Signals Deep Dive

Core Design

Each signal is a SignalState<T> instance with a version counter and subscriber set:

class SignalState<T> {
  private version = 0
  private subscribers = new Set<Effect>()

  get value(): T {
    track(this)
    return this._value
  }

  set value(newValue: T) {
    if (Object.is(this._value, newValue)) return
    this._value = newValue
    this.version++
    trigger(this)
  }
}

How Tracking Works

When a signal's .value is read inside an effect or computed, the signal registers the currently active effect as a subscriber. This creates a dependency graph that knows exactly which effects to notify when a signal changes.

Lazy Computed

computed values only recalculate when their .value is read. If nothing reads a computed, it never executes its derivation function — even if its dependencies change.

Batching Strategy

Multiple signal updates inside a batch() are coalesced into a single notification:

batch(() => {
  a.value = 1
  b.value = 2
  c.value = 3
  // effects run once after batch exits
})

Why Signals?

Signals outperform Virtual DOM diffing by eliminating the need to diff subtrees whose data hasn't changed. Nexa combines signals with VDOM to get the best of both approaches.