Next-generation React databases

Pietruszewski distinguishes online content apps, which may need only a cache, from offline-first apps that must let people create and manage data without a connection. Examples include note-taking, task management, and money management apps. These apps can accumulate many records, show the same information in different views, and sync across devices or users.

Keeping app data in Redux or MobX and loading it all into JavaScript can work for smaller apps, but Pietruszewski says it does not scale well. A team account in Nozbe might contain thousands or tens of thousands of projects, tasks, and comments, making startup slow. He argues that mobile performance should be judged by interaction time, including launch time, rather than database throughput alone.

WatermelonDB loads only the records needed for the current screen and lets native code handle database work. Pietruszewski also describes it as observable: when a task changes, related task lists and project counters can update automatically. The framework uses indexing, caching, batching, and native threading to reduce unnecessary work.

On React Native, WatermelonDB uses SQLite; on the web, it uses LokiJS in a web worker. It is not a sync engine, but it tracks records added, changed, or deleted so developers can connect a separate sync system. In a demo, the app launched in about 200 milliseconds on an iPhone X with 10,000 records, and in about 800 milliseconds on a Moto G5.

Pietruszewski shows how developers define tables, fields, and relationships, then expose them through JavaScript models. A post can belong to a blog and have many comments, while a blog-level query can find nasty comments across its posts. He points viewers to the WatermelonDB GitHub repository for installation, documentation, and demo code.