readthenwrite

Readthenwrite is a descriptive term used in computing to describe a pattern in which a value is first read from a data source and then a write is performed, often using the read data as input for the subsequent update. The pattern appears in databases, in-memory data structures, and distributed systems, and is frequently discussed in the context of concurrency control and data integrity.

In databases, readthenwrite can lead to race conditions if multiple transactions read the same value and then write based on that value without coordination. This can cause lost updates or write skew. Remedies include using higher isolation levels (such as serializable), applying optimistic concurrency control with version checks, or performing atomic operations that combine the read and the conditional write.

In memory and distributed caches, readthenwrite requires synchronization to ensure that concurrent readers and writers do not corrupt state. Without atomic primitives or proper locking, updates based on stale reads can produce inconsistent results.

Variants and related patterns include read-modify-write (RMW), where the modification is explicitly applied to the read value, and check-then-set or compare-and-swap (CAS) operations, which attempt the write only if the observed value remains unchanged. These approaches provide stronger guarantees than a plain readthenwrite.

Applications of the concept span configuration management, counters, feature flag updates, and state transitions in distributed services, where computing a new value from the current one is necessary but must be done safely to avoid conflicts and inconsistencies. See also concurrency control, serializability, and optimistic concurrency.