Diego Alonso

> Have we measured this to make a difference? No. It is based on the knowledge and interpretation of the definition of `evalMap`: ```Scala def evalMap[F2[x] >: F[x], O2](f: O...

Since the overhead of `evalMap` is a bit smaller now, this may not be needed.

@mrdziuban Good morning, and thanks for reporting this. One clarification: when you mention that it happens in the latest version `v3.2.7`, does that mean it does not happen in the...

@njwilson23 Good afternoon, Nat. Thanks for reporting this bug. We will try to look at it as soon as we can. In the meantime, to help us with testing, could...

I have just been running the executable, in the PR you opened, on top of the current `main`. It seems that the issue has been fixed in the four weeks...

@Daenyth In a recent Pull Request #2316, the `InspectableQueue` trait was removed from fs2, in favour of the `Queue` implementations from `cats-effect`. Given that, is this still needed?

@danicheg Here is some possible intuition / explanation as to what may be a significant difference: In both codes, the "observer" is pulling the bytes from the source stream, one...

@danicheg What would be the performance if the code was to do the following: - Keep the vector of source chunks, and keep the observer as it is, inserting the...

> also appending to an immutable vector leads to a vector allocation, doesn't it? Yeah, but only a part of it. IIRC internal implementation is a sort of tree, so...

Looking at the ByteVector implementation, one thing I can notice is that the sub-class `Buffer` does include an `AtomicLong`. Could it be that the updates to that "cell" add the...