WWW::Search and AutoSearch and WebSearch ======================================== WHAT IS NEW IN WWW::Search 2.09? (2000-02-08) ---------------------------------------------- overview: * new module WWW::Search::Test * new methods maintainer(), gui_query() * bug fixes for several backends (as usual) * some backends removed to their own modules For details, see the file ChangeLog and/or the pod of each module. WHAT IS WWW::Search? -------------------- WWW::Search is a collection of Perl modules which provide an API to WWW (and similar) search engines. Currently WWW::Search includes backends for variations of AltaVista, Dejanews, Infoseek, Lycos, Magellan, and WebCrawler, among others. Backends for some engines can be obtained separately, such as Excite, HotBot, and Yahoo. This distribution includes two applications built from this library: AutoSearch (a program to automate tracking of search results over time), and WebSearch, a small demonstration program to drive the library. WWW::Search does NOT try to emulate the default search that you would get with each search engine's GUI. WWW::Search performs the search in a way that is efficient and convenient for text processing. This might include getting "text-only" pages, making "OR" the default query term operator instead of "AND", ungrouping same-site results, making sure descriptions are turned on, and increasing the number of hits per page, among other tricks. Hopefully, the documentation for each backend will tell you about the more important tricks being used. Because WWW::Search depends on parsing the HTML output of web search engines it will fail if the search engine operators change their format (an unfortunately frequent occurrence). WWW::Search includes a test suite for many backends which verifies that they are functioning correctly. The test suite can be run by typing 'make test_parsing'; see under INSTALLATION below for details. As of the day of the release the current backend status is: AltaVista working AltaVista::AdvancedNews not working AltaVista::AdvancedWeb not working AltaVista::Careers working? (not in test suite) AltaVista::Intranet working AltaVista::News not working AltaVista::Web working AOL::Classifieds::Employment working (not in test suite) Crawler partially working? Deja/Dejanews working Dice working? (not in test suite) Euroseek working? (not in test suite) Excite::News working ExciteForWebServers not working Fireball working? (not in test suite) FolioViews working Google working Gopher not working? (not in test suite) GoTo working HeadHunter working? (not in test suite) HotFiles working Infoseek working Infoseek::Companies working Infoseek::Email not working Infoseek::News working Infoseek::Web working Livelink not working? (not in test suite) LookSmart working Lycos working Lycos::Pages defunct Lycos::Sites defunct Magellan working MetaCrawler working Metapedia working? (not in test suite) Monster working? (not in test suite) MSIndexServer not working? NetFind working NorthernLight working Null working OpenDirectory working PLweb not working Profusion defunct Search97 not working SFgate working Simple not working? (not in test suite) Snap working Verity not working (not in test suite) WebCrawler working Yahoo::Classifieds::Employment working? (not in test suite) ZDNet working ''Partially working'' indicates that some tests passed and some failed. The following backends are now registered at CPAN independently (not included with the WWW::Search release): Excite HotBot Yahoo WHAT IS AutoSearch? ------------------- WWW::Search's primary client is AutoSearch. AutoSearch performs a web-based search and puts the results set in a web page. It periodically updates this web page, indicating how the search changes over time. Sample output from AutoSearch can be found at . Output format is configurable. See the man page for AutoSearch details, or the DEMONSTRATION section below for quick-start instructions. REQUIREMENTS ------------ WWW::Search requires Perl5, the libwww-perl module suite, the URI module, and the HTML::Parser module. Some of the "not working" modules require the HTML::TreeBuilder module (so you can ignore warnings about TreeBuilder during the build). For information on Perl5, see . For all the modules, see to find a CPAN site near you. At the time of this release, the primary WWW::Search development and testing is under perl version 5.005_03 on Sun UltraSparc Solaris 7 and under ActiveState perl build 522 on Windows NT 4.0 service pack 6. WWW::Search has also been built and tested successfully on Win98J (that's Japanese) with ActiveState perl build 517. If you have successfully built and tested WWW::Search on any other (obscure) platform / version combination, please let me know! MartinThurn@iname.com AVAILABILITY ------------ The latest version of WWW::Search should always be available on CPAN. Here is the best URL for finding it: http://www.perl.com/CPAN-local/modules/by-module/WWW INSTALLATION ------------ In order to use this package you will need Perl version 5.002 or better. It is hightly recommended that you use CPAN.pm to install WWW::Search. It will automatically install all the prerequisite modules and put everything in the right places. On a unix system, just type perl -MCPAN -e 'install WWW::Search'. Otherwise, you can install WWW::Search as you would any perl module library, by running these commands in the WWW-Search-x.xx directory after unpacking the archive (and after installing all the prerequisite modules): perl Makefile.PL make test make install On Win32, maintenance and testing is done with Microsoft's nmake.exe; use 'nmake' instead of 'make' in the above sequence of commands. When you do `perl makefile.pl` on Win32, you might get warnings that a whole bunch of 'zero*.out' files are missing. This seems to be a bug in some versions of WinZip which refuse to extract empty files from the archive. You can ignore these warnings. If you want to install a private copy of WWW::Search in your home directory, then you should do the installation with something like these commands: perl Makefile.PL INSTALLDIRS=perl PREFIX=/my/perl/lib make test make pure_perl_install UNINST=1 Don't forget to add /my/perl/lib to your PERL5LIB environment variable (or use lib '/my/perl/lib'; or unshift @INC, '/my/perl/lib')! TESTING ------- The "make test_parsing" command compares expected output from WWW::Search with actual output. You can give arguments to the test_parsing program by using the TEST_ARGS macro. For example, the following command only runs the external queries for Dejanews: make test_parsing TEST_ARGS='-e Dejanews -x' To see all the available options, do this: make test_parsing TEST_ARGS='-help' The "test_parsing" utility detects two kinds of errors: - internal parsing: First it checks to make sure that your system computes the same results as my system based on some saved Web queries. This test should always pass for working backends; if it doesn't, send me mail. - external queries: Second, it makes real queries against the search engines and compares them with some saved results. External queries can fail for several reasons: - new pages have been added which match the test queries, or matching pages have been deleted, causing the page count to go too far out of whack from the expected number (not necessarily a bad thing) - changes in the web search engine output which break WWW::Search's parsers, usually resulting in no URLs being returned (a bad thing) If the external tests fail, please either investigate the error or send a description of the problem, a list of your operating system and all relevant perl version number, and the relevant output of "make test_parsing" to the maintainer of the backend for the search engine that fails. DISCUSSION, BUG REPORTS, AND IMPROVEMENTS ----------------------------------------- Feedback about WWW::Search is encouraged. If you're using it for a neat application, please let us know. If you'd like to (or have already) implement and publish a new backend for WWW::Search, let us know so we don't duplicate work. Feedback, bug reports, fixes, and new backends should be sent to Martin Thurn . When sending e-mail, please please put [WWW::Search] at the beginning of the subject line (or risk me losing the message in the pile). There is a mailing list for WWW::Search discussion. To subscribe, send "subscribe info-www-search" as the body of a message to . If you use WWW::Search at all, you should subscribe to the mailing list. Bug fixes are posted there as soon as they're fixed. Back-end-related bug reports ("search engine ABC doesn't work") should be sent to the author of the backend (backend authors are identified in the corresponding man page and in the output of `make test_parsing`). General bugs should be reported to . When submitting a bug report or request for help, please remember to include: - your operating system name and version - your version of perl - your version of WWW::Search - your version of the backend - the code you ran to produce the error (PLEASE cut-and-paste, do not just summarize!) - sample output showing the error (PLEASE cut-and-paste, do not just summarize!) DEMONSTRATION ------------- After installing the distribution, try: WebSearch '"Your Name Here"' or, if you are on Win32: WebSearch "\"Your Name Here\"" to see who's talking about you on the web. Then (in your web page directory), try: cd /path/to/your/web/pages AutoSearch -n me_on_the_web -s '"Your Name Here"' me or, if you are on Win32: cd /path/to/your/web/pages AutoSearch -n me_on_the_web -s "\"Your Name Here\"" me and the web page /path/to/your/web/pages/me/index.html will be created summarizing this information. If you are on UNIX you can add 0 3 * * 1 AutoSearch /path/to/your/web/pages/me to your crontab to update this search every week at 3:00 Monday morning, for example. DOCUMENTATION ------------- See `perldoc WWW::Search` after installation for an overview of the library. POD-style documentation is also included in all modules and programs, so you can do `perldoc WebSearch` and `perldoc AutoSearch` and `perldoc WWW::Search::AltaVista` after installation. FUTURE PLANS ------------ Some ideas: - use LWP::ParallelUA to speed up multiple backend search requests (I'm trying to decide what the API interface will look like; please send suggestions). Contact - updates to each backend that will force WWW::Search to perform the same search as the engine's web GUI (I'm looking for contributions of the precise arguments that will produce such a search for each engine; i.e. the hash that should be passed as the second argument to native_query). Contact - application-level proxy support (I'm looking for a contribution here from someone who uses/needs proxy support and can test it). Contact - more widespread use of result tags description, date, size, etc. across all backends - a freeze/restore interface to suspend and resume in-progress queries - more backends Contributions are always welcome. Send me e-mail if you plan a new backend and to discuss architectural changes (to avoid duplicating work). Contact SUPPORT AND CREDITS ------------------- The WWW::Search architecture is by John Heidemann with feedback from the other contributors. NOTE: This list is no longer updated; consult the on-line documentation and/or the output of `make test` to find out who is currently maintaining each component. PLATFORM SUPPORT: Unix John Heidemann Windows Jim Smyser (see ) APPLICATIONS: WebSearch John Heidemann AutoSearch William Scheding BACK-ENDS: AltaVista John Heidemann Dejanews Cesare Feroldi de Rosa and Martin Thurn Crawler Andreas Borchert Excite GLen Pringle and Martin Thurn ExciteForWebServers Paul Lindner Fireball Andreas Borchert FolioViews Paul Lindner Gopher Paul Lindner HotBot William Scheding and Martin Thurn HotFiles Jim Smyser Infoseek Cesare Feroldi de Rosa and Martin Thurn Livelink Paul Lindner Lycos William Scheding and John Heidemann, Martin Thurn Magellan Martin Thurn MSIndexServer Paul Lindner NorthernLight Jim Smyser Null Paul Lindner OpenDirectory Jim Smyser PLWeb Paul Lindner Profusion Jim Smyser Search97 Paul Lindner SFgate Paul Lindner Simple Paul Lindner Snap Jim Smyser Verity Paul Lindner WebCrawler Martin Thurn Yahoo William Scheding and Martin Thurn ZDNet Jim Smyser AutoSearch is based on an earlier implementation by Kedar Jog with advice from Joe Touch . Bugs and extensions (to the software and documentation) have been identified by William Scheding , T. V. Raman (proxy support), C. Feroldi , Larry Virden , Paul Lindner , Guy Decoux , R Chandrasekar (Mickey) , Martin Thurn , Chris Nandor , Martin Valldeby , Jim Smyser , Darren Stalder , Neil Bowers , Ave Wrigley , Andreas Borchert , Jim Smyser . Bugs have reported by Joseph McDonald , Juan Jose Amor , Bowen Dwelle , Vassilis Papadimos , Vidyut Luther , Chris P. Acantilado . COPYRIGHT --------- Copyright (c) 1996 University of Southern California. All rights reserved. Redistribution and use in source and binary forms are permitted provided that the above copyright notice and this paragraph are duplicated in all such forms and that any documentation, advertising materials, and other materials related to such distribution and use acknowledge that the software was developed by the University of Southern California, Information Sciences Institute. The name of the University may not be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED "AS IS" AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. Portions of this README are derived from the README for libwww-perl.