crawler-ninja

Crawler Ninja

This crawler aims to build custom solutions for crawling/scraping sites. For example, it can help to audit a site, find expired domains, build corpus, scrap texts, find netlinking spots, retrieve site ranking, check if web pages are correctly indexed, ...

This is just a matter of plugins ! :-) We plan to build generic & simple plugins but you are free to create your owns.

The best environment to run Crawler Ninja is a linux server.

Help & Forks welcomed ! or please wait ... work in progress !

How to install

$ npm install crawler-ninja --save

Crash course

###How to use an existing plugin ?

var crawler = require("crawler-ninja");
var logger  = require("crawler-ninja/plugins/log-plugin");

var c = new crawler.Crawler();
var log = new logger.Plugin(c);

c.on("end", function() {

    var end = new Date();
    console.log("End of crawl !, done in : " + (end - start));


});

var start = new Date();
c.queue({url : "http://www.mysite.com/"});

This script logs on the console all crawled pages thanks to the usage of the log-plugin component.

The Crawler component emits different kind of events that plugins can use (see below). When the crawl ends, the event 'end' is emitted.

###Create a new plugin

The following script show you the events callbacks that your have to implement for creating a new plugin.

This is not mandatory to implement all crawler events. You can also reduce the scope of the crawl by using the different crawl options (see below the section : option references).

// userfull lib for managing uri
var URI    = require('crawler/lib/uri');


function Plugin(crawler) {

    this.crawler = crawler;

    /**
     * Emits when the crawler found an error
     *
     * @param the usual error object
     * @param the result of the request (contains uri, headers, ...)
     */
    this.crawler.on("error", function(error, result) {

    });

    /**
     * Emits when the crawler crawls a resource (html,js,css, pdf, ...)
     *
     * @param result : the result of the crawled resource
     * @param the jquery like object for accessing to the HTML tags. Null is if the resource is not an HTML.
     * See the project cheerio : https://github.com/cheeriojs/cheerio
     */
    this.crawler.on("crawl", function(result,$) {

    });

    /**
     * Emits when the crawler founds a link in a page
     *
     * @param the page that contains the link
     * @param the link uri
     * @param the anchor text
     * @param true if the link is do follow
     */
    this.crawler.on("crawlLink", function(page, link, anchor, isDoFollow) {

    });


    /**
     * Emits when the crawler founds an image
     *
     * @param the page that contains the image
     * @param the image uri
     * @param the alt text
     */
    this.crawler.on("crawlImage", function(page, link, alt) {


    });

    /**
     * Emits when the crawler founds a redirect 3**
     *
     * @param the from url
     * @param the to url
     * @param statusCode : the exact status code : 301, 302, ...
     */
    this.crawler.on("crawlRedirect", function(from, to, statusCode) {

    });

}

module.exports.Plugin = Plugin;

Option references

The main crawler config options

You can pass these options to the Crawler() constructor like :

var c = new crawler.Crawler({
  externalLinks : true,
  scripts : false,
  images : false
});

• maxConnections : the number of connections used to crawl, default is 10.
• externalLinks : if true crawl external links, default = false.
• externalDomains : if true crawl the external domains. This option can crawl a lot of different linked domains.
• scripts : if true crawl script tags, default = true.
• links : if true crawl link tags, default = true.
• linkTypes : the type of the links tags to crawl (match to the rel attribute), default = ["canonical", "stylesheet"].
• images : if true crawl images, default = true.
• protocols : list of the protocols to crawl, default = ["http", "https"].
• timeout : timeout per requests in milliseconds, default = 20000.
• retries : number of retries if the request is on timeout, default = 3.
• retryTimeout : number of milliseconds to wait before retrying, default = 10000.
• maxErrors : number of timeout errors before forcing to decrease the crawl rate, default is 5. If the value is -1, there is no check.
• errorRates : an array of crawl rates to apply if there are no too many errors, default : [200,350,500] (in ms)
• skipDuplicates : if true skips URIs that were already crawled, default is true.
• rateLimits : number of milliseconds to delay between each requests , default = 0.
• depthLimit : the depth limit for the crawl, default is no limit.
• followRedirect : if true, the crawl will not return the 301, it will follow directly the redirection, default is false.
• userAgent : String, defaults to "node-crawler/[version]"
• referer : String, if truthy sets the HTTP referer header
• domainBlackList : The list of domain names (without tld) to avoid to crawl (an array of String). The default list is in the file : /default-lists/domain-black-list.js
• proxyList : The list of proxy to use for each crawler request (see below).

You can also use the mikeal's request options and will be directly passed to the request() method.

You can pass these options to the Crawler() constructor if you want them to be global or as items in the queue() calls if you want them to be specific to that item (overwriting global options)

Add your own crawl rules

If the predefined options are not sufficiants, you can customize which kind of links to crawl by implementing a callback function in the crawler config object. This is a nice way to limit the crawl scope in function of your needs. The following example crawls only dofollow links.

var c = new crawler.Crawler({
  // add here predefined options you want to override

  /**
   *  this callback is called for each link found in an html page
   *  @param  : the uri of the page that contains the link
   *  @param  : the uri of the link to check
   *  @param  : the anchor text of the link
   *  @param  : true if the link is dofollow
   *  @return : true if the crawler can crawl the link on this html page
   */
  canCrawl : function(htlmPage, link, anchor, isDoFollow) {
      return isDoFollow;
  }

});

Using proxies

Crawler.ninja can be configured to execute each http request through a proxy. It uses the npm package simple-proxies.

You have to install it in your project with the command :

$ npm install simple-proxies --save

Here is a code sample that uses proxies from a file :

var proxyLoader = require("simple-proxies/lib/proxyfileloader");
var crawler     = require("crawler-ninja");
var logger      = require("crawler-ninja/plugins/log-plugin");


var proxyFile = "proxies.txt";

// Load proxies
var config = proxyLoader.config()
                        .setProxyFile(proxyFile)
                        .setCheckProxies(false)
                        .setRemoveInvalidProxies(false);

proxyLoader.loadProxyFile(config, function(error, proxyList) {
    if (error) {
      console.log(error);

    }
    else {
       crawl(proxyList);
    }

});


function crawl(proxyList){
    var c = new crawler.Crawler({
        externalLinks : true,
        images : false,
        scripts : false,
        links : false, //link tags used for css, canonical, ...
        followRedirect : true,
        proxyList : proxyList
    });

    var log = new logger.Plugin(c);

    c.on("end", function() {

        var end = new Date();
        console.log("Well done Sir !, done in : " + (end - start));


    });

    var start = new Date();
    c.queue({url : "http://www.site.com"});
}

Using the crawl logger in your own plugin

The current crawl logger is based on Bunyan.

• It logs the all crawl actions & errors in the file ./logs/crawler.log.
• The list of errors can be found in ./logs/errors.log.

You can query the log file after the crawl (see the Bunyan doc for more informations) in order to filter errors or other info.

You can also use the current logger or create a new one in your own Plugin.

Use default loggers

var log = require("crawler-ninja./lib/logger.js").Logger;

log.info("log info");  // Log into crawler.log
log.debug("log debug"); // Log into crawler.log
log.error("log error"); // Log into crawler.log & errors.log
log.info({statusCode : 200, url: "http://www.google.com" }) // log a json

Create a new logger for your plugin

// Log into crawler.log
var log = require("crawler-ninja/lib/logger.js");

var myLog = log.createLogger("myLoggerName", "./logs/myplugin.log");

myLog.log({url:"http://www.google.com", pageRank : 10});

Please, feel free to read the code in log-plugin to get more info on how to log from you own plugin.

More features & flexibilities will be added in the upcoming releases.

Control the crawl rate

All sites cannot support an intensive crawl. This crawl provide 2 solutions to control the crawl rates :

• implicit : the crawl decrease the crawl rate if there are too many timeouts on a host. the crawl rate is controlled for each crawled hosts separately.
• explicit : you can specify the crawl rate in the crawler config. This setting is unique for all hosts.

Implicit setting

Without changing the crawler config, it will decrease the crawl rate after 5 timouts errors on a host. It will force a rate of 200ms between each requests. If new 5 timout errors still occur, it will use a rate of 350ms and after that a rate of 500ms between all requests for this host. If the timouts persist, the crawler will cancel the crawl on that host.

You can change the default values for this implicit setting (5 timout errors & rates = 200, 350, 500ms). Here is an example :

var crawler = require("crawler-ninja");
var logger  = require("crawler-ninja/plugins/log-plugin");

var c = new crawler.Crawler({
  // new values for the implicit setting
  maxErrors : 5,
  errorRates : [300, 600, 900]

});

var log = new logger.Plugin(c);

c.on("end", function() {

    var end = new Date();
    console.log("End of crawl !, done in : " + (end - start));


});

var start = new Date();
c.queue({url : "http://www.mysite.com/"});

Note that an higher value for maxErrors can decrease the number of analyzed pages. You can assign the value -1 to maxErrors in order to desactivate the implicit setting

Explicit setting

In this configuration, you are apply the same crawl rate for all requests on all hosts.

var crawler = require("crawler-ninja");
var logger  = require("crawler-ninja/plugins/log-plugin");

var c = new crawler.Crawler({
  rateLimits         : 200 //200ms between each request

});

var log = new logger.Plugin(c);

c.on("end", function() {

    var end = new Date();
    console.log("End of crawl !, done in : " + (end - start));


});

var start = new Date();
c.queue({url : "http://www.mysite.com/"});

If both settings are applied for one crawl, the implicit setting will be forced by the crawler after the "maxErrors".

Current Plugins

• Log
• Stat
• Audit

Rough todolist

• More & more plugins (in progress)
• Use Riak as default persistence layer/Crawler Store
• Multicore architecture and/or micro service architecture for plugins that requires a lot of CPU usage
• CLI for extracting data from the Crawl DB
• Build UI : dashboards, view project data, ...

ChangeLog

0.1.0

• crawler engine that support navigation through a.href, detect images, links tag & scripts.
• Add flexible parameters to crawl (see the section crawl option above) like the crawl depth, crawl rates, craw external links, ...
• Implement a basic log plugin & an SEO audit plugin.
• Unit tests.

0.1.1

• Add proxy support.
• Gives the possibility to crawl (or not) the external domains which is different than crawling only the external links. Crawl external links means to check the http status & content of the linked external resources which is different of expand the crawl through the entire external domains.

0.1.2

• Review Log component.
• set the default userAgent to NinjaBot.
• update README.

0.1.3

• avoid crash for long crawls.

0.1.4

• code refactoring in order to make a tentative of multicore proccess for making http requests

0.1.5

• remove the multicore support for making http requests due to the important overhead. Plan to use multicore for some intensive CPU plugins.
• refactor the rate limit and http request retries in case of errors.

0.1.6

• Review logger : use winston, different log files : the full crawl, errors and urls. Gives the possibility to create a specific logger for a plugin.

0.1.7

• Too many issues with winston, use Bunyan for the logs
• Refactor how to set the urls in the crawl option : simple url, an array of urls or of json option objects.
• Review the doc aka README
• Review how to manage the timeouts in function of the site to crawl. If too many timeouts for one domain, the crawler will change the settings in order to decrease request concurrency. If errors persist, the crawler will stop to crawl this domain.
• Add support for a blacklist of domains.