Provided by: libwww-search-perl_2.50.80-1_all bug

NAME

       WWW::Search - Virtual base class for WWW searches

SYNOPSIS

         use WWW::Search;
         my $sEngine = 'AltaVista';
         my $oSearch = new WWW::Search($sEngine);

DESCRIPTION

       This class is the parent for all access methods supported by the "WWW::Search" library.
       This library implements a Perl API to web-based search engines.

       See README for a list of search engines currently supported, and for a lot of interesting
       high-level information about this distribution.

       Search results can be limited, and there is a pause between each request to avoid
       overloading either the client or the server.

   Sample program
       Here is a sample program:

         my $sQuery = 'Columbus Ohio sushi restaurant';
         my $oSearch = new WWW::Search('AltaVista');
         $oSearch->native_query(WWW::Search::escape_query($sQuery));
         $oSearch->login($sUser, $sPassword);
         while (my $oResult = $oSearch->next_result())
           {
           print $oResult->url, "\n";
           } # while
         $oSearch->logout;

       Results are objects of type "WWW::SearchResult" (see WWW::SearchResult for details).  Note
       that different backends support different result fields.  All backends are required to
       support title and url.

SEE ALSO

       For specific search engines, see WWW::Search::TheEngineName (replacing TheEngineName with
       a particular search engine).

       For details about the results of a search, see WWW::SearchResult.

METHODS AND FUNCTIONS FOR SEARCHERS

       new To create a new WWW::Search, call

               $oSearch = new WWW::Search('SearchEngineName');

           where SearchEngineName is replaced with a particular search engine.  For example:

               $oSearch = new WWW::Search('Yahoo');

           If no search engine is specified, a default (currently 'Null::Empty') will be chosen
           for you.

       version
           Returns the value of the $VERSION variable of the backend engine, or
           $WWW::Search::VERSION if the backend does not contain $VERSION.

       maintainer
           Returns the value of the $MAINTAINER variable of the backend engine, or
           $WWW::Search::MAINTAINER if the backend does not contain $MAINTAINER.

       installed_engines
           Returns a list of the names of all installed backends.  We can not tell if they are
           up-to-date or working, though.

             use WWW::Search;
             my @asEngines = sort &WWW::Search::installed_engines();
             local $" = ', ';
             print (" + These WWW::Search backends are installed: @asEngines\n");
             # Choose a backend at random (yes, this is rather silly):
             my $oSearch = WWW::Search->new($asEngines[rand(scalar(@asEngines))]);

       native_query
           Specify a query (and optional options) to the current search object.  Previous query
           (if any) and its cached results (if any) will be thrown away.  The option values and
           the query must be escaped; call WWW::Search::escape_query() to escape a string.  The
           search process is not actually begun until "results()" or "next_result()" is called
           (lazy!), so native_query does not return anything.

           Example:

             $oSearch->native_query('search-engine-specific+escaped+query+string',
                                   { option1 => 'able', option2 => 'baker' } );

           The hash of options following the query string is optional.  The query string is
           backend-specific.  There are two kinds of options: options specific to the backend,
           and generic options applicable to multiple backends.

           Generic options all begin with 'search_'.  Currently a few are supported:

           search_url
               Specifies the base URL for the search engine.

           search_debug
               Enables backend debugging.  The default is 0 (no debugging).

           search_parse_debug
               Enables backend parser debugging.  The default is 0 (no debugging).

           search_to_file FILE
               Causes the search results to be saved in a set of files prefixed by FILE.  (Used
               internally by the test-suite, not intended for general use.)

           search_from_file FILE
               Reads a search from a set of files prefixed by FILE.  (Used internally by the
               test-suite, not intended for general use.)

           Some backends may not implement these generic options, but any which do implement them
           must provide these semantics.

           Backend-specific options are described in the documentation for each backend.  In most
           cases the options and their values are packed together to create the query portion of
           the final URL.

           Details about how the search string and option hash are interpreted might be found in
           the search-engine-specific manual pages (WWW::Search::SearchEngineName).

       gui_query
           Specify a query to the current search object; the query will be performed with the
           engine's default options, as if it were typed by a user in a browser window.

           Same arguments as "native_query()" above.

           Currently, this feature is supported by only a few backends; consult the documentation
           for each backend to see if it is implemented.

       cookie_jar
           Call this method (anytime before asking for results) if you want to communicate cookie
           data with the search engine.  Takes one argument, either a filename or an
           HTTP::Cookies object.  If you give a filename, WWW::Search will attempt to read/store
           cookies there (by passing the filename to HTTP::Cookies::new).

             $oSearch->cookie_jar('/tmp/my_cookies');

           If you give an HTTP::Cookies object, it is up to you to save the cookies if/when you
           wish.

             use HTTP::Cookies;
             my $oJar = HTTP::Cookies->new(...);
             $oSearch->cookie_jar($oJar);

           If you pass in no arguments, the cookie jar (if any) is returned.

             my $oJar = $oSearch->cookie_jar;
             unless (ref $oJar) { print "No jar" };

       date_from
           Set/get the start date for limiting the query by a date range.  See the documentation
           for each backend to find out if date ranges are supported.

       date_to
           Set/get the end date for limiting the query by a date range.  See the documentation
           for each backend to find out if date ranges are supported.

       env_proxy
           Enable loading proxy settings from environment variables.  The proxy URL will be read
           from $ENV{http_proxy}.  The username for authentication will be read from
           $ENV{http_proxy_user}.  The password for authentication will be read from
           $ENV{http_proxy_pwd}.

           If you don't want to put passwords in the environment, one solution would be to
           subclass LWP::UserAgent and use $ENV{WWW_SEARCH_USERAGENT} instead (see user_agent
           below).

           env_proxy() must be called before the first retrieval is attempted.

           Example:

             $ENV{http_proxy     } = 'http://my.proxy.com:80';
             $ENV{http_proxy_user} = 'bugsbun';
             $ENV{http_proxy_pwd } = 'c4rr0t5';
             $oSearch->env_proxy('yes');  # Turn on with any true value
             ...
             $oSearch->env_proxy(0);  # Turn off with zero
             ...
             if ($oSearch->env_proxy)  # Test

       http_proxy
           Set up an HTTP proxy (for connections from behind a firewall).

           Takes the same arguments as LWP::UserAgent::proxy().

           This routine should be called before calling any of the result functions (any method
           with "result" in its name).

           Example:

             # Turn on and set address:
             $oSearch->http_proxy(['http','ftp'] => 'http://proxy:8080');
             # Turn off:
             $oSearch->http_proxy('');

       http_proxy_user, http_proxy_pwd
           Set/get HTTP proxy authentication data.

           These routines set/get username and password used in proxy authentication.
           Authentication is attempted only if all three items (proxy URL, username and password)
           have been set.

           Example:

               $oSearch->http_proxy_user("myuser");
               $oSearch->http_proxy_pwd("mypassword");
               $oSearch->http_proxy_user(undef);   # Example for no authentication

               $username = $oSearch->http_proxy_user();

       maximum_to_retrieve
           Set the maximum number of hits to return.  Queries resulting in more than this many
           hits will return the first hits, up to this limit.  Although this specifies a maximum
           limit, search engines might return less than this number.

           Defaults to 500.

           Example:
               $max = $oSearch->maximum_to_retrieve(100);

           You can also spell this method "maximum_to_return".

       maximum_to_return
           Synonym for maximum_to_retrieve

       timeout
           The maximum length of time any portion of the query should take, in seconds.

           Defaults to 60.

           Example:
               $oSearch->timeout(120);

       login
           Backends which need to login to the search engine should implement this function.
           Takes two arguments, user and password.  Return nonzero if login was successful.
           Return undef or 0 if login failed.

       logout
           Backends which need to logout from the search engine should implement this function.

       approximate_result_count
           Some backends indicate how many results they have found.  Typically this is an
           approximate value.

       approximate_hit_count
           This is an alias for approximate_result_count().

       results
           Return all the results of a query as an array of WWW::SearchResult objects.

           Note: This might take a while, because a web backend will keep asking the search
           engine for "next page of results" over and over until there are no more next pages,
           and THEN return from this function.

           If an error occurs at any time during query processing, it will be indicated in the
           response().

           Example:

               @results = $oSearch->results();
               # Go have a cup of coffee while the previous line executes...
               foreach $oResult (@results)
                 {
                 print $oResult->url(), "\n";
                 } # foreach

       next_result
           Call this method repeatedly to return each result of a query as a WWW::SearchResult
           object.  Example:

               while ($oResult = $oSearch->next_result())
                 {
                 print $oResult->url(), "\n";
                 } # while

           When there are no more results, or if an error occurs, next_result() will return
           undef.

           If an error occurs at any time during query processing, it will be indicated in the
           response().

       seek_result($offset)
           Set which result should be returned next time "next_result()" is called.  Results are
           zero-indexed.

           The only guaranteed valid offset is 0, which will replay the results from the
           beginning.  In particular, seeking past the end of the current cached results probably
           will not do what you might think it should.

           Results are cached, so this does not re-issue the query or cause IO (unless you go off
           the end of the results).  To re-do the query, create a new search object.

           Example:

               $oSearch->seek_result(0);

       response
           Returns an HTTP::Response object which resulted from the most-recently-sent query.
           Errors can be detected like this:

               if (! $oSearch->response->is_success)
                 {
                 print STDERR "Error:  " . $oSearch->response->as_string() . "\n";
                 } # if

           Note to backend authors: even if the backend does not involve the web, it should
           return an HTTP::Response object.

       submit
           This method can be used to submit URLs to the search engines for indexing.  Consult
           the documentation for each backend to find out if it is implemented there, and if so
           what the arguments are.

           Returns an HTTP::Response object describing the result of the submission request.
           Consult the documentation for each backend to find out the meaning of the response.

       opaque
           This function provides an application a place to store one opaque data element (or
           many, via a Perl reference).  This facility is useful to (for example), maintain
           client-specific information in each active query when you have multiple concurrent
           queries.

       escape_query
           Escape a query.  Before queries are sent to the internet, special characters must be
           escaped so that a proper URL can be formed.  This is like escaping a URL, but all non-
           alphanumeric characters are escaped and and spaces are converted to "+"s.

           Example:

               $escaped = WWW::Search::escape_query('+hi +mom');
               # $escaped is now '%2Bhi+%2Bmom'

           See also "unescape_query()".  NOTE that this is not a method, it is a plain function.

       unescape_query
           Unescape a query.  See "escape_query()" for details.

           Example:

               $unescaped = WWW::Search::unescape_query('%22hi+mom%22');
               # $unescaped eq q{"hi mom"}

           NOTE that this is not a method, it is a plain function.

       strip_tags
           Given a string, returns a copy of that string with HTML tags removed.  This should be
           used by each backend as they insert the title and description values into the search
           results objects.

           NOTE that this is not a method, it is a plain function.

       is_http_proxy
           Returns true if proxy information is available.

METHODS AND FUNCTIONS FOR BACKEND PROGRAMMERS

       reset_search
           Resets internal data structures to start over with a new search (on the same engine).

       is_http_proxy_auth_data
           Returns true if all authentication data (proxy URL, username, and password) are
           available.

       agent_name($sName)
           If your search engine rejects certain browser, you can trick it into thinking you're
           any browser type you want.  See below under user_agent().

       agent_email($sName)
       user_agent($NON_ROBOT)
           This internal routine creates a user-agent for derived classes that query the web.  If
           any non-false argument is given, a normal LWP::UserAgent (rather than a LWP::RobotUA)
           is used.

           Returns the user-agent object.

           If a backend needs the low-level LWP::UserAgent or LWP::RobotUA to have a particular
           name, $oSearch->agent_name() and possibly $oSearch->agent_email() should be called to
           set the desired values *before* calling $oSearch->user_agent().

           If the environment variable WWW_SEARCH_USERAGENT has a value, it will be used as the
           class for a new user agent object.  This class should be a subclass of LWP::UserAgent.
           For example,

             $ENV{WWW_SEARCH_USERAGENT} = 'My::Own::UserAgent';
             # If this env.var. has no value,
             # LWP::UserAgent or LWP::RobotUA will be used.
             $oSearch = new WWW::Search('MyBackend');
             $oSearch->agent_name('MySpider');
             if ($iBackendWebsiteRequiresNonRobot)
               {
               $oSearch->user_agent('non-robot');
               }
             else
               {
               $oSearch->agent_email('me@here.com');
               $oSearch->user_agent();
               }

           Backends should use robot-style user-agents whenever possible.

       http_referer
           Get / set the value of the HTTP_REFERER variable for this search object.  Some search
           engines might only accept requests that originated at some specific previous page.
           This method lets backend authors "fake" the previous page.  Call this method before
           calling http_request.

             $oSearch->http_referer('http://prev.engine.com/wherever/setup.html');
             $oResponse = $oSearch->http_request('GET', $url);

       http_method
           Get / set the method to be used for the HTTP request.  Must be either 'GET' or 'POST'.
           Call this method before calling http_request.  (Normally you would set this during
           _native_setup_search().)  The default is 'GET'.

             $oSearch->http_method('POST');

       http_request($method, $url)
           Submit the HTTP request to the world, and return the response.  Similar to
           LWP::UserAgent::request.  Handles cookies, follows redirects, etc.  Requires that
           http_referer already be set up, if needed.

       next_url
           Get or set the URL for the next backend request.  This can be used to save the
           WWW::Search state between sessions (e.g. if you are showing pages of results to the
           user in a web browser).  Before closing down a session, save the value of next_url:

             ...
             $oSearch->maximum_to_return(10);
             while ($oSearch->next_result) { ... }
             my $urlSave = $oSearch->next_url;

           Then, when you start up the next session (e.g. after the user clicks your "next"
           button), restore this value before calling for the results:

             $oSearch->native_query(...);
             $oSearch->next_url($urlSave);
             $oSearch->maximum_to_return(20);
             while ($oSearch->next_result) { ... }

           WARNING: It is entirely up to you to keep your interface in sync with the number of
           hits per page being returned from the backend.  And, we make no guarantees whether
           this method will work for any given backend.  (Their caching scheme might not enable
           you to jump into the middle of a list of search results, for example.)

       split_lines
           This internal routine splits data (typically the result of the web page retrieval)
           into lines in a way that is OS independent.  If the first argument is a reference to
           an array, that array is taken to be a list of possible delimiters for this split.  For
           example, Yahoo.pm uses <p> and <dd><li> as "line" delimiters for convenience.

       generic_option
           This internal routine checks if an option is generic or backend specific.  Currently
           all generic options begin with 'search_'.  This routine is not a method.

       _native_setup_search
           Do some backend-specific initialization.  It will be called with the same arguments as
           native_query().

       setup_search
           This internal routine does generic Search setup.  It calls "_native_setup_search()" to
           do backend-specific setup.

       need_to_delay
           A backend should override this method in order to dictate whether user_agent_delay()
           needs to be called before the next HTTP request is sent.  Return any perlish true or
           zero value.

       user_agent_delay
           According to what need_to_delay() returns, user_agent_delay() will be called between
           requests to remote servers to avoid overloading them with many back-to-back requests.

       absurl
           An internal routine to convert a relative URL into a absolute URL.  It takes two
           arguments, the 'base' url (usually the search engine CGI URL) and the URL to be
           converted.  Returns a URI object.

       retrieve_some
           An internal routine to interface with "_native_retrieve_some()".  Checks for overflow.

       _native_retrieve_some
           Fetch the next page of results from the web engine, parse the results, and prepare for
           the next page of results.

           If a backend defines this method, it is in total control of the WWW fetch, parsing,
           and preparing for the next page of results.  See the WWW::Search::AltaVista module for
           example usage of the _native_retrieve_some method.

           An easier way to achieve this in a backend is to inherit _native_retrieve_some from
           WWW::Search, and do only the HTML parsing.  Simply define a method _parse_tree which
           takes one argument, an HTML::TreeBuilder object, and returns an integer, the number of
           results found on this page.  See the WWW::Search::Yahoo module for example usage of
           the _parse_tree method.

           A backend should, in general, define either _parse_tree() or _native_retrieve_some(),
           but not both.

           Additional features of the default _native_retrieve_some method:

           Sets $self->{_prev_url} to the URL of the page just retrieved.

           Calls $self->preprocess_results_page() on the raw HTML of the page.

           Then, parses the page with an HTML::TreeBuilder object and passes that populated
           object to $self->_parse_tree().

           Additional notes on using the _parse_tree method:

           The built-in HTML::TreeBuilder object used to parse the page has store_comments turned
           ON.  If a backend needs to use a subclassed or modified HTML::TreeBuilder object, the
           backend should set $self->{'_treebuilder'} to that object before any results are
           retrieved.  The best place to do this is at the end of _native_setup_search.

             my $oTree = new myTreeBuilder;
             $oTree->store_pis(1);  # for example
             $self->{'_treebuilder'} = $oTree;

           When _parse_tree() is called, the $self->next_url is cleared.  During parsing, the
           backend should set $self->next_url to the appropriate URL for the next page of
           results.  (If _parse_tree() does not set the value, the search will end after parsing
           this page of results.)

           When _parse_tree() is called, the URL for the page being parsed can be found in
           $self->{_prev_url}.

       result_as_HTML
           Given a WWW::SearchResult object, formats it human-readable with HTML.

       preprocess_results_page
           A filter on the raw HTML of the results page.  This allows the backend to alter the
           HTML before it is parsed, such as to correct for known problems, HTML that can not be
           parsed correctly, etc.

           Takes one argument, a string (the HTML webpage); returns one string (the same HTML,
           modified).

           This method is called from within _native_retrieve_some (above) before the HTML of the
           page is parsed.

           See the WWW::Search::Ebay distribution 2.07 or higher for example usage.

       test_cases (DEPRECATED)
           Deprecated.

           Returns the value of the $TEST_CASES variable of the backend engine.

       hash_to_cgi_string (DEPRECATED)
           Given a reference to a hash of string => string, constructs a CGI parameter string
           that looks like 'key1=value1&key2=value2'.

           If the value is undef, the key will not be added to the string.

           At one time, for testing purposes, we asked backends to use this function rather than
           piecing the URL together by hand, to ensure that URLs are identical across platforms
           and software versions.  But this is no longer necessary.

           Example:

               $self->{_options} = {
                                    'opt3' => 'val3',
                                    'search_url' => 'http://www.deja.com/dnquery.xp',
                                    'opt1' => 'val1',
                                    'QRY' => $native_query,
                                    'opt2' => 'val2',
                                   };
               $self->{_next_url} = $self->{_options}{'search_url'} .'?'.
                                    $self->hash_to_cgi_string($self->{_options});

IMPLEMENTING NEW BACKENDS

       "WWW::Search" supports backends to separate search engines.  Each backend is implemented
       as a subclass of "WWW::Search".  WWW::Search::Yahoo provides a good sample backend.

       A backend must have the routine "_native_setup_search()".  A backend must have the routine
       "_native_retrieve_some()" or "_parse_tree()".

       "_native_setup_search()" is invoked before the search.  It is passed a single argument:
       the escaped, native version of the query.

       "_native_retrieve_some()" is the core of a backend.  It will be called periodically to
       fetch URLs.  It should retrieve several hits from the search service and add them to the
       cache.  It should return the number of hits found, or undef when there are no more hits.

       Internally, "_native_retrieve_some()" typically sends an HTTP request to the search
       service, parses the HTML, extracts the links and descriptions, then saves the URL for the
       next page of results.  See the code for the "WWW::Search::AltaVista" module for an
       example.

       Alternatively, a backend can define the method "_parse_tree()" instead of
       "_native_retrieve_some()".  See the "WWW::Search::Ebay" module for a good example.

       If you implement a new backend, please let the authors know.

BUGS AND DESIRED FEATURES

       The bugs are there for you to find (some people call them Easter Eggs).

       Desired features:

       A portable query language.
           A portable language would easily allow you to move queries easily between different
           search engines.  A query abstraction is non-trivial and unfortunately will not be done
           any time soon by the current maintainer.  If you want to take a shot at it, please let
           me know.

AUTHOR

       John Heidemann <johnh@isi.edu> Maintained by Martin Thurn, "mthurn@cpan.org",
       http://www.sandcrawler.com/SWB/cpan-modules.html <http://www.sandcrawler.com/SWB/cpan-
       modules.html>.

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.