@haaxor1689/nil
v0.1.6
Published
TypeScript-first binary data parsing library with static type inference
Downloads
13
Readme
TypeScript-first binary data parsing library with static type inference. Heavily inspired by Zod and Restructure.
Installation
npm install @haaxor1689/nil
Basic usage
Creating a simple string schema
import { n } from '@haaxor1689/nil';
// Create schema for string with length of 14 characters
const mySchema = n.string(14);
// Parse into Uint8Array
const buffer = await mySchema.toBuffer('Hello world!!!');
// Prase from Uint8Array
const parsed = await mySchema.fromBuffer(buffer);
Creating an object schema
const User = n.object({
username: n.string(4),
age: n.int8(),
active: n.bool()
});
// Extract the output type
type User = n.output<typeof User>;
// { username: string; age: number; active: boolean; }
const buffer = await User.toBuffer({ username: 'Jane', age: 30, active: true });
// Prase from Uint8Array
await User.fromBuffer(buffer);
Primitives
import { n } from '@haaxor1689/nil';
// boolean
n.bool();
// number
n.int8();
n.uint8();
n.int16();
n.uint16();
n.int32();
n.uint32();
n.float();
n.double();
// bigint
n.int64();
n.uint64();
Endianness
By default, all numbers are assumed to be in little-endian byte order. You can change this by using the .be()
option:
import { n } from '@haaxor1689/nil';
// Will be read in big-endian byte order
n.int32().be();
.be()
is not available forbool
and is silently ignored forint8
anduint8
schemas.
Objects
Since we are dealing with binary data, there are no optional properties, and the order of the attributes matters. All values are read in the order they were declared in.
import { n } from '@haaxor1689/nil';
// Declare object schema with given shape
const User = n.object({
rank: n.uint16(),
active: n.int32()
});
// Extract the output type
type User = n.output<typeof User>;
// Equivalent to
type User = {
rank: number;
active: boolean;
};
The above object schema would be equivalent to this C struct definition:
struct User {
unsigned short age;
bool active;
}
Array-like types
All array-like types must have a known size. It can either be provided as a constant, or by referencing value from surrounding context.
import { n } from '@haaxor1689/nil';
// Constant size
n.buffer(10); // Uint8Array
n.string(10); // string
n.array(n.int16(), 10); // number[]
// Size defined from context
n.object({
itemCount: n.int16(),
items: n.array(n.int16(), ['itemCount'])
});
// Nested context
n.object({
channels: n.int8(),
itemCount: n.int16(),
items: n.array(
n.object({
color: n.array(n.uint8(), ['..', '..', 'channels'])
}),
['itemCount']
)
});
// Set size to fill source
n.buffer('fill');
n.array(n.int16(), 'fill');
Note that any dynamically sized array-like type will fill the whole remaining space in the buffer so they should always be at the end.
Length in bytes
The .bytes()
option can be used to interpret a given length in bytes instead of the count of elements.
import { n } from '@haaxor1689/nil';
// Size will be 256 bytes
n.buffer(256).bytes();
n.string(256).bytes();
n.array(n.int8, 256).bytes();
Enums
You can load C enum values as a string literal union. Only default numbered C enums are supported now.
import { n } from '@haaxor1689/nil';
// Declare enum schema with given options
const Level = n.enum(n.int8(), ['LOW', 'MEDIUM', 'HIGH']);
// Extract the output type
type Level = n.output<typeof Level>;
// Equivalent to
type Level = 'LOW' | 'MEDIUM' | 'HIGH';
The above enum schema would be equivalent to this C enum definition:
enum Level {
LOW,
MEDIUM,
HIGH
}
.options
You can access the tuple used to create a given enum with .options
.
import { n } from '@haaxor1689/nil';
// Declare enum schema with given options
const Level = n.enum(n.int8(), ['LOW', 'MEDIUM', 'HIGH']);
Level.options; // ["LOW", "MEDIUM", "HIGH"]
Undefined
If you need a placeholder that represents 0 bytes in the binary data, you can use the undefined type for that:
import { n } from '@haaxor1689/nil';
// Declare object schema with given shape
const User = n.object({
empty: n.undefined(), // represents 0 bytes in buffer
active: n.int32()
});
// Extract the output type
type User = n.output<typeof User>;
// Equivalent to
type User = {
empty: undefined;
active: boolean;
};
Schema methods
All Nil schemas contain these methods.
.transform
.transform(
afterDecode: (v: Input, ctx?: ParseContext) => Promise<Output>,
beforeEncode: (v: Output, ctx?: ParseContext) => Promise<Input>
)
You can provide custom transformation functions for your schemas that will change the output both when parsing from the raw buffer and creating a buffer from the JS object.
import { n } from '@haaxor1689/nil';
// Define transform that handles calculating `itemCount`
const MySchema = n
.object({
itemCount: n.int16(),
items: n.array(n.int8(), ['itemCount'])
})
.transform(
v => v.items, // keep only raw items
v => ({ itemCount: v.length, items: v }) // calculate itemCount
);
// Inferred output type is `number[]`
type MySchema = z.output<typeof MySchema>;
// Resulting buffer will start with correct `itemCount` number
await MySchema.toBuffer([1, 2, 3, 4]);
You can also access the current context when creating transformations to reference other attributes from the parent type (if any). The easiest way to do this is by using the resolvePath
helper function.
import { n } from '@haaxor1689/nil';
const MySchema = n.object({
hasAlpha: n.bool(),
data: n.array(n.int8(), 'fill').transform(
(v, ctx) => {
n.resolvePath(['hasAlpha'], ctx); // will hold value of `hasAlpha` attribute from parent object
return v;
},
(v, ctx) => {
n.resolvePath(['hasAlpha'], ctx); // will hold value of `hasAlpha` attribute from parent object
return v;
}
)
});
.fromBuffer
.fromBuffer(data: Uint8Array): Promise<Output>
Tries to parse given buffer into output type of used schema. Throws errors on failure.
.toBuffer
.toBuffer(value: Output): Promise<Uint8Array>
Tries to serialize a given object into a buffer.
TODO
- Literal types
- Unions
- Better error handling
- Dynamic length null-terminated strings
- Don't allow smaller arrays than their defined constant size
- Tests