Skip to content

Latest commit

 

History

History
226 lines (176 loc) · 7.79 KB

custom-http-client.md

File metadata and controls

226 lines (176 loc) · 7.79 KB

Custom HTTP Clients for the Twilio Node Helper Library

If you are working with the Twilio Node.js Helper Library, and you need to modify the HTTP requests that the library makes to the Twilio servers, you’re in the right place.

The helper library uses axios, a promise-based HTTP client, to make requests. You can also provide your own httpClient to customize requests as needed.

The following example shows a typical request without a custom httpClient.

const client = require('twilio')(accountSid, authToken);

client.messages
  .create({
    to: '+15555555555',
    from: '+15555555551',
    body: 'Ahoy default requestClient!',
  })
  .then((message) => console.log(`Message SID ${message.sid}`))
  .catch((error) => console.error(error));

Out of the box, the helper library creates a default RequestClient for you, using the Twilio credentials you pass to the init method. If you have your own RequestClient, you can pass it to any Twilio REST API resource action you want. Here’s an example of sending an SMS message with a custom client called MyRequestClient.

// require the Twilio module and MyRequestClient
const twilio = require('twilio');
const MyRequestClient = require('./MyRequestClient');

// Load environment variables
require('dotenv').config();

// Twilio Credentials
const accountSid = process.env.ACCOUNT_SID;
const authToken = process.env.AUTH_TOKEN;

const client = twilio(accountSid, authToken, {
  // Custom HTTP Client with a one minute timeout
  httpClient: new MyRequestClient(60000),
});

client.messages
  .create({
    to: '+15555555555',
    from: '+15555555551',
    body: 'Ahoy, custom requestClient!',
  })
  .then((message) => console.log(`Message SID ${message.sid}`))
  .catch((error) => console.error(error));

Create your custom Twilio RestClient

When you take a closer look at the constructor for twilio.restClient, you see that the httpClient parameter is a RequestClient. This class provides the client to the Twilio helper library to make the necessary HTTP requests.

Now that you can see how all the components fit together, you can create our own RequestClient:

'use strict';

const _ = require('lodash');
const qs = require('qs');
const axios = require('axios');

/**
 * Custom HTTP Client
 * Based on: /twilio/lib/base/RequestClient.js
 */
class MyRequestClient {
  constructor(timeout) {
    this.timeout = timeout;
  }

  request(opts) {
    opts = opts || {};

    if (!opts.method) {
      throw new Error('http method is required');
    }

    if (!opts.uri) {
      throw new Error('uri is required');
    }

    // Axios auth option will use HTTP Basic auth by default
    if (opts.username && opts.password) {
      this.auth = {
        username: opts.username,
        password: opts.password,
      };
    }

    // Options for axios config
    const options = {
      url: opts.uri,
      method: opts.method,
      headers: opts.headers,
      auth: this.auth,
      timeout: this.timeout,
    };

    // Use 'qs' to support x-www-form-urlencoded with axios
    // Construct data request body option for axios config
    if (!_.isNull(opts.data)) {
      options.headers = { 'content-type': 'application/x-www-form-urlencoded' };
      options.data = qs.stringify(opts.data, { arrayFormat: 'repeat' });
    }

    // Use 'qs' to support x-www-form-urlencoded with axios
    // Construct URL params option for axios config
    if (!_.isNull(opts.params)) {
      options.params = opts.params;
      options.paramsSerializer = (params) => {
        return qs.stringify(params, { arrayFormat: 'repeat' });
      };
    }

    return axios(options)
      .then((response) => {
        if (opts.logLevel === 'debug') {
          console.log(`response.statusCode: ${response.status}`);
          console.log(`response.headers: ${JSON.stringify(response.headers)}`);
        }
        return {
          statusCode: response.status,
          body: response.data,
        };
      })
      .catch((error) => {
        console.error(error);
        throw error;
      });
  }
}

module.exports = MyRequestClient;

Add a custom timeout

One common need to alter the HTTP request is to set a custom timeout. In the code sample using a custom client, you will see httpClient: new MyRequestClient(60000) where 60000 is the custom timeout value — one minute in milliseconds.

To make this reusable, here’s a class that you can use to create this MyRequestClient whenever you need one. This class is based on the default RequestClient provided by the helper library, and uses axios to make requests.

In this example, we are using some environmental variables loaded at the program's startup to retrieve our credentials:

These settings are located in a .env file like so:

ACCOUNT_SID=ACxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
AUTH_TOKEN= your_auth_token

Here’s the full console program that sends a text message and shows how it all can work together. It loads the .env file for us. The timeout value, 60,000 milliseconds, will be used by axios to set the custom timeout.

// require the Twilio module and MyRequestClient
const twilio = require('twilio');
const MyRequestClient = require('./MyRequestClient');

// Load environment variables
require('dotenv').config();

// Twilio Credentials
const accountSid = process.env.ACCOUNT_SID;
const authToken = process.env.AUTH_TOKEN;

const client = twilio(accountSid, authToken, {
  // Custom HTTP Client with a one minute timeout
  httpClient: new MyRequestClient(60000),
});

client.messages
  .create({
    to: '+15555555555',
    from: '+15555555551',
    body: 'Ahoy, custom requestClient!',
  })
  .then((message) => console.log(`Message SID ${message.sid}`))
  .catch((error) => console.error(error));

Call Twilio through a proxy server

The most common need to alter the HTTP request is to connect and authenticate with an enterprise’s proxy server. The Node.JS Helper library now supports this using the HTTP_PROXY environment variable. The Twilio Node.js Helper library uses the https-proxy-agent package to connect with the proxy you assign to the environment variable.

HTTP_PROXY=http://127.0.0.1:8888

If you prefer to use your custom RequestClient to connect with a proxy, you could follow the pattern outlined in the code samples on this page. For example, axios supports the use of a proxy with its proxy option. The axios proxy takes an object with protocol, host, and method options. These can be passed to your MyRequestClient to be used by axios.

// Pass proxy settings to client constructor
const client = twilio(accountSid, authToken, {
  // Custom HTTP Client
  httpClient: new MyRequestClient(60000, {
      protocol: 'https',
      host: '127.0.0.1',
      port: 9000,
    }
  ),
});

// Update class to accept a proxy
class MyRequestClient {
  constructor(timeout, proxy){
    this.timeout = timeout,
    this.proxy = proxy
  }

  const options = {
    proxy: this.proxy,
    // other axios options...
  }
}

What else can this technique be used for?

Now that you know how to inject your own httpClient into the Twilio API request pipeline, you can use this technique to add custom HTTP headers and authorization to the requests (perhaps as required by an upstream proxy server).

You could also implement your own httpClient to mock the Twilio API responses. With a custom httpClient, you can run your unit and integration tests quickly without the need to make a connection to Twilio. In fact, there’s already an example online showing how to do exactly that with Node.js and Prism.

We can't wait to see what you build!