search.cgi -- Example Perl program for searching with Swish-e and SWISH::API
This is a very simple program that shows how to use the SWISH::API module in a
CGI script or mod_perl handler using Template-Toolkit to generate output. This
program is intended for programmers that want to create a custom search
this script does not have many features, and provides no
external configuration (with the execption of a few config options under
mod_perl). So don't ask why it doesn't do something. The point is that this
script is used as a starting point that YOU customize.
You must have swish-e and the SWISH::API module installed. See the README and
INSTALL documents in the swish-e distribution. As of this writing SWISH::API
is part of the swish-e distribution, but in the future may be provided as a
separate package (provided on the CPAN). In either case SWISH::API is a
separate installation procedure from installing swish-e. The Storable module
is also required if using mod_perl.
This program does require that some modules are installed from CPAN. You will
need Template-Toolkit and HTML::FillInForm (which depends on HTML::Parser).
How those are installed depends on your computer's packaging system.
You will need a web server, obviously. The discussion below assumes Apache is
used. If you are using MS IIS take note that IIS works differently in a number
script and related templates are installed when swish-e is
is installed in $prefix/lib/swish-e/ and
templates are installed in $prefix/share/swish-e/templates/. $prefix is
/usr/local by default but can be changed when running the swish-e
script. Upon installation search.cgi
is updated with
correct paths to your perl binary and
When running as a CGI script search.cgi
is copied or symlinked to the
location of your CGI scripts (or any directory that allows CGI scripts). By
default, the search.cgi
script looks for the index index.swish-e
in the current directory (that's what the web server considers the current
directory). On Apache running mod-cgi that's the same place as the script. On
IIS it's not. If your index is elsewhere you will need to modify the script.
The script works by parsing the query, calling SWISH::API to run the actual
search, then calls Template-Toolkit to generate the ouput.
The script calls the search.tt
template. This template generates the
query form and the search results. The search.tt
template uses a
Template-Toolkit "WRAPPER" function to wrap the search form and
results in your site's design. This design is in the page_layout
template. The idea is if you use Template-Toolkit to manage your entire site
then your entire site would be formatted by the same page_layout
template. The page_layout
template calls two other templates
to generate a common header and
footer for the site. Those are just demonstrating Template-Toolkit's features.
page only defines the basic structure of the site. The
true design of the site is managed by style sheets. style.css
the basic layout and markup.css
sets fonts and colors.
Note: these style sheets are included directly in the output of the CGI script.
In production the style sheets would be stored as separate style sheet files
and imported by the browser instead of directly included in the search results
See the section MOD_PERL below for more on templates.
Highlighting of search terms is provided by the SWISH::PhraseHighlight module.
That is a very slow module, so you may wish to disable it if you expect a lot
Enough talking, sometimes it's nice to see a complete example. Below swish-e is
installed in the default location (/usr/local). The "$" is a normal
user prompt, where "#" is a root prompt. Use ./configure --prefix to
install in another location (e.g. if you do not have root access).
Download and install swish-e
$ wget -q http://swish-e.org/Download/latest.tar.gz
$ tar zxf latest.tar.gz
$ cd swish-e-2.x.x
$ (./configure && make) >/dev/null
$ make check
# make install
$ cd perl
$ perl Makefile.PL && make && make test
# make install
Install requried Perl modules. You can install via RPMs, Debs or directly from
the CPAN or by using the CPAN shell.
# perl -MCPAN -e 'install Template'
# perl -MCPAN -e 'install HTML::FillInForm'
Now setup the script in someplace that allows CGI scripts.
$ cd $HOME/apache
$ ln -s /usr/local/lib/swish-e/search.cgi .
$ cat .htaccess
deny from all
allow from all
Create an index
$ cat swish.config
IndexOnly .htm .html
StoreDescription HTML* <body>
metanames swishtitle swishdocpath
$ swish-e -c swish.config -i /usr/share/doc/apache-doc/manual
Test the index and the CGI script:
$ swish-e -w apache -m1 | grep hits
# Number of hits: 152
$ lynx -dump http://localhost/apache/search.cgi?query=apache | grep hits
Showing page 1 (1 - 10 of 152 hits) Next
'hits' => 152,
Now, the above isn't very helpful because the Apache documentation indexed is
not in the web space. You would likely index content available on your web
Perl CGI script must be compiled for each request. SpeedyCGI is a tool to speed
up scripts by running them persistently. To run search.cgi
SpeedyCGI install the program (you can Google, right?) and then change the
first line of search.cgi
to run the speedy
This script can be run directly as a mod_perl handler, and the same code can be
used to run multiple sites by using separate Location directives and passing
in a "site id." The script caches in memory different configurations
based on this site id.
Below is a complete httpd.conf file. It requires an Apache httpd that has
mod_perl compiled in statically. It runs mod_perl on a high port (port 5000)
listening to all interfaces.
For testing I put this config file in a directory along with search.cgi
but that's just done to make the example simple (i.e. so I don't have to show
any absolute paths). Normally the httpd.conf and the swish.cgi
"module" would be in separate locations.
# httpd.conf -- test file for search.cgi as mod_perl handler
LoadModule mime_module /usr/lib/apache/1.3/mod_mime.so
push @PerlSetVar, [
index => Apache->server_root_relative( 'index.swish-e'),
$DocumentRoot = Apache->server_root_relative;
# Define this site
PerlSetVar site_id othersite
PerlSetVar title "Some other Site"
The server is started using this command:
$ /usr/sbin/apache-perl -d $(pwd) -f $(pwd)/httpd.conf
which says to use the current directory as the ServerRoot. (See comments below.)
Stop the server like:
$ kill `cat swish_httpd.pid`
Then access either:
A few Notes:
I like test configurations to not care where things are located. Thus, the above
httpd.conf does a few tricks in the "Perl Section" shown.
First, mod_perl, unlike CGI, doesn't set the working directory. So, the index
file name must be absolute. This is accomplished by a PerlSetVar entry
building the index file name from the ServerRoot.
Second, the DocumentRoot is set to the same as the ServerRoot. The DocumentRoot
needs to be set so search.cgi can figure out the path to the script (for
creating next and previous links).
Third, the script is loaded by a "require" statement. This works only
because the current directory "." is in Perl's @INC path at Apache
start up time and search.cgi
is also in the current directory.
Normally, set PERL5LIB on server startup or use a "use lib" line in
your startup.pl file to point to the location of search.cgi.
The "PerlSetVar" lines pass config information into the script. Note
that they can be set globally or specific to a given Location.
The following config options are currently available:
- The site_id options allow caching of configurations on a
per-site basis. It's overkill in this example, but normally you might have
expensive configuration processes that you might want to do only once.
But, since there is caching by this id it's a good id to set a site_id if
using more than one Location directive.
- This specifies the index file to use. The index file needs
to be absolute as discussed above. Example:
PerlSetVar index /usr/share/swish/site.index
- This options sets the title that's passed into the
- Sets the file name of the template use to generate the
form. This might be useful if you want an "advanced" form, for
- This can be used to update the path where templates are
searched. Useful if you wish to override templates.
- This allow changing the default number of results shown per
Not much support is provided. But what support is provided is ONLY provided via
the Swish-e discussion list.
Copyright 2003, 2004 Bill Moseley. All rights reserved.
This program is free software; you can redistribute it and/or modify it under
the same terms as Perl itself.
SWISH::API, Template, HTML::FillInForm