construct-js

construct-js is a library for creating byte level data structures.

npm i construct-js

Features

Signed and unsigned fields, up to 64-bit
Nested structs
Pointer and SizeOf fields
Different struct alignments, up to 64-bit, including packed structs. Padding can be added before or after the data
Ability to specify endianness per field
String support - both raw and null-terminated
Outputs to the standard Uint8Array type, which can be used in the browser and node
Getting and setting data in fields
Fast computation for the size of a field or complete struct
Written in TypeScript - providing static typing in both JS and TS (dependant on editor support)
Less than 3.5KiB after minification and gzip

Table of contents

High Level Overview

construct-js is all about creating a low-cost and performant abstraction for working with structured, binary data. If you've ever found yourself trying to manually assemble binary data in an ArrayBuffer - stuffing data, calculating sizes, scratching your head over endianness and offsets, then this is likely the tool for you.

construct-js allows you to specify and manipulate binary data using an expressive API made up of standard primitives that you may be used to in lower level languages - such as structs, pointers, sizeof operators, and standard sized signed and unsigned integer types.

Why?

Why not? I mean - I think there is genuine utility, but even if there wasn't, it would simply be an interesting project to undertake.

In terms of actual utility, as the web and distributed services evolve, web pages and JavaScript are taking on increasing more diverse and complex task, as well as connecting to more elaborate and varied services. Typically communication channels between different services use simple interchange formats like JSON over HTTP or WebSockets, but for a variety of reasons this is not always ideal. A large part of the reason this is so widespread is that JavaScript traditionally hasn't had good facilities or abstractions for creating byte-level data structures. Now, however, with the advent and standardisation of Typed Arrays and BigInt, this is no longer the case. construct-js allows developers to write expressive descriptions using standard native types like numbers, strings, and regular arrays - and outputs to an efficient Uint8Array format for interchange with the network, filesystem, or even across execution environments like WebAssembly.

Examples

There are more examples in the examples folder, showing the some more possibilities - including array fields, explicit endianness, etc.

The following example builds a (just about) valid* zip archive with one file inside - helloworld.txt.

^{*At least when unzipped using the unzip command. Some GUI programs seem to have less success}

import * as fs from 'fs/promises';
import {RawString, U16, U32, Struct, Pointer32, Endian} from 'construct-js';

const data = RawString('helloworldhelloworldhelloworldhelloworldhelloworldhelloworldhelloworldhelloworld');
const dataSize = data.computeBufferSize();
const filename = RawString('helloworld.txt');
const filenameSize = filename.computeBufferSize();

// Create a stub for the top level struct that can be referenced by other structs
const zipFile = Struct('ZipFile');

const sharedHeaderInfo = Struct('sharedHeaderInfo')
  .field('minVersion', U16(10))
  .field('gpFlag', U16(0))
  .field('compressionMethod', U16(0))
  .field('lastModifiedTime', U16(0))
  .field('lastModifiedDate', U16(0))
  .field('crc32', U32(0))
  .field('compressedSized', U32(dataSize))
  .field('uncompressedSized', U32(dataSize))
  .field('filenameSize', U16(filenameSize))
  .field('extraFieldLength', U16(0));

const localHeader = Struct('localHeader')
  .field('header', U32(0x04034b50))
  .field('sharedHeaderInfo', sharedHeaderInfo)
  .field('filename', filename);

const centralDirectory = Struct('centralDirectory')
  .field('header', U32(0x02014b50))
  .field('madeByVersion', U16(10))
  .field('sharedHeaderInfo', sharedHeaderInfo)
  .field('fileCommentSize', U16(0))
  .field('diskNumber', U16(0))
  .field('internalFileAttributes', U16(0))
  .field('externalFileAttributes', U32(0))
  .field('relativeOffset', U32(0))
  .field('filename', filename);

const endOfCentralDirectory = Struct('endOfCentralDirectory')
  .field('header', U32(0x06054b50))
  .field('diskNumber', U16(0))
  .field('centralDirDiskStart', U16(0))
  .field('numberOfCentralDirsOnDisk', U16(1))
  .field('totalNumberOfCentralDirs', U16(1))
  .field('centralDirSize', U32(0))
  .field('offsetToStart', Pointer32(zipFile, 'centralDirectory'))
  .field('commentLength', U16(0));

// Finalise the top level struct
zipFile
  .field('localHeader', localHeader)
  .field('data', data)
  .field('centralDirectory', centralDirectory)
  .field('endOfCentralDirectory', endOfCentralDirectory);

const fileBuffer = zipFile.toUint8Array();

fs.writeFile('./test.zip', fileBuffer).then(() => {
  console.log('Done writing zip file.');
});

Changelog

1.0.0

Full rewrite in TypeScript
Added 64-Bit support
Added alignment support
Breaking changes with the pre 1.0.0 releases

0.7.0

Add TypeScript definition file

0.6.1

Add .value() method to fields.

0.6.0

Refactor the user-facing API to remove endianness flags in fields and instead create a field for little endian and big endian variations.

0.5.0

Added NullTerminatedString field
Fixed a bug in getDeep that allowed requesting nonsense values in the path

0.4.2

Added Pointer8, Pointer16, Pointer32, SizeOf8, SizeOf16 and SizeOf32 fields

0.4.0

Removed concept of endianness from Structs. All endianness information comes directly from the Fields themselves
Removed deprecated Fields
Renamed I8, I16, I32, I8s, I16s, I32s -> S8, S16, S32, S8s, S16s, S32s

0.3.0

Allow the bit ordering to be specified for BitStructs

API

Struct

Struct(name: string, alignment = StructAlignment.Packed, paddingDirection = AlignmentPadding.AfterData)

Creates a Struct object. alignment specifies how much (if any) padding should be applied to the fields in order for them to align to a fixed byte boundary. paddingDirection specifies where the extra bytes should be added (before or after the data).

field

.field(name: string, value: ConstructDataType)

Adds a field to the struct. name is used to lookup the field using methods like struct.get(name). value must be either a Struct or one of the other data types provided by construct-js.

get

.get<T extends ConstructDataType>(name: string)

Returns the field with that name. Note: When using TypeScript, this value must be cast to the correct type, either using the generic or with the as keyword:

const s = Struct('example').field('first', U8(0));

s.get<DataType<typeof U8>>('first');

getOffset

.getOffset(name: string)

Returns the byte offset within the struct of the field with that name.

getDeep

.getDeep(path: string)

Returns the field within multiple structs, where path is a . separated string. Note: When using TypeScript, this value must be cast to the correct type, either using the generic or with the as keyword:

const struct = Struct('firstStruct')
  .field('aRawString', RawString('ABC'));

const struct2 = Struct('secondStruct')
  .field('deeperStruct', struct);

struct2.getDeep<DataType<RawString>>('deeperStruct.aRawString');

getDeepOffset

.getDeepOffset(path: string)

Returns the byte offset within multiple structs, where path is a . separated string.

computeBufferSize

.computeBufferSize()

Returns the size of the struct in bytes.

toUint8Array

.toUint8Array()

Returns a Uint8Array representation of the Struct.

Fields

Field Interfaces

Fields implement the IField interface, and optionally the IValue interface:

IField

interface IField {
  computeBufferSize(): number;
  toUint8Array(): Uint8Array;
}

IValue

interface IValue<T> {
  set(value: T): void;
  get(): T;
}

U8

U8(value: number) implements IField, IValue<number>

A single 8-bit unsigned value.

U16

U16(value: number, endian = Endian.Little) implements IField, IValue<number>

A single 16-bit unsigned value, in either big or little endian byte order.

U32

U32(value: number, endian = Endian.Little) implements IField, IValue<number>

A single 32-bit unsigned value, in either big or little endian byte order.

U64

U64(value: bigint, endian = Endian.Little) implements IField, IValue<bigint>

A single 64-bit unsigned value, in either big or little endian byte order. Note: Values for 64-bit fields must be specified as bigint.

I8

I8(value: number) implements IField, IValue<number>

A single 8-bit signed value.

I16

I16(value: number, endian = Endian.Little) implements IField, IValue<number>

A single 16-bit signed value, in either big or little endian byte order.

I32

I32(value: number, endian = Endian.Little) implements IField, IValue<number>

A single 32-bit signed value, in either big or little endian byte order.

I64

I64(value: bigint, endian = Endian.Little) implements IField, IValue<bigint>

A single 64-bit signed value, in either big or little endian byte order. Note: Values for 64-bit fields must be specified as bigint.

RawString

RawString(string) implements IField, IValue<string>

A collection of 8-bit unsigned values, interpreted directly from the string provided.

NullTerminatedString

NullTerminatedString(string) implements IField, IValue<string>

A collection of 8-bit unsigned values, interpreted directly from the string provided. This field appends a single 0x00 byte to the end of the data.

U8s

U8s(values: number[]) implements IField, IValue<number[]>

A collection of 8-bit unsigned values.

U16s

U16s(values: number[], endian = Endian.Little) implements IField, IValue<number[]>

A collection of 16-bit unsigned values, in either big or little endian byte order.

U32s

U32s(values: number[], endian = Endian.Little) implements IField, IValue<number[]>

A collection of 32-bit unsigned values, in either big or little endian byte order.

U64s

U64s(values: bigint[], endian = Endian.Little) implements IField, IValue<bigint[]>

A collection of 64-bit unsigned values, in either big or little endian byte order. Note: Values for 64-bit fields must be specified as bigint.

I8s

I8s(values: number[]) implements IField, IValue<number[]>

A collection of 8-bit signed values.

I16s

I16s(values: number[], endian = Endian.Little) implements IField, IValue<number[]>

A collection of 16-bit signed values, in either big or little endian byte order.

I32s

I32s(values: number[], endian = Endian.Little) implements IField, IValue<number[]>

A collection of 32-bit signed values, in either big or little endian byte order.

I64s

I64s(values: bigint[], endian = Endian.Little) implements IField, IValue<bigint[]>

A collection of 64-bit signed values, in either big or little endian byte order. Note: Values for 64-bit fields must be specified as bigint.

Pointer8

Pointer8(struct: Struct, path: string) implements IField

Pointer8 takes a Struct and a path, and represents an 8-bit pointer (offset) to the field specified by the path in the provided struct.

Pointer16

Pointer16(struct: Struct, path: string, endian = Endian.Little) implements IField

Pointer16 takes a Struct and a path, and represents an 16-bit pointer (offset) to the field specified by the path in the provided struct.

Pointer32

Pointer32(struct: Struct, path: string, endian = Endian.Little) implements IField

Pointer32 takes a Struct and a path, and represents an 32-bit pointer (offset) to the field specified by the path in the provided struct.

Pointer64

Pointer64(struct: Struct, path: string, endian = Endian.Little) implements IField

Pointer64 takes a Struct and a path, and represents an 64-bit pointer (offset) to the field specified by the path in the provided struct.

SizeOf8

SizeOf8(target: ConstructDataType) implements IField

SizeOf8 takes a Struct or a Field, and represents the size of the Struct or the Field as an 8-bit unsigned integer.

SizeOf16

SizeOf16(target: ConstructDataType, endian = Endian.Little) implements IField

SizeOf16 takes a Struct or a Field, and represents the size of the Struct or the Field as an 16-bit unsigned integer.

SizeOf32

SizeOf32(target: ConstructDataType, endian = Endian.Little) implements IField

SizeOf32 takes a Struct or a Field, and represents the size of the Struct or the Field as an 32-bit unsigned integer.

SizeOf64

SizeOf64(target: ConstructDataType, endian = Endian.Little) implements IField

SizeOf64 takes a Struct or a Field, and represents the size of the Struct or the Field as an 64-bit unsigned integer.