idempotency

Idempotency is a property of an operation in which applying the operation once or multiple times yields the same result as applying it once. In computing, an idempotent operation should have no additional effect after the first application, even if it is executed again, provided the system state is unchanged between executions. This concept helps reason about safety in retries, error handling, and distributed processing.

In web APIs and HTTP, idempotence is a key design consideration. The HTTP GET, PUT, and DELETE methods are defined as idempotent: sending the same request multiple times should produce the same state as sending it once (for example, retrieving a resource, updating it to a given value, or removing it). POST is not guaranteed to be idempotent and is typically used to create new resources; retries may cause duplicates unless the server implements idempotency through mechanisms such as idempotency keys. Idempotent design allows clients to retry failed requests without risking inconsistent or unintended side effects.

In databases and distributed systems, idempotency is used to manage retries and fault tolerance. Upsert operations (insert or update) can be idempotent, because applying the operation with the same input yields the same final state. Non-idempotent actions, such as generating a new sequential identifier or issuing a one-time side effect, require additional controls (for example, idempotency tokens, external tracking, or compensating actions) to be safely retried. Idempotency keys are commonly employed to distinguish distinct requests that should not produce duplicate effects when retried.

In practice, achieving idempotency involves careful API and operation design, clear state management, and, when necessary, reconciliation mechanisms to ensure repeated executions do not alter outcomes beyond the initial effect.