# 
watr [](https://npmjs.org/watr) [](https://github.com/dy/watr/actions/workflows/test.js.yml) [](https://github.com/dy/watr/actions/workflows/spec.yml)
_Light & fast WAT compiler_
* [feature](https://webassembly.org/features/?categories=tools) & [spec](https://webassembly.github.io/spec/core/text/index.html)-complete, zero deps
* [compile](./docs.md#compilesource), [polyfill](./docs.md#polyfillast-options), [optimize](./docs.md#optimizeast-options), [parse](./docs.md#parsesource-options), [prettify](./docs.md#printtree-options), [minify](./docs.md#printtree-options)
* instant wasm, JS interop, CLI, [source maps](./docs.md#source-maps), clear errors
**[site](https://dy.github.io/watr)** / **[docs](./docs.md)** / **[playground](https://dy.github.io/watr/play/)**
## Usage
```js
import watr, { compile, parse, print } from 'watr'
// compile to binary
const src = '(func (export "f") (result f64) (f64.const 1))'
const binary = compile(src)
const module = new WebAssembly.Module(binary)
const { f } = new WebAssembly.Instance(module).exports
// optional, heavy transforms ship as separate entries — compose them
import optimize from 'watr/optimize' // fold constants, treeshake, eliminate dead code …
import polyfill from 'watr/polyfill' // newer features → MVP
compile(optimize(polyfill(src)))
// parse
parse('(i32.const 42)') // ['i32.const', '42'], with a .loc source offset
parse('(i32.const 42)', { loc: false }) // omit offsets for generated WAT
// print
print('(module(func(result i32)i32.const 42))') // (module\n (func (result i32)\n ...
// instant wasm function
const { add } = watr`(func (export "add") (param i32 i32) (result i32)
(i32.add (local.get 0) (local.get 1))
)`
add(2, 3) // 5
// instant wasm: interpolate, auto-import ...
const { test } = watr`(func (export "test") (call ${console.log} (i32.const 42)))`
test() // logs 42
```
## CLI
```sh
npx watr input.wat # → input.wasm
npx watr input.wat -o out.wasm # custom output
npx watr input.wat --print # pretty-print
npx watr input.wat --minify # minify
npx watr input.wat --optimize # fold, treeshake, inline, coalesce, …
npx watr input.wat --polyfill # newer features → MVP
npx watr input.wat --source-map # + input.wasm.map from ;;@ file:line:col comments
```
## Metrics
* **watr** — ~44 KB minified (~14 KB gzipped), 3,123 op/s
* [spec/wast.js](https://github.com/WebAssembly/spec) — 216 KB, 1,509 op/s
* [wabt](https://github.com/WebAssembly/wabt) — 282 KB, 859 op/s
* [binaryen](https://github.com/WebAssembly/binaryen) — 1,100 KB, 473 op/s
* [wat-compiler](https://github.com/stagas/wat-compiler) — ~152 KB (+ wabt dep), 355 op/s
### Optimizer vs binaryen (wasm-opt 128)
| | size (total) | time (batch) | footprint |
|---|---|---|---|
| **watr/optimize** | **19,734 B** | **~110 ms** in-process | 149 KB min (46 KB gz) |
| `wasm-opt -Oz` | 19,852 B | 990 ms CLI | ~1.1 MB js / native binary |
| `wasm-opt -O3` | 22,302 B | — | — |
Measured on [test/example](./test/example) (21 modules, `optimize(src)` defaults vs `wasm-opt -all`).
Smaller than `-O3` on every module and than `-Oz` in aggregate (19,734 vs 19,852 B); ties or beats `-Oz` on 18 of 20 files.
On a 5.5 MB module watr's output is 1.7% SMALLER than `-Oz` (261,345 vs 265,977 B — cross-function outlining);
binaryen's native core is faster there (0.44 s vs ~2.6 s optimize + 0.19 s compile) — watr's edge is output size, batch/in-process use
with no process spawn, and a ~25× smaller footprint.
## Used by
* [jz](https://github.com/dy/jz) – minimal static JS subset
ॐ