experimentalalphaAOT JavaScript
Maligator is an experimental AOT JavaScript compiler, exploring its identity in public.
It already produces configurable native executables. It is not yet battle-tested, uniformly fast, or trying to be a drop-in Node replacement. The alpha is here to make the progress visible—and invite better ideas.
Progress, not promises
The graphs use committed project data. They exist to show movement and regressions—not to manufacture an “Nx faster” headline. Weeks without a trustworthy full result stay empty.
Authoritative compiled/normal baseline from
scripts/test262.json. Strict and sloppy results are
folded together; the chart always uses a true 0–100% scale.
| Date | Pass | Passed | Failed | Skipped | Commit |
|---|
Fresh-process median to load Maligator’s compiler source, compile it, and emit C. Startup and file I/O count; the C compiler and linker do not. Erasable TypeScript syntax is stripped before timing because the native compiler does not yet include that frontend. Node runs the same prepared source as a neutral reference.
| Date | Maligator | Node | Runs | Environment | Commit |
|---|
The honest alpha
A compatibility percentage is useful evidence, not a product identity. These are the capabilities that make Maligator interesting today—and the constraints that still define using it.
Compile JavaScript and erasable TypeScript applications ahead of time into standalone native programs.
Select eval, Realms, RegExp, Temporal, Intl services, web APIs, curated Node APIs, and Maligator’s own surface.
Embed files or selected directory trees in the executable and materialize the exact captured bytes at runtime.
An eval-enabled build embeds the compiler and executes dynamic JavaScript through the interpreter; eval-free builds leave that payload out.
More than fifty thousand applicable tests pass in the current strict-and-sloppy compiled baseline.
Maligator is only now being used in real project work. That is exposing bugs, rough edges, and assumptions quickly.
The Node surface is intentionally curated: enough compatibility to build useful things, not a promise to reproduce all of Node.js.
Some focused paths are promising. Other ordinary JavaScript constructs remain much slower than Node. Performance is uneven.
The project does not yet know whether its strongest identity will be tiny deployment, configurable runtimes, self-hosting, performance, or something not found yet.
The current packaged alpha targets macOS arm64 and remains unsigned. Other deployment targets are not presented as ready.
Dogfooding
The homepage is one self-contained HTML file. A small Maligator program embeds it as an asset and serves it from the resulting executable behind plain HTTP—ready for a future TLS-terminating reverse proxy.
import { readFileSync } from "node:fs";
const html = readFileSync(
mal.assets.materialize("site"),
"utf8",
);
Mal.serve({
port: Number(process.env.PORT ?? "3000"),
fetch() {
return new Response(html, {
headers: { "Content-Type": "text/html; charset=utf-8" },
});
},
});export default defineBuild({
assets: {
site: { type: "file", path: "website/index.html" },
},
engine: { eval: false, regexp: false },
surface: { webPlatform: true, node: true },
});Ideas welcome
Follow the progress, reply with an idea, or open an issue with something concrete. The alpha’s most useful outcome is discovering the project’s identity through real use.