Currently, client stream receives a request with type impl tonic::IntoStreamingRequest<Message=...>, it seems not expose a interface to manually fail this streaming. A real world user case is to use client streaming to send a file, each message represent a chunk of this file. It is possible that the source of file may meet error, e.g, read from a disk file meet IO errors, thus we may want to manually terminate the streaming with a fail status so the server side can know that this is not s successful call.
I saw the internal pat of the client streaming will also transfer this stream to Stream<Item=<Result<Message, tonic::Status>>> at:
thus, it is better to expose the Result to the input type? Why not directly the request type as impl tonic::IntoStreamingRequest<Message=Result<ReqType, tonic::Status>>?
Currently, client stream receives a request with type
impl tonic::IntoStreamingRequest<Message=...>, it seems not expose a interface to manually fail this streaming. A real world user case is to use client streaming to send a file, each message represent a chunk of this file. It is possible that the source of file may meet error, e.g, read from a disk file meet IO errors, thus we may want to manually terminate the streaming with a fail status so the server side can know that this is not s successful call.I saw the internal pat of the client streaming will also transfer this stream to
Stream<Item=<Result<Message, tonic::Status>>>at:https://github.com/hyperium/tonic/blob/d312dcc0ec362cb12f6e54072622761d7466a650/tonic/src/codec/encode.rs#L38-L54
thus, it is better to expose the Result to the input type? Why not directly the request type as
impl tonic::IntoStreamingRequest<Message=Result<ReqType, tonic::Status>>?