• Stars
    star
    216
  • Rank 176,481 (Top 4 %)
  • Language
  • Created over 2 years ago
  • Updated 26 days ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

JavaScript Structs: Fixed Layout Objects

JavaScript Structs: Fixed Layout Objects and Some Synchronization Primitives

Stage: 1

Author: Shu-yu Guo (@syg)

Champion: Shu-yu Guo (@syg), Ron Buckton (@rbuckton), Asumu Takikawa (@takikawa), Keith Miller (@kmiller68)

Introduction

Structs

Structs are declarative sealed objects. There are two kinds of structs: plain structs and shared structs. Plain structs behave as if they were sealed objects. Shared structs have additional restrictions and can be concurrently accessed from different agents.

All structs have the following properties:

  • Opaque storage like plain objects. Not aliasable via ArrayBuffer or SharedArrayBuffer.
  • Instances have all fields initialized in one shot, then sealed. The engine must be able to fix a layout that is unchanging. This implies that all superclasses must also be structs.
  • Transitively immutable [[Prototype]] slot. A struct instance's [[Prototype]] slot is immutable, as are the [[Prototype]] slot of every object on its prototype chain.

Shared structs have the following additional properties:

  • Only data fields are allowed. Getters, setters, and methods are disallowed.
  • All superclasses must be shared structs.
  • Shared structs only reference other primitives or shared objects.
  • Shared structs have a null [[Prototype]].
  • Shared structs do not have a .constructor property.
  • Shared struct constructors have [Symbol.hasInstance] to support instanceof.

This proposal is intended to be minimal.

Structs can be designed with or without novel syntax. For brevity of presentation, examples are given using class syntax with a struct or shared struct qualifier.

A minimal plain struct example.

struct class Box {
  constructor(x) { this.x = x; }
  x;
}

let box = new Box();
box.x = 42;  // x is declared
assertThrows(() => { box.y = 8.8; });       // structs are sealed
assertThrows(() => { box.__proto__ = {} }); // structs are sealed

A minimal shared struct example.

// main.js
shared struct class SharedBox {
  constructor(x) { this.x = x; }
  x;
}

let sharedBox = new SharedBox();
let sharedBox2 = new SharedBox();

sharedBox.x = 42;          // x is declared and rhs is primitive
sharedBox.x = sharedBox2;  // x is declared and rhs is shared
assertThrows(() => { sharedBox.x = {}; }) // rhs is not a shared struct

let worker = new Worker('worker.js');
worker.postMessage({ sharedBox });

sharedBox.x = "main";      // x is declared and rhs is primitive
console.log(sharedBox.x);
// worker.js
onmessage = function(e) {
  let sharedBox = e.data.sharedBox;
  sharedBox.x = "worker";  // x is declared and rhs is primitive
  console.log(sharedBox.x);
};

The above program is permitted to print out any interleaving:

  • main main
  • main worker
  • worker worker
  • worker main

Shared fixed-length arrays

Shared fixed-length arrays are the closed counterpart to Arrays, as structs are the closed counterpart to ordinary JS objects.

Shared fixed-length arrays are always shared. Structured data sharing requires some primitive notion of collections, on top of which more sophisticated collections can be built.

While there is nothing in principle preventing addition of a non-shared fixed-length array, the use case is unclear. Where sharing across agents is not needed, ordinary Array instances are more flexible and already performant.

Shared fixed-length arrays have the following property:

  • Length is required at construction time.
  • Instances cannot be resized.
  • Elements can only be other primitives other shared objects.
  • Shared arrays have a null [[Prototype]].
  • Shared arrays do not have a .constructor property.
  • The shared array constructor has [Symbol.hasInstance] to support instanceof.

Shared fixed-length arrays do not need novel syntax. For brevity of presentation, examples are given using the SharedFixedArray constructor.

// main.js
let sharedArray = new SharedFixedArray(10);
assert(sharedArray.length === 10);

let worker = new Worker('worker.js');
worker.postMessage({ sharedArray });

sharedArray[0] = "main";
console.log(sharedArray[0]);
// worker.js
onmessage = function(e) {
  let sharedArray = e.data.sharedArray;
  sharedArray[0] = "worker";
  console.log(sharedArray[0]);
};

Just like the struct example, above program is permitted to print out any interleaving:

  • main main
  • main worker
  • worker worker
  • worker main

Synchronization primitives

Non-recursive mutexes and conditional variables are well-understood synchronization primitives. Structured data sharing are well served by these higher-level synchronization primitives beyond Atomics.wait and Atomics.notify.

Pending future work on method and code sharing, mutexes and conditional variables are currently proposed as opaque, prototypeless shared objects to be used with static methods.

Minimal examples below.

// Creates a new mutex.
let mutex = new Atomics.Mutex;

// This would block the current agent if the lock is held
// by another thread. This cannot be used on the agents
// whose [[CanBlock]] is false.
Atomics.Mutex.lock(mutex, function runsUnderLock() {
  // Do critical section stuff
});

// tryLock is like lock but returns true if the lock was acquired
// without blocking, and false is the lock is held by another
// thread.
Atomics.Mutex.tryLock(mutex, function runsUnderLock() {
  // Do critical section stuff
});
let cv = new Atomics.Condition;

Atomics.Mutex.lock(mutex, () => {
  // This blocks the current agent, and cannot be used on the agents
  // whose [[CanBlock]] is false. The passed in mutex must be locked.
  Atomics.Condition.wait(cv, mutex);
});

// Waiters can notified with a count. A count of undefined or
// +Infinity means "all waiters".
let count = 1;
let numWaitersWokenUp = Atomics.Condition.notify(cv, count);

Extending Atomics.Mutex to support using

The Explicit Resource Management proposal adds using declarations that perform lexically scoped resource management. The Atomics.Mutex API can be extended to better support using. For example,

  • Atomics.Mutex.lock(mutex) can be overloaded to lock mutex
  • Atomics.Mutex.unlock(mutex) can be added
  • The [Symbol.dispose] own property can be added to all Atomics.Mutex instances

Asynchronous locking and waiting

See ASYNC-LOCKING-WAITING.md for lockAsync and waitAsync.

Motivation and requirements

Shared memory for parallelism

This proposal seeks to enable more shared memory parallelism for a more parallel future. Like other shared memory features in JavaScript, it is high in expressive power and high in difficulty to use correctly. This proposal is both intended as an incremental step towards higher-level, easier-to-use (e.g. data-race free by construction) parallelism abstractions as well as an escape hatch for expert programmers who need the expressivity.

The two design principles that this proposal follows are:

  1. Syntax that looks atomic ought to be atomic. (For example, the dot operator on shared structs should only access an existing field and does not tear.)
  2. There are no references from shared objects to non-shared objects. The shared and non-shared heaps are conceptually separate, with references only going one way.

WasmGC interoperability

The WasmGC proposal adds fixed layout, garbage-collected objects to Wasm. While the details of the type system of these objects are yet to be nailed down, interoperability with JavaScript will be important.

WasmGC objects have opaque storage and are not aliased by linear memory, so they cannot be exposed as all Wasm memory is exposed today via ArrayBuffers. We propose structs to be the reflection of WasmGC objects in JS.

WasmGC objects exposed to JS should behave the same as structs, modulo extra type checking that WasmGC require that JS structs do not. JS structs is also a good foundation for reflecting into Wasm as WasmGC objects, but that is currently left as future work as it may need a typed field extensions to be worthwhile.

Further, WasmGC itself will eventually have multithreading. It behooves us to maintain a single memory model between JavaScript and Wasm as we have today, even with higher-level object abstractions.

Predictable instance performance

Objects that are declared with a fixed layout help engines to have more predictable performance. A fixed layout object also lays groundwork for future refinement, such as typed fields.

Out-of-scope

Value semantics, immutability, and operator overloading

This proposal does not intend to explore the space of objects with value semantics, including immutability and operator overloading. Structs have identity like other objects and are designed to be used like other objects. Value semantics is a sufficient departure that it may be better solved with other proposals that focus on that space.

Sophisticated type systems

This proposal does not intend to explore sophisticated type and runtime guard systems. It is even more minimal than the closest spiritual ancestor, the Typed Objects proposal, in that we do not propose integral types for sized fields. (Typed and sized fields are reserved for future work.)

Binary data overlay views

This proposal does not intend to explore the space of overlaying structured views on binary data in an ArrayBuffer. This is a requirement arising from the desire for WasmGC integration, and WasmGC objects are similarly opaque.

Structured overlays are fundamentally about aliasing memory, which we feel is both a different problem domain and sufficiently solvable today in userland. For example, see buffer-backed objects.

Proposal

This proposal can be developed with or without novel syntax. It is presented with novel syntax below.

Plain structs

Plain structs are declared with the contextual struct keyword in front of a class declaration.

struct class expressions parse the same as plain class expressions.

During evaluation of a struct class expression, the following checks are performed.

  • If there is an extends clause and the superclass is not a struct class, throw a TypeError

When a struct class constructor is invoked, it creates instances with the following properties:

  • All instance fields, including those from any superclasses, are defined and initialized to undefined before the newly constructed instance escapes to the constructor function (aka "one-shot")
  • Instances are sealed
  • Instances' [[Prototype]] slot is immutable after initialization

Shared structs

Shared structs are declared with the contextual shared struct keywords in front of a class declaration. Intuitively, they behave as very restricted structs.

shared struct class expressions throw an early error if the following forms are encountered.

  • method
  • getter
  • setter
  • field initializers

During evaluation of a struct class expression, the following checks are performed.

  • If there is an extends clause and the superclass is not a shared struct class, throw a TypeError

When a struct class constructor is invoked, it creates instances with the following properties:

  • All instance fields are defined and initialized to undefined before the newly constructed instance escapes to the constructor function (aka "one-shot")
  • Constructors have [Symbol.hasInstance] to support instanceof
  • Instances are sealed
  • Instances' [[Prototype]] slot contains null
  • Instances do not have a .constructor property
  • Instances are shared with instead of copied to other agents
  • Instances' identities are preserved when communicated between agents
  • When a value is assigned to an instance field, if it is neither a primitive nor a shared object or is a Symbol, throw a TypeError
  • Instances' field accesses are unordered shared memory accesses and must not tear (i.e. another agent must not observe a partially written value)

The following Atomics methods will be extended to accept shared struct instances as the first argument and a field name as the second argument, to support sequentially consistent accesses.

  • Atomics.load
  • Atomics.store
  • Atomics.exchange
  • Atomics.compareExchange

Note that the arithmetic Atomics methods like Atomics.add are not included because there isn't widespread ISA support for atomic read-modify-write of floating point values. The workaround is to compute locally and store with Atomics.compareExchange.

Shared fixed-length arrays

Shared fixed-length arrays are constructed using the SharedFixedArray(len) constructor. They are considered shared objects.

  • Constructors have [Symbol.hasInstance] to support instanceof
  • Instances are sealed
  • Instances' [[Prototype]] slot contains null
  • Instances do not have a .constructor property
  • Instances have an immutable .length property
  • Instances are shared with instead of copied to other agents
  • Instances' identities are preserved when communicated between agents
  • When a value is assigned to an instance field, if it is neither a primitive nor a shared object or is a Symbol, throw a TypeError
  • Instances' element accesses are unordered shared memory accesses and must not tear (i.e. another agent must not observe a partially written value)

The following Atomics methods will be extended to accept SharedFixedArray instances as the first argument and an index as the second argument, to support sequentially consistent accesses.

  • Atomics.load
  • Atomics.store
  • Atomics.exchange
  • Atomics.compareExchange

Note that the arithmetic Atomics methods like Atomics.add are not included because there isn't widespread ISA support for atomic read-modify-write of floating point values. The workaround is to compute locally and store with Atomics.compareExchange.

Atomics.Mutex

Non-recursive mutexes are constructed using the Atomics.Mutex constructor. They are considered shared objects.

  • Atomics.Mutex has a [Symbol.hasInstance] to support instanceof
  • Instances are sealed
  • Instances have no properties
  • Instances are shared with instead of copied to other agents
  • Instances' identities are preserved when communicated between agents

The following static methods exist on Atomics.Mutex

  • Atomics.Mutex.lock(mutex, funToRunUnderLock): Acquire mutex by blocking. Once acquired, invoke funToRunUnderLock, then release the lock.
  • Atomics.Mutex.lockAsync(mutex, funToRunUnderLock): Acquire mutex asynchronously. Once acquired, enqueue funToRunUnderLock as a task to run. Release the lock after the enqueued task is finished.
  • Atomics.Mutex.tryLock(mutex, funToRunUnderLock): Try to acquire mutex, returning undefined if already locked. Once acquired, invoke funToRunUnderLock, then release the lock.

Atomics.Condition

Condition variables are constructed using the Atomics.Condition constructor. They are considered shared objects.

  • Atomics.Condition has a [Symbol.hasInstance] to support instanceof
  • Instances are sealed
  • Instances have no properties
  • Instances are shared with instead of copied to other agents
  • Instances' identities are preserved when communicated between agents

The following static methods exist on Atomics.Condition

  • Atomics.Condition.wait(cv, mutex, [ timeout ]): Block the agent until cv is notified or until timeout milliseconds have passed. mutex must be locked. It is atomically released when the agent blocks and is reacquired after notification. timeout defaults to Infinity. Throw a TypeError of the agent's [[CanBlock]] is false.
  • Atomics.Condition.waitAsync(cv, mutex, [ timeout ]): Returns a promise that is fulfilled when cv is notified or until timeout milliseconds have passed. mutex must be locked. It is released after the promise is constructed.
  • Atomics.Condition.notify(cv, count): Notify count number of cv's waiters. A count of Infinity notifies all waiters.

Implementation guidance

Immutable shapes

Structs are declared with fixed layout up front. Engines should make an immutable shape for such objects. Optimizers can optimize field accesses without worrying about deopts.

Shared structs: make sure fields are pointer-width and aligned

Shared structs should store fields such that underlying architectures can perform atomic stores and loads. This usually means the fields should be at least pointer-width and aligned.

Shared structs: strings will be difficult

Except for strings, sharing primitives in the engine is usually trivial, especially for NaN-boxing implementations.

Strings in production engines have in-place mutation to transition representation in order to optimize for different use cases (e.g. ropes, slices, canonicalized, etc). Sharing strings will likely be the most challenging part of the implementation.

It is possible to support sharing strings by copying-on-sharing, but may be too slow. If possible, lockfree implementations of in-place mutations above is ideal.

Synchronization primitives: they must be moving GC-safe

Production engines use moving garbage collectors, such as generational collectors and compacting collectors. If JS synchronization primitives are implemented under the hood as OS-level synchronization primitives, those primitives most likely depend on an unchanging address in memory and are not moving GC-safe.

Engines can choose to pin these objects and make them immovable.

Engines can also choose to implement synchronization primitives entirely in userspace. For example, WebKit's ParkingLot is a userspace implementation of Linux futexes. This may have other benefits, such as improved and tuneable performance.

Future work

Code sharing

Code sharing is not part of this proposal and thus shared structs cannot have methods. This may prove to be unergonomic enough that we bring code sharing in scope. Doing so would likely require a new kind of function cannot close over non-shared objects.

This proposal future-proofs by having shared struct instances throw when touching the [[Prototype]] slot. This will be relaxed with a future proposal when code sharing becomes possible.

See CODE-SHARING-IDEAS.md for a collection of ideas.

Typed fields

In the future, it may be sensible for more efficient memory representation ("packing") to also declare the type and size of fields. It is omitted from this proposal in that it is not a requirement for none of the primary motivations. At the same time, starting without types lets us add them incrementally in the future.

Shared structs: fast cloning

Since structs have fixed layout with an immutable [[Prototype]] slot and no accessors, they are amenable to fast cloning.

More Repositories

1

proposals

Tracking ECMAScript Proposals
17,177
star
2

ecma262

Status, process, and documents for ECMA-262
HTML
14,437
star
3

proposal-pipeline-operator

A proposal for adding a useful pipe operator to JavaScript.
HTML
7,380
star
4

proposal-pattern-matching

Pattern matching syntax for ECMAScript
HTML
5,341
star
5

proposal-optional-chaining

HTML
4,952
star
6

proposal-type-annotations

ECMAScript proposal for type syntax that is erased - Stage 1
JavaScript
4,090
star
7

proposal-temporal

Provides standard objects and functions for working with dates and times.
HTML
3,135
star
8

proposal-observable

Observables for ECMAScript
JavaScript
3,032
star
9

proposal-signals

A proposal to add signals to JavaScript.
TypeScript
2,668
star
10

proposal-decorators

Decorators for ES6 classes
2,640
star
11

proposal-record-tuple

ECMAScript proposal for the Record and Tuple value types. | Stage 2: it will change!
HTML
2,423
star
12

test262

Official ECMAScript Conformance Test Suite
JavaScript
2,073
star
13

proposal-dynamic-import

import() proposal for JavaScript
HTML
1,859
star
14

proposal-bind-operator

This-Binding Syntax for ECMAScript
1,736
star
15

proposal-class-fields

Orthogonally-informed combination of public and private fields proposals
HTML
1,720
star
16

proposal-async-await

Async/await for ECMAScript
HTML
1,577
star
17

proposal-object-rest-spread

Rest/Spread Properties for ECMAScript
HTML
1,496
star
18

proposal-shadowrealm

ECMAScript Proposal, specs, and reference implementation for Realms
HTML
1,365
star
19

proposal-nullish-coalescing

Nullish coalescing proposal x ?? y
HTML
1,233
star
20

proposal-iterator-helpers

Methods for working with iterators in ECMAScript
HTML
1,220
star
21

proposal-top-level-await

top-level `await` proposal for ECMAScript (stage 4)
HTML
1,082
star
22

proposal-partial-application

Proposal to add partial application to ECMAScript
HTML
1,002
star
23

proposal-do-expressions

Proposal for `do` expressions
HTML
990
star
24

agendas

TC39 meeting agendas
JavaScript
952
star
25

proposal-binary-ast

Binary AST proposal for ECMAScript
945
star
26

proposal-built-in-modules

HTML
886
star
27

proposal-async-iteration

Asynchronous iteration for JavaScript
HTML
854
star
28

proposal-explicit-resource-management

ECMAScript Explicit Resource Management
JavaScript
671
star
29

proposal-operator-overloading

JavaScript
610
star
30

proposal-string-dedent

TC39 Proposal to remove common leading indentation from multiline template strings
HTML
588
star
31

proposal-bigint

Arbitrary precision integers in JavaScript
HTML
560
star
32

proposal-set-methods

Proposal for new Set methods in JS
HTML
557
star
33

proposal-import-attributes

Proposal for syntax to import ES modules with assertions
HTML
538
star
34

ecmascript_simd

SIMD numeric type for EcmaScript
JavaScript
536
star
35

proposal-slice-notation

HTML
515
star
36

proposal-change-array-by-copy

Provides additional methods on Array.prototype and TypedArray.prototype to enable changes on the array by returning a new copy of it with the change.
HTML
509
star
37

ecma402

Status, process, and documents for ECMA 402
HTML
506
star
38

notes

TC39 meeting notes
JavaScript
496
star
39

proposal-class-public-fields

Stage 2 proposal for public class fields in ECMAScript
HTML
489
star
40

proposal-iterator.range

A proposal for ECMAScript to add a built-in Iterator.range()
HTML
464
star
41

proposal-uuid

UUID proposal for ECMAScript (Stage 1)
JavaScript
462
star
42

proposal-throw-expressions

Proposal for ECMAScript 'throw' expressions
JavaScript
425
star
43

proposal-module-expressions

HTML
424
star
44

proposal-UnambiguousJavaScriptGrammar

413
star
45

proposal-decimal

Built-in decimal datatype in JavaScript
HTML
408
star
46

proposal-array-grouping

A proposal to make grouping of array items easier
HTML
407
star
47

proposal-async-context

Async Context for JavaScript
HTML
406
star
48

proposal-weakrefs

WeakRefs
HTML
404
star
49

proposal-error-cause

TC39 proposal for accumulating errors
HTML
378
star
50

proposal-ecmascript-sharedmem

Shared memory and atomics for ECMAscript
HTML
376
star
51

proposal-cancelable-promises

Former home of the now-withdrawn cancelable promises proposal for JavaScript
Shell
376
star
52

proposal-relative-indexing-method

A TC39 proposal to add an .at() method to all the basic indexable classes (Array, String, TypedArray)
HTML
351
star
53

proposal-first-class-protocols

a proposal to bring protocol-based interfaces to ECMAScript users
350
star
54

proposal-global

ECMAScript Proposal, specs, and reference implementation for `global`
HTML
346
star
55

proposal-private-methods

Private methods and getter/setters for ES6 classes
HTML
344
star
56

proposal-numeric-separator

A proposal to add numeric literal separators in JavaScript.
HTML
327
star
57

proposal-private-fields

A Private Fields Proposal for ECMAScript
HTML
320
star
58

proposal-object-from-entries

TC39 proposal for Object.fromEntries
HTML
317
star
59

proposal-module-declarations

JavaScript Module Declarations
HTML
314
star
60

proposal-promise-allSettled

ECMAScript Proposal, specs, and reference implementation for Promise.allSettled
HTML
314
star
61

tc39.github.io

Get involved in specifying JavaScript
HTML
313
star
62

proposal-regex-escaping

Proposal for investigating RegExp escaping for the ECMAScript standard
JavaScript
309
star
63

proposal-await.ops

Introduce await.all / await.race / await.allSettled / await.any to simplify the usage of Promises
HTML
308
star
64

proposal-logical-assignment

A proposal to combine Logical Operators and Assignment Expressions
HTML
302
star
65

proposal-export-default-from

Proposal to add `export v from "mod";` to ECMAScript.
HTML
297
star
66

proposal-promise-finally

ECMAScript Proposal, specs, and reference implementation for Promise.prototype.finally
HTML
278
star
67

proposal-asset-references

Proposal to ECMAScript to add first-class location references relative to a module
268
star
68

proposal-cancellation

Proposal for a Cancellation API for ECMAScript
HTML
262
star
69

proposal-json-modules

Proposal to import JSON files as modules
HTML
259
star
70

proposal-promise-with-resolvers

HTML
255
star
71

proposal-string-replaceall

ECMAScript proposal: String.prototype.replaceAll
HTML
254
star
72

proposal-export-ns-from

Proposal to add `export * as ns from "mod";` to ECMAScript.
HTML
241
star
73

proposal-ses

Draft proposal for SES (Secure EcmaScript)
HTML
217
star
74

proposal-intl-relative-time

`Intl.RelativeTimeFormat` specification [draft]
HTML
215
star
75

proposal-flatMap

proposal for flatten and flatMap on arrays
HTML
215
star
76

proposal-json-parse-with-source

Proposal for extending JSON.parse to expose input source text.
HTML
204
star
77

ecmarkup

An HTML superset/Markdown subset source format for ECMAScript and related specifications
TypeScript
201
star
78

proposal-promise-any

ECMAScript proposal: Promise.any
HTML
198
star
79

proposal-decorators-previous

Decorators for ECMAScript
HTML
184
star
80

proposal-smart-pipelines

Old archived draft proposal for smart pipelines. Go to the new Hack-pipes proposal at js-choi/proposal-hack-pipes.
HTML
181
star
81

proposal-defer-import-eval

A proposal for introducing a way to defer evaluate of a module
HTML
174
star
82

proposal-array-filtering

A proposal to make filtering arrays easier
HTML
171
star
83

proposal-optional-chaining-assignment

`a?.b = c` proposal
168
star
84

proposal-array-from-async

Draft specification for a proposed Array.fromAsync method in JavaScript.
HTML
167
star
85

proposal-extractors

Extractors for ECMAScript
JavaScript
166
star
86

proposal-upsert

ECMAScript Proposal, specs, and reference implementation for Map.prototype.upsert
HTML
165
star
87

proposal-ptc-syntax

Discussion and specification for an explicit syntactic opt-in for Tail Calls.
HTML
165
star
88

how-we-work

Documentation of how TC39 operates and how to participate
161
star
89

proposal-collection-methods

HTML
160
star
90

proposal-Array.prototype.includes

Spec, tests, reference implementation, and docs for ESnext-track Array.prototype.includes
HTML
157
star
91

proposal-error-stacks

ECMAScript Proposal, specs, and reference implementation for Error.prototype.stack / System.getStack
HTML
156
star
92

proposal-promise-try

ECMAScript Proposal, specs, and reference implementation for Promise.try
HTML
154
star
93

proposal-hashbang

#! for JS
HTML
148
star
94

proposal-resizablearraybuffer

Proposal for resizable array buffers
HTML
145
star
95

proposal-import-meta

import.meta proposal for JavaScript
HTML
145
star
96

proposal-intl-segmenter

Unicode text segmentation for ECMAScript
HTML
145
star
97

proposal-extensions

Extensions proposal for ECMAScript
HTML
143
star
98

proposal-seeded-random

Proposal for an options argument to be added to JS's Math.random() function, and some options to start it with.
HTML
143
star
99

proposal-intl-duration-format

141
star
100

proposal-regexp-unicode-property-escapes

Proposal to add Unicode property escapes `\p{…}` and `\P{…}` to regular expressions in ECMAScript.
HTML
134
star