The lightest JavaScript plugin system / plugin manager / messagebus for the map/reduce world
Using npm:
npm install --save iplugYou want to use plugins to dynamically extend the functionality of your code.
The core part of your application loads your plugins and creates one or more message buses to communicate with them, through a number of topics. Plugins can register their interest to any number of topics. When you want to defer control to plugins you emit a message on a topic and all registered plugins will be invoked, sequentially or in parallel.
main.js
import iplug from 'iplug'
import module1 from './plugins/module1.js'
import module2 from './plugins/module2.js'
const modules = {
module1,
module2,
}
const plugins = iplug(modules).init()
function main() {
const input = 'test payload'
const output = plugins.chain('test:message', input)
console.log(output)
}
main()Each plugin module can export:
- A manifest object:
export default { // ... 'message': config => data => ... }
- a function that can take an optional argument and returns a manifest object
export default (moduleConfig) => { // keep state here for better testability return { 'message': config => data => ... // ... } }
A manifest object is one whose keys are messagebus messages we want to register to, and handlers to be called when those are emitted
type Handler<T> = (config: T) => (data: any) => any;
type Manifest<T> = {
[message: string]: Handler<T>;
};example-plugin.js
export default {
'category1:message1': config => data => `result for a call to category1:message1 - payload was ${data}`,
'category2:message2': config => data => `result for a call to category1:message1 - payload was ${data}`,
}There are two ways you can call plugins: in sequence, or in parallel.
Calling plugins in sequence means starting with an optional initial value, and chaining it through each, so the output of one becomes the input of the next, and the final result is the output of the last. This is useful when your plugins can be somewhat aware of each-other and the output of any may be the final one.
main.js
// this returns the output of the last plugin in the chain
plugins.chain(<message> [, initial data])
plugins.reduce(<message> [, initial data])
plugins(<message>, [initial data])reduce is an alias for chain.
For chained calls you can also omit both the chain or reduce keywords and just call:
Calling plugins in parallel means passing the same optional initial value, then collecting the results of each together, which come out as an array, perhaps for further processing. This is useful if you want to run many plugins in parallel, especially async ones, or that need to run in isolation from the input or the output of the other plugins.
main.js
// this returns an array of your plugins' output
plugins.map(<message> [, initial data])
plugins.all(<message> [, initial data])
plugins.parallel(<message> [, initial data])all is an alias for map
You want to use plugins to moderate content before rendering it, by passing it through a number of plugins, each of which has to approve the content.
module.js
import {moderation} from './moderation.js
const pluginsList = [moderation]
const config = { }
const plugins = iplug(pluginsList, config)
const initialData[] = await fetch('/api/getMessages').then(x=>x.json())
const result = plugins('moderate', data)moderation.js
export default {
'moderate': config => {
const blackList = await fetch('/word-blacklist').then(x=>x.json())
return data[] => data.map(str => blackList.forEach(word => str.replace(word, '###redacted###')))
}
}You may want each plugin to emit more data over time, effectively exposing an Observable interface back to the main application
plugin.js
import { Observable } from 'rxjs'
const sourceStream = Observable(...)
export default {
'getdata': config => data => sourceStream,
}app.js
import { merge } from 'rxjs'
// Get an Observable from each plugin.
const streams = streamingPlugins.map('getdata')
merge(streams)
.subscribe(doSomething)You can pass an observable to each of your plugins and get one back to enable two-way communication over time
echo.js
import { Observable } from 'rxjs'
const sourceStream = Observable(...)
export default {
'duplex': config => { inputStream } => inputStream.map(inputMessage=>`This is a reply to ${inputMessage}.`,
}app.js
import { Subject } from 'rxjs'
import { merge } from 'rxjs'
const outputStream = new Subject()
const streams = streamingPlugins.map('duplex', outputStream)
merge(streams)
.subscribe(doSomething)This extra step allows some plugins to perform some one-time (perhaps slow) initialisation and return a "production" function that's geared up for faster, repeated executions. It's often best to perform initialisation in the main handler function to enable multiple instances of the same plugin to be used in different isolated contexts (multiple message buses in the same application).
plugin.js
// some global initialisation can go here, but the state will be shared!
// ...
export default {
'message:handler': config => {
// perform some more initialisation here when you want more isolation.
// const slow = slowOperation
return data => {
// put your performance-optimised code here, to run multiple times
}
}
}main.js
// The following two message buses are meant to run independently
const messageBus1 = plugins()
const messageBus2 = plugins()
// The following two calls will be run in isolation from each-other
messageBus1.chain('message:handler')
messageBus2.chain('message:handler')Writing unit tests for your plugins should be just as simple as calling the function they export for a particular event/topic.
import plugin from '/plugins/double-it.js'
describe('plugin1', () => {
describe('when handling a "test:topic"', () => {
it('doubles its input', () => {
const fn = plugin['test:topic']()
expect(fn(2)).toEqual(4)
});
});
});Following is an example unit test for a hypothetical plugin that returns 0, written for Jest, but easily adaptable to other test frameworks.
import plugin from '/plugins/plugin2.js'
import fixture from '/plugins/plugin2.fixture.js'
describe('plugin2', () => {
describe('when handling a "test:event"', () => {
it('returns 0', () => {
const fn = plugin['test:event']()
const result = fn(fixture)
expect(result).toEqual(0)
});
});
});Global variables, even inside an ES6 module, can pose various challenges to unit testing. Consider the follwing plugin:
const globalState = get_some_state()
export default {
'message:handler': config => data => globalState(data),
}The problem here is sometimes it can be hard for test frameworks to mock or stub globalState in order to force a certain behaviour.
What you may experience is the first time a unit test runs, the globalState may be mocked as expected, but at subsequent runs, re-mocking or re-stubbing may just not work, failing the tests.
A solution to this problem is creating a plugin that exports a function, which in turn will return everything else.
export default function() {
const globalState = get_some_state()
return {
'message:handler': config => data => globalState(data),
}
}This way, no global state will remain between test runs.
import initModule from '/plugins/plugin.js'
describe('plugin', () => {
describe('when handling a "test:event"', () => {
it('returns 0', () => {
// Loading the plugin from a test will need this one extra line
const plugin = initModule()
const fn = plugin['test:event']()
const result = fn(fixture)
expect(result).toEqual(0)
});
});
});You can find more examples in the respective folder