What is the difference between flatMap and switchMap methods in reactive programming?

In reactive programming, the methods flatMap and switchMap are used for transforming and managing data streams, but they have different semantics:

• flatMap: creates a new inner stream for each incoming element and merges all these streams into one. All inner streams operate in parallel, and results from them can arrive in any order.

• switchMap: upon receiving a new element, cancels (unsubscribes from) the previous inner stream and switches to the new one. Thus, only one inner stream — the latest — is active at any time.

Example in Kotlin with RxJava:

observable.flatMap { value ->
    someAsyncOperation(value) // returns Observable
}

observable.switchMap { value ->
    someAsyncOperation(value) // the previous call is canceled when a new value arrives
}

Use flatMap when you want to process all requests in parallel, and switchMap when you need to ignore old requests when new ones appear (for example, in autocomplete search).