WebSPHINX:
A Personal, Customizable Web Crawler

Contents

About WebSPHINX

WebSPHINX ( Website-Specific Processors for HTML INformation eXtraction) is a Java class library and interactive development environment for web crawlers. A web crawler (also called a robot or spider) is a program that browses and processes Web pages automatically.

WebSPHINX consists of two parts: the Crawler Workbench and the WebSPHINX class library.

Crawler Workbench

The Crawler Workbench is a graphical user interface that lets you configure and control a customizable web crawler. Using the Crawler Workbench, you can:

WebSPHINX class library

The WebSPHINX class library provides support for writing web crawlers in Java. The class library offers a number of features:

Download

First, you need Java 1.2 or later installed on your computer. If you're not sure, try running java -version. If you need to install Java on Windows, Linux, or Solaris, go directly to Sun; for other platforms, consult the list of Java ports.

If your computer has AFS access, run java -jar /afs/cs.cmu.edu/user/rcm/www/websphinx/websphinx.jar

If you don't have AFS, you'll need to download this JAR file:

websphinx.jar
and then run java -jar websphinx.jar

The Crawler Workbench will appear in a new window.

Examples

Here are some things to try in the Workbench.

Visualize part of the Web as a graph
This crawler retrieves the WebSPHINX pages you've been reading and renders them as a graph of pages and links.

Crawl
 the subtree
URL
 http://www.cs.cmu.edu/~rcm/websphinx
Action
 none


Visualize part of the Web as a graph

 
Save pages to disk
This crawler retrieves the SCS Technical Reports index, which consists of about 30 pages, and saves it to a directory on your local disk.

Crawl
 the subtree
URL
 http://reports-archive.adm.cs.cmu.edu/cs.html
Action
 save
Directory
 ./scs-techreports


Save pages to disk



Concatenate pages for printing
This crawler concatenates all the pages in Bob Harper's Introduction to SML into a single massive page, suitable for printing.

Crawl
 the subtree
URL
 http://www.cs.cmu.edu/~rwh/introsml/
Action
 concatenate
File
 ./intro-sml.html

Concatenate pages for printing



Extract images from a set of pages
This crawler surfs over a few pages of Dr. Fun logos and generates a new page containing all the logos found.

Crawl
 the subtree
URL
 http://sunsite.unc.edu/Dave/logos.htm
Action
 extract
Pattern
 <a>(?{logo}<img>)<p>(?{caption})</a>
File ./dr-fun.html


Extract images from a set of pages

Frequently Asked Questions

Who is WebSPHINX intended for?
WebSPHINX is designed for advanced web users and Java programmers who want to crawl over a small part of the web (such as a single web site) automatically.
Can I get the source code?
Yes, WebSPHINX is open source, covered by an Apache-style license (see Source Code).
Where can I find documentation or examples for programming WebSPHINX crawlers in Java?
Some examples can be found in our WWW7 paper (see Publications), and JavaDoc API documentation for the class library is also available (see Documentation).
Can I use WebSPHINX to crawl the entire Web, like search engines do?
WebSPHINX isn't designed for enormous crawls like that. Search engines typically use distributed crawlers running on farms of PCs with a fat network pipe and a distributed filesystem or database for managing the crawl frontier and storing page data.  WebSPHINX is intended more for personal use, to crawl perhaps a hundred or a thousand web pages.  If you want to use WebSPHINX for large crawls, you should definitely read the next question about memory usage.
My WebSPHINX crawler is running out of RAM.  How can I control its memory use?
By default, WebSPHINX retains all the pages and links that it has crawled until you clear the crawler.  This can use up memory quickly, especially if you're crawling more than a few hundred pages.  Here are some tricks for changing the defaults and keeping memory under control.  (Note that these tricks only apply when you're writing your own crawler in Java, not when you're using the Crawler Workbench.)
  1. Use Page.discardContent() to throw away (stop referencing) a page's content when you're done with it, so that it can be reclaimed by the garbage collector. This method preserves the page's array of outgoing Links, however, so you'll still have the crawl graph if you need it.
  2. Disconnect the crawl graph entirely by breaking references between links and pages, so that every Page and Link object can be reclaimed once the crawler has finished visiting this. To do this, call page.getOrigin().setPage(null) whenever you're done processing a page.
  3. Another kind of memory bloat is caused by the implementation of   java.lang.String.substring().  Calling s.substring() does not make a copy of the characters in the substring.  Instead, it returns a special String that points to the substring within s. As a result, if you use substring() to grab a short part of a 10KB web page, you're keeping a reference to the whole 10KB.  If you need to call substring() on page content and want to keep the substring around but not the original page, you should make a copy of the substring using new String (s.toCharArray()).
  4. If all else fails, and you're using the Sun JDK, you can use the -mx option (called -Xmx in recent JDKs) to increase the maximum limit of heap memory.
The WWW7 paper mentions a "CategoryClassifier", but I can't find it in the source code.  Where can I get it?
The CategoryClassifier was part of an earlier web-crawling system, SPHINX, developed at Compaq SRC.  The original SPHINX code belongs to Compaq SRC and was never released.  WebSPHINX is an open-source reimplementation of the SPHINX interface.  CategoryClassifier was not part of this reimplementation because CategoryClassifier depended on some other software that belongs to SRC.
The search engine classifiers don't work.
Most of the search engine classifiers were written in 1998.  Search engines have changed the format of their results many times since then, so the classifiers are out of date.
My web crawler needs to use a web proxy, user authentication, cookies, a special user-agent, etc. What do I do?
WebSPHINX uses the built-in Java classes URL and URLConnection to fetch web pages.  If you're running the Crawler Workbench inside a browser, that means your crawler uses the proxy, authentication, cookies, and user-agent of the browser, so if you can visit the site manually, then you can crawl it.  If you're running your crawler from the command line, however, you'll have to configure Java to set up your proxy, authentication, user-agents, and so forth.

Source Code

The crawler library is open source, licensed under an Apache-style license. The latest release is version 0.5, released on July 8, 2002. See the change history to find out what's new.

Download the source code here:

websphinx.zip

WebSPHINX is Copyright © 1998-2002 - Carnegie Mellon University.


Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.

2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

THIS SOFTWARE IS PROVIDED BY CARNEGIE MELLON UNIVERSITY ``AS IS'' AND  ANY EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL CARNEGIE MELLON UNIVERSITY NOR ITS EMPLOYEES BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

This product includes software developed by the Apache Software Foundation (http://www.apache.org/).  In particular, WebSPHINX includes the Apache jakarta-regexp regular expression library, version 1.2. The (unmodified) source code for this library is included in the WebSPHINX source code. Redistribution is allowed under the terms of the Apache Public License.

Documentation

Online API documentation (generated by Javadoc)

For offline access to the API documentation, download:

websphinx-doc.zip

Publications

WebSPHINX was inspired by an earlier system, SPHINX, developed over summer 1997 at Compaq System Research Center (then part of Digital). For more information about SPHINX, see the paper:

Robert C. Miller and Krishna Bharat. SPHINX: A Framework for Creating Personal, Site-Specific Web Crawlers. In Proceedings of WWW7, Brisbane Australia, April 1998.

WebSPHINX is a ground-up reimplementation of the SPHINX interface, but some features described in the paper were omitted in the reimplementation (namely, the category classifier).

Related Links

Some other projects are using WebSPHINX:

A few Java toolkits worthy of mention:

There are several crawling toolkits with goals similar to WebSPHINX.

Several web sites, and even a few books, describe the crawlers and robots that already roam the Web:

Crawler writers should be aware of robot ethics:

Send comments or questions to Rob Miller ( rcm@cs.cmu.edu)