A client-gRPC streaming written in TypeScript for browser applications. The focus is on simplicity, reliability, and extensibility. It provides convenient features to automatically reconnect and convert data to object.
Features
- Dependency-free and small in size
- When the connection is lost, it can optionally be configured to
- Automatically try to reconnect in a smart way
- Easy to override metadata and request
- Support uses the
backofffactor to apply between attempts. - Builder-class for easy initialization and configuration
- High test-coverage and in-code documentation
- Enables you to easily modify and extend the code
Installation
In your project-root:
🚀 Install with npm
npm install @nixjs23n6/grpc-stream-client
🚀 Install with yarn
yarn add @nixjs23n6/grpc-stream-client
In your project-code:
import { StreamBuilder } from '@nixjs23n6/grpc-stream-client';
Usage
Example proto:
export class ChatClient {
constructor (hostname: string,
credentials?: null | { [index: string]: string; },
options?: null | { [index: string]: any; });
room(
request: JoinRoomRequest,
metadata?: grpcWeb.Metadata
): grpcWeb.ClientReadableStream<RoomEvent>;
}
Initialization
Create a new instance with the provided StreamBuilder:
const client = new ChatClient()
const stream = new StreamBuilder(client, 'room').build();
Request
Add a request into builder:
const client = new ChatClient()
const stream = new StreamBuilder(client, 'room').withRequest(new google_protobuf_empty_pb.Empty()).build();
Metadata
Add a metada into buider:
const client = new ChatClient()
let deadline = new Date()
deadline.setSeconds(deadline.getSeconds() + BaseTimeoutStreamingRequest).toString()
const authorization = 'eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJzdWIiOiIxMjM0NTY3ODkwIiwibmFtZSI6IkpvaG4gRG9lIiwiaWF0IjoxNTE2MjM5MDIyfQ.SflKxwRJSMeKKF2QT4fwpMeJf36POk6yJV_adQssw5c'
const stream = new StreamBuilder(client, 'room').withMetadata({ deadline, authorization }).build();
Events & Callbacks
There are five events which can be subscribed to through callbacks:
export enum StreamEvents {
data = 'data', // A data was received
metadata = 'metadata', // A metadata of the connection
error = 'error', // An error occurred
end = 'end', // Connection is ended
retry = 'retry', // A try to re-connect is made
}
The callbacks are called with the issuing streaming-instance and the causing event as arguments:
const client = new ChatClient()
const stream = new StreamBuilder(client, 'room')
.onData((i, data) => console.log('[Stream] data', data)))
.onEnd((i) => console.log('[Stream] Ended'))
.onError((i, error) => console.log('[Stream] Error', error))
.onRetry((i, e) => console.log('[Stream] Retry', e))
.build();
You can remove event-listener with removeEventListener:
let stream: Stream
/* ... */
stream.removeEventListener(StreamEvents.data, dataEventListener);
Retry advanced
You can update metadata or request when using onRetry method
const client = new ChatClient()
const stream = new StreamBuilder(client, 'room')
.onRetry(async (instance) => {
const token = await requestNewAccessToken()
instance.metaData = {
authorization: token
deadline
}
})
.build();
Reconnect & Backoff
If you want the gRpc streaming to automatically try to re-connect when the connection is lost, you can provide it with a Backoff.
The streaming will use the Backoff to determine how long it should wait between re-tries. There are currently three
Backoff-implementations. You can also implement your own by inheriting from the Backoff-interface.
To use the Backoff feature, install package:
yarn add @nixjs23n6/backoff-typescript
// Or
npm install @nixjs23n6/backoff-typescript
ConstantBackoff
The ConstantBackoff will make the streaming wait a constant time between each connection retry. To use the ConstantBackoff
with a wait-time of 1 second:
const client = new ChatClient()
const stream = new StreamBuilder(client, 'room')
.withBackoff(new ConstantBackoff(1000)) // 1000ms = 1s
.build();
LinearBackoff
The LinearBackoff linearly increases the wait-time between connection-retries until an optional maximum is reached.
To use the LinearBackoff to initially wait 0 seconds and increase the wait-time by 1 second with every retry until
a maximum of 8 seconds is reached:
const client = new ChatClient()
const stream = new StreamBuilder(client, 'room')
.withBackoff(new LinearBackoff(0, 1000, 8000))
.build();
ExponentialBackoff
The ExponentialBackoff doubles the backoff with every retry until a maximum is reached. This is modelled after the binary
exponential-backoff algorithm used in computer-networking. To use the ExponentialBackoff that will produce the series
[100, 200, 400, 800, 1600, 3200, 6400]:
const client = new ChatClient()
const stream = new StreamBuilder(client, 'room')
.withBackoff(new ExponentialBackoff(100, 7))
.build();
Top comments (0)