Node.js Backend Development Bootcamp · Leçon

Les événements comme source de vérité et le journal en ajout uniquement

Remplacez l’état mutable par un flux d’événements immuable et reconstruisez l’état en rejouant les événements.

Leçon 1 sur 413 étapes

Les événements comme source de vérité et le journal en ajout uniquement est une leçon Node.js Backend Development Bootcamp gratuite sur CoddyKit. Ceci est la leçon 1 sur 4. Tu peux lire la leçon complète ci-dessous gratuitement — puis la pratiquer en direct dans le navigateur avec un éditeur de code intégré et un tuteur IA 24/7. Elle fait partie du parcours d'apprentissage Node.js Backend Development Bootcamp, et ta progression se synchronise sur le web et l'application CoddyKit. Le cours Node.js Backend Development Bootcamp comprend 4 leçons au total.

Certaines parties de cette leçon n'ont pas encore été traduites et s'affichent en anglais.

The Problem with Mutable State

In a classic CRUD backend, you store the current state of an entity and overwrite it on every change. A user's balance is one row; when money moves, you UPDATE the number in place.

This is convenient but lossy. Once you overwrite the old value, the history is gone. You cannot answer:

How did the balance reach this value?
When and why did each change happen?
What did the state look like last Tuesday?

Event Sourcing flips this: instead of storing the latest state, you store the sequence of facts that produced it.

// Classic mutable CRUD: history is destroyed on every write
let account = { id: 'acc-1', balance: 100 };

function deposit(amount) {
  account.balance += amount; // old value is gone forever
}

function withdraw(amount) {
  account.balance -= amount; // no record of why or when
}

deposit(50);
withdraw(30);
console.log(account); // { id: 'acc-1', balance: 120 }
// We know the result, but not the journey.

Events as the Source of Truth

In Event Sourcing, an event is an immutable record of something that already happened. Events are named in the past tense: MoneyDeposited, MoneyWithdrawn, AccountOpened.

The event log becomes the source of truth. The current state is no longer stored directly — it is a derived value you compute by replaying events.

Events are facts: they cannot be changed or deleted.
State is an opinion: a projection of the facts at a point in time.

Each event captures intent and context, not just the resulting number.

// An event is an immutable, past-tense fact
const events = [
  { type: 'AccountOpened',  data: { accountId: 'acc-1', owner: 'Ada' }, at: '2026-01-01T09:00:00Z' },
  { type: 'MoneyDeposited', data: { accountId: 'acc-1', amount: 50 },  at: '2026-01-02T10:15:00Z' },
  { type: 'MoneyWithdrawn', data: { accountId: 'acc-1', amount: 30 },  at: '2026-01-03T11:30:00Z' }
];

// Each entry is a fact that already happened. Nothing is overwritten.
console.log(`Stored ${events.length} immutable facts.`);

The Append-Only Log

The store that holds events is an append-only log. There are only two real operations:

Append a new event to the end.
Read events in order, usually for a given stream (e.g. one account).

There is no UPDATE and no DELETE. This single constraint gives you a complete, ordered, tamper-evident audit trail for free.

Because the log is ordered, the position (or version) of each event matters: replaying in the same order always yields the same state.

// A minimal in-memory append-only log
class EventLog {
  constructor() { this.events = []; }

  append(event) {
    // Only ever push to the end — never mutate or remove
    const stored = { ...event, position: this.events.length + 1 };
    this.events.push(Object.freeze(stored));
    return stored;
  }

  read() {
    return [...this.events]; // ordered copy
  }
}

const log = new EventLog();
log.append({ type: 'AccountOpened', data: { accountId: 'acc-1' } });
log.append({ type: 'MoneyDeposited', data: { accountId: 'acc-1', amount: 50 } });
console.log(log.read());

Reconstructing State by Replaying

If state is derived, how do we get it back? We replay: start from an empty state and apply each event in order. This function is often called apply, evolve, or a reducer.

Notice the shape: (state, event) => newState. It is exactly a reducer — the same idea as Array.prototype.reduce.

The reducer is pure: same events in, same state out.
It handles each event type and returns a new state object.

function applyEvent(state, event) {
  switch (event.type) {
    case 'AccountOpened':
      return { id: event.data.accountId, owner: event.data.owner, balance: 0 };
    case 'MoneyDeposited':
      return { ...state, balance: state.balance + event.data.amount };
    case 'MoneyWithdrawn':
      return { ...state, balance: state.balance - event.data.amount };
    default:
      return state; // ignore unknown events
  }
}

const events = [
  { type: 'AccountOpened',  data: { accountId: 'acc-1', owner: 'Ada' } },
  { type: 'MoneyDeposited', data: { amount: 50 } },
  { type: 'MoneyWithdrawn', data: { amount: 30 } }
];

const state = events.reduce(applyEvent, null);
console.log(state); // { id: 'acc-1', owner: 'Ada', balance: 20 }

Streams and Aggregates

You rarely replay every event in the system. Events are grouped into streams, one per entity — for example account-acc-1. The entity reconstructed from a stream is called an aggregate.

To load an aggregate:

Read only that stream's events (filtered by streamId).
Replay them through the reducer.
Return the resulting in-memory state.

Keeping streams small keeps replay fast and keeps consistency boundaries clear.

function loadAggregate(allEvents, streamId, reducer) {
  return allEvents
    .filter(e => e.streamId === streamId)
    .sort((a, b) => a.position - b.position)
    .reduce(reducer, null);
}

const allEvents = [
  { streamId: 'account-acc-1', position: 1, type: 'AccountOpened',  data: { accountId: 'acc-1', owner: 'Ada' } },
  { streamId: 'account-acc-2', position: 1, type: 'AccountOpened',  data: { accountId: 'acc-2', owner: 'Lin' } },
  { streamId: 'account-acc-1', position: 2, type: 'MoneyDeposited', data: { amount: 75 } }
];

const reducer = (s, e) => {
  if (e.type === 'AccountOpened') return { id: e.data.accountId, balance: 0 };
  if (e.type === 'MoneyDeposited') return { ...s, balance: s.balance + e.data.amount };
  return s;
};

console.log(loadAggregate(allEvents, 'account-acc-1', reducer));

Commands vs Events

A crucial distinction: a command is a request to do something (it may be rejected); an event is the record that it happened (it cannot be undone).

Withdraw is a command — imperative, present tense, may fail.
MoneyWithdrawn is an event — past tense, a settled fact.

The aggregate's job is to take the current state plus a command, run business rules, and decide which event(s) to append — or reject the command entirely.

function decide(state, command) {
  switch (command.type) {
    case 'Withdraw':
      if (command.amount > state.balance) {
        throw new Error('Insufficient funds'); // command rejected
      }
      return [{ type: 'MoneyWithdrawn', data: { amount: command.amount } }];
    case 'Deposit':
      return [{ type: 'MoneyDeposited', data: { amount: command.amount } }];
    default:
      throw new Error('Unknown command: ' + command.type);
  }
}

const state = { balance: 40 };
console.log(decide(state, { type: 'Withdraw', amount: 30 }));
try { decide(state, { type: 'Withdraw', amount: 100 }); }
catch (e) { console.log('Rejected:', e.message); }

The Decide / Evolve Cycle

Put the pieces together and you get the core write flow of an event-sourced aggregate:

Load: replay the stream to rebuild current state.
Decide: run the command against that state to produce new events.
Append: write the new events to the log.
Evolve: the same reducer that loaded state also keeps it current.

The decide function never writes; the evolve reducer never validates. This separation keeps the domain logic clean and testable.

function evolve(state, event) {
  if (event.type === 'MoneyDeposited') return { ...state, balance: state.balance + event.data.amount };
  if (event.type === 'MoneyWithdrawn') return { ...state, balance: state.balance - event.data.amount };
  return state;
}
function decide(state, cmd) {
  if (cmd.type === 'Deposit') return [{ type: 'MoneyDeposited', data: { amount: cmd.amount } }];
  if (cmd.type === 'Withdraw' && cmd.amount <= state.balance)
    return [{ type: 'MoneyWithdrawn', data: { amount: cmd.amount } }];
  throw new Error('Invalid command');
}

let history = [{ type: 'MoneyDeposited', data: { amount: 100 } }];
let state = history.reduce(evolve, { balance: 0 });   // load
const newEvents = decide(state, { type: 'Withdraw', amount: 60 }); // decide
history = [...history, ...newEvents];                 // append
state = newEvents.reduce(evolve, state);              // evolve
console.log(state); // { balance: 40 }

Optimistic Concurrency with Expected Version

Two requests may try to change the same aggregate at once. Because the log is append-only, we guard writes with the expected version: the position the writer believes the stream is at.

When appending, you say "I expect this stream to be at version N." If another writer already advanced it, the append fails and the caller retries by reloading.

No locks held across the request.
Conflicts are detected, never silently lost.

class VersionedStore {
  constructor() { this.streams = new Map(); }

  append(streamId, expectedVersion, newEvents) {
    const current = this.streams.get(streamId) || [];
    if (current.length !== expectedVersion) {
      throw new Error(
        `Concurrency conflict: expected v${expectedVersion}, got v${current.length}`
      );
    }
    this.streams.set(streamId, [...current, ...newEvents]);
    return current.length + newEvents.length;
  }
}

const store = new VersionedStore();
store.append('acc-1', 0, [{ type: 'AccountOpened' }]);   // ok -> v1
try {
  store.append('acc-1', 0, [{ type: 'MoneyDeposited' }]); // stale version
} catch (e) { console.log(e.message); }
store.append('acc-1', 1, [{ type: 'MoneyDeposited' }]);  // correct -> v2
console.log('Final version:', store.streams.get('acc-1').length);

Snapshots: Replay Without Re-reading Everything

Replaying thousands of events on every load gets slow. A snapshot is a cached copy of the aggregate state at a known version. To load, you start from the snapshot and only replay events after it.

Important: snapshots are an optimization, not a source of truth. You can delete every snapshot and still rebuild perfect state from the events. The log remains authoritative.

Store snapshot + the version it represents.
On load: hydrate from snapshot, then replay the tail.

function loadWithSnapshot(snapshot, events, evolve) {
  // snapshot = { state, version } or null
  let state = snapshot ? snapshot.state : { balance: 0 };
  const fromVersion = snapshot ? snapshot.version : 0;
  return events
    .filter(e => e.position > fromVersion)
    .reduce(evolve, state);
}

const evolve = (s, e) =>
  e.type === 'MoneyDeposited' ? { balance: s.balance + e.data.amount } : s;

const events = [
  { position: 1, type: 'MoneyDeposited', data: { amount: 100 } },
  { position: 2, type: 'MoneyDeposited', data: { amount: 50 } },
  { position: 3, type: 'MoneyDeposited', data: { amount: 25 } }
];
const snapshot = { state: { balance: 150 }, version: 2 };
console.log(loadWithSnapshot(snapshot, events, evolve)); // { balance: 175 }

Schema Evolution and Upcasting

Events live forever, so their shape will outlive the code that wrote them. You can never edit an old event in place, but the new code must still understand it.

The standard tool is upcasting: a function that transforms an old event version into the current shape at read time, before it reaches the reducer.

Add a version field to each event type.
Provide new fields with sensible defaults; rename via the upcaster.
Never mutate stored events — transform a copy on the way in.

// v1 had `amount` (cents implied); v2 adds explicit `currency`
function upcast(event) {
  if (event.type === 'MoneyDeposited' && (event.version || 1) === 1) {
    return {
      ...event,
      version: 2,
      data: { ...event.data, currency: 'USD' } // default for legacy events
    };
  }
  return event;
}

const legacy = { type: 'MoneyDeposited', data: { amount: 50 } };
console.log(upcast(legacy));
// { type: 'MoneyDeposited', data: { amount: 50, currency: 'USD' }, version: 2 }

Persisting the Log in Node.js

In production the append-only log is backed by durable storage: a dedicated store like EventStoreDB, or a relational table used as a log. A typical Postgres design:

One events table with (stream_id, version, type, data jsonb, recorded_at).
A unique constraint on (stream_id, version) — this is what enforces optimistic concurrency at the database level.
Inserts only; the application code never issues UPDATE or DELETE on it.

The same load/decide/append cycle runs on top, just with SQL behind the store interface.

// Sketch of an append against a Postgres-backed log (pg client `db`).
// The UNIQUE (stream_id, version) constraint rejects concurrent duplicates.
async function appendEvents(db, streamId, expectedVersion, events) {
  const client = await db.connect();
  try {
    await client.query('BEGIN');
    let version = expectedVersion;
    for (const e of events) {
      version += 1;
      await client.query(
        `INSERT INTO events (stream_id, version, type, data)
         VALUES ($1, $2, $3, $4)`,
        [streamId, version, e.type, JSON.stringify(e.data)]
      );
    }
    await client.query('COMMIT');
    return version;
  } catch (err) {
    await client.query('ROLLBACK'); // unique violation => concurrency conflict
    throw err;
  } finally {
    client.release();
  }
}

Quick Check

Consider an event-sourced account aggregate.

Réponds-y dans la leçon interactive

Recap

You replaced mutable state with an immutable event stream:

Events are past-tense, immutable facts; the append-only log that holds them is the source of truth.
State is derived, not stored — you rebuild it by replaying events through a pure reducer: (state, event) => newState.
Events are grouped into streams, one per aggregate.
The write cycle is load → decide → append → evolve; commands may be rejected, events never are.
Expected version gives optimistic concurrency; a UNIQUE(stream_id, version) constraint enforces it in the database.
Snapshots speed up replay but are optional; upcasting keeps old events readable as schemas evolve.

Next, you'll build read models (projections) and complete the CQRS picture.

Gratuit pour commencer

Apprends JavaScript avec un tuteur IA — gratuit

Écris et exécute du vrai code dans ton navigateur, obtiens de l'aide instantanée d'un tuteur IA disponible 24h/24, et reprends là où tu t'es arrêté sur le web ou dans l'app.

Cours: 22
Leçons: 92

Commencer gratuitement Voir le cours complet

Questions Fréquemment Posées

La leçon « Les événements comme source de vérité et le journal en ajout uniquement » est-elle gratuite ?

Oui — le texte complet de « Les événements comme source de vérité et le journal en ajout uniquement » est gratuit à lire ici sur le web. Pour la pratiquer de manière interactive (un éditeur de code intégré et un tuteur IA 24/7) et déverrouiller le reste du cours Node.js Backend Development Bootcamp, passe à CoddyKit PRO. Le cours Node.js Backend Development Bootcamp comprend 4 leçons au total.

Qu'est-ce que j'apprendrai dans « Les événements comme source de vérité et le journal en ajout uniquement » ?

Remplacez l’état mutable par un flux d’événements immuable et reconstruisez l’état en rejouant les événements. Tu pratiques Node.js Backend Development Bootcamp avec du code pratique que tu exécutes directement dans le navigateur, et un tuteur IA 24/7 répond à tes questions au fur et à mesure que tu avances dans la leçon.

Dois-je avoir de l'expérience pour commencer Node.js Backend Development Bootcamp ?

Aucune expérience préalable n'est requise. Node.js Backend Development Bootcamp sur CoddyKit est structuré pour les débutants jusqu'aux apprenants avancés, donc tu peux commencer ici ou depuis le début et avancer à ton rythme. Ceci est la leçon 1 sur 4.

Combien de temps prend la leçon « Les événements comme source de vérité et le journal en ajout uniquement » ?

La plupart des leçons CoddyKit prennent environ 5–10 minutes. Chacune est courte et interactive, tu progresses régulièrement et tu repiques exactement où tu t'es arrêté sur le web et l'app.

Peux-tu écrire et exécuter du code dans cette leçon Node.js Backend Development Bootcamp ?

Oui. Chaque leçon Node.js Backend Development Bootcamp inclut un éditeur de code intégré, tu écris et exécutes du vrai code directement dans ton navigateur et tu reçois des retours IA instantanés — aucune configuration locale requise.

Toutes les leçons de ce cours

Les événements comme source de vérité et le journal en ajout uniquement
Agrégats, commandes et modélisation des événements métier
Créer des modèles de lecture et des projections
Instantanés, gestion des versions et évolution du schéma d’événements

← Retour à Node.js Backend Development Bootcamp