Jayson
Jayson is a JSON-RPC 2.0 and 1.0 compliant server and client written in JavaScript for node.js that aims to be as simple as possible to use.
Table of contents
Features
Example
A basic JSON-RPC 2.0 server via HTTP:
Server example in examples/simple_example/server.js:
const jayson = require('jayson');
const server = new jayson.Server({
add: function(args, callback) {
callback(null, args[0] + args[1]);
}
});
server.http().listen(3000);
Client example in examples/simple_example/client.js invoking add
on the above server:
const jayson = require('jayson');
const client = jayson.Client.http({
port: 3000
});
client.request('add', [1, 1], function(err, response) {
if(err) throw err;
console.log(response.result);
});
Installation
Install the latest version of jayson from npm by executing npm install jayson
in your shell. Do a global install with npm install --global jayson
if you want the jayson
client CLI in your PATH.
Changelog (only notable milestones/changes)
- 4.1.0
- New server option
maxBatchLength
- 4.0.0
- Remove
lodash
dependency which should halve bundle size. There might be minor incompatibilities if you pass funky object or array types to jayson methods.
- 3.6.4
- Websocket client and server support
- 3.6.1
- JSON-RPC 2.0 notifications no longer have id property unless overridden
- 3.3.3
- Promise support for browser client
- TypeScript declaration for promise browser client
- TypeScript declaration for browser client
- 3.3.0
- Remove URL parsing when passing a string option to the TLS and TCP client, string options are instead treated as an IPC path
- 3.0.0
- Can pass a context object to handlers
- Breaking:
collect
option removed from jayson.Server/Method
. JSON-RPC params to handlers are now always in the first argument.
- 2.1.0
- Experimental typescript support
- 2.0.6
- 2.0.0
- Added support for promises
- Breaking:
collect: true
is now the default option for a new jayson.Server
and jayson.Method
- 1.2.0
- 1.1.1
- More http server events
- Remove fork server and client
- Add server routing
- 1.0.11
Add support for a HTTPS client
- 1.0.9
Add support for TCP servers and clients
CLI client
There is a basic CLI client in bin/jayson.js
and it should be available as jayson
in your shell if you installed the package globally. Run jayson --help
to see how it works.
Requirements
Jayson does not have any special dependencies that cannot be resolved with a simple npm install
or yarn install
.
Class documentation
In addition to this document, a comprehensive class documentation made with jsdoc is available at jayson.tedeh.net.
Running tests
- Change directory to the repository root
- Install the development packages by executing
npm install --dev
- Run the tests with
npm run test
- Run the typescript tests with
npm run test-tsc
- Run the coverage tests with
npm run coverage
Typescript
Since v2.1.0
there is typescript support available with jayson.
If you encounter any problems with the type definitions, see the Contributing section.
Usage
Client
The client is available as the Client
or client
property of require('jayson')
.
Client interface description
Name | Description |
---|
Client | Base class |
Client.tcp | TCP sub-class |
Client.tls | TLS sub-class |
Client.http | HTTP sub-class |
Client.https | HTTPS sub-class |
Client.browser | Standalone class |
Client.websocket | Websocket sub-class |
Every client supports these options:
Option | Default | Type | Description |
---|
reviver | undefined | Function | JSON.parse reviver |
replacer | undefined | Function | JSON.stringify replacer |
generator | RFC4122 generator | Function | Generates a String for request ID. |
version | 2 | Number | JSON-RPC version to support (1 or 2) |
notificationIdNull | false | Boolean | Since 3.6.1. When true "id" property of a request will be set to null when version 2. |
Client.http
Uses the same options as http.request in addition to these options:
Option | Default | Type | Description |
---|
encoding | utf8 | String | Determines the encoding to use |
headers | undefined | Object | Extend the headers sent by the client |
Client.http Events
The HTTP client will emit the following events:
Event | When | Arguments | Notes |
---|
http request | Created an HTTP request | 1. Instance of http.ClientRequest | |
http response | Received an HTTP response | 1. Instance of http.IncomingMessage 2. Instance of http.ClientRequest | |
http error | Underlying stream emits error | 1. Error | |
http timeout | Underlying stream emits timeout | | Automatically causes the request to abort |
It is possible to pass a string URL as the first argument. The URL will be run through url.parse. Example:
const jayson = require('jayson');
const client = jayson.Client.http('http://localhost:3000');
Client.https
Uses the same options as https.request in addition to the same options as Client.http
. This means it is also possible
to pass a string URL as the first argument and have it interpreted by url.parse.
Will emit the same custom events as Client.http
.
Client.tcp
Uses the same options as net.connect.
Client.tcp Events
Since version 3.5.1
The TCP client will emit the following events:
Event | When | Arguments | Notes |
---|
tcp socket | TCP socket is opened | 1. net.Socket | Can be used to setup timeouts |
tcp error | TCP socket emits error | 1. Error emit by net.Socket | |
Client.tls
Uses the same options as tls.connect.
Client.tls Events
Since version 3.5.1
The TLS client will emit the following events:
Event | When | Arguments | Notes |
---|
tcp socket | TCP socket is opened | 1. net.Socket | Can be used to setup timeouts |
tcp error | TCP socket emits error | 1. Error emit by net.Socket | |
Client.browser
The browser client is a simplified version of the regular client for use browser-side. It does not have any dependencies on node.js core libraries, but does depend on the uuid
package for generating request ids. It also does not know how to "send" a request to a server like the other clients.
Because it does not depend on any core libraries, the browser client is not an instance of JaysonClient
or EventEmitter
and therefore does not emit any of the normal request events that the other clients do.
To use the browser client, require('jayson/lib/client/browser')
and pass a calling/transport function as the first argument. The transport function receives a JSON-RPC string request and is expected to callback with a string response received from the server (not JSON) or an error (not a JSON-RPC error).
The reason for dealing with strings is to support the reviver
and replacer
options like the other clients.
This client example in examples/browser_client/client.js below uses node-fetch in the transport function, but a dropin replacement for use in an actual browser could instead use whatwg-fetch.
The browser client has a separate TypeScript type declaration available in jayson/lib/client/browser/index.d.ts
which depends on the main Jayson type declaration.
'use strict';
const jaysonBrowserClient = require('jayson/lib/client/browser');
const fetch = require('node-fetch');
const callServer = function(request, callback) {
const options = {
method: 'POST',
body: request,
headers: {
'Content-Type': 'application/json',
}
};
fetch('http://localhost:3000', options)
.then(function(res) { return res.text(); })
.then(function(text) { callback(null, text); })
.catch(function(err) { callback(err); });
};
const client = new jaysonBrowserClient(callServer, {
});
client.request('multiply', [5, 5], function(err, error, result) {
if(err) throw err;
console.log(result);
});
Client.websocket
Since v3.6.4
Experimental websocket client that wraps around an isomorphic-ws
instance. Will listen to every received (JSON) message and see if it matches any of the currently outstanding requests made, in which case the callback of that outstanding request will fire. If you do not provide the timeout
option it will wait forever. Has a promise-based equivalent receiving the same options, and a companion jayson server where you can find an example.
Has the following options:
Option | Default | Type | Description |
---|
url | undefined | String | First argument to require('isomorphic-ws') if options.ws not set |
ws | undefined | require('isomorphic-ws') instance | WebSocket instance |
timeout | undefined | Number | Timeout in ms before callbacking with an error |
If you want to "unwrap" the isomorphic-ws
instance you can use the Client.websocket.prototype.unlisten
which stops listening for messages on the isomorphic-ws
instance.
Notifications
Notification requests are for cases where the reply from the server is not important and should be ignored. This is accomplished by setting the id
property of a request object to null
.
Client example in examples/notifications/client.js doing a notification request:
const jayson = require('jayson');
const client = jayson.Client.http({
port: 3000
});
client.request('ping', [], null, function(err) {
if(err) throw err;
console.log('ok');
});
Server example in examples/notifications/server.js:
const jayson = require('jayson');
const server = new jayson.Server({
ping: function(args, callback) {
callback();
}
});
server.http().listen(3000);
Notes
- Any value that the server returns will be discarded when doing a notification request.
- Omitting the third argument
null
to Client.prototype.request
does not generate a notification request. This argument has to be set explicitly to null
for this to happen. - Network errors and the like will still reach the callback. When the callback is invoked (with or without error) one can be certain that the server has received the request.
- See the Official JSON-RPC 2.0 Specification for additional information on how Jayson handles notifications that are erroneous.
- Since 3.6.1 When making a JSON-RPC 2.0 notification request the "id" property will be omitted in the request object. In previous versions it was set to
null
against the recommendation of the official specification. This behaviour can be overridden with the notificationIdNull
option.
Batches
A batch request is an array of individual requests that are sent to the server as one. Doing a batch request is very simple in Jayson and consists of constructing an array of individual requests (created by not passing a callback to Client.prototype.request
) that is then itself passed to Client.prototype.request
.
Combined server/client example in examples/batch_request/index.js:
const jayson = require('jayson');
const server = new jayson.Server({
add: function(args, callback) {
callback(null, args[0] + args[1]);
}
});
const client = new jayson.Client(server);
const batch = [
client.request('does_not_exist', [10, 5]),
client.request('add', [1, 1]),
client.request('add', [0, 0], null)
];
client.request(batch, function(err, errors, successes) {
if(err) throw err;
console.log('errors', errors);
console.log('successes', successes);
});
client.request(batch, function(err, responses) {
if(err) throw err;
console.log('responses', responses);
});
Notes
- See the Official JSON-RPC 2.0 Specification for additional information on how Jayson handles different types of batches, mainly with regards to notifications, request errors and so forth.
- There is no guarantee that the results will be in the same order as request Array
request
. To find the right result, compare the ID from the request with the ID in the result yourself.
Client callback syntactic sugar
When the length (number of arguments) of a client callback function is either 2 or 3 it receives slightly different values when invoked.
- 2 arguments: first argument is an error or
null
, second argument is the response object as returned (containing either a result
or a error
property) or null
for notifications. - 3 arguments: first argument is an error or null, second argument is a JSON-RPC
error
property or null
(if success), third argument is a JSON-RPC result
property or null
(if error).
When doing a batch request with a 3-length callback, the second argument will be an array of requests with a error
property and the third argument will be an array of requests with a result
property.
Client events
A client will emit the following events (in addition to any special ones emitted by a specific interface):
Event | When | Arguments | Notes |
---|
request | About to dispatch a request | 1: Request object | |
response | Received a response | 1: Request object 2: Response object received | |
Server
The server classes are available as the Server
or server
property of require('jayson')
.
The server also sports several interfaces that can be accessed as properties of an instance of Server
.
Server interface description
Name | Description |
---|
Server | Base interface for a server that supports receiving JSON-RPC requests |
Server.tcp | TCP server that inherits from net.Server |
Server.tls | TLS server that inherits from tls.Server |
Server.http | HTTP server that inherits from http.Server |
Server.https | HTTPS server that inherits from https.Server |
Server.websocket | Websocket server that uses isomorphic-ws Server |
Server.middleware | Method that returns a Connect/Express compatible middleware function |
Servers supports these options:
Option | Default | Type | Description |
---|
reviver | null | Function | JSON.parse reviver |
replacer | null | Function | JSON.stringify replacer |
router | null | Function | Return the function for method routing |
useContext | false | Boolean | Passed to methodConstructor options |
params | undefined | Array/Object/null | Passed to methodConstructor options |
methodConstructor | jayson.Method | Function | Server functions are made an instance of this class |
version | 2 | Number | JSON-RPC version to support (1 or 2) |
maxBatchLength | Infinity | Number | Maximum batch requests allowed |
Server.tcp
Uses the same options as the base class. Inherits from net.Server.
Server.tls
Uses the same options as the base class. Inherits from tls.Server.
Server.http
Uses the same options as the base class. Inherits from http.Server.
Server.http Events
Event | When | Arguments | Notes |
---|
http request | Incoming HTTP request | 1. Instance of http.IncomingMessage | |
http response | About to send a HTTP response | 1. Instance of http.ServerResponse 2. Instance of http. IncomingMessage | |
Server.https
Uses the same options as the base class. Inherits from https.Server and jayson.Server.http
. For information on how to configure certificates, see the documentation on https.Server.
Will emit the same custom events as Server.http
.
Server.middleware
Uses the same options as the base class. Returns a function that is compatible with Connect or Express. Will expect the request to be req.body
, meaning that the request body must be parsed (typically using connect.bodyParser
) before the middleware is invoked.
The middleware supports the following options:
Option | Default | Type | Description |
---|
end | true | Boolean | If set to false causes the middleware to next() instead of res.end() when finished. |
Middleware example in examples/middleware/server.js:
const jayson = require('jayson');
const jsonParser = require('body-parser').json;
const connect = require('connect');
const app = connect();
const server = new jayson.Server({
add: function(args, callback) {
callback(null, args[0] + args[1]);
}
});
app.use(jsonParser());
app.use(server.middleware());
app.listen(3000);
Server.websocket
Websocket server that either wraps around a provided require('isomorphic-ws').Server
instance or creates one from scratch. Expects every incoming message on every connection to be a valid JSON-RPC call.
The websocket server supports the following options in addition to the base class:
Option | Default | Type | Description |
---|
wss | undefined | require('isomorphic-ws').Server | If not provided will be created |
Websocket server example in examples/websocket/server.js:
const jayson = require('jayson');
const server = new jayson.Server({
add: function (args, done) {
const sum = args.reduce((sum, val) => sum + val, 0);
done(null, sum);
},
});
const wss = server.websocket({
port: 12345,
});
Websocket client example in examples/websocket/client.js:
const jayson = require('jayson');
const client = jayson.Client.websocket({
url: 'ws://localhost:12345',
});
client.ws.on('open', function () {
client.request('add', [1,2,3,4], function (err, result) {
console.log(err, result);
client.ws.close();
});
});
Many interfaces at the same time
A Jayson server can use many interfaces at the same time.
Server example in examples/many_interfaces/server.js that listens to both http
and a https
requests:
const jayson = require('jayson');
const server = new jayson.Server();
const http = server.http();
const https = server.https({
});
http.listen(80, function() {
console.log('Listening on *:80');
});
https.listen(443, function() {
console.log('Listening on *:443');
});
Using the server as a relay
Passing an instance of a client as a method to the server makes the server relay incoming requests to wherever the client is pointing to. This might be used to delegate computationally expensive functions into a separate server or to abstract a cluster of servers behind a common interface.
Frontend server example in examples/relay/server_public.js listening on *:3000
:
const jayson = require('jayson');
const server = new jayson.Server({
add: jayson.Client.http({
port: 3001
})
});
server.http().listen(3000);
Backend server example in examples/relay/server_private.js listening on *:3001
:
const jayson = require('jayson');
const server = new jayson.Server({
add: function(args, callback) {
callback(null, args[0] + args[1]);
}
});
server.http().listen(3001);
Every request to add
on the public server will now relay the request to the private server. See the client example in examples/relay/client.js.
Method routing
Passing a property named router
in the server options will enable you to write your own logic for routing requests to specific functions.
Server example with custom routing logic in examples/method_routing/server.js:
Note: You are strongly recommended to check method names using Object.prototype.hasOwnProperty
to prevent a denial-of-service attack against your Jayson server when names such as __defineGetter__
are used in JSON-RPC calls.
const jayson = require('jayson');
const methods = {
add: function(args, callback) {
callback(null, args[0] + args[1]);
}
};
const server = new jayson.Server(methods, {
router: function(method, params) {
const fn = this._methods.hasOwnProperty(method) ? this._methods[method] : null;
if(typeof fn === 'function') {
return fn;
}
if(method === 'add_2') {
const fn = server.getMethod('add').getHandler();
return new jayson.Method(function(args, done) {
args.unshift(2);
fn(args, done);
});
}
}
});
server.http().listen(3000);
Client example in examples/method_routing/client.js invoking add_2
on the above server:
const jayson = require('jayson');
const client = jayson.Client.http({
port: 3000
});
client.request('add_2', [3], function(err, response) {
if(err) throw err;
console.log(response.result);
});
Server example of nested routes where each property is separated by a dot (you do not need to use the router option for this):
const jayson = require('jayson');
const methods = {
foo: {
bar: function(callback) {
callback(null, 'ping pong');
}
},
math: {
add: function(args, callback) {
callback(null, args[0] + args[1]);
}
}
};
const map = Object.keys(methods).reduce(collapse('', '.', methods), {});
const server = new jayson.Server(map);
function collapse(stem, sep, obj) {
return function(map, key) {
const prop = stem ? stem + sep + key : key;
const value = obj[key];
if(typeof value === 'function') map[prop] = value;
else if(typeof value === 'object' && value !== null) map = Object.keys(value).reduce(collapse(prop, sep, value), map);
return map;
}
}
Notes
- If
router
does not return anything, the server will respond with a Method Not Found
error. - The
Server.prototype
methods method
, methods
, removeMethod
and hasMethod
will not use the router
method, but will operate on the internal Server.prototype._methods
map. - The
router
method is expected to return instances of jayson.Method
(>=1.2.0)
Method definition
You can also define server methods inside a wrapping object named jayson.Method
. This allows additional options about the method to be specified. Using this wrapper - explicitly or implicitly (via server options) - makes it trivial to have your method accept a variable amount of arguments.
The method class is available as the Method
or method
property of require('jayson')
. It supports these options:
Option | Default | Type | Description |
---|
handler | | Function | The actual function that will handle a JSON-RPC request to this method |
useContext | false | Boolean | When true, the handler will receive a context object as the second argument |
params | null | Array|Object|null | Force JSON-RPC parameters to be of a certain type |
Server example showcasing most features and options in examples/method_definitions/server.js:
const jayson = require('jayson');
const methods = {
sum: function(args, done) {
done(null, sum(args));
},
context: jayson.Method(function(args, context, done) {
done(null, context);
}, {useContext: true}),
sumDefault: jayson.Method(function(args, done) {
const total = sum(args);
done(null, total);
}, {
params: {a: 2, b: 5}
}),
isArray: new jayson.Method({
handler: function(args, done) {
const result = Array.isArray(args);
done(null, result);
},
params: Array
})
};
const server = new jayson.Server(methods, {
useContext: false,
params: Array
});
server.http().listen(3000);
function sum(list) {
return Object.keys(list).reduce(function(sum, key) { return sum + list[key]; }, 0);
}
Client example in examples/method_definitions/client.js:
const jayson = require('jayson');
const client = jayson.Client.http({
port: 3000
});
client.request('sum', [3, 5, 9, 11], function(err, response) {
if(err) throw err;
console.log(response.result);
});
client.request('sum', {a: 2, b: 3, c: 4}, function(err, response) {
if(err) throw err;
console.log(response.result);
});
client.request('sumDefault', {b: 10}, function(err, response) {
if(err) throw err;
console.log(response.result);
});
client.request('isArray', {a: 5, b: 2, c: 9}, function(err, response) {
if(err) throw err;
console.log(response.result);
});
client.request('context', {hello: 'world'}, function(err, response) {
if(err) throw err;
console.log(response.result);
});
Server events
In addition to events that are specific to certain interfaces, all servers will emit the following events:
Event | When | Arguments | Notes |
---|
request | Interpretable non-batch request received | 1: Request object | |
response | Returning a response | 1: Request object 2: Response object | |
batch | Interpretable batch request received | 1. Array of requests | Emits request for every part |
Server Errors
If you should like to return an error from an method request to indicate a failure, remember that the JSON-RPC 2.0 specification requires the error to be an Object
with a code (Integer/Number)
to be regarded as valid. You can also provide a message (String)
and a data (Object)
with additional information. Example:
const jayson = require('jayson');
const server = new jayson.Server({
i_cant_find_anything: function(args, callback) {
const error = {code: 404, message: 'Cannot find ' + args.id};
callback(error);
},
i_cant_return_a_valid_error: function(callback) {
callback({message: 'I forgot to enter a code'});
}
});
Predefined Errors
It is also possible to cause a method to return one of the predefined JSON-RPC 2.0 error codes using the server helper function Server.prototype.error
inside of a server method. Example:
const jayson = require('jayson');
const server = new jayson.Server({
invalid_params: function(args, callback) {
const error = this.error(-32602);
callback(error);
}
});
You can even override the default messages:
const jayson = require('jayson');
const server = new jayson.Server({
error_giver_of_doom: function(callback) {
callback(true)
}
});
server.errorMessages[Server.errors.INTERNAL_ERROR] = 'I has a sad. I cant do anything right';
Server CORS
Jayson does not include functionality for supporting CORS requests natively but it is easy to use a CORS-enabling middleware
like cors. An example of this can be found in examples/cors/server.js:
const jayson = require('jayson');
const cors = require('cors');
const connect = require('connect');
const jsonParser = require('body-parser').json;
const app = connect();
const server = new jayson.Server({
myNameIs: function(args, callback) {
callback(null, 'Your name is: ' + args.name);
}
});
app.use(cors({methods: ['POST']}));
app.use(jsonParser());
app.use(server.middleware());
app.listen(3000);
Server Context
Since version 3.0.0
You can provide an optional context object to JSON-RPC method handlers. This can be used to give extra data to a handler such as request headers, authentication tokens, and so on.
This feature is unlocked by having jayson.Method
accepts a boolean option called useContext
. It always defaults to false
for backwards compatibility. When it is set to true
the method handler that jayson.Method
wraps will always receive a context object as the second argument. The object can be given as the third argument to jayson.Server.prototype.call
.
Server example in examples/context/server.js:
const jayson = require('jayson');
const jsonParser = require('body-parser').json;
const express = require('express');
const app = express();
const server = new jayson.Server({
getHeaders: function(args, context, callback) {
callback(null, context.headers);
},
oldMethod: new jayson.Method(function(args, callback) {
callback(null, {});
}, {
useContext: false
})
}, {
useContext: true
});
app.use(jsonParser());
app.use(function(req, res, next) {
const context = {headers: req.headers};
server.call(req.body, context, function(err, result) {
if(err) return next(err);
res.send(result || {});
});
});
app.listen(3001);
Client example in examples/context/client.js:
const jayson = require('jayson');
const client = jayson.Client.http({
port: 3001
});
client.request('getHeaders', {}, function(err, response) {
if(err) throw err;
console.log(response.result);
});
Notes
jayson.Server
also accepts useContext
as an option, and passes the value on to the jayson.Method
constructor. This option can be overriden on a per-method basis as shown above.- Individual requests in a JSON-RPC batch will all receive the exact same context object in their handler - take care not to mutate it
- If a falsy context value is given to
jayson.Server.prototype.call
, an empty object will be created - None of the current jayson server transports (http, https, tls, tcp, middleware) can make use of the context object. You will need to rig your own transport implementation, like the one above based on an
express
http server. See the FAQ for more info about this.
Revivers and Replacers
JSON lacks support for representing types other than the simple ones defined in the JSON specification. Fortunately the JSON methods in JavaScript (JSON.parse
and JSON.stringify
) provide options for custom serialization/deserialization routines. Jayson allows you to pass your own routines as options to both clients and servers.
Simple example transferring the state of an object between a client and a server:
Shared code between the server and the client in examples/reviving_and_replacing/shared.js:
'use strict';
const Counter = exports.Counter = function(value) {
this.count = value || 0;
};
Counter.prototype.increment = function() {
this.count += 1;
};
exports.replacer = function(key, value) {
if(value instanceof Counter) {
return {$class: 'counter', $props: {count: value.count}};
}
return value;
};
exports.reviver = function(key, value) {
if(value && value.$class === 'counter') {
const obj = new Counter();
for(const prop in value.$props) obj[prop] = value.$props[prop];
return obj;
}
return value;
};
Server example in examples/reviving_and_replacing/server.js:
const jayson = require('jayson');
const shared = require('./shared');
const options = {
reviver: shared.reviver,
replacer: shared.replacer
};
const server = new jayson.Server({
increment: function(args, callback) {
args.counter.increment();
callback(null, args.counter);
}
}, options);
server.http().listen(3000);
A client example in examples/reviving_and_replacing/client.js invoking "increment" on the server:
const jayson = require('jayson');
const shared = require('./shared');
const client = jayson.Client.http({
port: 3000,
reviver: shared.reviver,
replacer: shared.replacer
});
const params = {
counter: new shared.Counter(2)
};
client.request('increment', params, function(err, response) {
if(err) throw err;
const result = response.result;
console.log(
result instanceof shared.Counter,
result.count,
params.counter === result
);
});
Notes
- Instead of using a replacer, it is possible to define a
toJSON
method for any JavaScript object. Unfortunately there is no corresponding method for reviving objects (that would not work, obviously), so the reviver always has to be set up manually.
Named parameters
It is possible to specify named parameters when doing a client request by passing an Object instead of an Array.
Client example in examples/named_parameters/client.js:
const jayson = require('jayson');
const client = jayson.Client.http({
port: 3000
});
client.request('add', {b: 1, a: 2}, function(err, response) {
if(err) throw err;
console.log(response.result);
});
Server example in examples/named_parameters/server.js:
const jayson = require('jayson');
const server = new jayson.Server({
add: function(params, callback) {
callback(null, params.a + params.b);
}
});
server.http().listen(3000);
Notes
- If requesting methods on a Jayson server, arguments left out will be
undefined
- Too many arguments or arguments with invalid names will be ignored
- It is assumed that the last argument to a server method is the callback and it will not be filled with something else
- Parsing a function signature and filling in arguments is generally not recommended and should be avoided
Promises
Since version 2.0.0
A separate tree that does limited usage of the ES6 Promise object is available. The internal API remains callback based, with the addition that promises may be used for two things:
- Returning a Promise when requesting a JSON-RPC method using a Client
- Returning a Promise inside of a Server method
To use the separate tree, do a require('jayson/promise')
instead of require('jayson')
.
Server example in examples/promise/server.js showing how to return a Promise
in a server method:
const jayson = require('jayson/promise');
const server = new jayson.Server({
add: async function(args) {
const sum = Object.keys(args).reduce(function(sum, key) { return sum + args[key]; }, 0);
return sum;
},
rejection: async function(args) {
throw server.error(501, 'not implemented');
}
});
server.http().listen(3000);
Client example in examples/promise/client.js showing how to do a request:
const jayson = require('jayson/promise');
const client = jayson.Client.http({
port: 3000
});
const reqs = [
client.request('add', [1, 2, 3, 4, 5]),
client.request('rejection', [])
];
Promise.all(reqs).then(function(responses) {
console.log(responses[0].result);
console.log(responses[1].error);
});
Notes
- JSON-RPC errors will not result in rejection of the Promise. It is however possible that a future version will include a client setting to have JSON-RPC errors result in rejection. Please note that network errors and the like will result in rejection.
- A
Promise
is considered to have been returned from a server method if the returned object has a property then
that is a function.
Promise Batches
Since version 2.0.5
Sometimes you may want to return raw requests from a promise client. This needs to be handled differently, because PromiseClient.prototype.request
would normally always be expected to return a Promise which we in this case don't want.
To solve this, we need to set the fourth parameter to PromiseClient.prototype.request
explicitly to false
in order to not return a Promise.
Client example in examples/promise_batches/client.js showing how to properly execute a batch request:
const jayson = require('jayson/promise');
const client = jayson.Client.http({
port: 3000
});
const batch = [
client.request('add', [1, 2, 3, 4, 5], undefined, false),
client.request('add', [5, 6, 7, 8, 9], undefined, false),
];
client.request(batch).then(function(responses) {
console.log(responses[0].result);
console.log(responses[1].result);
});
Notes
- The third parameter to
PromiseClient.prototype.request
above is explicitly set to undefined
- this parameter would normally represent the desired ID of the call. Remember that null
would mean a notification (which does not return a response) and other falsy values may actually be used as ids. Setting undefined
ensures that the id is generated automatically.
Promise Browser Client
A browser client that has no dependencies on node.js core libraries is available too. It works similar to how the regular callback-style Browser Client works. Here is an example:
'use strict';
const jaysonPromiseBrowserClient = require('jayson/promise/lib/client/browser');
const fetch = require('node-fetch');
const callServer = function(request) {
const options = {
method: 'POST',
body: request,
headers: {
'Content-Type': 'application/json',
}
};
return fetch('http://localhost:3000', options).then(res => res.text());
};
const client = new jaysonPromiseBrowserClient(callServer, {
});
client.request('multiply', [5, 5]).then(function(result) {
console.log(result);
}, function(err) {
console.error(err);
});
Please refer to the regular browser client section of the README for more information.
FAQ
How can I pass HTTP headers/session/etc into my JSON-RPC request handler?
Support for method context added in version 3.0.0
See Server context section.
What is the recommended way to use jayson?
Using the provided clients and servers for http, https, tls, tcp and the express middleware is fine and works well for most use cases. However, sometimes issues like these crop up (quotes below are not directly from issue posters):
These are not issues with jayson, but stem from the fact that JSON-RPC 2.0 specification is transport agnostic and these kind of behaviours are not defined by that specification. The clients provided by jayson for http, https, tls, tcp are made to work and tested with their corresponding jayson server implementation. Any other compatibility with any other server or client is accidental when it comes to details of the transport layer. With that said, jayson is made to be 100 % compatible with the JSON-RPC 2.0 specification and compatibility with other non-jayson servers or clients when it comes to the application layer is pretty much guaranteed.
The library author tedeh therefore recommends that if you have particular needs when it comes to the transport layer you create an implementation satisfying these details yourself. Doing this is actually quite simple.
Example of a http server built with express in examples/faq_recommended_http_server/server.js:
const jayson = require('jayson');
const jsonParser = require('body-parser').json;
const express = require('express');
const app = express();
const server = new jayson.Server({
add: function(numbers, callback) {
const sum = Object.keys(numbers).reduce(function(sum, key) { return sum + numbers[key]; }, 0);
callback(null, sum);
}
});
app.use(jsonParser());
app.use(function(req, res, next) {
const request = req.body;
server.call(request, function(err, response) {
if(err) {
if(err instanceof Error) return next(err);
res.status(400);
res.send(err);
return;
}
if(response) {
res.send(response);
} else {
res.status(204);
res.send('');
}
});
});
app.listen(3001);
Using some of the utilities provided and exported by jayson, creating a client offering the same kind of flexibility is also simple. Example of a compatible http client built with superagent in examples/faq_recommended_http_server/client.js:
const jayson = require('jayson');
const request = require('superagent');
const requestBody = jayson.Utils.request('add', [1,2,3,4], undefined, {
version: 2,
});
request.post('http://localhost:3001')
.timeout({response: 5000, deadline: 60000})
.send(requestBody)
.end(function(err, response) {
if(err) {
if(!err.status) throw err;
const body = err.response.body;
if(body && body.error && jayson.Utils.Response.isValidError(body.error, 2)) {
console.err(body.error);
return;
}
throw err;
}
const body = response.body;
if(!jayson.Utils.Response.isValidResponse(body, 2)) {
console.err(body);
}
if(body.error) {
console.err(body.error);
} else {
console.log(body.result);
}
});
Contributing
Highlighting issues or submitting pull
requests on Github is most welcome.
Please make sure to follow the style of the project, and lint your code with npm run lint
before submitting a patch.
Submitting issues or pull requests with the Typescript type definitions
You are required to provide an easily reproducible code sample of any errors with the Typescript type definitions so that they can be added to the typescript test file in typescript/test.ts. Better yet, issue a pull request adding a test there yourself that shows up when running the package.json
script test-tsc
.