Yjs vs Traditional Databases
Yjs is a conflict-free, replicated data type, which allows for distributed, collaborative editing and offline synchronization, with automatic conflict resolution.
It can also be used as a database backend.
Recently, I was deciding between Yjs and the Prisma + tRPC + SQLite combination. As it was a small todo-style app, I settled on Yjs.
Here are the factors I considered for this decision.
| YJS | tRPC + Prisma + SQLite | |
|---|---|---|
| Initial Load | Fast, as client has copy of entire DB and only updates are synced | Slow, as entire DB needs to be sent over. Fast, if paginated. |
| Concurrent edits | Supported | Not supported |
| Full-text Search | Yes, client-side | If paginated, server-side (requires PostgresQL). Otherwise, client-side. |
| Auth | Supported | Supported |
| Offline Sync | Yes, with PWA | No |
| Enforcing Relations | Difficult (e.g. two users, one deletes the parent, the other deletes the child, and they sync) | Yes |
| Enforcing Schema | Client-side | Server-side |
| Best use cases | Small databases | Large databases |