WWW::Search(3) User Contributed Perl Documentation WWW::Search(3)NAMEWWW::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.
perl v5.14.1 2011-07-21 WWW::Search(3)
