Twine is an IPC request library built around the concept of explicitness. Twine is heavily inspired by Netflix's Ribbon.

Installation

Browsers (npm)

npm install --save @inmar/twine-browser

Note: The browser version depends on fetch. If using in a browser that doesn't support fetch consider using a polyfill like whatwg-fetch.

Node (npm)

npm install --save @inmar/twine-node

Usage

Commonly Used Methods

There are many components and methods provided by Twine out-of-the-box. Below are the commonly used methods to construct HTTP(S) calls.

More methods can be found in the source. There is inline documentation for all methods.

Resource Service

• usingHTTP
• usingHTTPS
• createRequestTemplate

Request Template

• withURITemplate
• withMethod
• withHeader
• withBearerToken
• withEndpointTimeout
• handleWhen
• withRetryStrategy
• createRequest

Request

• withHeader
• withParameters
• withBody
• execute

Example

const twine = require('@inmar/twine-node')
const {
  Receives,
  RetryStrategy
} = twine

//Note: You only need to construct the service once.
const jsonPlaceholderService = twine.createResourceService('jsonplaceholder.typicode.com')
  .usingHTTPS()

//Note: You only need to construct the template once.
const getTodoTemplate = jsonPlaceholderService.createRequestTemplate('get-todo')
  .withoutInstrumentation()
  .withURITemplate('/todos/{id}')
  .withMethod('GET')
  .withTimeout(100)
  .receivesJSON()
  .handleWhen(200, [
    Receives.json,
    resp => resp.title
  ])
  .handleWhen(404, () => ({
    error: 'not-found'
  }))
  .withRetryStrategy(new RetryStrategy()
    .maxAutoRetries(2)
    .delayRetryForMilliseconds(250)
    .escalateRetryDelayWith(previousDelay => previousDelay + 100)
    .fallbackTo(() => ({ 
      error: 'retries-exhausted'
    }))
  )

export default async function retrieveTodo(todoId) {
  //Each time we want to make a request, we simply create one from the template and execute it.
  const response = await getTodoTemplate.createRequest()
     .withParameters({id: todoId})
     .execute()
 
  if (!response.error) {
    return response
  }
  
  if (response.error === 'not-found') {
    return null
  }
  
  throw new Error(`Failed to retrieve todo due to error: ${response.error}` )
}

Concepts

There are a number of important concept to keep in mind when integrating Twine into your application.

Pipelines

A pipeline is a collection of components that execute sequentially in the order they are added to the pipeline, and are split into 3 key sections: Resource Services, Request Templates, Requests. The combination of the components from these 3 sections make up a complete pipeline that is executed when Twine runs.

Components

Components are modules or plugins that modify or add functionality to the pipeline when they are added to it. Components can be written by consumers of Twine to augment pipelines in ways not already included in the library.

A more extensive explanation with examples is available on the wiki.

Resource Service

A Resource Service represents a network resource to which requests can be made. The most common use-case is an HTTP(S) service, however any type of resource can function as a Resource Service as long as Twine has been provided a usingXXX component to understand it. (Example: PostgreSQL as a Resource Service)

These are the entry points for creating a Twine Request Pipeline.

Request Template

Request Templates are children to a Resource Service and represent the instructions telling Twine how to repeatably make a specific request to that service.

Request

A Request is the Twine representation of a built Request Template and is the terminal at which the request can have final options and modifiers attached before twine executes the entire pipeline to contact the service defined by the Resource Service

Contribute

This repository is a mono-repo containing the user-facing libraries @inmar/twine-browser and @inmar/twine-node. It also contains the code shared between the two platform implementations as @inmar/twine-core.

All npm related actions like npm install, npm publish, etc are handled via a mono-repo supporting library called lerna. Lerna is used to manage synchronized version bumps and package interdependence via symlinks for local development.

As such, to develop this repository, lerna becomes an integral part for dependency management and publishing.

Important Commands

• lerna bootstrap
• lerna add
• lerna version
• lerna publish
  • Note, publishing should not be done manually. It will be handled by the CI pipeline given that the package versions are bumped.

Initial Setup

1. Pull the repository

git clone https://github.com/inmar/twine.js

2. Run npm install in the root of the repository to install lerna. As part of the root npm install, a post-install hook will instruct lerna to bootstrap the repository via lerna link and lerna bootstrap. This will install and setup all packages and their dependencies.