.println(neg); // Int32[-1, -10, -20, -30, -100]
cb.close();

Julia.close();

To deal with strings requires some tricks, see the example below:

import { Julia, JuliaArray, JuliaFunction, safeCString } from "jlbun";

Julia.init();

const jsFunc = (x: number) => {
  return safeCString(x.toString());
};

const cb = JuliaFunction.from(jsFunc, {
  returns: "cstring",
  args: ["i32"],
});

const arr = JuliaArray.from(new Int32Array([1, 10, 20, 30, 100]));
Julia.Base["sort!"].callWithKwargs({ by: cb, rev: true }, arr);
Julia.println(arr); // Int32[30, 20, 100, 10, 1]
cb.close();

Julia.close();

Run Julia with multiple threads

To use multiple threads in Julia, you need to set the JULIA_NUM_THREADS environment variable.

With export JULIA_NUM_THREADS=2 set, the following program should output 2 instead of 1:

import { Julia } from "jlbun";

Julia.init();

Julia.eval("println(Threads.nthreads())");

Julia.close();

A more advanced example:

import { Julia, JuliaFunction, JuliaTask, JuliaValue } from "jlbun";

Julia.init();

const promises: Promise<JuliaValue>[] = [];

const func = Julia.eval(`function ()
  ans = 0;
  for i in 1:1000
    ans += i
  end
  ans
end
`) as JuliaFunction;

for (let i = 0; i < Julia.nthreads; i++) {
  promises.push(JuliaTask.from(func).schedule(i).value);
}

const results = (await Promise.all(promises)).map(promise => promise.value);
console.log(results); // [ 500500n, 500500n, 500500n, 500500n, 500500n, 500500n, 500500n, 500500n ]

Julia.close();

TODO

• [x] Tuple
• [x] NamedTuple
  • [x] Keyword arguments of Julia functions
• [ ] Range
• [x] Dict
• [x] Set
• [x] Pair