transaction-correctness

Turso ensures transactional correctness and durability by appending database writes to a write-ahead log (WAL) and committing them in a controlled manner. This WAL mechanism guarantees that logged writes are safely persisted before being applied to the main database.

Checkpoint variants like PASSIVE, FULL, RESTART, and TRUNCATE differ in how aggressively they transfer WAL frames back to the main database and manage subsequent read/write access. These checkpoint types balance performance and consistency during concurrent database operations.

You handle read and write concurrency by relying on Turso's per-connection and shared state management alongside WAL checkpointing. This ensures snapshot consistency and isolate concurrent operations without corrupting the database state during active transactions.

You should use different checkpoint types like PASSIVE, FULL, or TRUNCATE based on your system's need to balance performance and consistency. PASSIVE allows active transactions to continue, while FULL, RESTART, and TRUNCATE enforce stricter WAL frame transfer rules for crash recovery.

Yes, Turso supports snapshot consistency for concurrent database access by managing per-connection and shared state through its WAL mechanism. This ensures that concurrent reads and writes maintain a consistent view of the database even during active checkpointing.