fsf
v1.0.8
Published
A typesafe websocket RPC library which thins the borders between clients and servers.
Downloads
203
Maintainers
Readme
Functions Without Borders - Fonctions sans Frontières (fsf)
A typesafe websocket RPC library which thins the borders between clients and servers.
While defining a server's APIs using TS, you can automatically define the frontend's function contracts. These functions can be called "natively", as if they're already present in the frontend itself.
Usage
Installation
Install the package
npm i fsf
Common
Define an interface for your api which is accessable in your client and server source code (e.g. by using yarn workspaces).
export type API = {
hello: () => string;
world: () => string;
sum: (x: number, y: number) => number;
};
Client
Note: On the client side, all functions return a Promise with the result by default, because of the asynchronous nature of sockets. So, all passed functions are also modified to return a Promise.
import { Client } from "fsf";
import { API } from "../api";
const client = Client<API>("http://localhost:8080");
const { listFiles, searchMovie } = client;
const main = async () => {
console.log(await client.hello());
// passing multiple parameters to the function
console.log(await client.sum(12, 20));
console.log(await listFiles());
// passing a string parameter
console.log(await searchMovie("kong"));
};
main();
Server
import { Server } from "fsf";
import { API } from "../api";
import listFiles from "./apis/listFiles";
import searchMovie from "./apis/searchMovie";
const api: API = {
hello: () => "Hello World!",
sum: (x, y) => x + y,
// Make an API call to movies API
searchMovie: searchMovie,
// Fetch all files on server
listFiles: listFiles,
errorFunction: (a: any) => a.b,
};
Server(8080, api);
Error Handling
At the moment, any error on the server-side is sent to std:error
and thrown on the client side.
Try running /example/client/throwsError.ts
to check it out.
What's Not Supported?
- Passing functions as a parameter. This would require stringifying the function on the frontend and running
eval
on it on the backend, which is an UNSAFE OPERATION.
How does it work internally?
In short, the library depends on Websockets, Object Proxies, and Typescript generics to work. In detail:
1. Websockets
We use socket.io for fast and reliable socket connections. Websockets can be lighter than HTTP requests when a large number of connections are needed. Also, they have a smaller code footprint than HTTP requests. Their usage is anyways abstracted away in the codebase, and they can be replaced with any other technology if needed.
2. Object Proxies
The framework utilizes Object Proxies get control over the client object. Any function call made on a property of the client object (or on a deconstructed property), like
client.functionOne();
// or
const { functionOne } = client;
functionOne();
is handled by a get
property which has been set on the Object Proxy here.
You can go through the code to see how it uses the property name and parameters to make a socket call to the server.
3. Typescript Generics
All of the auto-complete goodness that the framework provides throughout the app depends on Typescript generics. On the server side, the type is directly applied on the API object,
const api: API = { ...yourApi };
while on the client side it's passed to the Client
initializer.
const client = Client<API>(endpoint);
The client function is actually a generic, which accepts the type provided by the user and applies Promise
to the return type of each of them. It's a very Typescript-specific piece of code but you can read it here.
Contributing
Pull requests are welcome. You'll probably find lots of improvements to be made.
Open issues for feedback, requesting features, reporting bugs or discussing ideas.
Special thanks to @mkrhere for hand-holding me through most of the code.