Skip to content

Latest commit

 

History

History
582 lines (406 loc) · 20.7 KB

README.md

File metadata and controls

582 lines (406 loc) · 20.7 KB
Raw

Node.js Web Crawler

npm npm GitHub issues David David Travis

Supercrawler is a Node.js web crawler. It is designed to be highly configurable and easy to use.

When Supercrawler successfully crawls a page (which could be an image, a text document or any other file), it will fire your custom content-type handlers. Define your own custom handlers to parse pages, save data and do anything else you need.

Features

  • Link Detection. Supercrawler will parse crawled HTML documents, identify links and add them to the queue.
  • Robots Parsing. Supercrawler will request robots.txt and check the rules before crawling. It will also identify any sitemaps.
  • Sitemaps Parsing. Supercrawler will read links from XML sitemap files, and add links to the queue.
  • Concurrency Limiting. Supercrawler limits the number of requests sent out at any one time.
  • Rate limiting. Supercrawler will add a delay between requests to avoid bombarding servers.
  • Exponential Backoff Retry. Supercrawler will retry failed requests after 1 hour, then 2 hours, then 4 hours, etc. To use this feature, you must use the database-backed or Redis-backed crawl queue.
  • Hostname Balancing. Supercrawler will fairly split requests between different hostnames. To use this feature, you must use the Redis-backed crawl queue.

How It Works

Crawling is controlled by the an instance of the Crawler object, which acts like a web client. It is responsible for coordinating with the priority queue, sending requests according to the concurrency and rate limits, checking the robots.txt rules and despatching content to the custom content handlers to be processed. Once started, it will automatically crawl pages until you ask it to stop.

The Priority Queue or UrlList keeps track of which URLs need to be crawled, and the order in which they are to be crawled. The Crawler will pass new URLs discovered by the content handlers to the priority queue. When the crawler is ready to crawl the next page, it will call the getNextUrl method. This method will work out which URL should be crawled next, based on implementation-specific rules. Any retry logic is handled by the queue.

The Content Handlers are functions which take content buffers and do some further processing with them. You will almost certainly want to create your own content handlers to analyze pages or store data, for example. The content handlers tell the Crawler about new URLs that should be crawled in the future. Supercrawler provides content handlers to parse links from HTML pages, analyze robots.txt files for Sitemap: directives and parse sitemap files for URLs.

Get Started

First, install Supercrawler.

npm install supercrawler --save

Second, create an instance of Crawler.

var supercrawler = require("supercrawler");

// 1. Create a new instance of the Crawler object, providing configuration
// details. Note that configuration cannot be changed after the object is
// created.
var crawler = new supercrawler.Crawler({
  // By default, Supercrawler uses a simple FIFO queue, which doesn't support
  // retries or memory of crawl state. For any non-trivial crawl, you should
  // create a database. Provide your database config to the constructor of
  // DbUrlList.
  urlList: new supercrawler.DbUrlList({
    db: {
      database: "crawler",
      username: "root",
      password: secrets.db.password,
      sequelizeOpts: {
        dialect: "mysql",
        host: "localhost"
      }
    }
  }),
  // Tme (ms) between requests
  interval: 1000,
  // Maximum number of requests at any one time.
  concurrentRequestsLimit: 5,
  // Time (ms) to cache the results of robots.txt queries.
  robotsCacheTime: 3600000,
  // Query string to use during the crawl.
  userAgent: "Mozilla/5.0 (compatible; supercrawler/1.0; +https://github.com/brendonboshell/supercrawler)",
  // Custom options to be passed to request.
  request: {
    headers: {
      'x-custom-header': 'example'
    }
  }
});

Third, add some content handlers.

// Get "Sitemaps:" directives from robots.txt
crawler.addHandler(supercrawler.handlers.robotsParser());

// Crawl sitemap files and extract their URLs.
crawler.addHandler(supercrawler.handlers.sitemapsParser());

// Pick up <a href> links from HTML documents
crawler.addHandler("text/html", supercrawler.handlers.htmlLinkParser({
  // Restrict discovered links to the following hostnames.
  hostnames: ["example.com"]
}));

// Match an array of content-type
crawler.addHandler(["text/plain", "text/html"], myCustomHandler);

// Custom content handler for HTML pages.
crawler.addHandler("text/html", function (context) {
  var sizeKb = Buffer.byteLength(context.body) / 1024;
  logger.info("Processed", context.url, "Size=", sizeKb, "KB");
});

Fourth, add a URL to the queue and start the crawl.

crawler.getUrlList()
  .insertIfNotExists(new supercrawler.Url("http://example.com/"))
  .then(function () {
    return crawler.start();
  });

That's it! Supercrawler will handle the crawling for you. You only have to define your custom behaviour in the content handlers.

Crawler

Each Crawler instance represents a web crawler. You can configure your crawler with the following options:

Option Description
urlList Custom instance of UrlList type queue. Defaults to FifoUrlList, which processes URLs in the order that they were added to the queue; once they are removed from the queue, they cannot be recrawled.
interval Number of milliseconds between requests. Defaults to 1000.
concurrentRequestsLimit Maximum number of concurrent requests. Defaults to 5.
robotsEnabled Indicates if the robots.txt is downloaded and checked. Defaults to true.
robotsCacheTime Number of milliseconds that robots.txt should be cached for. Defaults to 3600000 (1 hour).
robotsIgnoreServerError Indicates if 500 status code response for robots.txt should be ignored. Defaults to false.
userAgent User agent to use for requests. This can be either a string or a function that takes the URL being crawled. Defaults to Mozilla/5.0 (compatible; supercrawler/1.0; +https://github.com/brendonboshell/supercrawler).
request Object of options to be passed to request. Note that request does not support an asynchronous (and distributed) cookie jar.

Example usage:

var crawler = new supercrawler.Crawler({
  interval: 1000,
  concurrentRequestsLimit: 1
});

The following methods are available:

Method Description
getUrlList Get the UrlList type instance.
getInterval Get the interval setting.
getConcurrentRequestsLimit Get the maximum number of concurrent requests.
getUserAgent Get the user agent.
start Start crawling.
stop Stop crawling.
addHandler(handler) Add a handler for all content types.
addHandler(contentType, handler) Add a handler for a specific content type. If contentType is a string, then (for example) 'text' will match 'text/html', 'text/plain', etc. If contentType is an array of strings, the page content type must match exactly.

The Crawler object fires the following events:

Event Description
crawlurl(url) Fires when crawling starts with a new URL.
crawledurl(url, errorCode, statusCode, errorMessage) Fires when crawling of a URL is complete. errorCode is null if no error occurred. statusCode is set if and only if the request was successful. errorMessage is null if no error occurred.
urllistempty Fires when the URL list is (intermittently) empty.
urllistcomplete Fires when the URL list is permanently empty, barring URLs added by external sources. This only makes sense when running Supercrawler in non-distributed fashion.

DbUrlList

DbUrlList is a queue backed with a database, such as MySQL, Postgres or SQLite. You can use any database engine supported by Sequelize.

If a request fails, this queue will ensure the request gets retried at some point in the future. The next request is schedule 1 hour into the future. After that, the period of delay doubles for each failure.

Options:

Option Description
opts.db.database Database name.
opts.db.username Database username.
opts.db.password Database password.
opts.db.sequelizeOpts Options to pass to sequelize.
opts.db.table Table name to store URL queue. Default = 'url'
opts.recrawlInMs Number of milliseconds to recrawl a URL. Default = 31536000000 (1 year)

Example usage:

new supercrawler.DbUrlList({
  db: {
    database: "crawler",
    username: "root",
    password: "password",
    sequelizeOpts: {
      dialect: "mysql",
      host: "localhost"
    }
  }
})

The following methods are available:

Method Description
insertIfNotExists(url) Insert a Url object.
upsert(url) Upsert Url object.
getNextUrl() Get the next Url to be crawled.

RedisUrlList

RedisUrlList is a queue backed with Redis.

If a request fails, this queue will ensure the request gets retried at some point in the future. The next request is schedule 1 hour into the future. After that, the period of delay doubles for each failure.

It also balances requests between different hostnames. So, for example, if you crawl a sitemap file with 10,000 URLs, the next 10,000 URLs will not be stuck in the same host.

Options:

Option Description
opts.redis Options passed to ioredis.
opts.delayHalfLifeMs Hostname delay factor half-life. Requests are delayed by an amount of time proportional to the number of pages crawled for a hostname, but this factor exponentially decays over time. Default = 3600000 (1 hour).
opts.expiryTimeMs Amount of time before recrawling a successful URL. Default = 2592000000 (30 days).
opts.initialRetryTimeMs Amount of time to wait before first retry after a failed URL. Default = 3600000 (1 hour)

Example usage:

new supercrawler.RedisUrlList({
  redis: {
    host: "127.0.0.1"
  }
})

The following methods are available:

Method Description
insertIfNotExists(url) Insert a Url object.
upsert(url) Upsert Url object.
getNextUrl() Get the next Url to be crawled.

FifoUrlList

The FifoUrlList is the default URL queue powering the crawler. You can add URLs to the queue, and they will be crawled in the same order (FIFO).

Note that, with this queue, URLs are only crawled once, even if the request fails. If you need retry functionality, you must use DbUrlList.

The following methods are available:

Method Description
insertIfNotExists(url) Insert a Url object.
upsert(url) Upsert Url object.
getNextUrl() Get the next Url to be crawled.

Url

A Url represents a URL to be crawled, or a URL that has already been crawled. It is uniquely identified by an absolute-path URL, but also contains information about errors and status codes.

Option Description
url Absolute-path string url
statusCode HTTP status code or null.
errorCode String error code or null.

Example usage:

var url = new supercrawler.Url({
  url: "https://example.com"
});

You can also call it just a string URL:

var url = new supercrawler.Url("https://example.com");

The following methods are available:

Method Description
getUniqueId Get the unique identifier for this object.
getUrl Get the absolute-path string URL.
getErrorCode Get the error code, or null if it is empty.
getStatusCode Get the status code, or null if it is empty.

handlers.htmlLinkParser

A function that returns a handler which parses a HTML page and identifies any links.

Option Description
hostnames Array of hostnames that are allowed to be crawled.
urlFilter(url, pageUrl) Function that takes a URL and returns true if it should be included.

Example usage:

var hlp = supercrawler.handlers.htmlLinkParser({
  hostnames: ["example.com"]
});
var hlp = supercrawler.handlers.htmlLinkParser({
  urlFilter: function (url) {
    return url.indexOf("page1") === -1;
  }
});

handlers.robotsParser

A function that returns a handler which parses a robots.txt file. Robots.txt file are automatically crawled, and sent through the same content handler routines as any other file. This handler will look for any Sitemap: directives, and add those XML sitemaps to the crawl.

It will ignore any files that are not /robots.txt.

If you want to extract the URLs from those XML sitemaps, you will also need to add a sitemap parser.

Option Description
urlFilter(sitemapUrl, robotsTxtUrl) Function that takes a URL and returns true if it should be included.

Example usage:

var rp = supercrawler.handlers.robotsParser();
crawler.addHandler("text/plain", supercrawler.handlers.robotsParser());

handlers.sitemapsParser

A function that returns a handler which parses an XML sitemaps file. It will pick up any URLs matching sitemapindex > sitemap > loc, urlset > url > loc.

It will also handle a gzipped file, since that it part of the sitemaps specification.

Option Description
urlFilter Function that takes a URL (including sitemap entries) and returns true if it should be included.

Example usage:

var sp = supercrawler.handlers.sitemapsParser();
crawler.addHandler(supercrawler.handlers.sitemapsParser());

Changelog

2.0.0

  • [Added] crawledurl event to contain the error message, thanks hjr3.
  • [Changed] sitemapsParser to apply urlFilter on the sitemaps entries, thanks hjr3.
  • [Added] Crawler to take userAgent option as a function, thanks hjr3.

1.7.2

  • [Fixed] Update DbUrlList to use symbol operators, thanks hjr3.

1.7.1

1.7.0

  • [Changed] Crawler#addHandler can now take an array of content-type to match, thanks taina0407.

1.6.0

  • [Added] Added opts.db.table option to DbUrlList (adversinc).
  • [Added] Added recrawlInMs option to DbUrlList (adversinc).
  • [Added] Added the urlFilter option to htmlLinkParser (adversinc).

1.5.0

  • [Added] Added the robotsEnabled (default true) option to allow the robots.txt check to be disabled (cbess).

1.4.0

  • [Added] Added the robotsIgnoreServerError option to accept a robots.txt 500 error code as "allow all" rather than "deny all" (default), thanks cbess.

1.3.3

  • [Fix] Updated dependencies, thanks cbess.

1.3.1

  • [Fix] htmlLinkParser should detect links matching the area[href] selector.

1.3.0

  • [Added] Crawler fires the crawledurl event the crawl of a specific URL is complete (whether successful or not).

1.2.0

  • [Added] Crawler fires the urllistcomplete event when the UrlList is permanently empty (compare with urllistempty, which may fire intermittently).

1.1.0

  • [Added] Ability to provide custom options to the request library.

1.0.0

  • [Fixed] Removed warnings from unit tests.
  • [Changed] Updated dependencies.
  • [Changed] Make API stable - release 1.0.0.

0.16.1

  • [Fixed] Treats 410 the same as 404 for robots.txt requests.

0.16.0

  • [Added] Support for gzipContentTypes option to sitemapsParser. Example: gzipContentTypes: 'application/gzip' and gzipContentTypes: ['application/gzip'].

0.15.1

  • [Fixed] Support for multiple "User-agent" lines in robots.txt files

0.15.0

  • [Added] Redis based queue.

0.14.0

  • [Added] Crawler emits redirect, links and httpError events.

### 0.13.1

  • [Fixed] DbUrlList doesn't fetch the existing record from the database unless there was an error.

0.13.0

  • [Added] errorMessage column on urls table that gives more information about, e.g., a handlers error that occurred.

0.12.1

  • [Fixed] Downgrade to cheerio 0.19, to fix a memory leak issue.

0.12.0

  • [Change] Rather than calling content handlers with (body, url), they are now called with a single context argument. This allows you to pass information forwards via handlers. For example, you might cache the cheerio parsing so you don't parse with every content handler.

0.11.0

  • [Added] Event called handlersError is emitted if any of the handlers returns an error.

0.10.4

  • [Fixed] Shortend urlHash field to 40 characters, in case tables are using utf8mb4 collations for strings.

0.10.3

  • [Fixed] URLs are now crawled in a random order. Improved the getNextUrl function of DbUrlList to use a more optimized query.

0.10.2

  • [Fixed] When content handler throws an exception / rejects a Promise, it will be marked as an error. (And scheduled for a retry if using DbUrlList).

0.10.1

  • [Fixed] Request sends Accept-Encoding: gzip, deflate header, so the responses arrive compressed (saving data transfer).

0.10.0

  • [Added] Support for a custom URL filter on the robotsParser function.

0.9.1

  • [Fixed] Performance improvement for sitemaps parser. Very large sitemap previous took 25 seconds, now takes 1-2 seconds.

0.9.0

  • [Added] Support for a custom URL filter on the sitemapsParser function.

0.8.0

  • [Changed] Sitemaps parser now extracts <xhtml:link rel="alternate"> URLs, in addition to the <loc> URLs.

0.7.0

  • [Added] Support for optional insertIfNotExistsBulk method which can insert a large list of URLs into the crawl queue.
  • [Changed] DbUrlList supports the bulk insert method.

0.6.1

  • [Fix] Support sitemaps with content type application/gzip as well as application/x-gzip.

0.6.0

  • [Added] Crawler fires the urllistempty and crawlurl events. It also captures the RangeError event when the URL list is empty.

0.5.0

  • [Changed] htmlLinkParser now also picks up link tags where rel=alternate.

0.4.0

  • [Changed] Supercrawler no longer follows redirects on crawled URLs. Supercrawler will now add a redirected URL to the queue as a separate entry. We still follow redirects for the /robots.txt that is used for checking rules; but not for /robots.txt added to the queue.

0.3.3

  • [Fix] DbUrlList to mark a URL as taken, and ensure it never returns a URL that is being crawled in another concurrent request. This has required a new field called holdDate on the url table

### 0.3.2

  • [Fix] Time-based unit tests made more reliable.

0.3.1

  • [Added] Support for Travis CI.

0.3.0

  • [Added] Content type passed as third argument to all content type handlers.
  • [Added] Sitemaps parser to extract sitemap URLs and urlset URLs.
  • [Changed] Content handlers receive Buffers rather than strings for the first argument.
  • [Fix] Robots.txt checking to work for the first crawled URL. There was a bug that caused robots.txt to be ignored if it wasn't in the cache.

0.2.3

  • [Added] A robots.txt parser that identifies Sitemap: directives.

0.2.2

  • [Fixed] Support for URLs up to 10,000 characters long. This required a new urlHash SHA1 field on the url table, to support the unique index.

0.2.1

  • [Added] Extensive documentation.

0.2.0

  • [Added] Status code is updated in the queue for successfully crawled pages (HTTP code < 400).
  • [Added] A new error type error.RequestError for all errors that occur when requesting a page.
  • [Added] DbUrlList queue object that stores URLs in a SQL database. Includes exponetial backoff retry logic.
  • [Changed] Interface to DbUrlList and FifoUrlList is now via methods insertIfNotExists, upsert and getNextUrl. Previously, it was just insert (which also updated) and upsert, but we need a way to differentiate between discovered URLs which should not update the crawl state.

0.1.0

  • [Added] Crawler object, supporting rate limiting, concurrent requests limiting, robots.txt caching.
  • [Added] FifoUrlList object, a first-in, first-out in-memory list of URLs to be crawled.
  • [Added] Url object, representing a URL in the crawl queue.
  • [Added] htmlLinkParser, a function to extract links from crawled HTML documents.