resiliency4j with reactor: transformDeferred or transform

[trung]

Hi, I saw that the example uses transformDeferred(RetryOperator.of(..)) instead of transform(...).
My thinking is that: with transformDeferred(), retry operator will be applied separately to individual subscriber. Whereas transform() will share the retry operator among subscribers.

For example, my reactive API returns a Flux. If this Flux is being resubscribed across requests then transformDeferred() makes sense, otherwise if Flux is created for each request, then transform() would be ok.

My question is:

1. Is my understanding above correct?
2. In practice, will the reactive-enabled server resubscribe the Flux across multiple requests? Or will it create a new Flux for each request?

Many thanks,

Trung

[RobWin]

Hi,
I think this question is better raised in the Reactor GitHub repo.

[RassulYunussov]

Hi @RobWin!
I think the question is valid and it would be good to have a thorough explanation why that method is preferred instead of a transform.
As within documentation it is explicitly demonstrated, that transformDeferred is used everywhere: https://resilience4j.readme.io/docs/examples-1

I'd love to see that explanation.
Thank you!