onek

Onek

⚡️ 1.7KB full-featured state management inspired by MobX and Solid.js, batteries included ⚡️

Onek (reads as one-kay or on-ek) is a simple but powerful state management library for React based on a solid foundation of functional reactive data structures from MobX and Solid.js. It provides everything needed for managing state in complex React applications, all in a less than 2KB package.

Features

• 🚀 Reactive Observable and Computed Values - Inspired by MobX, Solid.js and Preact Signals, Onek delivers the same capabilities in a more compact package. Its goal is to provide a full-featured solution in a small size.
• 🎭 Both MobX and Solid.js Flavors - Onek is not restricted to a single way of creating observable values and defining models. Feel free to choose and mix the styles that best fit your needs.
• 🤔 Not Opinionated - Unlike Redux, Onek doesn't tell you how to structure your models. You can use global state, relational or object-oriented models - whatever you need to do your task.
• 👁 Transparency - Onek guarantees no data glitches. It ensures that every computed value is up-to-date and optimally cached, so you don't have to worry about inconsistent states during complex operations.
• 💧 No Memory Leaks - Unlike MobX, Onek doesn't need a flag for computed values to keep them alive for caching and invalidation. Onek seamlessly switches between reactive and immutable caching strategies, achieving both optimal caching and no memory leaks.
• 🧩 Single Hook - Onek offers a simple hook for integrating with React components, making your components reactive without any hassle.
• 🔀 Concurrent React Features - With out-of-the-box support for React 18 concurrent rendering, Onek allows you to optimize your app's responsiveness without any extra work.
• 🤓 Built-in Shallow Equality - Onek includes built-in shallow equality for objects, arrays, Map, and Set, covering most use cases and allowing for easy rendering optimization.
• 💾 Compatibility - Onek's core only needs ES6 Set and Map. This makes it a good choice for older browsers. Also, if you don't need concurrent React features, there is a compatibility package for React 16.8.0.
• 💯 100% Test Coverage - all core functionality is fully covered by complex and exhaustive tests.
• ⭐️ Fully TypeScript - The source code and the tests are all in TypeScript with strict rules and no hacks. This adds another proof of implementation correctness and type safety.
• ☯️ Beauty Inside - Beyond its powerful features, Onek shines in its simplicity. The implementation of complex logic in Onek is done using simple data structures and uses minimal possible amount of class fields. The core of Onek is done in an expressive object-oriented way, enabling great extensibility. All this makes Onek source easy to read and understand. As Antoine de Saint-Exupéry said, "Perfection is achieved, not when there is nothing more to add, but when there is nothing left to take away."
• 📦 ...and all in a less than 2KB package

Table of contents

• Installation
• Introduction
  • Show me the code!
  • Observable values
  • Computed values
  • Using with React
  • Actions and transactions
  • Reactions
  • Making models
    • Solid.js flavor
    • MobX flavor
• Examples
  • Counter
  • Counter list
  • Todo list
• Recipes
  • Configuring reaction scheduler
  • Catching exceptions in reactions
  • Memory leaks: why not?
• API Documentation

Installation

yarn add onek

npm install --save onek

Show me the code

Here's an example of a counter app that showcases all the main features of Onek with React:

import { action, computed, observable } from "onek";
import { useObserver } from "onek/react";

// defined observable value
const [count, setCount] = observable(0);

// define computed values derived from the observable
const canIncrease = computed(() => count() < 10);

const canDecrease = computed(() => count() > 0);

// defined actions that manipulate observable values
const increase = action(() => {
  if (canIncrease()) {
    setCount((count) => count + 1);
  }
});

const decrease = action(() => {
  if (canDecrease()) {
    setCount((count) => count - 1);
  }
});

const Counter = () => {
  // get observer instance
  const observer = useObserver();

  // wrap your render code with the observer to make it reactive
  return observer(() => (
    <div>
      <p>Count: {count()}</p>
      <button disabled={!canDecrease()} onClick={decrease}>
        -
      </button>
      <button disabled={!canIncrease()} onClick={increase}>
        +
      </button>
    </div>
  ));
};

// two counters rendered in sync
root.render(
  <>
    <Counter />
    <Counter />
  </>
);

See it on CodeSandbox

Introduction

Note: in this section Solid.js flavor will be used. If you want examples of MobX flavor, check out the MobX flavor section.

Observable values

If you're familiar with React's useState hook, you're already halfway to understanding Onek's observable function. Like the useState hook, it accepts an initial value and returns a tuple of value getter and setter. The difference is that the value getter is a function that returns the value instead of the value itself:

import { observable } from "onek";

// create observable value
const [greeting, setGreeting] = observable("hello!");

// set value directly
setGreeting("hola!");

// set value with updater function
setGreeting((oldGreeting) => oldGreeting + "!!!");

greeting() === "hola!!!!";

Please note that while it's similar to React's useState, it shouldn't be used in a React component. In this case, use the useObservable hook described in Using with React section.

Extra: equality check argument

observable supports an equality check function as a second argument. This function can be used to prevent unnecessary updates when the value hasn't effectively changed. You can also use true to use the built-in shallowEquals implementation:

import { shallowEquals } from "onek";

const [greetings, setGreetings] = observable(["hello"], true);
// or equivalently
const [greetings, setGreetings] = observable(["hello"], shallowEquals);

// setting an equal value doesn't trigger updates
setNumber(["hello"]);

Built-in shallowEquals covers plain objects, arrays, Map and Set equality, but if you need something else (like lodash isEqual), just pass it as the second argument.

Extra: storing functions in observable

In Onek, you can store functions directly in an observable. This is useful for cases where you need to store callback or computation functions. To do this, pass true as the second argument to the setter function:

// create an observable for a callback function
const [callback, setCallback] = observable(() => console.log("hello!"));

// stores the callback as is
setCallback(() => console.log("hola!"), true);

Computed values

A computed value is like useMemo in React - it's cached and returns the cached value afterward. All accessed observable or other computed values are automatically tracked, there is no need to specify a dependency list. Changes to these tracked values automatically invalidate the cached value, which is recalculated on the next access to the computed:

import { computed } from "onek";

const loudGreeting = computed(() => greeting().toUpperCase());

loudGreeting() === "HOLA!!!!";

setGreeting("hi!");

loudGreeting() === "HI!";

Extra: equality check argument

Just like with observable, you can also provide an equality check function as a second argument to computed (or true for default shallowEquals implementation). This allows you to control when the computed value is considered to have changed and needs to notify its subscribers about it. In case the equality check function returns true, the output of the computed remains referentially equal to the old one:

// create observable with an array of numbers
const [numbers, setNumbers] = observable([1, 2, 3, 4]);

// create a computed value that returns sorted array
const sortedNumbers = computed(() => [...numbers()].sort(), true);

const result = sortedNumbers();

console.log(result); // output: [1,2,3,4]

// the array is different, but sorted result is the same
setNumbers([4, 3, 2, 1]);

sortedNumbers() === result; // result is referrentially the same

The primary goal of the equality check argument is to manage and limit side effects, such as updates to React components or executions of reaction functions. These side effects might occur due to changes in the source observable or computed values. By using an equality check, you can ensure that these side effects are triggered only when the result of the computed function changes substantially, rather than being activated by every minor change to the source values. This approach can be particularly useful when the source values change frequently, but the computed result does not.

Using with React

Using observable and computed in React components is as simple as:

import { computed, observable, useObserver } from "onek";

const [greeting, setGreeting] = observable("hello!");

const loudGreeting = computed(() => greeting().toUpperCase());

const LoudGreeting = () => {
  const observer = useObserver();

  return observer(() => <p>{loudGreeting()}</p>);
};

const GreetingInput = () => {
  const observer = useObserver();

  return observer(() => (
    <input
      type="text"
      onChange={(e) => setGreeting(e.target.value)}
      value={greeting()}
    />
  ));
};

root.render(
  <>
    <GreetingInput />
    <LoudGreeting />
  </>
);

useObserver hook has no arguments and returns an observer function. You can wrap your component code with it or pass it to observable and computed getters to get the component update on their changes. Reading observable values outside of the observer fn or without passing it to getters won't subscribe the component to changes:

const [value, setValue] = observable(1);

const Component = () => {
  const observer = useObserver();

  observer(() => value()); // component will rerender on value change
  observer(value); // correct, will rerender as well
  value(observer); // also corrrect
  value(); // no rerender on value change
};

Actions and transactions

Actions automatically batch updates to observable values, and also make access to observable getters untracked - so if your action is called inside a component's render function or a reaction it won't make it re-render on accessed values change.

Important note: by default, all changes to observable values are batched until the end of the current microtask. To run reactions synchronously on the transaction end, please read the Changing reaction scheduler section.

const [x, setX] = observable(1);
const [y, setY] = observable(2);

const updateValues = action((value) => {
  const xValue = x(); // access to x is not tracked by reaction or component

  setX(0); // these two updates are batched,
  setY(xValue + value); // so components will see updated values at once
});

updateValues(100);

A transaction is the same, except it's executed immediately and doesn't make values access untracked:

import { tx } from "onek";

tx(() => {
  setX(100);
  setY(200);
});

To get the same behavior as the action use utx (Untracked transaction) instead:

const result = utx(() => {
  setX(1000);
  setY(2000);

  return x() + y(); // access is untracked
});

Async operations

Just define an action with async function:

const [data, setData] = observable(null);
const [fetching, setFetching] = observable(false);
const [error, setError] = observable(null);

const fetchData = action(async () => {
  try {
    setFetching(true);
    const responseData = await axios.get("url");
    setData(responseData);
  } catch (err) {
    setError(err);
  } finally {
    setFetching(false);
  }
});

await fetchData();

By default, Onek uses a microtask scheduler for reactions, so updates to observables are batched until the current microtask end. This means both data and fetching values will be consistent when any side effects run.

Extra: async operations for synchronous scheduler

You can configure Onek to use synchronous reaction scheduler that will execute side effects synchronously after each transaction ends. In this case, you need to use action for promise handlers or utx for code blocks in async function, i.e.:

const fetchData = action(() => {
  setFetching(true);

  return axios
    .get("url")
    .then(
      action((data) => {
        setFetching(false);
        setData(data);
      })
    )
    .catch(
      action((err) => {
        setFetching(false);
        setError(err);
      })
    );
});

or with async functions:

const fetchData = action(async () => {
  setFetching(true);

  try {
    const data = await axios.fetch("url");
    utx(() => {
      setFetching(false);
      setData(data);
    });
  } catch (err) {
    utx(() => {
      setFetching(false);
      setError(err);
    });
  }
});

Reactions

A reaction is a way to react to observable or computed changes without involving React. It's the same as the autorun function from MobX:

import { reaction } from "onek";

// will print "Greeting is HOLA!!!!"
const disposer = reaction(() => {
  console.log("Greeting is " + greeting());
});

setGreeting("Привет!"); // prints "Greeting is Привет!"

disposer();

setGreeting("Hello!"); // doesn't print anymore

disposer.run(); // prints "Greeting is Hello!" again

Extra: reaction destructor

The return value of the reaction body might be a reaction destructor - a function that is called before each reaction run and on disposer call:

const [topic, setTopic] = observable("something");

const disposer = reaction(() => {
  const currentTopic = topic();

  subscribeToTopic(currentTopic, callback);

  return () => {
    unsubscribeFromTopic(currentTopic, callback);
  };
});

setTopic("different"); // calls destructor function before executing reaction

disposer(); // unsubscribes from topic and won't run anymore

Making models

Solid.js flavor

To compose your observable and computed values into a single model, you can use the following pattern:

function makeModel(initialValue) {
  const [value, setValue] = observable(initialValue);
  const doubleValue = computed(() => value() * 2);

  return {
    value,
    setValue,
    doubleValue,
  };
}

A downside of this approach is that it's required to explicitly return all model getters/setters/actions, which can be cumbersome for large models. Also, it requires defining a convenient TypeScript type for the model:

type Model = ReturnType<typeof makeModel>;

MobX flavor

Another flavor for making Onek models is MobX flavor. It requires importing an additional tiny (~300 bytes) package:

import { makeObservable } from "onek/mobx";

class Model {
  constructor(initialValue) {
    this.value = observable.prop(initialValue);
    this.double = computed.prop(() => this.value * 2);

    makeObservable(this);
  }
}

const model = new Model(10);

// read observable and computed values
model.value === 10;
model.double === 20;

// set observable value
model.value = 100;

It defines observable and computed values on the class and then calls makeObservable to create getters/setters on the class instance. The only difference in defining the values on the class is that you need to use .prop modifier on observable/computed factories. Otherwise, the usage of MobX models is equivalent to Solid.js ones - just read the values inside observer function to make a component re-render on changes.

Note: it's safe to call makeObservable more than once on a class instance. This makes it work for inheritance cases where both base and inherited classes have observable values.

Examples?

Simple counter

Simple counter - Actions and models

import { action, observable, useObserver } from "onek";

const makeCounter = (initial) => {
  const [count, setCount] = observable(initial);
  const inc = action(() => setCount((count) => count + 1));
  const dec = action(() => setCount((count) => count - 1));
  const reset = action(() => setCount(initial));

  return { count, inc, dec, reset };
};

const Counter = ({ counter }) => {
  const observer = useObserver();

  const { count, inc, dec, reset } = counter;

  return observer(() => (
    <>
      <button onClick={inc}>+</button>
      <button onClick={dec}>-</button>
      <button onClick={reset}>Reset</button>
      Count: {count()}
    </>
  ));
};

const counter = makeCounter(0);

root.render(<Counter counter={counter} />);

Counter list

Counter list with stats - Model composition and computed data

import { observable, computed, action, useObserver } from "onek";
import { makeCounter, Counter } from "./Counter";

const makeCountersList = () => {
  const [counters, setCounters] = observable([]);

  const countersCount = computed(() => counters().length);
  const countersSum = computed(() =>
    counters().reduce((sum, counter) => sum + counter.count(), 0)
  );

  const addCounter = action(() => {
    const counter = makeCounter(0);
    setCounters((counters) => [...counters, counter]);
  });

  const removeCounter = action((counter) => {
    setCounters((counters) =>
      counters.filter((_counter) => _counter !== counter)
    );
  });

  const resetAll = action(() => {
    counters().forEach((counter) => counter.reset());
  });

  return {
    counters,
    countersCount,
    countersSum,
    addCounter,
    removeCounter,
    resetAll,
  };
};

const CounterStats = ({ count, sum }) => {
  const observer = useObserver();

  return observer(() => (
    <>
      <p>Total count: {count()}</p>
      <p>Total sum: {sum()}</p>
    </>
  ));
};

const CountersList = ({ model }) => {
  const observer = useObserver();

  return observer(() => (
    <div>
      <CounterStats count={model.countersCount} sum={model.countersSum} />
      <button onClick={model.addCounter}>Add</button>
      <button onClick={model.resetAll}>Reset all</button>
      {model.counters().map((counter) => (
        <div>
          <Counter counter={counter} />
          <button onClick={() => model.removeCounter(counter)}>Remove</button>
        </div>
      ))}
    </div>
  ));
};

const countersList = makeCountersList();

root.render(<CountersList model={countersList} />);

Todo List

Todo List - Complex multi-component app

import { action, computed, observable, useObserver } from "onek";

let id = 0;

export const makeTodo = (todoText) => {
  const [text, setText] = observable(todoText);
  const [done, setDone] = observable(false);

  const toggleDone = action(() => {
    setDone((done) => !done);
  });

  return {
    id: id++,
    text,
    done,
    setText,
    toggleDone,
  };
};

export const makeTodoList = () => {
  const [text, setText] = observable("");
  const [todos, setTodos] = observable([], true);
  const [filter, setFilter] = observable("ALL");

  const doneTodos = computed(() => {
    return todos().filter((todo) => todo.done());
  });

  const undoneTodos = computed(() => {
    return todos().filter((todo) => !todo.done());
  });

  const visibleTodos = computed(() => {
    switch (filter()) {
      case "ALL":
        return todos();
      case "DONE":
        return doneTodos();
      case "UNDONE":
        return undoneTodos();
    }
  }, true);

  const addTodo = action(() => {
    const todo = makeTodo(text());
    setTodos((todos) => [...todos, todo]);
    setText("");
  });

  const removeTodo = action((todo) => {
    setTodos((todos) => todos.filter((_todo) => _todo !== todo));
  });

  const clearDone = action((todo) => {
    setTodos(undoneTodos());
  });

  return {
    text,
    setText,
    todos,
    filter,
    visibleTodos,
    setFilter,
    addTodo,
    removeTodo,
    clearDone,
  };
};

const FILTER_OPTIONS = [
  { name: "All", value: "ALL" },
  { name: "Done", value: "DONE" },
  { name: "Undone", value: "UNDONE" },
];

const NewTodoInput = ({ model }) => {
  const observer = useObserver();

  const { text, setText, addTodo } = model;

  return observer(() => (
    <div>
      <input onChange={(e) => setText(e.target.value)} value={text()} />
      <button onClick={addTodo} disabled={text().length === 0}>
        Add
      </button>
    </div>
  ));
};

const TodoListFilter = ({ model }) => {
  const observer = useObserver();

  return observer(() => (
    <select
      value={model.filter()}
      onChange={(e) => model.setFilter(e.target.value)}
    >
      {FILTER_OPTIONS.map(({ name, value }) => (
        <option key={value} value={value}>
          {name}
        </option>
      ))}
    </select>
  ));
};

const Todo = ({ model }) => {
  const observer = useObserver();

  return observer(() => (
    <div className="todo">
      <label>
        <input
          type="checkbox"
          checked={model.done()}
          onChange={model.toggleDone}
        />
        <span
          style={{ textDecoration: model.done() ? "line-through" : "none" }}
        >
          {model.text()}
        </span>
      </label>
    </div>
  ));
};

export const TodoList = ({ model }) => {
  const observer = useObserver();

  return observer(() => (
    <div className="todo-list">
      <button onClick={model.clearDone}>Clear done</button>
      <TodoListFilter model={model} />
      <NewTodoInput model={model} />
      {model.visibleTodos().map((todo) => (
        <Todo key={todo.id} model={todo} />
      ))}
    </div>
  ));
};

Recipes

Optimizing React re-renders with check flag

The check flag in observable and computed constructors can be used to optimize React re-renders. It's useful when the result of your computed changes much less frequently than the source observable values. In this case, you can use the check flag to prevent unnecessary re-renders:

// todos can change frequently
const [todos, setTodos] = observable([]);

// but the result of this computed changes much less frequently
const isEmpty = computed(() => todos().length === 0, true);

Reaction scheduler

The reaction scheduler is a function that's called at the end of the first transaction executed after the previous scheduler run. It has one argument - a runner function that should somehow be " scheduled" to run. The default implementation of the scheduler is a microtask Promise-based scheduler:

const reactionScheduler = (runner) => Promise.resolve().then(runner);

configure({ reactionScheduler });

This is a good compromise between speed and developer experience, but sometimes you might want to run all reactions synchronously at the transaction end (for example, this is done in the Onek test suite):

const reactionScheduler = (runner) => runner();

configure({ reactionScheduler });

Another alternative to the default microtask scheduler is a macrotask scheduler:

const reactionScheduler = (runner) => setTimeout(runner, 0);

configure({ reactionScheduler });

Reaction exception handler

The default exception handler for auto-run reactions is just console.error. It can be configured by the reactionExceptionHandler option:

configure({
  reactionExceptionHandler: (exception) => {
    // some exception handling logic
  },
});

Memory leaks: why not?

Onek does not have memory leaks while maintaining optimal caching for computed values. There is no keepAlive option like in MobX, and here's why. When a computed value has lost its last subscriber or is being read in an untracked context without existing subscribers, it enters a **passive ** state. This state means the computed is no longer referenced by any observable or other computed, but still holds references to its dependencies, so it can check later if some of them changed.

How is change detection possible without guarantees that values stored in observable and computed value are immutable? The answer is simple: along with the value, observable and computed store a revision - an immutable plain object that is new each time an observable or computed is updated. This allows the implementation of reselect-like logic of checking computed dependencies with very small overhead and preserves cached values without any memory leaks.

API Documentation

Interfaces

Here are some general interfaces used in the following documentation:

import { ComputedImpl, ObservableImpl } from "./types";

type ISubscriber = ComputedImpl | ReactionImpl;

interface Getter<T> {
  (subscriber?: ISubscriber): T;
}

interface ObservableGetter<T> extends Getter<T> {
  instance: IObservable<T>;
}

interface ComputedGetter<T> extends Getter<T> {
  instance: IComputed<T>;

  destroy(): void;
}

interface Setter<T> {
  (value?: T | UpdaterFn<T>, asIs?: boolean): void;
}

type CheckFn<T> = (prev: T, next: T) => boolean;
type UpdaterFn<T> = (prevValue: T) => T;

observable

function observable<T>(
  value: T,
  checkFn?: boolean | CheckFn<T>
): readonly [ObservableGetter<T>, Setter<T>];

Creates a getter and setter for reactive value. The value argument is the value stored in the observable instance, and the checkFn is a function that's used for checking if the new value from the setter is the same as the old one.

The getter is a function that can accept an ISubscriber - return value of useObserver hook or the value of instance attribute of a computed getter.

The setter function can accept a value argument that can be of a generic type or an updater function that returns a value of the generic type.

The second argument to the setter function is an asIs boolean that indicates if the value should be stored as is without interpreting it as an updater function.

The setter also can be called without arguments - this will mark the observable as changed without changing its value. This can be useful when you mutate the observable value directly without changing the reference to it.

computed

function computed<T>(
  fn: () => T,
  checkFn?: boolean | CheckFn<T>
): ComputedGetter<T>;

Creates a getter for a computed value. The fn argument is a function that returns the computed. The checkFn argument is a function that's used for checking if the new value from the setter is the same as the old one. It can be true to use the built-in shallowEquals implementation.

reaction

type Destructor = (() => void) | null | undefined | void;

type Disposer = (() => void) & { run: () => void };

function reaction(fn: () => Destructor, manager?: () => void): Disposer;

Creates a reaction that runs the fn function and subscribes to all accessed observables and computed values. The fn function can return a reaction destructor - a function that's called before each reaction run and on the disposer call. If the manager function is specified, it's called instead of the reaction body. It should schedule the reaction to run later.

action

function action<Args extends any[], T>(
  fn: (...args: Args) => T
): (...args: Args) => T;

Creates an action that runs the fn function and batches all updates to observables and computed.

tx

function tx(fn: () => void): void;

Executes the fn function immediately and batches all updates to observables and computed.

utx

function utx<T>(fn: () => T, subscriber = null): T;

Executes the fn function immediately and batches all updates to observables and computed. The difference from tx is that it makes all observable and computed values accessed inside the fn untracked, so they won't make the component rerender or reaction run.

untracked

function untracked<Args extends any[], T>(
  fn: (...args: Args) => T
): (...args: Args) => T;

Creates a function that runs the fn function and makes all observable and computed values accessed inside the fn untracked, so they won't make the component rerender or reaction run.

useObserver

function useObserver(): ISubscriber | undefined;

Returns the ISubscriber instance that is also a function. The resulting function can be used to wrap your component code with it or be passed as an argument to observable and computed getters to make the component rerender on their changes. Also, it can be called with observable or computed as an argument, it will also make the component rerender on their changes.

shallowEquals

function shallowEquals<T>(prev: T, next: T): boolean;

Returns true if prev and next are equal. Supports plain objects, arrays, Map and Set.

License

MIT

Author

Eugene Daragan