effect-batching

Define requests with Request.Class and resolve them with RequestResolver.make, which receives all collected requests as a batch. Concurrent Effect.request calls are automatically collected and resolved together in a single resolver invocation.

Use SqlResolver.ordered, grouped, findById, or void from effect/unstable/sql to create schema-validated batched SQL resolvers. Concurrent requests are combined into single SQL statements using IN clauses, with automatic transaction grouping.

Yes, requests use structural equality via the Equal trait, so two identical requests within a batch window are deduplicated automatically. In-flight deduplication also attaches new callers to pending results when caching is enabled.

QueryFailure occurs when a resolver fails to complete every entry in the batch. Each entry must be completed with entry.completeUnsafe using Exit.succeed, Exit.fail, or Exit.die, including entries with no matching results.

Batching requires concurrent execution. Effect.forEach without a concurrency option runs sequentially, so each request resolves individually. Pass { concurrency: "unbounded" } to enable concurrent execution and trigger batching.

Apply RequestResolver.withCache with a capacity and optional LRU or FIFO strategy to cache results in memory. For TTL control, use RequestResolver.asCache to convert the resolver into a Cache instance with timeToLive configuration.