Research
Security News
Quasar RAT Disguised as an npm Package for Detecting Vulnerabilities in Ethereum Smart Contracts
Socket researchers uncover a malicious npm package posing as a tool for detecting vulnerabilities in Etherium smart contracts.
@thi.ng/wasm-api-bindgen
Advanced tools
Polyglot bindings code generators for hybrid JS & WebAssembly projects
This project is part of the @thi.ng/umbrella monorepo.
Polyglot bindings code generators for hybrid JS & WebAssembly projects. This is a support package for @thi.ng/wasm-api.
Without any additional help, current data exchange between a WebAssembly module and the JS/TS host application is restricted to simple numeric values. Not even strings can be directly passed between the two worlds. For even the most basic non-Hello-World style applications this is very cumbersome and insufficient.
The package provides an extensible code generation framework to simplify the bilateral design & exchange of data structures shared between the WASM & JS host env. Currently, code generators for the following languages are supplied:
These code generators derive their outputs from a single source of truth, a user provided JSON file of shared type definitions and optional additional configuration, e.g. to configure string behavior and/or provide custom user code to inject into the generated source code. Please see the @thi.ng/wasm-api-dom support package for a more thorough realworld example...
Currently, the code generators support top level types: enums, function pointers, structs and unions. See API docs for supported options & further details:
Enum
EnumValue
(individual enum value spec)Field
(individual spec for values contained in structs/unions)FuncPointer
Struct
Union
TopLevelType
Struct field types can be any of the supported WASM primitives or other user
defined types in the same JSON spec. In all cases, each field's base type can be
customized via the tag
, len
, const
and sentinel
options.
i8
, i16
, i32
, i64
, u8
, u16
, u32
, u64
f32
, f64
The following types are always available too, but are treated specially in some or all languages (explained in more detail further below):
opaque
- pointer to opaque data (e.g. void*
in C or *anyopaque
in Zig)string
- configurable string abstraction (see dedicated section in this readme)Base type | Tag | Length | Const | Equiv Zig type signature | Description |
---|---|---|---|---|---|
Foo | Foo | A single Foo | |||
Foo | array | N | [N]Foo | Array of N Foo | |
Foo | slice | []Foo | Slice of Foo (1) | ||
Foo | slice | true | []const Foo | Slice of readonly Foo (1) | |
Foo | ptr | *Foo | Pointer to a single Foo | ||
Foo | ptr | true | *const Foo | Pointer to a single readonly Foo | |
Foo | ptr | N | *[N]Foo | Pointer to N Foo | |
Foo | ptr | 0 | [*]Foo | Pointer to multiple Foo (3) | |
Foo | ptr | 0 | true | [*]const Foo | Pointer to multiple readonly Foo (3) |
f32 | vec | N | @Vector(N, f32) | Vector of N f32 (2) |
Most low-level languages deal with strings very differently and alas there's no
general standard. Some have UTF-8/16 support, others don't. In some languages
(incl. C & Zig), strings are stored (by default) as zero terminated char
sequence, in others they aren't... It's outside the scope of this package to
provide an allround out-of-the-box solution. The WasmBridge
runtime API
provides read & write accessors to obtain JS strings from UTF-8 encoded WASM
memory. See
getString()
and
setString()
for details.
The code generators check a global stringType
option to interpret the built-in
string
type of a struct field in different ways:
ptr
(default): Considers a string as C-style char*
pointer
(zero-terminated, but without any explicit length)slice
: Considers strings as Zig-style slices (i.e. pointer + length)Regardless of implementation choice (and in opposite fashion to all other
types), the default for strings is const
aka readonly... If mutable strings
are required, set const
field option to false
.
This is the default behavior/implementation for string
:
See
C/C++
and
Zig
types for definitions of StringPtr
and ConstStringPtr
et al...
Base type | Tag | Length | Const | Equiv Zig type signature | Description |
---|---|---|---|---|---|
string | ConstStringPtr | Single readonly string | |||
string | false | StringPtr | Single mutable string | ||
string | array | N | [N]ConstStringPtr | Array of N readonly strings | |
string | array | N | false | [N]StringPtr | Array of N mutable strings |
string | slice | ConstStringPtrSlice | Slice of readonly strings | ||
string | slice | false | StringPtrSlice | Slice of mutable strings | |
string | ptr | *ConstStringPtr | Pointer to a single readonly string | ||
string | ptr | false | *StringPtr | Pointer to a single mutable string | |
string | ptr | N | *[N]ConstStringPtr | Pointer to N readonly strings | |
string | ptr | N | false | *[N]StringPtr | Pointer to N mutable strings |
If the global stringType
option is set to slice
:
Base type | Tag | Length | Const | Equiv Zig type signature | Description |
---|---|---|---|---|---|
string | ConstString | Single readonly string | |||
string | false | String | Single mutable string | ||
string | array | N | [N]ConstString | Array of N readonly strings | |
string | array | N | false | [N]String | Array of N mutable strings |
string | slice | ConstStringSlice | Slice of readonly strings | ||
string | slice | false | StringSlice | Slice of mutable strings | |
string | ptr | *ConstString | Pointer to a single readonly string | ||
string | ptr | false | *String | Pointer to a single mutable string | |
string | ptr | N | *[N]ConstString | Pointer to N readonly strings | |
string | ptr | N | false | *[N]String | Pointer to N mutable strings |
In many languages a "slice" is a typed view of a memory region: A coupling of a
pointer to a start item and a given length (number of items). Of the languages
currently targeted by this package, only Zig has native support for this
concept, however forbids using slices in so-called extern struct
s (which are
the struct type used for interop and required for guaranteed memory layouts).
Therefore, all slices used here will be emulated using simple auto-generated wrapper structs, like:
// C
typedef struct { char* ptr; size_t len; } String;
// Zig
pub const String = extern struct { ptr: [*:0]u8, len: usize };
The TypeScript codegen will emit slices as JS arrays and doesn't support manipulation of a slice itself at current.
Should there ever be a need for manual padding inside a struct or union
definition, the following field spec can be used: { pad: N }
, where N is the
number of bytes to use for the empty space... Names for these special purpose
fields will be autogenerated and all other field options are ignored.
The package provides a detailed schema to aid the authoring of type definitions
(and provide inline documentation) via editors with JSON schema integration. The
schema is distributed as part of the package and located in
/schema/wasm-api-types.json
.
For VSCode, you can add this snippet to your workspace
settings
to apply the schema to any typedefs.json
files:
"json.schemas": [
{
"fileMatch": ["**/typedefs.json"],
"url": "./node_modules/@thi.ng/wasm-api-bindgen/schema/wasm-api-types.json"
}
]
The package includes a small CLI wrapper to invoke the code generator(s) from JSON type definitions and to write the generated source code(s) to different files:
$ npx @thi.ng/wasm-api
█ █ █ │
██ █ │
█ █ █ █ █ █ █ █ │ @thi.ng/wasm-api-bindgen 0.1.0
█ █ █ █ █ █ █ █ █ │ Multi-language data bindings code generator
█ │
█ █ │
usage: wasm-api-bindgen [OPTS] JSON-INPUT-FILE(S) ...
wasm-api-bindgen --help
Flags:
-d, --debug enable debug output & functions
--dry-run enable dry run (don't overwrite files)
Main:
-a FILE, --analytics FILE output file path for raw codegen analytics
-c FILE, --config FILE JSON config file with codegen options
-l ID[,..], --lang ID[,..] [multiple] target language: "c11", "ts", "zig" (default: ["ts","zig"])
-o FILE, --out FILE [multiple] output file path
-s TYPE, --string TYPE Force string type implementation: "slice", "ptr"
By default, the CLI generates sources for TypeScript and Zig (in this order!). Order is important, since the output file paths must be given in the same order as the target languages. It's recommended to be explicit with this. An example invocation looks like:
wasm-api-bindgen \
--config config.json \
--lang ts -o src/generated.ts \
--lang zig -o src.zig/generated.zig \
typedefs.json
The structure of the config file is as follows (all optional):
{
"global": { ... },
"c11": { ... },
"ts": { ... },
"zig": { ... },
}
More details about possible
global
,
c
and
ts
and
zig
config
options & values.
All code generators have support for custom code prologues & epilogues which can
be specified via the above options. These config options exist for both non-CLI
& CLI usage. For the latter, these custom code sections can also be loaded from
external files by specifying their file paths using @
as prefix, e.g.
{
"ts": { "pre": "@tpl/prelude.ts" },
"zig": { "pre": "@tpl/prelude.zig", "post": "@tpl/epilogue.zig" },
}
The following example defines 1x enum, 2x structs and 1x union. Shown here are the JSON type definitions and the resulting source codes:
⬇︎ CLICK TO EXPAND EACH CODE BLOCK ⬇︎
[
{
"name": "EventType",
"type": "enum",
"tag": "u8",
"doc": "Supported event types",
"values": [
{ "name": "mouse", "value": 1, "doc": "Any kind of mouse event" },
{ "name": "key", "doc": "Key down/up event" },
"misc"
]
},
{
"name": "MouseEvent",
"type": "struct",
"tag": "extern",
"doc": "Example struct",
"fields": [
{ "name": "type", "type": "EventType" },
{ "name": "pos", "type": "u16", "tag": "array", "len": 2 }
]
},
{
"name": "KeyEvent",
"type": "struct",
"tag": "extern",
"doc": "Example struct",
"fields": [
{ "name": "type", "type": "EventType" },
{ "name": "key", "type": "string", "doc": "Name of key which triggered event" },
{ "name": "modifiers", "type": "u8", "doc": "Bitmask of active modifier keys" }
]
},
{
"name": "Event",
"type": "union",
"tag": "extern",
"fields": [
{ "name": "mouse", "type": "MouseEvent" },
{ "name": "key", "type": "KeyEvent" }
]
}
]
/**
* Generated by @thi.ng/wasm-api-bindgen at 2022-11-23T16:55:22.188Z
* DO NOT EDIT!
*/
// @ts-ignore possibly includes unused imports
import { MemorySlice, Pointer, WasmStringPtr, WasmTypeBase, WasmTypeConstructor } from "@thi.ng/wasm-api";
/**
* Supported event types
*/
export enum EventType {
/**
* Any kind of mouse event
*/
MOUSE = 1,
/**
* Key down/up event
*/
KEY,
MISC,
}
/**
* Example struct
*/
export interface MouseEvent extends WasmTypeBase {
type: EventType;
/**
* WASM type: [2]u16
*/
pos: Uint16Array;
}
export const $MouseEvent: WasmTypeConstructor<MouseEvent> = (mem) => ({
get align() {
return 2;
},
get size() {
return 6;
},
instance: (base) => {
return {
get __base() {
return base;
},
get __bytes() {
return mem.u8.subarray(base, base + 6);
},
get type(): EventType {
return mem.u8[base];
},
set type(x: EventType) {
mem.u8[base] = x;
},
get pos(): Uint16Array {
const addr = (base + 2) >>> 1;
return mem.u16.subarray(addr, addr + 2);
},
};
}
});
/**
* Example struct
*/
export interface KeyEvent extends WasmTypeBase {
type: EventType;
/**
* Name of key which triggered event
*/
key: WasmStringPtr;
/**
* Bitmask of active modifier keys
*
* WASM type: u8
*/
modifiers: number;
}
export const $KeyEvent: WasmTypeConstructor<KeyEvent> = (mem) => ({
get align() {
return 4;
},
get size() {
return 12;
},
instance: (base) => {
let $key: WasmStringPtr | null = null;
return {
get __base() {
return base;
},
get __bytes() {
return mem.u8.subarray(base, base + 12);
},
get type(): EventType {
return mem.u8[base];
},
set type(x: EventType) {
mem.u8[base] = x;
},
get key(): WasmStringPtr {
return $key || ($key = new WasmStringPtr(mem, (base + 4), true));
},
get modifiers(): number {
return mem.u8[(base + 8)];
},
set modifiers(x: number) {
mem.u8[(base + 8)] = x;
},
};
}
});
export interface Event extends WasmTypeBase {
mouse: MouseEvent;
key: KeyEvent;
}
export const $Event: WasmTypeConstructor<Event> = (mem) => ({
get align() {
return 4;
},
get size() {
return 12;
},
instance: (base) => {
return {
get __base() {
return base;
},
get __bytes() {
return mem.u8.subarray(base, base + 12);
},
get mouse(): MouseEvent {
return $MouseEvent(mem).instance(base);
},
set mouse(x: MouseEvent) {
mem.u8.set(x.__bytes, base);
},
get key(): KeyEvent {
return $KeyEvent(mem).instance(base);
},
set key(x: KeyEvent) {
mem.u8.set(x.__bytes, base);
},
};
}
});
//! Generated by @thi.ng/wasm-api-bindgen at 2022-11-23T16:55:22.190Z
//! DO NOT EDIT!
const std = @import("std");
const wasm = @import("wasmapi");
/// Supported event types
pub const EventType = enum(u8) {
/// Any kind of mouse event
mouse = 1,
/// Key down/up event
key,
misc,
};
/// Example struct
pub const MouseEvent = extern struct {
type: EventType,
pos: [2]u16,
};
/// Example struct
pub const KeyEvent = extern struct {
type: EventType,
/// Name of key which triggered event
key: wasm.ConstStringPtr,
/// Bitmask of active modifier keys
modifiers: u8,
};
pub const Event = extern union {
mouse: MouseEvent,
key: KeyEvent,
};
On the TypeScript/JS side, the memory-mapped wrappers (e.g. $Event
)
can be used in combination with the WasmBridge
to obtain fully typed views
(according to the generated types) of the underlying WASM memory. Basic usage is
like:
import { WasmBridge } from "@thi.ng/wasm-api";
import { $Event, EventType } from "./generated.ts";
const bridge = new WasmBridge();
// bridge initialization omitted here (see other examples)
// ...
// Create an instance using the bridge's memory views
// and mapping a `Event` union from given address
// (e.g. obtained from an exported WASM function/value)
const event = $Event(bridge).instance(0x10000);
// then use like normal JS object
event.mouse.pos
// Uint16Array(2) [100, 200]
// IMPORTANT: any modifications like this are directly
// applied to the underlying WASM memory...
event.mouse.pos[0] = 300;
// ...or
event.mouse.pos.set([1, 2]);
// buffer overflow protection
event.mouse.pos.set([1, 2, 3]);
// Uncaught RangeError: offset is out of bounds
event.mouse.type === EventType.MOUSE
// true
IMPORTANT: Field setters are currently only supported for single values, incl. enums, strings, structs, unions. The latter two will always be copied by value (mem copy). Currently, array, multi-pointers and slices do not provide write access (from the JS side)...
ALPHA - bleeding edge / work-in-progress
Search or submit any issues for this package
yarn add @thi.ng/wasm-api-bindgen
ES module import:
<script type="module" src="https://cdn.skypack.dev/@thi.ng/wasm-api-bindgen"></script>
For Node.js REPL:
# with flag only for < v16
node --experimental-repl-await
> const wasmApiBindgen = await import("@thi.ng/wasm-api-bindgen");
Package sizes (gzipped, pre-treeshake): ESM: 6.29 KB
TODO
Please also see further examples in the @thi.ng/wasm-api main readme and the various (commented) example projects linked above.
Karsten Schmidt
If this project contributes to an academic publication, please cite it as:
@misc{thing-wasm-api-bindgen,
title = "@thi.ng/wasm-api-bindgen",
author = "Karsten Schmidt",
note = "https://thi.ng/wasm-api-bindgen",
year = 2022
}
© 2022 Karsten Schmidt // Apache Software License 2.0
FAQs
Polyglot bindings code generators (TS/JS, Zig, C11) for hybrid WebAssembly projects
The npm package @thi.ng/wasm-api-bindgen receives a total of 19 weekly downloads. As such, @thi.ng/wasm-api-bindgen popularity was classified as not popular.
We found that @thi.ng/wasm-api-bindgen demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 0 open source maintainers collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Research
Security News
Socket researchers uncover a malicious npm package posing as a tool for detecting vulnerabilities in Etherium smart contracts.
Security News
Research
A supply chain attack on Rspack's npm packages injected cryptomining malware, potentially impacting thousands of developers.
Research
Security News
Socket researchers discovered a malware campaign on npm delivering the Skuld infostealer via typosquatted packages, exposing sensitive data.