Build a searchable projection from past events

Add a searchable purchase history that includes purchases made before the feature existed.

Every PurchaseRecorded event carries a purchase ID and its amount. Give the new feature a fresh consumer name and start at the beginning of the retained event log:

Java

public record PurchaseRecorded(String purchaseId, int amountInCents) {}

@Consumer(name = "purchase-history", minIndex = 0)
public class PurchaseHistory {
    @HandleEvent
    void handle(PurchaseRecorded purchase) {
        Fluxzero.index(purchase, purchase.purchaseId(), "purchases").join();
    }
}

Kotlin

data class PurchaseRecorded(val purchaseId: String, val amountInCents: Int)

@Consumer(name = "purchase-history", minIndex = 0)
class PurchaseHistory {
    @HandleEvent
    fun handle(purchase: PurchaseRecorded) {
        Fluxzero.index(purchase, purchase.purchaseId, "purchases").join()
    }
}

Register PurchaseHistory as an application handler. It first processes retained history, then continues with new events. Restarts resume from the consumer’s saved position. Repeated delivery replaces the same purchase document.

Use the result

Java

var purchases = Fluxzero.search("purchases").fetch(20, PurchaseRecorded.class);

Kotlin

val purchases = Fluxzero.search<PurchaseRecorded>("purchases").fetch(20, PurchaseRecorded::class.java)

The first results appear while the consumer catches up. A log can only replay messages it still retains.

If a feature needs Model state as it was when an event happened, use an injected Model or Graph as explained in Temporal Graphs. For an existing feature that needs correction, see live and historical processing.