Quick start

Hello world

Save the following as hello.wren:

System.print("hello, world")

Run it:

wlift hello.wren
// hello, world

That's the whole loop. wlift parses the source, builds MIR, runs the bytecode interpreter, and tiers up to the Cranelift JIT for any hot function. No build step, no config file.

A first class

class Counter {
  construct new(start) {
    _n = start
  }

  next {
    _n = _n + 1
    return _n
  }

  toString { "Counter(%(_n))" }
}

var c = Counter.new(0)
System.print(c.next)         // 1
System.print(c.next)         // 2
System.print(c)              // Counter(2)

Wrenlift accepts the stock Wren syntax verbatim, plus a few small additions documented in the reference.

REPL

Omit the file argument to drop into the REPL:

$ wlift
> var greet = Fn.new {|name| "hello, %(name)" }
> greet.call("wren")
"hello, wren"
> .exit

CLI flags you'll reach for

Flag	What it does
--mode interpreter	Skip the JIT entirely. Useful for debugging.
--mode tiered (default)	Interpret cold code, JIT-compile hot functions on the fly.
--mode jit	Compile everything up front before executing.
--aot <out>	Build a self-contained native binary. See below.
--bundle <out>	Pack a source tree into a .hatch distribution package.
--inspect	Print the manifest + section list of a .hatch file.
--dump-mir · --dump-opt · --dump-asm	Dump MIR / optimised MIR / generated machine code. Compiler debug.
--gc-stats	Print GC counters after the program finishes.
--gc <strategy>	generational (default), arena, or mark-sweep.

AOT — ship a single binary

wlift --aot compiles a source tree into a self-contained executable that runs without the wlift runtime installed.

wlift hello.wren --aot ./hello
./hello
// hello, world

The output is a real ELF / Mach-O / PE binary — ship it the same way you'd ship any other native program. Cross-compile by passing a target triple to wlift --aot (see the platform guides for per-OS notes on packaging and code signing).

Running a published package

Packages from the Hatch registry ship as .hatch bundles. Pass one directly to wlift:

wlift my-app.hatch

For a project with dependencies, use the hatch CLI:

$ hatch run path/to/my-app
  loading @hatch:http@0.3.4
  loading @hatch:json@0.1.4
  > standalone · serving :8080

Next steps

• Run Wrenlift in the browser — the wlift.js harness API.
• Embed in a Rust app — add wren_lift as a crate.
• Built-in reference — what's available beyond stock Wren.