Web Scraping Scotch: The Node Way

Publikováno: 14.5.2018

A lot of new web technologies and design paradigms have emerged in the last couple of years. Some programming languages are beginning to gain increasing popularity. It's very likely to have heard a...

Celý článek

A lot of new web technologies and design paradigms have emerged in the last couple of years. Some programming languages are beginning to gain increasing popularity. It's very likely to have heard about concepts like responsive design, hybrid mobile/desktop apps, progressive web apps(PWAs), single page applications(SPAs), server-side rendered(SSR) apps, serverless architechture, the list goes on.

While every modern web developer aims at getting up to speed with these technologies, there are a few less popular web concepts and techniques that are quite useful - one of which is web scraping. In this tutorial, we will take a look at web scraping and practical ways we can harness the technique.

What is Web Scraping?

In very simple terms, web scraping is the technique of extracting data from websites. This data can further be stored in a database or any other storage system for analysis or other uses. While extracting data from websites can be done manually, web scraping usually refers to an automated and less tedious process.

Web scraping may seem very trivial, but it is the technique used by most bots and web crawlers for data extraction. There are different techniques that can be employed for web scraping. However, in this tutorial, we will use a technique that involves DOM parsing a webpage.

Scotch Scraping

In this tutorial, we will use web scraping to extract some data from the Scotch website. Scotch does not provide an API for fetching the profiles and tutorials/posts of authors. However, we will be building a very simple API for fetching the profiles and tutorials/posts of Scotch authors.

Here is a screenshot of a simple demo app created based on the API we will be build in this tutorial. You can see the app on Heroku and the source code on GitHub.

Demo App Screenshot

Prerequisites

Advanced JavaScript and ES6 Syntax

Web scraping can be done in virtually any programming language that has support for HTTP and XML or DOM parsing. In this tutorial, we will focus on web scraping using JavaScript in a Node.js server environment. Hence, an advanced knowledge of JavaScript is required to fully understand the code snippets.

Also in this tutorial, there is heavy usage of ES6 syntax as shown through the repeated use of arrow functions, destructuring, block-scope variables, template literals, rest parameters, spread operator, default parameters, object enhancements and promises. Hence, adequate familiarity with ES6 syntax is required to fully understand the code snippets.

We will also use a couple of ES7 features such as async functions and the await operator. Hence, knowledge of asynchronous functions and working with Promises is required.

jQuery Familiarity

Some familiarity with the jQuery DOM library is required to completely understand some of the code snippets in this tutorial. This is because the Cheerio package we will be using in this tutorial is based on the jQuery DOM manipulation and traversing style. Check the jQuery API Documentation to learn more about jQuery.

Functional Programming

In this tutorial, we will employ the functional programming paradigm in building our desired API. As such, we will create so many specialized functions, and also apply a couple of functional programming concepts such as pure functions(immutability concept), higher-order functions and composition. Hence, you will be better-off if you already have prior understanding of functional programming concepts. You can learn more about functional programming concepts here.

Core Dependencies

Before you begin, ensure that you have Node and npm or yarn installed on your machine. Since we will use a lot of ES6/7 syntax in this tutorial, I recommend you use the following versions of Node and npm for complete ES6/7 support: Node(8.9.0 or higher) and npm(5.2.0 or higher).

Here are the core packages we will be using:

  1. Cheerio - Cheerio is a fast, flexible, and lean implementation of core jQuery designed specifically for the server. It makes DOM parsing very easy.

  2. Axios - Axios is a promise based HTTP client for the browser and Node.js. It will enable us fetch page contents through HTTP requests.

  3. Express - Express is a minimal and flexible Node.js web application framework that provides a robust set of features for web and mobile applications.

  4. Lodash - Lodash is a modern JavaScript utility library delivering modularity, performance & extras. It makes JavaScript easier by taking the hassle out of working with arrays, numbers, objects, strings, etc.

Getting Started

Installing Dependencies

Create a new directory for the application and run the following command to install the required dependencies for the app.

# Create a new directory
mkdir scotch-scraping

# cd into the new directory
cd scotch-scraping

# Initiate a new package and install app dependencies
npm init -y
npm install express morgan axios cheerio lodash

Setting up the Express server application

We will go ahead to setup a simple HTTP server application using Express. Create a server.js file in the root directory of your application and add the following code snippet to setup the server:

/* server.js */

// Require dependencies
const logger = require('morgan');
const express = require('express');

// Create an Express application
const app = express();

// Configure the app port
const port = process.env.PORT || 3000;
app.set('port', port);

// Load middlewares
app.use(logger('dev'));

// Start the server and listen on the preconfigured port
app.listen(port, () => console.log(`App started on port ${port}.`));

Modify npm scripts

Finally, we will modify the "scripts" section of the package.json file to look like the following snippet:

"scripts": {
  "start": "node server.js"
}

We have gotten all we need to start building our application. If you run the command npm start in your terminal now, it will start up the application server on port 3000 if it is available. However, we cannot access any route yet since we are yet to add routes to our application. Let's start building some helper functions we will need for web scraping.

Helper Functions

As stated earlier, we will create a couple of helper functions that will used in several parts of our application. Create a new app directory in your project root. Create a new file named helpers.js in the just created directory and add the following content to it:

/* app/helpers.js */

const _ = require('lodash');
const axios = require("axios");
const cheerio = require("cheerio");

As you can see we are simply requiring the dependencies we will need for our helper functions. Let's go ahead and add the helper functions.

Utility Helper Functions

We will start by creating some utility helper functions. Add the following snippet to the app/helpers.js file.

/* app/helpers.js */

///////////////////////////////////////////////////////////////////////////////
// UTILITY FUNCTIONS
///////////////////////////////////////////////////////////////////////////////

/**
 * Compose function arguments starting from right to left
 * to an overall function and returns the overall function
 */
const compose = (...fns) => arg => {
  return _.flattenDeep(fns).reduceRight((current, fn) => {
    if (_.isFunction(fn)) return fn(current);
    throw new TypeError("compose() expects only functions as parameters.");
  }, arg);
};

/**
 * Compose async function arguments starting from right to left
 * to an overall async function and returns the overall async function
 */
const composeAsync = (...fns) => arg => {
  return _.flattenDeep(fns).reduceRight(async (current, fn) => {
    if (_.isFunction(fn)) return fn(await current);
    throw new TypeError("compose() expects only functions as parameters.");
  }, arg);
};

/**
 * Enforces the scheme of the URL is https
 * and returns the new URL
 */
const enforceHttpsUrl = url =>
  _.isString(url) ? url.replace(/^(https?:)?\/\//, "https://") : null;

/**
 * Strips number of all non-numeric characters
 * and returns the sanitized number
 */
const sanitizeNumber = number =>
  _.isString(number)
    ? number.replace(/[^0-9-.]/g, "")
    : _.isNumber(number) ? number : null;

/**
 * Filters null values from array
 * and returns an array without nulls
 */
const withoutNulls = arr =>
  _.isArray(arr) ? arr.filter(val => !_.isNull(val)) : [];

/**
 * Transforms an array of ({ key: value }) pairs to an object
 * and returns the transformed object
 */
const arrayPairsToObject = arr =>
  arr.reduce((obj, pair) => ({ ...obj, ...pair }), {});

/**
 * A composed function that removes null values from array of ({ key: value }) pairs
 * and returns the transformed object of the array
 */
const fromPairsToObject = compose(arrayPairsToObject, withoutNulls);

Let's go through the functions one at a time to understand what they do.

  • compose() - This is a higher-order function that takes one or more functions as its arguments and returns a composed function. The composed function has the same effect as invoking the functions passed in as arguments from right to left, passing the result of a function invocation as argument to the next function each time.

    If any of the arguments passed to compose() is not a function, the composed function will throw an error whenever it is invoked. Here is a code snippet that describes how compose() works.
/**
* -------------------------------------------------
* Method 1: Functions in sequence
* -------------------------------------------------
*/
function1( function2( function3(arg) ) );

/**
* -------------------------------------------------
* Method 2: Using compose()
* -------------------------------------------------
* Invoking the composed function has the same effect as (Method 1)
*/
const composedFunction = compose(function1, function2, function3);

composedFunction(arg);
  • composeAsync() - This function works in the same way as the compose() function. The only difference being that it is asynchronous. Hence, it is ideal for composing functions that have asynchronous behaviour - for example, functions that return promises.

  • enforceHttpsUrl() - This function takes a url string as argument and returns the url with https scheme provided the url begins with either https://, http:// or //. If the url is not a string then null is returned. Here is an example.
enforceHttpsUrl('scotch.io'); // returns => 'scotch.io'
enforceHttpsUrl('//scotch.io'); // returns => 'https://scotch.io'
enforceHttpsUrl('http://scotch.io'); // returns => 'https://scotch.io'
  • sanitizeNumber() - This function expects a number or string as argument. If a number is passed to it, it returns the number. However, if a string is passed to it, it removes non-numeric characters from the string and returns the sanitized string. For other value types, it simply returns null. Here is an example.
sanitizeNumber(53.56); // returns => 53.56
sanitizeNumber('-2oo,40'); // returns => '-240'
sanitizeNumber('badnumber.decimal'); // returns => '.'
  • withoutNulls() - This function expects an array as argument and returns a new array that only contains the non-null items of the original array. Here is an example.
withoutNulls([ 'String', [], null, {}, null, 54 ]); // returns => ['String', [], {}, 54]
  • arrayPairsToObject() - This function expects an array of ({ key: value }) objects, and returns a transformed object with the keys and values. Here is an example.
const pairs = [ { key1: 'value1' }, { key2: 'value2' }, { key3: 'value3' } ];

arrayPairsToObject(pairs); // returns => { key1: 'value1', key2: 'value2', key3: 'value3' }
  • fromPairsToObject() - This is a composed function created using compose(). It has the same effect as executing:
arrayPairsToObject( withoutNulls(array) );

Request and Response Helper Functions

Add the following to the app/helpers.js file.

/* app/helpers.js */

/**
 * Handles the request(Promise) when it is fulfilled
 * and sends a JSON response to the HTTP response stream(res).
 */
const sendResponse = res => async request => {
  return await request
    .then(data => res.json({ status: "success", data }))
    .catch(({ status: code = 500 }) =>
      res.status(code).json({ status: "failure", code, message: code == 404 ? 'Not found.' : 'Request failed.' })
    );
};

/**
 * Loads the html string returned for the given URL
 * and sends a Cheerio parser instance of the loaded HTML
 */
const fetchHtmlFromUrl = async url => {
  return await axios
    .get(enforceHttpsUrl(url))
    .then(response => cheerio.load(response.data))
    .catch(error => {
      error.status = (error.response && error.response.status) || 500;
      throw error;
    });
};

Here, we have added two new functions: sendResponse() and fetchHtmlFromUrl(). Let's try to understand what they do.

  • sendResponse() - This is a higher-order function that expects an Express HTTP response stream(res) as its argument and returns an async function. The returned async function expects a promise or a thenable as its argument(request).

    If the request promise resolves, then a successful JSON response is sent using res.json(), containing the resolved data. If the promise rejects, then an error JSON response with an appropriate HTTP status code is sent. Here is how it can be used in an Express route:
app.get('/path', (req, res, next) => {
  const request = Promise.resolve([1, 2, 3, 4, 5]);
  sendResponse(res)(request);
});

Making a GET request to the /path endpoint will return this JSON response:

{
  "status": "success",
  "data": [1, 2, 3, 4, 5]
}
  • fetchHtmlFromUrl() - This is an async function that expects a url string as its argument. First, it uses axios.get() to fetch the content of the URL(which returns a promise). If the promise resolves, it uses cheerio.load() with the returned content to create a Cheerio parser instance, and then returns the instance. However, if the promise rejects, it throws an error with an appropriate status code.

    The Cheerio parser instance that is returned by this function will enable us extract the data we require. We can use it in much similar ways as we use the jQuery instance returned by calling $() or jQuery() on a DOM target.

DOM Parsing Helper Functions

Let's go ahead to add some additional functions to help us with DOM parsing. Add the following content to the app/helpers.js file.

/* app/helpers.js */

///////////////////////////////////////////////////////////////////////////////
// HTML PARSING HELPER FUNCTIONS
///////////////////////////////////////////////////////////////////////////////

/**
 * Fetches the inner text of the element
 * and returns the trimmed text
 */
const fetchElemInnerText = elem => (elem.text && elem.text().trim()) || null;

/**
 * Fetches the specified attribute from the element
 * and returns the attribute value
 */
const fetchElemAttribute = attribute => elem =>
  (elem.attr && elem.attr(attribute)) || null;

/**
 * Extract an array of values from a collection of elements
 * using the extractor function and returns the array
 * or the return value from calling transform() on array
 */
const extractFromElems = extractor => transform => elems => $ => {
  const results = elems.map((i, element) => extractor($(element))).get();
  return _.isFunction(transform) ? transform(results) : results;
};

/**
 * A composed function that extracts number text from an element,
 * sanitizes the number text and returns the parsed integer
 */
const extractNumber = compose(parseInt, sanitizeNumber, fetchElemInnerText);

/**
 * A composed function that extracts url string from the element's attribute(attr)
 * and returns the url with https scheme
 */
const extractUrlAttribute = attr =>
  compose(enforceHttpsUrl, fetchElemAttribute(attr));

module.exports = {
  compose,
  composeAsync,
  enforceHttpsUrl,
  sanitizeNumber,
  withoutNulls,
  arrayPairsToObject,
  fromPairsToObject,
  sendResponse,
  fetchHtmlFromUrl,
  fetchElemInnerText,
  fetchElemAttribute,
  extractFromElems,
  extractNumber,
  extractUrlAttribute
};

We've added a few more functions. Here are the functions and what they do:

  • fetchElemInnerText() - This function expects an element as argument. It extracts the innerText of the element by calling elem.text(), it trims the text of surrounding whitespaces and returns the trimmed inner text. Here is an example.
const $ = cheerio.load('<div class="fullname">  Glad Chinda </div>');
const elem = $('div.fullname');

fetchElemInnerText(elem); // returns => 'Glad Chinda'
  • fetchElemAttribute() - This is a higher-order function that expects an attribute as argument and returns another function that expects an element as argument. The returned function extracts the value of the given attribute of the element by calling elem.attr(attribute). Here is an example.
const $ = cheerio.load('<div class="username" title="Glad Chinda">@gladchinda</div>');
const elem = $('div.username');

// fetchTitle is a function that expects an element as argument
const fetchTitle = fetchElemAttribute('title');

fetchTitle(elem); // returns => 'Glad Chinda'
  • extractFromElems() - This is a monster function although it does a very simple job. It is a higher-order function that returns another monster higher-order function. Here, we have used a functional programming technique known as currying to create a sequence of functions each requiring just one argument. Here is the sequence of arguments:
extractorFunction -> transformFunction -> elementsCollection -> cheerioInstance

extractFromElems() makes it possible to extract data from a collection of similar elements using an extractor function, and also transform the extracted data using a transform function. The extractor function receives an element as argument, while the transform function receives an array of values as argument.

Let's say we have a collection of elements, each containing the name of a person as innerText. We want to extract all these names and return them in an array, all in uppercase. Here is how we can do this using extractFromElems():

const $ = cheerio.load('<div class="people"><span>Glad Chinda</span><span>John Doe</span><span>Brendan Eich</span></div>');

// Get the collection of span elements containing names
const elems = $('div.people span');

// The transform function
const transformUpperCase = values => values.map(val => String(val).toUpperCase());

// The arguments sequence: extractorFn => transformFn => elemsCollection => cheerioInstance($)
// fetchElemInnerText is used as extractor function
const extractNames = extractFromElems(fetchElemInnerText)(transformUpperCase)(elems);

// Finally pass in the cheerioInstance($)
extractNames($); // returns => ['GLAD CHINDA', 'JOHN DOE', 'BRENDAN EICH']
  • extractNumber() - This is a composed function that expects an element as argument and tries to extract a number from the innerText of the element. It does this by composing parseInt(), sanitizeNumber() and fetchElemInnerText(). It has the same effect as executing:
parseInt( sanitizeNumber( fetchElemInnerText(elem) ) );
  • extractUrlAttribute() - This is a composed higher-order function that expects an attribute as argument and returns another function that expects an element as argument. The returned function tries to extract the URL value of an attribute in the element and returns it with the https scheme. Here is a snippet that shows how it works:
// METHOD 1
const fetchAttribute = fetchElemAttribute(attr);
enforceHttpsUrl( fetchAttribute(elem) );

// METHOD 2: Using extractUrlAttribute()
const fetchUrlAttribute = extractUrlAttribute(attr);
fetchUrlAttribute(elem);

Finally, we export all the helper functions we have created using module.exports. Now that we have our helper functions, we can proceed to the web scraping part of this tutorial.

Getting Ready to go Scotchy

Create a new file named scotch.js in the app directory of your project and add the following content to it:

/* app/scotch.js */

const _ = require('lodash');

// Import helper functions
const {
  compose,
  composeAsync,
  extractNumber,
  enforceHttpsUrl,
  fetchHtmlFromUrl,
  extractFromElems,
  fromPairsToObject,
  fetchElemInnerText,
  fetchElemAttribute,
  extractUrlAttribute
} = require("./helpers");

// scotch.io (Base URL)
const SCOTCH_BASE = "https://scotch.io";

///////////////////////////////////////////////////////////////////////////////
// HELPER FUNCTIONS
///////////////////////////////////////////////////////////////////////////////

/**
 * Resolves the url as relative to the base scotch url
 * and returns the full URL
 */
const scotchRelativeUrl = url =>
  _.isString(url) ? `${SCOTCH_BASE}${url.replace(/^\/*?/, "/")}` : null;

/**
 * A composed function that extracts a url from element attribute,
 * resolves it to the Scotch base url and returns the url with https
 */
const extractScotchUrlAttribute = attr =>
  compose(enforceHttpsUrl, scotchRelativeUrl, fetchElemAttribute(attr));

As you can see, we imported lodash as well as some of the helper functions we created earlier. We also defined a constant named SCOTCH_BASE that contains the base URL of the Scotch website. Finally, we added two helper functions:

  • scotchRelativeUrl() - This function takes a relative url string as argument and returns the URL with the pre-configured SCOTCH_BASE prepended to it. If the url is not a string then null is returned. Here is an example.
scotchRelativeUrl('tutorials'); // returns => 'https://scotch.io/tutorials'
scotchRelativeUrl('//tutorials'); // returns => 'https://scotch.io///tutorials'
scotchRelativeUrl('http://domain.com'); // returns => 'https://scotch.io/http://domain.com'
  • extractScotchUrlAttribute() - This is a composed higher-order function that expects an attribute as argument and returns another function that expects an element as argument. The returned function tries to extract the URL value of an attribute in the element, prepends the pre-configured SCOTCH_BASE to it and returns it with the https scheme. Here is a snippet that shows how it works:
// METHOD 1
const fetchAttribute = fetchElemAttribute(attr);
enforceHttpsUrl( scotchRelativeUrl( fetchAttribute(elem) ) );

// METHOD 2: Using extractScotchUrlAttribute()
const fetchUrlAttribute = extractScotchUrlAttribute(attr);
fetchUrlAttribute(elem);

Scotch Extraction Functions

We want to be able to extract the following data for any Scotch author:

  • profile (name, role, avatar, etc)
  • social links (facebook, twitter, github, etc)
  • stats (total views, total posts, etc)
  • posts

If you recall, the extractFromElems() helper function we created earlier requires an extractor function for extracting content from a collection of similar elements. We are going to define some extractor functions in this section.

Extracting Social Links

First, we will create an extractSocialUrl() function for extracting the social network name and URL from a social link <a> element. Here is the DOM structure of the social link <a> element expected by extractSocialUrl().

<a href="https://github.com/gladchinda" target="_blank" title="GitHub">
  <span class="icon icon-github">
    <svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" version="1.1" id="Capa_1" x="0px" y="0px" width="50" height="50" viewBox="0 0 512 512" style="enable-background:new 0 0 512 512;" xml:space="preserve">
      ...
    </svg>
  </span>
</a>

Calling the extractSocialUrl() function should return an object that looks like the following:

{ github: 'https://github.com/gladchinda' }

Let's go on to create the function. Add the following content to the app/scotch.js file.

/* app/scotch.js */

///////////////////////////////////////////////////////////////////////////////
// EXTRACTION FUNCTIONS
///////////////////////////////////////////////////////////////////////////////

/**
 * Extract a single social URL pair from container element
 */
const extractSocialUrl = elem => {

  // Find all social-icon <span> elements
  const icon = elem.find('span.icon');

  // Regex for social classes
  const regex = /^(?:icon|color)-(.+)$/;

  // Extracts only social classes from the class attribute
  const onlySocialClasses = regex => (classes = '') => classes
      .replace(/\s+/g, ' ')
      .split(' ')
      .filter(classname => regex.test(classname));

  // Gets the social network name from a class name
  const getSocialFromClasses = regex => classes => {
    let social = null;
    const [classname = null] = classes;

    if (_.isString(classname)) {
      const [, name = null] = classname.match(regex);
      social = name ? _.snakeCase(name) : null;
    }

    return social;
  };

  // Extract the href URL from the element
  const href = extractUrlAttribute('href')(elem);

  // Get the social-network name using a composed function
  const social = compose(
    getSocialFromClasses(regex),
    onlySocialClasses(regex),
    fetchElemAttribute('class')
  )(icon);

  // Return an object of social-network-name(key) and social-link(value)
  // Else return null if no social-network-name was found
  return social && { [social]: href };

};

Let's try to understand how the extractSocialUrl() function works:

  1. First, we fetch the <span> child element with an icon class. We also define a regular expression that matches social-icon class names.

  2. We define onlySocialClasses() higher-order function that takes a regular expression as its argument and returns a function. The returned function takes a string of class names separated by spaces. It then uses the regular expression to extract only the social class names from the list and returns them in an array. Here is an example:
const regex = /^(?:icon|color)-(.+)$/;
const extractSocial = onlySocialClasses(regex);
const classNames = 'first-class another-class color-twitter icon-github';

extractSocial(classNames); // returns [ 'color-twitter', 'icon-github' ]
  1. Next, we define getSocialFromClasses() higher-order function that takes a regular expression as its argument and returns a function. The returned function takes an array of single class strings. It then uses the regular expression to extract the social network name from the first class in the list and returns it. Here is an example:
const regex = /^(?:icon|color)-(.+)$/;
const extractSocialName = getSocialFromClasses(regex);
const classNames = [ 'color-twitter', 'icon-github' ];

extractSocialName(classNames); // returns 'twitter'
  1. Afterwards, we extract the href attribute URL from the element. We also extract the social network name from the <span> icon element using a composed function created by composing getSocialFromClasses(regex), onlySocialClasses(regex) and fetchElemAttribute('class').

  2. Finally, we return an object with the social network name as key and the href URL as value. However, if no social network was fetched, then null is returned. Here is an example of the returned object:
{ twitter: 'https://twitter.com/gladchinda' }

Extracting Posts and Stats

We will go ahead to create two additional extraction functions namely: extractPost() and extractStat(), for extracting posts and stats respectively. Before we create the functions, let's take a look at the DOM structure of the elements expected by these functions.

Here is the DOM structure of the element expected by extractPost().

<div class="card large-card" data-type="post" data-id="2448">
  <a href="/tutorials/password-strength-meter-in-angularjs" class="card__img lazy-background" data-src="https://cdn.scotch.io/7540/iKZoyh9WSlSzB9Bt5MNK_post-cover-photo.jpg">
    <span class="tag is-info">Post</span>
  </a>
  <h2 class="card__title">
    <a href="/tutorials/password-strength-meter-in-angularjs">Password Strength Meter in AngularJS</a>
  </h2>
  <div class="card-footer">
    <a class="name" href="/@gladchinda">Glad Chinda</a>
    <a href="/tutorials/password-strength-meter-in-angularjs" title="Views">
      ????️ <span>24,280</span>
    </a>
    <a href="/tutorials/password-strength-meter-in-angularjs#comments-section" title="Comments">
      ???? <span class="comment-number" data-id="2448">5</span>
    </a>
  </div>
</div>

Here is the DOM structure of the element expected by extractStat().

<div class="profile__stat column is-narrow">
  <div class="stat">41,454</div>
  <div class="label">Pageviews</div>
</div>

Add the following content to the app/scotch.js file.

/* app/scotch.js */

/**
 * Extract a single post from container element
 */
const extractPost = elem => {
  const title = elem.find('.card__title a');
  const image = elem.find('a[data-src]');
  const views = elem.find("a[title='Views'] span");
  const comments = elem.find("a[title='Comments'] span.comment-number");

  return {
    title: fetchElemInnerText(title),
    image: extractUrlAttribute('data-src')(image),
    url: extractScotchUrlAttribute('href')(title),
    views: extractNumber(views),
    comments: extractNumber(comments)
  };
};

/**
 * Extract a single stat from container element
 */
const extractStat = elem => {
  const statElem = elem.find(".stat")
  const labelElem = elem.find('.label');

  const lowercase = val => _.isString(val) ? val.toLowerCase() : null;

  const stat = extractNumber(statElem);
  const label = compose(lowercase, fetchElemInnerText)(labelElem);

  return { [label]: stat };
};

The extractPost() function extracts the title, image, URL, views and comments of a post by parsing the children of the given element. It uses a couple of helper functions we created earlier to extract data from the appropriate elements.

Here is an example of the object returned from calling extractPost().

{
  title: "Password Strength Meter in AngularJS",
  image: "https://cdn.scotch.io/7540/iKZoyh9WSlSzB9Bt5MNK_post-cover-photo.jpg",
  url: "https://scotch.io//tutorials/password-strength-meter-in-angularjs",
  views: 24280,
  comments: 5
}

The extractStat() function extracts the stat data contained in the given element. Here is an example of the object returned from calling extractStat().

{ pageviews: 41454 }

Fetching the Scotch Author Profile

Now we will proceed to define the extractAuthorProfile() function that extracts the complete profile of the Scotch author. Add the following content to the app/scotch.js file.

/* app/scotch.js */

/**
 * Extract profile from a Scotch author's page using the Cheerio parser instance
 * and returns the author profile object
 */
const extractAuthorProfile = $ => {

  const mainSite = $('#site__main');
  const metaScotch = $("meta[property='og:url']");
  const scotchHero = mainSite.find('section.hero--scotch');
  const superGrid = mainSite.find('section.super-grid');

  const authorTitle = scotchHero.find(".profile__name h1.title");
  const profileRole = authorTitle.find(".tag");
  const profileAvatar = scotchHero.find("img.profile__avatar");
  const profileStats = scotchHero.find(".profile__stats .profile__stat");
  const authorLinks = scotchHero.find(".author-links a[target='_blank']");
  const authorPosts = superGrid.find(".super-grid__item [data-type='post']");

  const extractPosts = extractFromElems(extractPost)();
  const extractStats = extractFromElems(extractStat)(fromPairsToObject);
  const extractSocialUrls = extractFromElems(extractSocialUrl)(fromPairsToObject);

  return Promise.all([
    fetchElemInnerText(authorTitle.contents().first()),
    fetchElemInnerText(profileRole),
    extractUrlAttribute('content')(metaScotch),
    extractUrlAttribute('src')(profileAvatar),
    extractSocialUrls(authorLinks)($),
    extractStats(profileStats)($),
    extractPosts(authorPosts)($)
  ]).then(([ author, role, url, avatar, social, stats, posts ]) => ({ author, role, url, avatar, social, stats, posts }));

};

/**
 * Fetches the Scotch profile of the given author
 */
const fetchAuthorProfile = author => {
  const AUTHOR_URL = `${SCOTCH_BASE}/@${author.toLowerCase()}`;
  return composeAsync(extractAuthorProfile, fetchHtmlFromUrl)(AUTHOR_URL);
};

module.exports = { fetchAuthorProfile };

The extractAuthorProfile() function is very straight-forward. We first use $(the cheerio parser instance) to find a couple of elements and element collections.

Next, we use the extractFromElems() helper function together with the extractor functions we created earlier in this section (extractPost, extractStat and extractSocialUrl) to create higher-order extraction functions. Notice how we use the fromPairsToObject helper function we created earlier as a transform function.

Finally, we use Promise.all() to extract all the required data, leveraging on a couple of helper functions we created earlier. The extracted data is contained in an array structure following this sequence: author name, role, Scotch link, avatar link, social links, stats and posts.

Notice how we use destructuring in the .then() promise handler to construct the final object that is returned when all the promises resolve. The returned object should look like the following:

{
  author: 'Glad Chinda',
  role: 'Author',
  url: 'https://scotch.io/@gladchinda',
  avatar: 'https://cdn.scotch.io/7540/EnhoZyJOQ2ez9kVhsS9B_profile.jpg',
  social: {
    twitter: 'https://twitter.com/gladchinda',
    github: 'https://github.com/gladchinda'
  },
  stats: {
    posts: 6,
    pageviews: 41454,
    readers: 31676
  },
  posts: [
    {
      title: 'Password Strength Meter in AngularJS',
      image: 'https://cdn.scotch.io/7540/iKZoyh9WSlSzB9Bt5MNK_post-cover-photo.jpg',
      url: 'https://scotch.io//tutorials/password-strength-meter-in-angularjs',
      views: 24280,
      comments: 5
    },
    ...
  ]
}

We also define the fetchAuthorProfile() function that accepts an author's Scotch username and returns a Promise that resolves to the profile of the author. For an author whose username is gladchinda, the Scotch URL is https://scotch.io/@gladchinda.

fetchAuthorProfile() uses the composeAsync() helper function to create a composed function that first fetches the DOM content of the author's Scotch page using the fetchHtmlFromUrl() helper function, and finally extracts the profile of the author using the extractAuthorProfile() function we just created.

Finally, we export fetchAuthorProfile as the only identifier in the module.exports object.

Finishing up with a Route

We are almost done with our API. We need to add a route to our server to enable us fetch the profile of any Scotch author. The route will have the following structure, where the author parameter represents the username of the Scotch author.

GET /scotch/:author

Let's go ahead and create this route. We will make a couple of changes to the server.js file. First, add the following to the server.js file to require some of the functions we need.

/* server.js */

// Require the needed functions
const { sendResponse } = require('./app/helpers');
const { fetchAuthorProfile } = require('./app/scotch');

Finally, add the route to the server.js file immediately after the middlewares.

/* server.js */

// Add the Scotch author profile route
app.get('/scotch/:author', (req, res, next) => {
  const author = req.params.author;
  sendResponse(res)(fetchAuthorProfile(author));
});

As you can see, we pass the author received from the route parameter to the fetchAuthorProfile() function to get the profile of the given author. We then use the sendResponse() helper method to send the returned profile as a JSON response.

We have successfully built our API using a web scraping technique. Go ahead and test the API by running npm start command on your terminal. Launch your favorite HTTP testing tool e.g Postman and test the API endpoint. If you followed all the steps correctly, you should have a result that looks like the following demo:

Scotch Scraping API Demo

Conclusion

In this tutorial, we have seen how we can employ web scraping techniques(especially DOM parsing) to extract data from a website. We used the Cheerio package to parse the content of a webpage using available DOM methods in a much similar fashion as the popular jQuery library. Note however that Cheerio has its limitations. You can achieve more advanced parsing using headless browsers like JSDOM and PhantomJS.

You can find the source code for the API we built in this tutorial on GitHub. I have also built a demo app based on the API from this tutorial as shown in the initial screenshot. You can see the app on Heroku and the source code on GitHub.

Nahoru
Tento web používá k poskytování služeb a analýze návštěvnosti soubory cookie. Používáním tohoto webu s tímto souhlasíte. Další informace