D1 IndexStore

D1IndexStore is the built-in IndexStore implementation for Cloudflare D1. It stores document identifiers and indexed field values as a rebuildable projection. Complete document content remains in the configured DocumentStore.

Dependency and construction

The D1 adapter is part of dev.nexcraft:r2d1:

implementation("dev.nexcraft:r2d1:<version>")

Configure the account, database, and API token explicitly:

import dev.nexcraft.r2d1.d1.D1Config;
import dev.nexcraft.r2d1.d1.D1IndexStore;

D1Config config = new D1Config(accountId, databaseId, apiToken);
D1IndexStore indexes = new D1IndexStore(config);

The values are placeholders. Use secret configuration for the API token. D1Config.toString() redacts the account, database, and token values.

REST transport

The public constructor uses Cloudflare D1’s REST API through Java’s reusable asynchronous HttpClient. It does not use a Cloudflare Worker transport, create an executor, create a scheduler, or provide automatic retries.

If the application owns an HttpClient, pass it to the two-argument constructor. Closing the store does not close that caller-owned client.

Initialize a collection

initialize(Class<?>) inspects the document annotations, creates or validates the managed collection schema, and returns a CompletionStage. PersistenceCollectionFactory supplies this initializer to the collection facade:

R2D1 database =
    R2D1.builder()
        .collectionFactory(
            new PersistenceCollectionFactory(
                documentStore, indexes, documentCodec, indexes::initialize))
        .build();

The collection is not used until initialization completes. Concurrent first initialization calls for the same collection share one stage, and a failed initialization is not retried automatically by the same D1IndexStore instance.

Query projection

The D1 v1 adapter supports indexed String, Long, Double, and Boolean values. Queries use explicitly indexed fields and the public comparison model. D1 never downloads R2 objects to filter them in application memory.

clear(String) removes only derived rows. Tables, columns, and managed indexes remain available for the replacement rows written by rebuildIndex().

Read Querying for the public operators and Consistency and Recovery for partial failures and recovery.