yahiko00 / AssemblyScript

A subset of TypeScript that compiles to WebAssembly.

AssemblyScript defines a subset of TypeScript that it compiles to WebAssembly. It aims to provide everyone with an existing background in TypeScript and standard JavaScript-APIs with a comfortable way to compile to WebAssembly, eliminating the need to switch between languages or to learn new ones just for this purpose.

Try it out in your browser: dcode.io/AssemblyScript

Contents

• How it works
  A few insights to get an initial idea.

• What to expect
  General remarks on design decisions and trade-offs.

• Example
  Basic examples to get you started.

• Usage
  An introduction to the environment and its provided functionality.

• Command line
  How to use the command line utility.

• API
  How to use the API programmatically.

• Additional documentation
  A list of available documentation resources.

• Building
  How to build the compiler and its components yourself.

How it works

Under the hood, AssemblyScript rewires TypeScript's compiler API to Binaryen's compiler backend. The compiler itself is written in (and based upon) TypeScript and no binary dependencies are required to get started.

Every AssemblyScript program is valid TypeScript syntactically, but not necessarily semantically. The definitions required to start developing in AssemblyScript are provided by assembly.d.ts. See also: Usage

The compiler is able to produce WebAssembly binaries (.wasm) as well as their corresponding text format (.wast). Both Binaryen's s-expression format and, with a little help of WABT, official linear text format are supported. See also: CLI

What to expect

The most prominent difference of JavaScript and any strictly typed language is that, in JavaScript, a variable can reference a value of any type. This implies that a JavaScript execution environment has to emit additional runtime checks whenever a variable is accessed. Modern JavaScript VMs shortcut the overhead introduced by this and similar dynamic features by generating case-specific code based on statistical information collected just in time (JIT), speeding up execution significantly. Similarily, developers shortcut the overhead of remembering each variable's type by using TypeScript. The combination of both also makes for a good match because it potentially aids the JIT compiler.

Because it has the ability to fall back to dynamic JavaScript features, TypeScript isn't a strictly typed language after all. For example, TypeScript supports omittable (i.e. someParameter?: number) function parameters resulting in a union type number | undefined at runtime, just like it also allows declaring union types explicitly. These constructs are incompatible with a strict, ahead of time (AOT) compiled type system unless additional runtime checks are emitted that'd usually execute slower than similar code running in a VM that has the ability to perform optimizations at runtime. Hence...

TL;DR

Instead of reimplementing TypeScript as closely as possible at the expense of performance, AssemblyScript tries to support its features as closely as reasonable while not supporting certain dynamic constructs intentionally:

• All types must be annotated to avoid possibly unwanted implicit type conversions
• Optional function parameters require an initializer expression
• Union types, any and undefined are not supported by design
• The result of logical && / || expressions is always bool

Also note that AssemblyScript is a rather new and ambitious project developed by one guy and a hand full of occasional contributors. Expect bugs and breaking changes. Prepare to fix stuff yourself and to send a PR for it, unless you like the idea enough to consider sponsoring development.

Example

export function add(a: int, b: double): short {
  return (a + (b as int)) as short;
}

Compiles to:

(module
 (type $iFi (func (param i32 f64) (result i32)))
 (memory $0 256)
 (export "memory" (memory $0))
 (export "add" (func $add))
 (func $add (type $iFi) (param $0 i32) (param $1 f64) (result i32)
  (return
   (i32.shr_s
    (i32.shl
     (i32.add
      (get_local $0)
      (i32.trunc_s/f64
       (get_local $1)
      )
     )
     (i32.const 16)
    )
    (i32.const 16)
   )
  )
 )
)

See the pre-configured example project for a quickstart.

Running a module

The stand-alone loader component provides an easy way to run and work with compiled WebAssembly modules:

$> npm install assemblyscript-loader

import load from "assemblyscript-loader"; // JS: var load = require("assemblyscript-loader").load;

load("path/to/module.wasm", {
  imports: {
    ...
  }
}).then(module => {
  ...
  // i.e. call module.exports.main()
});

Usage

$> npm install assemblyscript --save-dev

The environment is configured by either referencing assembly.d.ts directly or by using a tsconfig.json that simply extends tsconfig.assembly.json, like so:

{
  "extends": "./node_modules/assemblyscript/tsconfig.assembly.json",
  "include": [
    "./*.ts"
  ]
}

The tsconfig.json-approach is recommended to inherit other important settings as well.

Once configured, the following AssemblyScript-specific types become available:

Type	Alias	Native type	sizeof	Description
sbyte	int8	i32	1	An 8-bit signed integer.
byte	uint8	i32	1	An 8-bit unsigned integer.
short	int16	i32	2	A 16-bit signed integer.
ushort	uint16	i32	2	A 16-bit unsigned integer.
int	int32	i32	4	A 32-bit signed integer.
uint	uint32	i32	4	A 32-bit unsigned integer.
long	int64	i64	8	A 64-bit signed integer.
ulong	uint64	i64	8	A 64-bit unsigned integer.
uintptr	-	i32 / i64	4 / 8	A 32-bit unsigned integer when targeting 32-bit WebAssembly. A 64-bit unsigned integer when targeting 64-bit WebAssembly.
float	float32	f32	4	A 32-bit float.
double	float64	f64	8	A 64-bit float.
bool	-	i32	1	A 1-bit unsigned integer.
void	-	none	-	No return type

While generating a warning to avoid type confusion, the JavaScript types number and boolean resolve to double and bool respectively.

WebAssembly-specific operations are available as built-in functions that translate to the respective opcode directly:

• rotl(value: int, shift: int): int
  Performs the sign-agnostic rotate left operation on a 32-bit integer.
• rotll(value: long, shift: long): long
  Performs the sign-agnostic rotate left operation on a 64-bit integer.
• rotr(value: int, shift: int): int
  Performs the sign-agnostic rotate right operation on a 32-bit integer.
• rotrl(value: long, shift: long): long
  Performs the sign-agnostic rotate right operation on a 64-bit integer.
• clz(value: int): int
  Performs the sign-agnostic count leading zero bits operation on a 32-bit integer. All zero bits are considered leading if the value is zero.
• clzl(value: long): long
  Performs the sign-agnostic count leading zero bits operation on a 64-bit integer. All zero bits are considered leading if the value is zero.
• ctz(value: int): int
  Performs the sign-agnostic count tailing zero bits operation on a 32-bit integer. All zero bits are considered trailing if the value is zero.
• ctzl(value: long): long
  Performs the sign-agnostic count trailing zero bits operation on a 64-bit integer. All zero bits are considered trailing if the value is zero.
• popcnt(value: int): int
  Performs the sign-agnostic count number of one bits operation on a 32-bit integer.
• popcntl(value: long): long
  Performs the sign-agnostic count number of one bits operation on a 64-bit integer.
• abs(value: double): double
  Computes the absolute value of a 64-bit float.
• absf(value: float): float
  Computes the absolute value of a 32-bit float.
• ceil(value: double): double
  Performs the ceiling operation on a 64-bit float.
• ceilf(value: float): float
  Performs the ceiling operation on a 32-bit float.
• floor(value: double): double
  Performs the floor operation on a 64-bit float.
• floorf(value: float): float
  Performs the floor operation on a 32-bit float.
• sqrt(value: double): double
  Calculates the square root of a 64-bit float.
• sqrtf(value: float): float
  Calculates the square root of a 32-bit float.
• trunc(value: double): double
  Rounds to the nearest integer towards zero of a 64-bit float.
• truncf(value: float): float
  Rounds to the nearest integer towards zero of a 32-bit float.
• nearest(value: double): double
  Rounds to the nearest integer tied to even of a 64-bit float.
• nearestf(value: float): float
  Rounds to the nearest integer tied to even of a 32-bit float.
• min(left: double, right: double): double
  Determines the minimum of two 64-bit floats. If either operand is NaN, returns NaN.
• minf(left: float, right: float): float
  Determines the minimum of two 32-bit floats. If either operand is NaN, returns NaN.
• max(left: double, right: double): double
  Determines the maximum of two 64-bit floats. If either operand is NaN, returns NaN.
• maxf(left: float, right: float): float
  Determines the maximum of two 32-bit floats. If either operand is NaN, returns NaN.
• copysign(x: double, y: double): double
  Composes a 64-bit float from the magnitude of x and the sign of y.
• copysignf(x: float, y: float): float
  Composes a 32-bit float from the magnitude of x and the sign of y.
• reinterpreti(value: float): int
  Reinterprets the bits of a 32-bit float as a 32-bit integer.
• reinterpretl(value: double): long
  Reinterprets the bits of a 64-bit float as a 64-bit integer.
• reinterpretf(value: int): float
  Reinterprets the bits of a 32-bit integer as a 32-bit float.
• reinterpretd(value: long): double
  Reinterprets the bits of a 64-bit integer as a 64-bit double.
• current_memory(): int
  Returns the current memory size in units of pages. One page is 64kb.
• grow_memory(value: uint): int
  Grows linear memory by a given unsigned delta of pages. One page is 64kb. Returns the previous memory size in units of pages or -1 on failure.
• unreachable(): void
  Emits an unreachable operation that results in a runtime error when executed.

The following AssemblyScript-specific operations are implemented as built-ins as well:

• sizeof<T>(): uintptr
  Determines the byte size of the specified core or class type. Compiles to a constant.
• unsafe_cast<T1,T2>(value: T1): T2
  Casts a value of type T1 to a value of type T2. Useful for casting classes to pointers and vice-versa. Does not perform any checks.
• isNaN(value: double): bool
  Tests if a 64-bit float is a NaN.
• isNaNf(value: float): bool
  Tests if a 32-bit float is a NaN.
• isFinite(value: double): bool
  Tests if a 64-bit float is finite.
• isFinitef(value: float): bool
  Tests if a 32-bit float is finite.

These constants are present as immutable globals (note that optimizers might inline them):

• NaN: double
  NaN (not a number) as a 64-bit float.
• NaNf: float
  NaN (not a number) as a 32-bit float.
• Infinity: double
  Positive infinity as a 64-bit float.
• Infinityf: float
  Positive infinity as a 32-bit float.

By default, standard memory management routines based on dlmalloc and musl will be linked statically and can be configured to be exported to the embedder:

• malloc(size: uintptr): uintptr
  Allocates a chunk of memory of the specified size and returns a pointer to it.
• free(ptr: uintptr): void
  Frees a previously allocated chunk of memory by its pointer.
• memcpy(dest: uintptr, src: uintptr, size: uintptr): uintptr
  Copies data from one chunk of memory to another.
• memset(dest: uintptr, c: int, size: uintptr): uintptr
  Sets a chunk of memory to the provided value c. Usually used to reset it to all 0s.
• memcmp(vl: uintptr, vr: uintptr, n: uintptr): int
  Compares a chunk of memory to another. Returns 0 if both are equal, otherwise vl[i] - vr[i] at the first difference's byte offset i.

Linking in memory management routines adds about 11kb to a module. Once WebAssembly exposes the garbage collector natively, there'll be other options as well. Note that the new operator depends on malloc and will break when --no-malloc is specified (and no other malloc is present). Also note that calling grow_memory where malloc is present will most likely break malloc as it expects contiguous memory.

Type coercion requires an explicit cast where precision or signage is lost respectively is implicit where it is maintained. For example, to cast a double to an int:

function example(value: double): int {
  return value as int; // translates to the respective opcode
}

Global WebAssembly imports can be declared anywhere while WebAssembly exports are exported from the entry file (the file specified when calling asc or Compiler.compileFile). Aside from that, imports and exports work just like in TypeScript.

// entry.ts

import { myOtherExportThatDoesntBecomeAWebAssemblyExport } from "./imported";

declare function myImport(): void;

export function myExport(): void {
  myOtherExportThatDoesntBecomeAWebAssemblyExport();
}

Currently, imports can also be pulled from different namespaces by separating the namespace and the function with a $ character.

declare function Math$random(): double;

Naming a function start with no arguments and a void return type will automatically make it the start function that is being called on load even before returning to the embedder.

function start(): void {
  ...
}

Command line

The command line compiler asc works similar to TypeScript's tsc:

Syntax: asc [options] entryFile

Options:

 --config, -c       Specifies a JSON configuration file with command line options.
                    Will look for 'asconfig.json' in the entry's directory if omitted.

 --outFile, -o      Specifies the output file name. Emits text format if ending with .wast
                    (sexpr) or .wat (linear). Prints to stdout if omitted.

 --optimize, -O     Runs optimizing binaryen IR passes.

 --validate, -v     Validates the module.

 --quiet, -q        Runs in quiet mode, not printing anything to console.

 --target, -t       Specifies the target architecture:

                    wasm32  Compiles to 32-bit WebAssembly [default]
                    wasm64  Compiles to 64-bit WebAssembly

 --memoryModel, -m  Specifies the memory model to use / how to proceed with malloc etc.:

                    malloc        Bundles malloc etc. [default]
                    exportmalloc  Bundles malloc etc. and exports each
                    importmalloc  Imports malloc etc. from 'env'
                    bare          Excludes malloc etc. entirely

 --textFormat, -f   Specifies the format to use for text output:

                    sexpr   Emits s-expression syntax (.wast) [default]
                    linear  Emits official linear syntax (.wat)

                    Text format only is emitted when used without --textFile.

 --textFile         Can be used to save text format alongside a binary in one command.

 --help, -h         Displays this help message.

A configuration file (usually named asconfig.json) using the long option keys above plus a special key entryFile specifying the path to the entry file can be used to reuse options between invocations.

API

It's also possible to use the API programmatically:

• Compiler.compileFile(filename: string, options?: CompilerOptions): binaryen.Module | null
  Compiles the specified entry file to a WebAssembly module. Returns null on failure.

• Compiler.compileString(source: string, options?: CompilerOptions): binaryen.Module | null
  Compiles the specified entry file source to a WebAssembly module. Returns null on failure.

• Compiler.lastDiagnostics: typescript.Diagnostic[]
  Contains the diagnostics generated by the last invocation of compilerFile or compileString.

• CompilerOptions
  AssemblyScript compiler options.

  • silent: boolean
    Whether compilation shall be performed in silent mode without writing to console. Defaults to false.
  • treeShaking: boolean
    Whether to use built-in tree-shaking. Defaults to true. Disable this when building a dynamically linked library.
  • target: CompilerTarget | string
    Specifies the target architecture. Defaults to CompilerTarget.WASM32.
  • memoryModel: CompilerMemoryModel | string
    Specifies the memory model to use. Defaults to CompilerMemoryModel.MALLOC.

• CompilerTarget
  Compiler target.

  • WASM32
    32-bit WebAssembly target using uint pointers.
  • WASM64
    64-bit WebAssembly target using ulong pointers.

• CompilerMemoryModel
  Compiler memory model.

  • BARE
    Does not bundle any memory management routines.
  • MALLOC
    Bundles malloc, free, etc.
  • EXPORT_MALLOC
    Bundles malloc, free, etc. and exports each to the embedder.
  • IMPORT_MALLOC
    Imports malloc, free, etc. as provided by the embedder.

Example

import { Compiler, CompilerTarget, CompilerMemoryModel, typescript } from "assemblyscript";

const module = Compiler.compileString(`
export function add(a: int, b: int): int {
  return a + b;
}
`, {
  target: CompilerTarget.WASM32,
  memoryModel: CompilerMemoryModel.MALLOC,
  silent: true
});

console.error(typescript.formatDiagnostics(Compiler.lastDiagnostics));
if (!module)
  throw Error("compilation failed");

module.optimize();

if (!module.validate())
  throw Error("validation failed");

const textFile = module.emitText();
const wasmFile = module.emitBinary();

...

module.dispose();

Remember to call binaryen.Module#dispose() once you are done with a module to free its resources. This is necessary because binaryen.js has been compiled from C/C++ and doesn't provide automatic garbage collection.

Additional documentation

AssemblyScript

• Standard Library Documentation
• API Documentation

WebAssembly

• WebAssembly Design Documents

Building

Clone the GitHub repository and install the development dependencies:

$> git clone https://github.com/dcodeIO/AssemblyScript.git
$> cd AssemblyScript
$> npm install

Afterwards, to build the distribution files to dist/, run:

$> npm run build

To build the documentation to docs/api/, run:

$> npm run docs

Running the tests (ideally on node.js >= 8):

$> npm test

---

License: Apache License, Version 2.0