transaction-correctness

WAL checkpointing in Turso transfers frames from the write-ahead log back to the database file, supporting Passive, Full, Restart, and Truncate checkpoint types to manage log accumulation and free disk space.

Turso's WAL mode defines specific concurrency rules dictating when writers and readers can operate simultaneously, including precise checkpoint timing restrictions to prevent reader starvation and ensure data consistency.

Crash recovery in Turso involves replaying the write-ahead log to restore the database to a consistent state, ensuring all committed transactions are persisted and incomplete operations are rolled back.

Turso checkpoint types differ in aggressiveness: Passive waits for readers, Full blocks new readers, Restart additionally truncates the WAL, and Truncate removes the physical WAL file after checkpointing completes.

Turso supports strict write-ahead logging and data integrity by writing changes to the WAL before applying them to the main database, ensuring durability and enabling robust crash recovery procedures.

Checkpointing can block writes in Turso because concurrency rules require coordinating read and write access to the database file, temporarily pausing writers during the log transfer process.