core
Tiny, efficient, featured, and extensible core to handle reactivity right. The ultimate state manager. Build anything, from a small widget to a huge application.
included in @reatom/framework
The raw API description is below.
About
Reatom allows you to describe both simple and complex logic using three main components: atoms for data reference, actions for logic processing, and context (ctx
) for system isolation. This core is a perfect solution for building your own high-order library or an entire framework, with all the packages built on top of it.
Reatom is inspired by the React and Redux architectures. All processed data should be immutable, computations should be pure, and all side effects should be scheduled for a separate effects queue using ctx.schedule(callback)
. Only consistent data transactions should be applied. All prerequisites can be checked in this article: What is a state manager.
Installation
Usage
Let’s describe a simple example of a search input with a tip and a list of goods. This code is written in TypeScript, but you can also use JavaScript; a lot of types are inferred automatically. Pay your attention to the comments; they will help you to understand the core concepts.
All atoms and actions accept ctx
as their first argument to match and process all data inside it. It assists you greatly in many areas: testing, debugging, SSR, effects chain management, and logging. It is the most powerful feature of Reatom and is indispensable in complex scenarios. And it only requires three extra letters for each function call - super efficient!
As the entire data processing flows through the context, you can easily inspect it: ctx.subscribe(logs => console.log(logs))
or connect a separate logger to view all changes in your app with proper formatting.
Now, let’s outline some logic.
Here we just described the logic of a module which uses ctx
, but does not import it. This is because we want to use the same module in different contexts, such as view components and tests. It is a good architectural practice in itself.
So, we should connect an IO and our module together somewhere.
Do you want to see the docs for React adapter next?
Action handling (advanced)
It is better to keep atoms stupid and handle all logic inside actions. But sometimes you need to turn the direction of your code coupling and make atoms depend on an action. And you can do it!
An action is an atom with a temporal state, which is an array of all passed payloads. This state is cleared after the transaction ends; if you try to get
or spy
an action which wasn’t called, you will receive an empty array. But if the action was called, the array will contain some elements.
You need to know one rare tricky thing. If during a transaction you call an action and read its dependent atom a few times step by step,
ctx.get
will return an array of all passed payloads, butctx.spy
will return an array with only new elements that weren’t handled in this reducer during this transaction. To make this rare case correct, you should spy your dependencies in the same way each time, without conditions. In other words, for this case your dependencies list should be static.
API
atom
API
The atom()
function is a factory for an atomic-based reactive primitive. Atoms don’t store their data (state, listeners, dependencies) within themselves; they only provide a key to a cache in ctx (context). You can think of an atom as a prototype for a cache. One of the most powerful features of Reatom is that the cache is immutable, and it is recreated on each relative update. The immutability of the cache helps to process transactions and is extremely useful for debugging. Don’t worry, it is also quite efficient.
As Atom is a key, it should be mapped somewhere to its cache. ctx
has an internal WeakMap caches
, which store your data until there is a link to Atom. When you subscribe (connect) and unsubscribe (disconnect) from Atom, the state isn’t reset or deleted; it is still stored in the cache, which will be cleared by the GC only after the link to the Atom disappears from your closures. This behavior is the most intuitive and works just like any variable storing. So, if you define a global Atom available in a few of your modules, the state will always persist in memory during the application lifetime, whether you subscribed or unsubscribed for the Atom, which is useful. If you need to clear the state on disconnect or do other lifetime transformations, check the hooks package and withreset helper.
If you need to create a base mutable atom, just pass the initial value to atom
. Pass the atom name as a second argument (it is optional but strongly recommended). The resulted atom will be mutable (Mut
) with a callable signature (a function); you can mutate it by passing a context and a new value or a reducer function.
All atom state changes should be immutable.
You could create a computed derived atom by passing a function to atom
. The first argument of the passed reducer is a special kind of ctx
with a spy
function, which allows you to subscribe to the passed atom and receive its fresh state. The second argument is an optional previous state
, which you can initiate by defining a default value.
Note to TypeScript users: It is impossible to describe the reducer type with an optional generic state argument, which is returned from the function. If you use the second
state
argument, you should define its type; do not rely on the return type.
To store a function in Reatom state, just wrap it in a container, like
atom({ fn })
.
Reatom allows you to use native language features to describe your conditions, with all reactive dependencies reconnecting in real-time.
Moreover, you could dynamically create and manage atoms.
You could handle each update independently by passing a function to the spy
method. It is useful for action-reaction scenarios or if you need to handle a few concurrent updates.
atom.pipe
API
Pipe is a general chain helper, it applies an operator to the atom to map it to another thing. Classic operator interface is <T extends Atom>(options?: any) => (anAtom: T) => aNewThing
. The main reason is a readable and type-safe way to apply decorators.
withInit
allows you to configure the initial state of the atom reading, which is sometimes more predictable and safer for testing.
Operator with
prefix mean that the target atom will be changed somehow and the returned reference will the same. reatom/async uses operators a lot to configure the behavior of the effect by composition, which is good for tree-shaking. Check naming conventions and more examples in this guide
Btw, actions has pipe
too!
atom.onChange
API
All links and computations between atoms and actions are performed in a separate context. However, there can be many cases when you need to describe some logic between two things statically outside a context, such as an action trigger on a data change, etc. The onChange
hook allows you to define this common logic right in the place of your atoms definition.
onChange
returns an unsubscribe function which you should use if you are adding a hook dynamically to a global atom.
The key difference between a hook and a subscription is that the hook does not activate the connections.
action
API
Actions are atoms with temporal states, which live only during a transaction. The action state is an array of parameters and payloads. The array is needed to handle multiple action calls during a transaction batch. Action callbacks can change atoms or call other actions, but their dependencies will only be notified after the callback ends - that is what a batch means.
Possible usage:
Action state is Array<{ params: Array<any>, payload: any }>
, but action call returns the payload:
action.onCall
API
The same as atom.onChange, but with the relative arguments: payload
and params
.
ctx
API
ctx
is the main shell that holds the state for all atoms, and where all user and metadata reside. Each atom and action produces an immutable version of the context and it should not be mutated!
An important rule to note, even if you might not need it, is: don’t run one context inside another, such as ctx1.get(() => ctx2.get(anAtom)). Doing so will throw an error.
ctx.get
atom API
Get fresh atom state
get<T>(anAtom: Atom<T>): T
ctx.get
batch API
Start transaction and batch all updates, same as in action call
get<T>(cb: () => T): T
ctx.subscribe
atom API
Subscribe to atom new state. Passed callback called immediately and after each atom state change.
subscribe<T>(anAtom: Atom<T>, cb: (newState: T) => void): () => void
ctx.subscribe
log API
Subscribe to transaction end. Useful for logging.
subscribe(cb: (logs: Array<AtomCache>, error?: Error) => void): () => void
ctx.schedule
To achieve atomicity, each update (action call / atom mutation) starts a complex batch operation, which tries to optimize your updates and collect them into a new immutable log of new immutable cache snapshots. If some computation throws an error (like can't use property of undefined
) the whole update will be canceled, otherwise the new caches will be merged into the context internal caches
weak map. To achieve purity of computations and the ability to cancel them, all side-effects should be called separately in a different queue, after all computations. This is where schedule
comes in; it accepts an effect callback and returns a promise which will be resolved after the effect call or rejected if the transaction fails.
A unique feature of Reatom, especially in scheduling, is ability to define the target queue. The second argument of schedule
is a priority number:
-1
- rollback queue, useful when you need to do a side-effect during pure computations. Check the example below.0
- computations queue, schedule pure computation, which will call right after current batch.1
- the default near effect queue, used to schedule regular effects. The calling of these effects can be redefined (or delayed) using thecallNearEffect
option ofcreateCtx
.2
- lates effect queue, used to schedule subscribers. The calling of these effects can be redefined (or delayed) using thecallLateEffect
option ofcreateCtx
.
Read more in the lifecycle guild.
ctx.schedule
rollback API
Sometimes, you may want to perform a side-effect during clean calculations or need to store an artifact of an effect. To make it clean, you should describe a rollback (cleanup) function for the case of an unexpected error by passing -1
as the second argument of ctx.schedule
. Check out this example with a debounced action: