gRPC & High Performance APIs · 课时

流式 RPC：服务器、客户端与双向流

超越一元调用，学习 gRPC 的三种流式模式，了解如何通过一次调用发送消息序列。

流式 RPC：服务器、客户端与双向流 是 CoddyKit 上的免费 gRPC & High Performance APIs 课时。这是第 4 节课，共 4 节。你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。这是 gRPC & High Performance APIs 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 gRPC & High Performance APIs 课程共包含 4 节课。

本课时的部分内容尚未翻译，以英文显示。

Beyond Unary

A unary RPC is one request, one response. But many problems need a sequence of messages — feeds, uploads, chat.

gRPC offers three streaming modes built on HTTP/2 streams.

The Four Method Shapes

In a .proto service you can declare:

Unary: one in, one out.
Server streaming: one in, many out.
Client streaming: many in, one out.
Bidirectional: many in, many out.

Declaring Streams in Proto

The stream keyword marks a parameter as a stream.

service Feed {
  rpc Watch (WatchRequest) returns (stream Event);
  rpc Upload (stream Chunk) returns (UploadResult);
  rpc Chat (stream Message) returns (stream Message);
}

Server Streaming

The client sends one request; the server replies with many messages until it closes the stream.

Great for: live feeds, large result sets, progress updates.

// Server side (Go-style pseudocode)
func (s *server) Watch(req *WatchRequest, stream Feed_WatchServer) error {
  for _, e := range events {
    stream.Send(e)
  }
  return nil
}

Client Streaming

The client sends many messages, then the server returns a single response.

Great for: file uploads, batched ingestion, aggregations.

// Server reads the whole client stream, then replies once
func (s *server) Upload(stream Feed_UploadServer) error {
  for {
    chunk, err := stream.Recv()
    if err == io.EOF {
      return stream.SendAndClose(&UploadResult{})
    }
  }
}

Bidirectional Streaming

Both sides send streams independently over the same call. Messages can interleave freely.

Great for: chat, real-time collaboration, interactive protocols.

Why It's Efficient

All four shapes ride a single HTTP/2 stream — no new connection per message.

Multiplexing means many streaming calls share one connection without head-of-line blocking.

Flow Control

HTTP/2 provides built-in flow control, so a slow reader naturally backpressures a fast writer.

This prevents a streaming server from overwhelming a client that can't keep up.

Ending a Stream

Streams close explicitly:

Server streaming ends when the server returns.
Client streaming ends when the client signals end-of-stream and the server replies.
Bidirectional ends when both halves complete or an error/cancel occurs.

Errors & Cancellation

Either side can cancel or fail mid-stream. The peer receives a status code and should clean up.

Always handle EOF and error returns from Recv() / Send() to avoid leaks.

Choosing a Mode

Pick by data shape:

One-shot request/response: unary.
Push many results: server streaming.
Send many, get a summary: client streaming.
Continuous two-way: bidirectional.

Quick Check

Test your understanding of streaming RPCs.

Recap

You learned gRPC's streaming modes.

Unary, server streaming, client streaming, and bidirectional.
Mark streams with the stream keyword in proto.
All ride a single multiplexed HTTP/2 stream with built-in flow control.
Choose the mode that matches your data's shape.

常见问题解答

「流式 RPC：服务器、客户端与双向流」课时是免费的吗？

是的 — 「流式 RPC：服务器、客户端与双向流」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 gRPC & High Performance APIs 课程的其余内容，请升级到 CoddyKit PRO。 gRPC & High Performance APIs 课程共包含 4 节课。

「流式 RPC：服务器、客户端与双向流」这节课中我会学到什么？

超越一元调用，学习 gRPC 的三种流式模式，了解如何通过一次调用发送消息序列。你通过在浏览器中直接运行的动手代码来练习 gRPC & High Performance APIs，全天候 AI 导师会在你学习这节课的过程中回答你的问题。

学习 gRPC & High Performance APIs 需要有经验吗？

无需任何先前经验。CoddyKit 上的 gRPC & High Performance APIs 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。这是第 4 节课，共 4 节。

「流式 RPC：服务器、客户端与双向流」课时需要多长时间？

大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。

我能在这节 gRPC & High Performance APIs 课中编写并运行代码吗？

能。每节 gRPC & High Performance APIs 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。

此课程中的所有课时

Protobuf 模式定义
生成 gRPC 代码
简单的一元 gRPC 服务
流式 RPC：服务器、客户端与双向流

← 返回 gRPC & High Performance APIs