@signalis/core

signalis is a lightweight library for reactivity influenced by @preact/signals, solidjs, and reactively. It aims to expose a small set of highly composable, highly performant primitives for building reactive programs as simply as possible.

To that end, signalis exposes three core primitives: signals, derived values, and effects.

Signals

Signals are the core unit of reactivity in signalis (and, frankly, most reactivity systems!). A Signal is simply a box around a value that tells other things when its value has changed. Whenever a Signal changes, any part of the reactivity system that relies on it will update accordingly the next time that part of the system is accessed. In other words, all parts of the system will be kept in sync lazily: changing the value of a Signal doesn't trigger any additional computations, it simply tells the parts of the system that care about the Signal that they will need to recompute eventually.

To create a Signal, use the createSignal function:

createSignal<T>(value: T, isEqual?: false | (old: T, new: T) => boolean): Signal<T>

This is the form of createSignal that you will likely use more often than not. It accepts a value and (optionally) an equality function. By default, Signal will use === to determine when a particular Signal's value has changed (and hence, when it needs to notify that it has updated), but you can provide your own equality function if you need to customize this behavior. Additonally, if you'd like to make the Signal always notify when it's been, you can set the second argument to false.

// Basic usage
const count = createSignal(0);
const list = createSignal(['foo', 'bar', 'baz']);
const me = createSignal({ firstName: 'Chris', lastName: 'Freeman' });

// to read a signal, access its `value` property
count.value; // 0

// to update a signal, assign a new value to its `value` property
count.value = 1;

// Since signals use `===` equality, object and array updates need to be immutable
me.value = {
  ...me.value,
  firstName: 'Christopher',
};

// Pass a custom equality function to change how signals determine when they've changed
import { isEqual } from 'lodash/isEqual';
const meButMutable = createSignal(
  {
    firstName, 'Chris',
    lastName: 'Freeman'
  },
  isEqual
)

createSignal(): Signal<unknown>

In some cases, you don't actually care about the value of a signal, but instead simply need a way to tell other parts of the system about a change. In those cases, this form of createSignal can make that process easier. Calling createSignal with no arguments will return a Signal<unknown> with its equality function set to false so that it will notify its dependencies every time value is set to a value.

const notifier = createSignal();

notifier.value = null; // will notify all dependents

notifier.value = null; // will notify again

Derived

As you may have already realized, a Signal on its own isn't very useful without something that reacts to it. This is where Derived comes in. Derived is a readonly reactive value that, as its name might suggest, is derived from one or more other reactive values (which can be Signals, other Derived, or a combination of both). The value of a Derived will update any time one of its dependents change, and it will also notify its own dependents whenever it changes.

To create a Derived, use createDerived:

createDerived<T>(fn: () => T): T

createDerived accepts a callback that reads one or more reactive values, does some computation with them, and returns the result. createDerived should always be used to represent a reactive value, so it's important that the callback actually return something. Furthermore, the callback passed into createDerived should be a pure function (it shouldn't have any side effects). If you need to represent a reactive function, there's a different primitive for that (see createEffect below).

const firstName = createSignal('Chris');
const lastName = createSignal('Freeman');

const fullName = createDerived(() => `${firstName.value} ${lastName.value}`);

// Since Derived is a reactive value, you access its value the same way you would a Signal
fullName.value; // 'Chris Freeman'

// Derived can rely on other Derived
const fullNameBUTYELLING = createDerived(() => `${fullName.value.toUpperCase()}!!!!!!!!!!!!!`);

fullNameBUTYELLING.value; // 'CHRIS FREEMAN!!!!!!!!!!!!!'

// Setting the value of a Signal at the root of a chain of Derived will notify all of them that they may need to recompute
firstName.value = 'Christopher';

fullNameBUTYELLING.value; // 'CHRISTOPHER FREEMAN!!!!!!!!!!!!!'

In order to avoid doing any unnecessary work, Derived is very clever about when it should actually recompute. It will never recompute just because one of its sources changed. Something else has to first try and read its value property before it will even consider recomputing. Once something does access its value, Derived follows the heuristic below to determine if it should actually recompute:

• Have any of its direct dependencies changed?
  • If so, this is the most certain we can be that value of the Derived has also changed, so recompute immediately.
• Have any of its indirect dependencies changed?
  • If so, we know that it's possible we may need to recompute, but are not certain, so have our immediate dependencies look to see if they need to recompute, and then decide what we should do
• If neither the direct or indirect dependencies have changed, don't recompute and instead just return the previous value.

Here's an example to help clarify further:

const count = createSignal(1);

// This will recompute whenever count changes
const countIsOdd = createDerived(() => count.value % 2 !=== 0);

// This will recompute whenever countIsOdd changes
const oddOrEven = createDerived(() => {
  if (countIsOdd.value) {
    return 'odd';
  } else {
    return 'even';
  }
})


// Trigger the `countIsOdd` computation since it has never been read before.
countIsOdd.value // true, since count.value === 1

// Trigger the `oddOrEven` computation since it *also* has never been run before. However, since `countIsOdd` has already been read, and `count` hasn't changed, `countIsOdd` does *not* recompute.
oddOrEven.value // 'odd', since countIsOdd.value === true


count.value = 3; // Set count to another odd number

/**
 * Read `oddOrEven` again. At this point, `countIsOdd` has not recomputed since nothing else has accessed its `value` property since we updated `count`.
 *
 * `oddOrEven` knows that it *might* need to recompute since `count` changed, but it's not certain since it doesn't depend directly on `count`, so we first
 * check to see `countIsOdd` should recompute.
 *
 * `countIsOdd` sees that its direct dependency (`count`) has changed, so it recomputes and but since its value is still `true`, it tells `oddOrEven` that not to recompute.
 *
 * `oddOrEven`, upon being told that its direct dependencies haven't changed, returns its previous value ('odd') and skips computing
 *
 */
oddOrEven.value // still 'odd', no need to recompute since we know `countIsOdd` hasn't changed

Effects

While Derived is a reactive value, you can think of an effect as a reactive function. It reacts to changes the same way that Derived does, but rather than returning a value, it runs a computation. Unlike Derived, side effects are welcome (and, in fact, encouraged) in effects. Effects use the same heuristic as Derived for determining when to recompute, with the one exception that they evaluate their callback function eagerly, whereas Derived is lazy. (This makes sense when you consider that effects don't have values, and therefore there's no way to "read" it the way you would with a Derived).

Effects are useful any time you need to perform some kind of action in response to something else changing.

To create an effect, use createEffect:

createEffect(fn: () => void | (() => void)): () => void

createEffect takes a callback that reads some number of reactive values and does something with them. The callback should not return a value (it can return a value, but signalis won't do anything with it since effects aren't mean to represent values). createEffect returns a disposal function that can be called if/when you no longer need the effect anymore

By default, the disposal function will simply disable the effect and remove it from the reactivity tree, but you can also add additional cleanup behavior by return a function from the callback that contains whatever action you'd like to run when the effect is disposed.

Effects have a wide range of uses, let's take a look at a few examples.

In the most basic case, we simply set up an effect that runs every time a signal changes:

const count = createSignal(0);

createEffect(() => {
  console.log(`The value of count is: ${count.value}`);
});

// since effects are eager, we will immediately log

count.value = 1; // the effect logs 'The value of count is 1'

Effects can also respond to Derived, and will use the same logic as Derived when determining whether to recompute:

const count = createSignal(1);
const countIsOdd = createDerived(() => count.value % 2 !=== 0);

let message: string;

createEffect(() => {
  message = countIsOdd.value ? 'odd' : 'even';
});

console.log(message); // 'odd'

count.value = 2;

// effect recomputes since we've gone from odd to even
console.log(message); // 'even'

count.value = 4;

// effect does *not* recompute since we've gone from one even number to another
console.log(message) // 'even'

Since effects compute eagerly, it's important that we provide a way to clean them up in the event that we no longer need them. To do that, we can use the function createEffect returns:

const count = createSignal(0);

const dispose = createEffect(() => {
  console.log(`count.value is: ${count.value}`);

  return () => {
    console.log(`cleaning up! no more messages!`);
  };
}); // logs the count immediately

count.value = 1; // logs again

dispose(); // logs the clean up message

count.value = 2; // no log, since the effect is disposed now

Finally, we can return a function from the callback to customize our effect's cleanup behavior:

function createTimer() {
  const time = createSignal(0);

  let interval;

  const dispose = createEffect(() => {
    console.log(time.value);

    return () => {
      if (interval) {
        clearInterval(interval);
      }
      console.log('Stopped!');
    };
  });

  return {
    start() {
      interval = setInterval(() => {
        time.value++;
      });
    },
    stop() {
      dispose();
    },
  };
}

const timer = createTimer();

timer.start(); // effect starts logging every second, 1...2...3...4...etc.

timer.stop(); // interval gets cleared, the effect is cleaned up, and it logs 'Stopped!'

Resources

A Resource is a reactive abstraction built to help developers incorporate asynchronous values into reactive systems. Signalis' Resource is heavily inspired by resources in SolidJS, though its API is quite a bit different.

Resource comes in two flavors:

• A standalone reactive async request
• A reactive async request that depends on another reactive value

A Resource exposes four pieces of state:

• value: Signal<ValueType | undefined> - The value of the most recent async request (this is the most up to date value)
• last: ValueType | undefined - The value of the previous run of the async request
• loading: Signal<boolean> - Whether or not the Resource is currently in the process of fetching
• error: Signal<unknown> - A Signal whose value will be populated with the contents of an error that is caught during the fetcher's execution.

createResource<ValueType>(fetcher: () => Promise<ValueType>): Resource<ValueType>

The single argument version of createResource accepts a function that performs some kind of async operation and returns a Promise. When the Resource is created, it will invoke the fetcher function and then updates the Resource's value property once the async request is complete. Resource also has a refetch method that will re-run the fetcher function and trigger updates to the Resource's reactive properties accordingly

const postResource = createResource(() => fetch('myUrl.com').then((res) => res.json()));

let error = '';
let content = '';

if (postResource.loading.value) {
  content = 'loading';
} else if (postResource.error.value) {
  error = postResource.error.value;
} else {
  content = postResource.value;
}

// run the fetch request again
postResource.refetch();

createResource<SourceType, ValueType>(source: Signal<SourceType> | Derived<SourceType>, fetcher: (source: SourceType) => Promise<ValueType>): ResourceWithSource<ValueType, SourceType>

The two-argument version of createResource allows you to designate a reactive value as a "source" for the fetcher function, and returns a ResourceWithSource. Rather than calling the fetcher immediately, a ResourceWithSource will fire the fetcher as long as the source value is not null, undefined, or false. If the source is truthy to begin with, the fetcher function will be fired as soon as it's create. ResourceWithSource will also re-fire the fetcher function any time the source value changes, updating its value property with the result and notifying any dependents of the change. The fetcher function passed to ResourceWithSource will receive the source value as an argument.

const pageNumber = createSignal(false);
const postResource = createResource(
  pageNumber,
  (pageNumber: number) =>
    fetch(`api.com/posts/${pageNumber}`).then((res) => res.json())
);

// No HTTP request has happened yet since `pageNumber` is false.

pageNumber.value = 1;

// Now that pageNumber has been updated, `postResource` will make the HTTP request using the pageNumber and update its
// `value` property, notifying anything that depends on it.

let error = '';
let content = '';

if (postResource.loading.value) {
  content = 'loading';
} else if (postResource.error.value) {
  error = postResource.error.value;
} else {
  content = postResource.value;
}

// run the fetch request again
postResource.refetch();

Stores

While the primitives covered above can technically cover all of the use cases for reactivity that you might encounter, having to call .value on every single read or write of every Signal or Derived can become a bit cumbersome, especially if you're trying to model more complicated states. In cases where you need highly nested reactivity, you'll likely that find that a Store is a more ergonomic solution.

Stores are objects (or arrays!) that have been wrapped in a Proxy that lazily wraps the object's properties in Signals as they are accessed. This gives you more fine-grained reactivity than you would get by simply wrapping an entire object in a Signal, while also making it easier to work with by calling .value on each read and write behind the scenes so you don't have to.

createStore<T extends object>(v: T | Store<T>): Store<T>

To create a Store, simply pass an object or array to createStore. Once you have a Store, any time you access one of its properties, the store will make it reactive and give you the value. Note that this only works for JS primitives (e.g. strings, numbers, booleans, etc.) and regular objects and arrays.

createStore also handles this binding for you, which means that you can define getters and methods on the object being passed into the store and they'll work like you'd expect (see the example below for an example of what this looks like).

update<T extends object>(base: T, recipe: (draft: T) => void): T

While property access on stores behaves just like it would with a normal POJO or array, the process of writing to the store is a bit different. Stores can't be mutated directly, and will throw an error if you try. This is because Signalis needs to do some careful bookkeeping behind the scenes to make sure that store changes correctly trigger updates to all the other reactive entities that observe it.

To update a store, Signalis provides an update function that behaves similarly to both Immer's produce and Solid's produce (Note: you will find that stores in general are similar to Solid's stores. This is because our stores are heavily inspired by the great work Solid has already done!). update takes two arguments: the store you want to update, and a callback that receives a draft version of the store and executes all of the changes you want to make. Unlike produce, however, update mutates the store for you, you don't need to actually do anything the result of calling update. All of the changes you make to the draft in the callback will be reflected in the store when update is finished running. That said, update still does return the updated store itself in case you would like to use the result yourself.

For example, let's say we wanted to model a classic todo list. Here's what it would look with raw Signals vs. a Store:

// with signals
const todos = createSignal([]);
const todoCount = createDerived(() => todos.length);
const todoStore = createSignal({
  todos,
  todoCount,
});

// add a todo
todoStore.todos.value = [
  ...todoStore.todos.value,
  createSignal({ id: '1', text: 'Use a store for this' }),
];

// get all todo names
todoStore.todos.value.map((todo) => todo.value.text);

// with a Store
const todoStore = createStore({
  todos: [],

  // we can define getters and method right on the store and `createStore` which will make sure
  // they're bound correctly
  get todoCount() {
    return this.todos.length;
  },

  addTodo(newTodo) {
    // Since every reactive property in a store is also a Store, you can pick properties off the
    // root store and pass them to `update` just like you would any other Store.
    update(this.todos, (draft) => {
      draft.push(newTodo);
    });
  },

  listTodoNames() {
    this.todos.map((todo) => todo.text);
  },
});

Utility functions

isSignal(v: any): v is Signal<unknown>

Indicates whether or not the value passed in is a Signal.

isDerived(v: any): v is Derived<unknown>

Indicates whether or not the value passed in is a Derived.