Room Across Processes

Most Android apps run in a single process. However, Android does give developers the option to have their apps be split across multiple processes.

However, this can cause problems with stuff held in memory, as multiple processes will not share that memory. In the case of Room, one area where the problem crops up is with “invalidation tracking”: the piece of Room that knows to update your reactive query results (Flow, LiveData, Observable, etc.) when the data changes. By default, that does not work across process boundaries, but you can enable it via enableMultiInstanceInvalidation() on your RoomDatabase.Builder.

In this chapter, we will explore all of this in greater detail and demonstrate how it works.

Room and Invalidation Tracking

When you modify your Room database, active observers of reactive responses from @Query-annotated functions get fresh results delivered automatically.

Under the covers, that is powered by InvalidationTracker and related code. Room tracks:

• Which tables and views are referenced by active queries, and
• Which tables are affected by database modifications

When you modify the database, Room looks up the active queries that are tied to any tables that you affected. For those, Room re-executes the queries and emits new results via the Flow, LiveData, Observable, or whatever.

This is all done for you, almost by magic. But, in reality, it just sophisticated code from a sophisticated library.