WebSocket concatenateFragmentedFrames is a no-op

[kamilkloch]

Current implementation of optionallyConcatenateFrames returns the original stream:

tapir/server/http4s-server/src/main/scala/sttp/tapir/server/http4s/Http4sWebSockets.scala

Line 71 in fdfa3da

s.mapAccumulate(None: Accumulator) {

[adamw]

Do you see where the problem is? I guess it would be high time to add some tests for the pipeToBody specifically

[kamilkloch]

Accumutor is always None - lines

tapir/server/http4s-server/src/main/scala/sttp/tapir/server/http4s/Http4sWebSockets.scala

Line 76 in fdfa3da

case (Some(Left(acc)), f: WebSocketFrame.Binary) if f.finalFragment => (None, Some(f.copy(payload = acc ++ f.payload)))

through

tapir/server/http4s-server/src/main/scala/sttp/tapir/server/http4s/Http4sWebSockets.scala

Line 79 in fdfa3da

case (Some(Right(acc)), f: WebSocketFrame.Text) if !f.finalFragment => (Some(Right(acc + f.payload)), None)

are never invoked.

[kamilkloch]

Question, if such functionality is actually needed. http4s concatenates web socket frames by default, not sure about the other servers.

[adamw]

It is, not all servers do :) Though of course, can't tell right now which don't ;) I just remember this being the case ...

[kamilkloch]

But, astonishingly, nobody is complaining about non-concatenating the frames by tapir interpreter for other servers.

[adamw]

Yes, maybe these are unpopular servers, or people don't send fragmented frames that often. I don't know really :)