Hey. I'm looking for some clues about how exactly Room behaves when: • a query returns a

Flow

• there is an active subscriber of that

Flow

• a few changes in the observed table take place in the background • the subscriber of the

Flow

consumes the data slower than the updates are being made In particular, I would like to know if the

Flow

is going to emit every intermediate snapshot of the queried data or just the latest.

----data update A----------data update B-----------data update C-->
       \
        \
      collected snapshot A & processing --------------------------------->finished
                                                                              \
                                                                               \
                                                                            what's in the next collected snapshot? B or C?

The term you are looking for is backpressure: https://medium.com/google-developer-experts/backpressure-in-your-kotlin-flows-3eec980869c7

Yeah, I'm familiar with that term. I tested Room-generated

Flow

with`conflate()` yesterday, hoping to always get the latest possible result from Room. It helped a lot but, unfortunately, it still didn't work the way I expected. My case looked like this:

database.dataFlow()
  .conflate()
  .collect { values ->
    // some processing
    database.updateSomethingThatAffectsTheFlow(values)
}

It turned out that the change made by

database.updateSomethingThatAffectsTheFlow()

was not reflected in the next collected value. I had to wait for yet another value to see that change 😕 I suspect that Room didn't have enough time to compute the new value for the

Flow

before the subscriber became ready to receive the next update. Or something.