The type-safe Node.js and browser RPC library built on top of socket.io. Empowered with SOA governance and load balancing.
yarn add orpcshared.ts
export interface HelloServicePrototype extends ServicePrototype {
say(name: string): string;
}
export interface MyRPCSchema extends RPCSchema {
hello: HelloServicePrototype;
}server.ts
class HelloService implements HelloServicePrototype {
say(name: string): string {
return `Hello, ${name}!`;
}
}
let myRPC: MyRPCSchema = {
hello: new HelloService(),
};
const server = new RPCServer(myRPC);
server.start(8013);client.ts
let rpc = createClient<MyRPCSchema>('http://localhost:8013/');
let result = await rpc.hello.say('dizy'); // Type-safe
console.log(result); // 'Hello, dizy!'
ServicePrototypeandRPCSchemashould be shared between client and server.Serviceshould be implemented only on server.
A service prototype is an interface that defines the signature of the methods provided in the service. This is shared between both server and client side, so it will help type both our server implementation and client usage.
interface FileServicePrototype extends ServicePrototype {
list(dir: string): Promise<FileInfo[]>;
upload(file: Buffer, saveToDir: string): Promise<void>;
download(path: string): Promise<Buffer>;
delete(path: string): Promise<boolean>;
rename(oldPath: string, newPath: string): Promise<boolean>;
}In order to run multiple services on one single server, we build RPC schema to assemble these services into one bigger type. This is also shared between client and server.
interface MyRPCSchema extends RPCSchema {
file: FileServicePrototype;
system: SystemServicePrototype;
blog: BlogServicePrototype;
}Service is implemented in a form of Class. And it could be considered as a collection of methods ready to be remotely called. It is expected to implement the corresponding ServerPrototype on server.
class FileService implements FileServicePrototype{
async list(dir: string): Promise<FileInfo[]>{
let files = await //...implementation
return files;
}
//...
}Gateway is a reverse proxy that takes clients' calls and sends them to one of the known servers, then gets results and feed them back to clients. To clients, a Gateway behaves exactly like a RPCServer, except its performance compete with several servers combined. Gateway is especially useful for high-cpu-usage services. Also, Gateway supports SOA governance and load balancing, so it can make sure services are highly-available and calls are well-balanced in terms of distribution to servers.
let config = {
servers: [
{url: 'http://localhost:8021/', weight: 10},
{url: 'http://localhost:8022/', weight: 3},
{url: 'http://localhost:8023/', weight: 4},
], // Servers' info
loadBalancer: smoothWRRLoadBalancer, // built-in Smooth-Weighted-Round-Robin LoadBalancer
log: {debug: true}, // Turn on debug mode
};
let gateway = new Gateway(config);
gateway.start(8014);-
constructor (rpcSchema?: SchemaImplement<Schema>) -
methods:
-
start(port?: number, hostname?: string): voidportdefaults to80,hostnamedefaults to'localhost' -
stop(): void -
register(name: string, service: object): void -
register(name: string, service: ServiceConstructor, initializeForEachClient?: boolean): voidDynamically register new service.
initializeForEachClientdefaults tofalse -
unregister(name: string): booleanDynamically cancel registered service.
-
-
parameters:
-
url?: stringDefaults to
'http://localhost/'
-
-
returns:
RPCClient<RPCSchema>
-
usages:
let rpc = createClient<MyRPCSchema>(8080); let result = await rpc.myService.methodA(...args);
RPCClient should always be created by createClient. That way, service and schema type information will be attached to it.
-
constructor (url?: string) -
methods:
-
open(): voidOpen connection. Automatically called when created.
-
close(): void -
call(service: string, data: CallData): Promise<any> -
call(service: string, method: string, params: any[], options: CallOptions): Promise<any>Call remote service method.
-
let config: GatewayConfig = {
servers: [
{url: 'http://localhost:8015', weight: 10},
{url: 'http://localhost:8016', weight: 4},
{url: 'http://localhost:8017', weight: 6},
], // Servers' info
loadBalancer: smoothWRRLoadBalancer, // Optional: defaults to `evenLoadBalancer`.
downgrade: true, // Optional: defaults to `true`.
downgradeAtAnyError: false, // Whether any error thrown by server will cause its downgrade. Optional: defaults to `false`.
downgradeTolerantTime: 3, // How many tolerant times to fail before the server is downgraded to `0` weight. Optional: defaults to `3`.
downgradeDeadSleepTime: 5, // A many times can a `0` weight server be offered a chance to revive. Optional: defaults to `5`.
log: {
enable: true,
debug: true,
}, // Optional: Defaults to `{enable: true, debug: false}`
};-
constructor(config: GatewayConfig) -
methods:
-
start(port?: number, hostname?: string): voidStart gateway server.
portdefaults to80,hostnamedefaults to'localhost' -
stop(): void -
reloadConfig(config: GatewayConfig): void
-
-
usage:
see Core Concepts
MIT, see the LICENSE file for details.