Pierre Ricadat
Pierre Ricadat
The main difference with `ZIO_GRPC_BACKPRESSURE_QUEUE_SIZE` being -1 and 1000 is that when using 1000, we call `stream.buffer` which loses the chunking. With -1, we keep the chunking which is quite...
> > `xs` always has size 1, which is surprising to me but fair enough > > that's because `map` rechunks in `ZStream.fromIterable(0 until 100).map(HelloReply(_))` Isn't it because of `buffer`...
Small reproducer using zio only: ```scala import zio.* import zio.stream.* object Test extends ZIOAppDefault { val expected = Chunk.fromIterable(0 until 100) val s = ZStream.fromChunk(expected).buffer(16) def run = s.runCollect.map(_ ==...
https://github.com/zio/zio/issues/8699
I actually added https://github.com/scalapb/zio-grpc/pull/578 to not buffer at all 😆 But yeah `bufferChunks` sounds better than `buffer`.
The ordering issue on zio side was fixed so I think this is resolved.
Done in https://github.com/zio/zio/pull/9212
How about using zio-test? It's better to stick with one library.
Sorry I forgot about this. Could you rebase and pull the latest tags to your fork? It makes mima checks fail.
How about removing the parameter and apply that behavior if `apiPath == uploadPath` (building a single route)?