Web crawler for Node.JS
Switch branches/tags
Clone or download
antivanov Remember urls up to a normalization
  * Store normalized urls in the state, interpret non-normalized url
form as same url
  * Fix the bug in state.js when 1 url was left it would be removed from "beingCrawled" even if it was not finished
Latest commit 6cd340d Jun 22, 2018

README.md

js-crawler

Web crawler for Node.JS, both HTTP and HTTPS are supported.

Installation

npm install js-crawler

Usage

The crawler provides intuitive interface to crawl links on web sites. Example:

var Crawler = require("js-crawler").default;

new Crawler().configure({depth: 3})
  .crawl("http://www.google.com", function onSuccess(page) {
    console.log(page.url);
  });

The call to configure is optional, if it is omitted the default option values will be used.

onSuccess callback will be called for each page that the crawler has crawled. page value passed to the callback will contain the following fields:

  • url - URL of the page
  • content - body of the page (usually HTML)
  • status - the HTTP status code

Extra information can be retrieved from the rest of the page fields: error, response, body which are identical to the ones passed to the callback of request invocation of the Request module. referer field will reference the url of the page that lead the crawler to the current page.

Options-based API

Alternative APIs for passing callbacks to the crawl function.

var Crawler = require("js-crawler").default;

var crawler = new Crawler().configure({ignoreRelative: false, depth: 2});

crawler.crawl({
  url: "https://github.com",
  success: function(page) {
    console.log(page.url);
  },
  failure: function(page) {
    console.log(page.status);
  },
  finished: function(crawledUrls) {
    console.log(crawledUrls);
  }
});

Handling errors

It is possible to pass an extra callback to handle errors, consider the modified example above:

var Crawler = require("js-crawler").default;

new Crawler().configure({depth: 3})
  .crawl("http://www.google.com", function(page) {
    console.log(page.url);
  }, function(response) {
    console.log("ERROR occurred:");
    console.log(response.status);
    console.log(response.url);
    console.log(response.referer);
  });

Here the second callback will be called for each page that could not be accessed (maybe because the corresponding site is down). status may be not defined.

Knowing when all crawling is finished

Extra callback can be passed that will be called when all the urls have been crawled and crawling has finished. All crawled urls will be passed to that callback as an argument.

var Crawler = require("js-crawler").default;

new Crawler().configure({depth: 2})
  .crawl("http://www.google.com", function onSuccess(page) {
    console.log(page.url);
  }, null, function onAllFinished(crawledUrls) {
    console.log('All crawling finished');
    console.log(crawledUrls);
  });

Limiting the rate at which requests are made

maxRequestsPerSecond option

By default the maximum number of HTTP requests made per second is 100, but this can be adjusted by using the option maxRequestsPerSecond if one wishes not to use too much of network or, opposite, wishes for yet faster crawling.

var Crawler = require("js-crawler").default;

var crawler = new Crawler().configure({maxRequestsPerSecond: 2});

crawler.crawl({
  url: "https://github.com",
  success: function(page) {
    console.log(page.url);
  },
  failure: function(page) {
    console.log(page.status);
  }
});

With this configuration only at most 2 requests per second will be issued. The actual request rate depends on the network speed as well, maxRequestsPerSecond configures just the upper boundary.

maxRequestsPerSecond can also be fractional, the value 0.1, for example, would mean maximum one request per 10 seconds.

maxConcurrentRequests option

Even more flexibility is possible when using maxConcurrentRequests option, it limits the number of HTTP requests that can be active simultaneously. If the number of requests per second is too high for a given set of sites/network requests may start to pile up, then specifying maxConcurrentRequests can help ensure that the network is not overloaded with piling up requests.

Specifying both options

It is possible to customize both options in case we are not sure how performant the network and sites are. Then maxRequestsPerSecond limits how many requests the crawler is allowed to make and maxConcurrentRequests allows to specify how should the crawler adjust its rate of requests depending on the real-time performance of the network/sites.

var Crawler = require("js-crawler").default;

var crawler = new Crawler().configure({
  maxRequestsPerSecond: 10,
  maxConcurrentRequests: 5
});

crawler.crawl({
  url: "https://github.com",
  success: function(page) {
    console.log(page.url);
  },
  failure: function(page) {
    console.log(page.status);
  }
});
Default option values

By default the values are as follows:

maxRequestsPerSecond 100

maxConcurrentRequests 10

That is, we expect on average that 100 requests will be made every second and only 10 will be running concurrently, and every request will take something like 100ms to complete.

Reusing the same crawler instance for repeated crawling: forgetting crawled urls

By default a crawler instance will remember all the urls it ever crawled and will not crawl them again. In order to make it forget all the crawled urls the method forgetCrawled can be used. There is another way to solve the same problem: create a new instance of a crawler. Example https://github.com/antivanov/js-crawler/blob/master/examples/github_forgetting_crawled_urls.js

Supported options

  • depth - the depth to which the links from the original page will be crawled. Example: if site1.com contains a link to site2.com which contains a link to site3.com, depth is 2 and we crawl from site1.com then we will crawl site2.com but will not crawl site3.com as it will be too deep.

The default value is 2.

  • ignoreRelative - ignore the relative URLs, the relative URLs on the same page will be ignored when crawling, so /wiki/Quick-Start will not be crawled and https://github.com/explore will be crawled. This option can be useful when we are mainly interested in sites to which the current sites refers and not just different sections of the original site.

The default value is false.

  • userAgent - User agent to send with crawler requests.

The default value is crawler/js-crawler

  • maxRequestsPerSecond - the maximum number of HTTP requests per second that can be made by the crawler, default value is 100

  • maxConcurrentRequests - the maximum number of concurrent requests that should not be exceeded by the crawler, the default value is 10

  • shouldCrawl - function that specifies whether a url should be crawled/requested, returns true or false, argument is the current url the crawler considers for crawling

  • shouldCrawlLinksFrom - function that specifies whether the crawler should crawl links found at a given url, returns true or false, argument is the current url being crawled

Note: shouldCrawl determines if a given URL should be requested/visited at all, whereas shouldCrawlLinksFrom determines if the links on a given URL should be harvested/added to the crawling queue. Many users may find that using shouldCrawl is sufficient, as links from a page cannot be crawled if the page is never visited/requested in the first place. However there is a common use case for having shouldCrawlLinksFrom: if a user would like to check external links on a site for errors without crawling those external links, the user could create a shouldCrawlLinksFrom function that restricts crawling to the original url without visiting external links.

Examples:

shouldCrawl: the following will crawl subreddit index pages reachable on reddit.com from the JavaScript subreddit:

var Crawler = require("js-crawler").default;

var rootUrl = "http://www.reddit.com/r/javascript";

function isSubredditUrl(url) {
  return !!url.match(/www\.reddit\.com\/r\/[a-zA-Z0-9]+\/$/g);
}

var crawler = new Crawler().configure({
  shouldCrawl: function(url) {
    return isSubredditUrl(url) || url == rootUrl;
  }
});

crawler.crawl(rootUrl, function(page) {
  console.log(page.url);
});

The default value for each is a function that always returns true.

Development

Install dependencies

npm install

Running the build

npm run build

Unit tests

npm test

launches unit tests in the console mode

npm run test:tdd

launches a browser in which unit tests can be debugged

End-to-end tests

mocha and express are used to setup and run end-to-end tests

Make sure dependencies are installed (mocha is included)

npm install

Install express globablly

npm install -g express

Start the end-to-end target server

cd e2e
node server.js

Now the server runs on the port 3000. Run the end-to-end specs:

npm run e2e

License

MIT License (c) Anton Ivanov

Credits

The crawler depends on the following Node.JS modules: