NAME
Tie::Google - Single-variable access to Google search results
SYNOPSIS
my $KEYFILE = glob "~/.googlekey";
my ($g, @g, %g);
# Tied array interface
tie @g, "Tie::Google", $KEYFILE, "perl";
for my $r (@g) {
printf " * <a href='%s'>%s</a>\n",
$r->{'URL'}, $r->{'title'};
}
# Tied hash interface
tie %g, "Tie::Google", $KEYFILE;
for my $term (qw[ perl python ruby ]) {
my $res = $g{$term};
printf "%d results for '%s:\n", scalar @$res, $term;
for my $r (@$res) {
printf " * <a href='%s'>%s</a>\n",
$r->{'URL'}, $r->{'title'};
}
}
# Tied scalar interface: I Feel Lucky
use LWP::Simple qw(getprint);
tie $g, "Tie::Google", $KEYFILE, "perl";
getprint($g->{'URL'});
USING Tie::Google
Using tied variables can make searching Google much simpler for trivial
programs. "Tie::Google" presents a simple interface to Google's search
API, using "Net::Google" as the underlying transport mechanism. To use
"Tie::Google", you must already be registered with Google and have an
API key.
You can tie scalars, arrays, or hashes to the "Tie::Google" class; each
offers slightly different functionality, but all offer direct access to
Google search results. The basic syntax of all types is:
tie VAR, 'Tie::Google', $APIKEY, $QUERY, \%OPTIONS;
where:
VAR VAR is the variable name, which can be a scalar, array, or hash:
tie $g, "Tie::Google", $KEY, $term;
tie @g, "Tie::Google", $KEY, $term;
tie %g, "Tie::Google", $KEY, $term;
$APIKEY APIKEY is your Google API key or a file containing the key as
the only item on the first line; see http://apis.google.com for
details.
$QUERY QUERY is your actual search term(s), as a string. This can be
arbitrarily complex, up to the limits Google allows:
tie $g, "Tie::Google", $KEY,
"site:cnn.com allintitle:priest court -judas";
%OPTIONS
Any options specified in this hashref will be passed to the
Net::Google::Search instance. Available options include
"starts_at", "max_results", "ie", "oe", and "lr". See the
Net::Google manpage.
The Tied Array Interface
Tieing an array to "Tie::Google" gives you an array of search results.
How many search results are returned depends on the value of the
"max_results" option defined when the array was tied (or
$DEFAULT_BATCH_SIZE if "max_results" was not set), though extending the
array of results can be done by growing the array.
$#g = 20;
Will resize the result set to 20 results. If there are more than 20, the
ones on the end will be popped off; if there are less than 20, then more
will be retrieved.
"Tie::Google" supports all non-additive array operations, including
"shift", "pop", and the 3 argument form of "splice" (not the 4 argument
version). Specifically unallowed are "unshift", "push", and general
assigment.
See the section on "RESULTS" for details about the individual search
results.
The Tied Hash Interface
The tied hash interface is similar to the tied array interface, except
that there are a bunch of them. Asking for a key in the hash %g
initiates a search to Google, with the specified key as the search term:
my $results = $g{'perl apache'};
This initiates a search with "perl apache" as the query. $results is a
reference to an array of hashrefs (see the section on "RESULTS" for
details about said hashrefs).
Tied hashes support all hash functions, including "each", "keys", and
"values". Deleting from the hash is allowed (it removes the search
results for the deleted terms), but adding to the hash is not.
There can be many sets of search results stored in a tied hash. To see
what this looks like, try this:
use Data::Dumper;
my (%g, $KEY, $dummy);
tie %g, "Tie::Google", $KEY;
$dummy = $g{'perl'};
$dummy = $g{'python'};
$dummy = $g{'ruby'};
print Dumper(tied(%g));
Also, for comparison, try:
tie @g, "Tie::Google", $KEY, "perl";
print Dumper(tied(@g));
If "starts_at" or "max_results" are specified during the "tie", these
options are carried over into new searches (when a new key is requested
from the hash), so plan accordingly. If the "max_value" is set to 1000,
for example, then every access of a new key is going to contain 1000
elements, which will be pretty slow.
The Tied Scalar Interface
Do you feel lucky? If so, tie a scalar to "Tie::Google":
tie $g, "Tie::Google", $KEY, "python";
Will give you the top result. This is conceptually similar to using the
"I Feel Lucky" button on Google.com's front search interface.
RESULTS
All results (values returned from these tied variables) are hash
references; the contents of these hashrefs are based on the
"Net::Google::Response" class (see the Net::Google::Response manpage for
details). These elements currently are:
* title
* URL
* snippet
* cachedSize
* directoryTitle
* summary
* hostName
* directoryCategory
All keys are case sensitive, and return exactly what the
Net::Google::Response manpage says they do ("Tie::Google" does no
massaging of this data).
TODO / BUGS
This module is far from complete, or even fully thought out. TODO items
currently include:
* The tests currently suck, to the point of embarrassment. Don't
mention it, I'm a little sensitive about it.
* Some of the behaviors are kind of wonky. If anyone has any better
ideas, please let me know. Patches demanded^Wwelcome^.
* Tied arrays should get the next 10 results when you get to the end
of the array. Currently, you have to manually extend the array
using:
$#g = 100;
to get 100 search results.
Although this technique will have the unfortunate side-effect of
trying to iterate through all the results in Google's database.
* The tied hash interface should be implemented in terms of the tied
array interface. That is, the values associated with each key
(search term) should be a reference to an array tied to
"Tie::Google". I started doing it this way but it made my brain
hurt.
* Does there need to be a TIEHANDLE interface as well? Hmmm...
while (<$google>) {
...
* Should returned search results be data structures (they are
currently), the actual "Net::Google::Result" instances (where the
data is currently being derived from), or new objects in their own
right (e.g., "Tie::Google::Result")? I see advantages to each path:
data structures would be simpler, passing on "Result" objects
without modification would be faster, and using a new set of objects
allows new functionality to be added, for example useful
stringification.
SEE ALSO
the Net::Google manpage, the DBD::google manpage
AUTHOR
darren chamberlain (<darren@cpan.org>), with some prompting from Richard
Soderberg.