seek_result($offset)user_agent($NON_ROBOT) (PRIVATE)
WWW::Search - Virtual base class for WWW searches
require WWW::Search;
$search_engine = "AltaVista";
$search = new WWW::Search($search_engine);
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.
Here is a sample program:
my $search = new WWW::Search('AltaVista');
$search->native_query(WWW::Search::escape_query($query));
while (my $result = $search->next_result())
{
print $result->url, "\n";
}
Results are objects of type WWW::SearchResult
(see the WWW::SearchResult manpage for details).
Note that different backends support different result fields.
All backends are required to support title and url.
For specific search engines, see the WWW::Search::TheEngineName manpage (replacing TheEngineName with a particular search engine).
For details about the results of a search, see the WWW::SearchResult manpage.
Methods and functions marked as PRIVATE are in general only useful to backend programmers.
To create a new WWW::Search, call
$search = new WWW::Search('SearchEngineName');
where SearchEngineName is replaced with a particular search engine. For example:
$search = new WWW::Search('Google');
If no search engine is specified a default (currently 'AltaVista') will be chosen for you. The next step is usually:
$search->native_query('search-engine-specific+query+string');
Resets internal data structures to start over with a new search.
Returns the value of the $VERSION variable of the backend engine, or $WWW::Search::VERSION if the backend does not contain $VERSION.
Returns the value of the $MAINTAINER variable of the backend engine, or $WWW::Search::MAINTAINER if the backend does not contain $MAINTAINER.
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.
The query must be escaped; call the &WWW::Search::escape_query manpage to escape
a plain query. See native_query below for more information.
Currently, this feature is supported by only a few backends; consult the documentation for each backend to see if it is implemented.
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 the WWW::Search::escape_query() manpage
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:
$search->native_query('search-engine-specific+escaped+query+string',
{ option1 => 'able', option2 => 'baker' } );
The hash of options following the query string is optional. The query string is backend-specific. There are two kinds of options: options specific to the backend, and generic options applicable to multiple backends.
Generic options all begin with 'search_'. Currently a few are supported:
GET or POST) for HTTP-based queries.
The default is GET
Some backends may not implement these generic options, but any which do implement them must provide these semantics.
Backend-specific options are described in the documentation for each backend. In most cases the options and their values are packed together to create the query portion of the final URL.
Details about how the search string and option hash are interpreted might be found in the search-engine-specific manual pages (WWW::Search::SearchEngineName).
After native_query, the next step is usually:
while ($result = $search->next_result())
{
# do_something;
}
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 in turn 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);
Set-up an HTTP proxy (for connections from behind a firewall).
This routine should be called before calling any of the result functions (next_result or results).
Example:
$search->http_proxy("http://gateway:8080";);
Some backends indicate how many hits they have found. Typically this is an approximate value.
Return all the results of a query as an array of SearchResult objects.
Example:
@results = $search->results();
foreach $result (@results) {
print $result->url(), "\n";
};
On error, results() will return undef and set response()
to the HTTP response code.
Call this method repeatedly to return each result of a query as a SearchResult object. Example:
while ($result = $search->next_result())
{
print $result->url(), "\n";
}
On error, next_result() will return undef and set response()
to the HTTP response code.
Returns the an HTTP::Response object which resulted from the
most-recently-sent query (see the HTTP::Response manpage). If the query
returns no results (i.e. $search->results is undef), errors can
be reported like this:
my $response = $search->response();
if ($response->is_success) {
print "normal end of result list\n";
} else {
print "error: " . $response->as_string() . "\n";
}
Note to backend authors: even if the backend does not involve the web, it should return an HTTP::Response object.
seek_result($offset)Set which result next_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:
$search->seek_result(0);
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 = $search->maximum_to_retrieve(100);
You can also spell this method ``maximum_to_return''.
The maximum length of time any portion of the query should take, in seconds.
Defaults to 60.
Example: $search->timeout(120);
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.
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 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');
(Returns "%2Bhi+%2Bmom").
See also unescape_query.
NOTE that this is not a method, it is a plain function.
Unescape a query.
See escape_query for details.
Example: $unescaped = WWW::Search::unescape_query('%22hi+mom%22');
(Returns '"hi mom"').
NOTE that this is not a method, it is a plain function.
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 SearchResults.
NOTE that this is not a method, it is a plain function.
Deprecated.
Given a reference to a hash of string => string, constructs a CGI parameter string that looks like 'key1=value1&key2=value2'.
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});
user_agent($NON_ROBOT) (PRIVATE)This internal routine creates a user-agent for derived classes that query the web. If non-false argument $non_robot is given, a normal user-agent (rather than a robot-style user-agent) is used.
If a backend needs the low-level LWP::UserAgent or LWP::RobotUA to have a particular name, $oSearch->{'agent_name'} (and possibly $oSearch->{'agent_e_mail'}) should be set to the desired values before calling $oSearch->user_agent():
$oSearch = new WWW::Search('NewBackend');
$oSearch->{'agent_e_mail'} = $oSearch->{'agent_name'};
$oSearch->{'agent_name'} = 'Mozilla/5.5';
$oSearch->user_agent('non-robot');
Backends should use robot-style user-agents whenever possible.
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);
Return the response from an http request, handling debugging. Requires that user_agent already be set up, if needed. Requires that http_referer already be set up, if needed.
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.)
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.
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.
This internal routine does generic Search setup.
It calls native_setup_search to do backend specific setup.
Derived classes should call this between requests to remote servers to avoid overloading them with many, fast back-to-back requests.
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 or URI::URL object (whichever is being used by HTTP on your system).
An internal routine to interface with native_retrieve_some.
Checks for overflow.
Deprecated.
Returns the value of the $TEST_CASES variable of the backend engine.
WWW::Search supports backends to separate search engines.
Each backend is implemented as a subclass of WWW::Search.
the WWW::Search::AltaVista manpage provides a good sample backend.
A backend must have the two routines
native_retrieve_some and native_setup_search.
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, parse the HTML, extract the links and
descriptions, then save the URL for the next page of results. See the
code for the AltaVista implementation for an example.
native_setup_search is invoked before the search.
It is passed a single argument: the escaped, native version
of the query.
The front- and backends share a single object (a hash).
The backend can change any hash element beginning with underscore,
and {response} (an HTTP::Response code) and {cache}
(the array of WWW::SearchResult objects caching all results).
Again, look at one of the existing web search backends as an example.
If you implement a new backend, please let the authors know.
The bugs are there for you to find (some people call them Easter Eggs).
Desired features:
WWW::Search was written by John Heidemann, & was written by John Heidemann, ;johnh@isi.edu>.
WWW::Search is currently maintained by Martin Thurn, & is currently maintained by Martin Thurn, ;MartinThurn@iname.com>.
backends and applications for WWW::Search were originally written by John Heidemann, Wm. L. Scheding, Cesare Feroldi de Rosa, and GLen Pringle.
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.