@loke/http-rpc
v5.5.0
Published
## Breaking Changes for v5
Downloads
1,172
Readme
@loke/http-rpc
Breaking Changes for v5
- createRequestHandler now accepts a list of services.
- The root endpoint on a request handler now returns an array of metadata for each service
- Legacy mode can be enabled to still handle calls in the older
/methodName
format. - It is now preferred to use the
/service-name/methodName
format, even if a process only hosts 1 service.
Migrating from v4 to v5
- Pass in your services as an array, instead of 1-by-1.
- If using an older style setup with one service hosted at
/rpc
then enable legacy mode.
v4:
RequestHandler is directly exposed in v4
const lokeHttpRpc = require("@loke/http-rpc");
const myRpcService = lokeHttpRpc.createRequestHandler(
myService,
MY_SERVICE_META,
);
app.use("/rpc", myRpcService);
app.use(lokeHttpRpc.createErrorHandler({ log: (msg) => console.log(msg) }));
v5:
createRequestHandler adds the service name to the path where its exposed. ("/rpc/service-name"). This allows to handle multiple services with single handler.
const { createRequestHandler, createErrorHandler } = require("@loke/http-rpc");
// service will be exposed on /rpc/service-name
app.use(
"/rpc",
createRequestHandler([{ implementation: myService, meta: MY_SERVICE_META }]),
);
app.use(createErrorHandler({ log: (msg) => console.log(msg) }));
// or... service will be exposed on /rpc AND /rpc/service-name
// but will be limited to 1 service
app.use(
"/rpc",
createRequestHandler([{ implementation: myService, meta: MY_SERVICE_META }], {
legacy: true,
}),
);
app.use(createErrorHandler({ log: (msg) => console.log(msg) }));
Implementation Guide
const { createRequestHandler, createErrorHandler } = require("@loke/http-rpc");
const myService = {
async doStuff() {
return await Promise.resolve("stuff done");
},
moreStuff(stuffs) {
return "you wanted " + stuffs;
},
};
const MY_SERVICE_META = {
service: "my-service", // display name
help: "Documentation goes here",
multiArg: false, // defaults to false. If true accepts an array for arguments, if false an array will be assumed to be the first (and only) argument.
expose: [
// The methods to be exposed publicly
{
methodName: "moreStuff",
methodTimeout: 15000,
paramNames: ["stuffs"],
help: "This is a silly method",
},
],
};
const rpcHandler = createRequestHandler([
{
implementation: service,
meta: SERVICE_META,
},
]);
const errorLogger = (msg) => console.log(msg);
app.use("/rpc", rpcHandler);
app.use(createErrorHandler({ log: errorLogger }));
Then, if running on port 5000:
curl -X POST http://localhost:5000/rpc/doStuff
Also, to list runtime RPC metadata you can GET /rpc
curl -X GET http://localhost:5000/rpc
Schemas and Context
Since v5.1.0 we now support JTD Schemas for requests and responses validation (Via AJV).
For more in-depth information about using the schema see SCHEMAS.md
.
Since v5.3.0 we now support passing a context object to the service methods. Use serviceWithSchema
if you don't want to use the context.
import {
createRequestHandler,
createErrorHandler,
serviceWithSchema,
} from "@loke/http-rpc";
import { Context } from "@loke/context";
interface Thing {
name: string;
}
const myService = {
async doStuff(ctx: Context, args: {}) {
return await Promise.resolve("stuff done");
},
async getThing(ctx: Context, args: { name: string }): Promise<Thing> {
return { name: args.name };
},
};
// Type for definitions
type Defs = { Thing: Thing };
// instead of `typeof myService` you could also name a type like
// type Service = {}
const myRpcService = contextServiceWithSchema<typeof myService, Defs>(
myService,
{
name: "my-service",
logger: console,
// Record<string, JTD>
definitions: {
Thing: {
properties: {
name: { type: "string" },
},
},
},
methods: {
doStuff: {
help: "This is a silly method",
// JTD
requestTypeDef: {
properties: {},
},
responseTypeDef: { type: "string" },
},
getThing: {
help: "Get a thing",
requestTypeDef: {
properties: { name: { type: "string" } },
},
responseTypeDef: { ref: "Thing" },
},
},
},
);
const rpcHandler = createRequestHandler([myRpcService]);
const errorLogger = (msg) => console.log(msg);
app.use("/rpc", rpcHandler);
app.use(createErrorHandler({ log: errorLogger }));
Void result schema
Return types of void
should generally be avoided, but if you need to use them
you can use voidSchema
to define the schema for the response.
import { voidSchema } from "@loke/http-rpc";
const myService = {
async doSomething(): Promise<void> {
return;
},
};
const myRpcService = serviceWithSchema(myService, {
name: "my-service",
logger: console,
methods: {
doSomething: {
requestTypeDef: {
properties: {},
},
responseTypeDef: voidSchema,
},
},
});