NAME¶
swish.cgi -- Example Perl script for searching with the SWISH-E search engine.
DESCRIPTION¶
"swish.cgi" is a CGI script for searching with the SWISH-E search
engine version 2.1-dev and above. It returns results a page at a time, with
matching words from the source document highlighted, showing a few words of
content on either side of the highlighted word.
The script is highly configurable. Features include searching multiple (or
selectable) indexes, limiting searches to a subset of documents, sorting by a
number of different properties, and limiting results to a date range.
On unix type systems the swish.cgi script is installed in the directory
$prefix/lib/swish-e, which is typically /usr/local/lib/swish-e. This can be
overridden by the configure options --prefix or --libexecdir.
The standard configuration (i.e. not using a config file) should work with most
swish index files. Customization of the parameters will be needed if you are
indexing special meta data and want to search and/or display the meta data.
The configuration can be modified by editing this script directly, or by using
a configuration file (.swishcgi.conf by default). The script's configuration
file is described below.
You are strongly encouraged to get the default configuration working before
making changes. Most problems using this script are the result of
configuration modifications.
The script is modular in design. Both the highlighting code and output
generation is handled by modules, which are included in the
example/modules distribution directory and installed in the
$libexecdir/perl directory. This allows for easy customization of the output
without changing the main CGI script.
Included with the Swish-e distribution is a module to generate standard HTML
output. There's also modules and template examples to use with the popular
Perl templating systems HTML::Template and Template-Toolkit. This is very
useful if your site already uses one of these templating systems The
HTML::Template and Template-Toolkit packages are not distributed with Swish-e.
They are available from the CPAN (
http://search.cpan.org).
This scipt can also run basically unmodified as a mod_perl handler, providing
much better performance than running as a CGI script. Usage under mod_perl is
described below.
Please read the rest of the documentation. There's a "DEBUGGING"
section, and a "FAQ" section.
This script should work on Windows, but security may be an issue.
REQUIREMENTS¶
A reasonably current version of Perl. 5.00503 or above is recommended (anything
older will not be supported).
The Date::Calc module is required to use the date range feature of the script.
The Date::Calc module is also available from CPAN.
INSTALLATION¶
Here's an example installation session under Linux. It should be similar for
other operating systems.
For the sake of simplicity in this installation example all files are placed in
web server space, including files such as swish-e index and configuration
files that would normally not be made available via the web server. Access to
these files should be limited once the script is running. Either move the
files to other locations (and adjust the script's configuration) or use
features of the web server to limit access (such as with
.htaccess).
Please get a simple installation working before modifying the configuration
file. Most problems reported for using this script have been due to improper
configuration.
The script's default settings are setup for initial testing. By default the
settings expect to find most files and the swish-e binary in the same
directory as the script.
For
security reasons, once you have tested the script you will want to
change settings to limit access to some of these files by the web server
(either by moving them out of web space, or using access control such as
.htaccess). An example of using
.htaccess on Apache is given
below.
It's expected that swish-e has already been unpacked and the swish-e binary has
be compiled from source and "make install" has been run. If swish-e
was installed from a vendor package (such as from a RPM or Debian package) see
that pakage's documentation for where files are installed.
Example Installation:
- 1 Symlink or copy the swish.cgi.
- Symlink (or copy if your platform or webserver does not allow symlinks)
the swish.cgi script from the installation directory to a local directory.
Typically, this would be the cgi-bin directory or a location where CGI
script are located. In this example a new directory is created and the
script is symlinked.
~$ mkdir swishdir
~$ cd swishdir
~/swishdir$ ln -s /usr/local/lib/swish-e/swish.cgi
The installation directory is set at configure time with the --prefix or
--libexecdir options, but by default is in /usr/local/lib/swish-e.
- 2 Create an index
- Use an editor and create a simple configuration file for indexing your
files. In this example the Apache documentation is indexed. Last we run a
simple query to test that the index works correctly.
~/swishdir$ cat swish.conf
IndexDir /usr/local/apache/htdocs
IndexOnly .html .htm
DefaultContents HTML*
StoreDescription HTML* <body> 200000
MetaNames swishdocpath swishtitle
ReplaceRules remove /usr/local/apache/
If you do not have the Apache docs installed then pick another directory to
index such as /usr/share/doc.
Create the index.
~/swishdir$ swish-e -c swish.conf
Indexing Data Source: "File-System"
Indexing "/usr/local/apache/htdocs"
Removing very common words...
no words removed.
Writing main index...
Sorting words ...
Sorting 7005 words alphabetically
Writing header ...
Writing index entries ...
Writing word text: Complete
Writing word hash: Complete
Writing word data: Complete
7005 unique words indexed.
5 properties sorted.
124 files indexed. 1485844 total bytes. 171704 total words.
Elapsed time: 00:00:02 CPU time: 00:00:02
Indexing done!
Now, verify that the index can be searched:
~/swishdir$ swish-e -w install -m 1
# SWISH format: 2.1-dev-25
# Search words: install
# Number of hits: 14
# Search time: 0.001 seconds
# Run time: 0.040 seconds
1000 htdocs/manual/dso.html "Apache 1.3 Dynamic Shared Object (DSO) support" 17341
.
Let's see what files we have in our directory now:
~/swishdir$ ls -1
index.swish-e
index.swish-e.prop
swish.cgi
swish.conf
- 3 Test the CGI script
- This is a simple step, but often overlooked. You should test from the
command line instead of jumping ahead and testing with the web server. See
the "DEBUGGING" section below for more information.
~/swishdir$ ./swish.cgi | head
Content-Type: text/html; charset=ISO-8859-1
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<title>
Search our site
</title>
</head>
<body>
The above shows that the script can be run directly, and generates a correct
HTTP header and HTML.
If you run the above and see something like this:
~/swishdir >./swish.cgi
bash: ./swish.cgi: No such file or directory
then you probably need to edit the script to point to the correct location
of your perl program. Here's one way to find out where perl is located
(again, on unix):
~/swishdir$ which perl
/usr/local/bin/perl
~/swishdir$ /usr/local/bin/perl -v
This is perl, v5.6.0 built for i586-linux
...
Good! We are using a reasonably current version of perl.
Now that we know perl is at /usr/local/bin/perl we can adjust the
"shebang" line in the perl script (e.g. the first line of the
script):
~/swishdir$ pico swish.cgi
(edit the #! line)
~/swishdir$ head -1 swish.cgi
#!/usr/local/bin/perl -w
- 4 Test with the web server
- How you do this is completely dependent on your web server, and you may
need to talk to your web server admin to get this working. Often files
with the .cgi extension are automatically set up to run as CGI scripts,
but not always. In other words, this step is really up to you to figure
out!
This example shows creating a symlink from the web server space to
the directory used above. This will only work if the web server is
configured to follow symbolic links (the default for Apache).
This operation requires root access:
~/swishdir$ su -c "ln -s $HOME/swishdir /usr/local/apache/htdocs/swishdir"
Password: *********
If your account is on an ISP and your web directory is ~/public_html
the you might just move the entire directory:
mv ~/swishdir ~/public_html
Now, let's make a real HTTP request:
~/swishdir$ GET http://localhost/swishdir/swish.cgi | head -3
#!/usr/local/bin/perl -w
package SwishSearch;
use strict;
Oh, darn. It looks like Apache is not running the script and instead
returning it as a static page. Apache needs to be told that swish.cgi is a
CGI script.
.htaccess comes to the rescue:
~/swishdir$ cat .htaccess
# Deny everything by default
Deny From All
# But allow just CGI script
<files swish.cgi>
Options ExecCGI
Allow From All
SetHandler cgi-script
</files>
That "Deny From All" prevents access to all files (such as config
and index files), and only access is allowed to the swish.cgi
script.
Let's try the request one more time:
~/swishdir >GET http://localhost/swishdir/swish.cgi | head
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<title>
Search our site
</title>
</head>
<body>
<h2>
<a href="http://swish-e.org">
That looks better! Now use your web browser to test.
Now, you may note that the links are not valid on the search results page.
The swish config file contained the line:
ReplaceRules remove /usr/local/apache/
To make those links works (and assuming your web server will follow symbolic
links):
~/swishtest$ ln -s /usr/local/apache/htdocs
BTW - "GET" used above is a program included with Perl's LWP
library. If you do no have this you might try something like:
wget -O - http://localhost/swishdir/swish.cgi | head
and if nothing else, you can always telnet to the web server and make a
basic request.
~/swishtest$ telnet localhost 80
Trying 127.0.0.1...
Connected to localhost.
Escape character is '^]'.
GET /swishtest/swish.cgi http/1.0
HTTP/1.1 200 OK
Date: Wed, 13 Feb 2002 20:14:31 GMT
Server: Apache/1.3.20 (Unix) mod_perl/1.25_01
Connection: close
Content-Type: text/html; charset=ISO-8859-1
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<title>
Search our site
</title>
</head>
<body>
This may seem like a lot of work compared to using a browser, but browsers
are a poor tool for basic CGI debugging.
If you have problems check the "DEBUGGING" section below.
CONFIGURATION¶
If you want to change the location of the swish-e binary or the index file, use
multiple indexes, add additional metanames and properties, change the default
highlighting behavior, etc., you will need to adjust the script's
configuration settings.
Again, please get a test setup working with the default parameters before making
changes to any configuration settings. Better to debug one thing at a time...
In general, you will need to adjust the script's settings to match the index
file you are searching. For example, if you are indexing a hypermail list
archive you may want to make the script use metanames/properties of Subject,
Author, and, Email address. Or you may wish to provide a way to limit searches
to subsets of documents (e.g. parts of your directory tree).
To make things somewhat "simple", the configuration parameters are
included near the top of the swish.cgi program. That is the only place that
the individual parameters are defined and explained, so you will need to open
up the swish.cgi script in an editor to view the options. Further questions
about individual settings should be referred to the swish-e discussion list.
The parameters are all part of a perl "hash" structure, and the
comments at the top of the program should get you going. The perl hash
structure may seem a bit confusing, but it makes it easy to create nested and
complex parameters. Syntax is important, so cut-n-paste should be your best
defense if you are not a perl programmer.
By the way, Perl has a number of quote operators. For example, to quote a string
you might write:
title => 'Search My Site',
Some options take more than one parameter, where each parameter must be quoted.
For example:
metanames => [ 'swishdefault', 'swishtitle', 'swishdocpath' ],
which assigns an array ( [...] ) of three strings to the "metanames"
variable. Lists of quoted strings are so common in perl that there's a special
operator called "qw" (quote word) to save typing all those quotes:
metanames => [ qw/ swishdefault swishtitle swishdocpath / ],
or to use the parenthesis as the quote character (you can pick any):
metanames => [ qw( swishdefault swishtitle swishdocpath ) ],
There are two options for changing the configuration settings from their default
values: One way is to edit the script directly, or the other was is to use a
separate configuration file. In either case, the configuration settings are a
basic perl hash reference.
Using a configuration file is described below, but contains the same hash
structure.
There are many configuration settings, and some of them are commented out either
by using a "#" symbol, or by simply renaming the configuration
directive (e.g. by adding an "x" to the parameter name).
A very basic configuration setup might look like:
return {
title => 'Search the Swish-e list', # Title of your choice.
swish_binary => 'swish-e', # Location of swish-e binary
swish_index => 'index.swish-e', # Location of your index file
};
Or if searching more than one index:
return {
title => 'Search the Swish-e list',
swish_binary => 'swish-e',
swish_index => ['index.swish-e', 'index2'],
};
Both of these examples return a reference to a perl hash ( "return
{...}" ). In the second example, the multiple index files are set as an
array reference.
Note that in the example above the swish-e binary file is relative to the
current directory. If running under mod_perl you will need to use absolute
paths.
The script can also use the SWISH::API perl module (included with the swish-e
distribution in the
perl directory) to access the swish-e index. The
"use_library" option is used to enable the use of the SWISH::API
module:
return {
title => 'Search the Swish-e list',
swish_index => ['index.swish-e', 'index2'],
use_library => 1, # enable use of the SWISH::API module
};
The module must be available via the @INC array, like all Perl modules.
Using the SWISH::API module avoids the need to fork and execute a the swish-e
program. Under mod_perl you will may see a significant performance improvement
when using the SWISH::API module. Under normal CGI usage you will probably not
see any speed improvements.
Using A Configuration File
As mentioned above, configuration settings can be either set in the
swish.cgi script, or set in a separate configuration file. Settings in
a configuration file will override the settings in the script.
By default, the
swish.cgi script will attempt to read settings from the
file
.swishcgi.conf. For example, you might only wish to change the
title used in the script. Simply create a file called
.swishcgi.conf in
the same directory as the CGI script:
> cat .swishcgi.conf
# Example swish.cgi configuration script.
return {
title => 'Search Our Mailing List Archive',
};
The settings you use will depend on the index you create with swish:
return {
title => 'Search the Apache documentation',
swish_binary => 'swish-e',
swish_index => 'index.swish-e',
metanames => [qw/swishdefault swishdocpath swishtitle/],
display_props => [qw/swishtitle swishlastmodified swishdocsize swishdocpath/],
title_property => 'swishdocpath',
prepend_path => 'http://myhost/apachedocs',
name_labels => {
swishdefault => 'Search All',
swishtitle => 'Title',
swishrank => 'Rank',
swishlastmodified => 'Last Modified Date',
swishdocpath => 'Document Path',
swishdocsize => 'Document Size',
},
};
The above configuration defines metanames to use on the form. Searches can be
limited to these metanames.
"display_props" tells the script to display the property
"swishlastmodified" (the last modified date of the file), the
document size, and path with the search results.
The parameter "name_labels" is a hash (reference) that is used to give
friendly names to the metanames.
Here's another example. Say you want to search either (or both) the Apache 1.3
documentation and the Apache 2.0 documentation indexed seperately.
return {
title => 'Search the Apache Documentation',
date_ranges => 0,
swish_index => [ qw/ index.apache index.apache2 / ],
select_indexes => {
method => 'checkbox_group',
labels => [ '1.3.23 docs', '2.0 docs' ], # Must match up one-to-one to swish_index
description => 'Select: ',
},
};
Now you can select either or both sets of documentation while searching.
All the possible settings are included in the default configuration located near
the top of the
swish.cgi script. Open the
swish.cgi script with
an editor to look at the various settings. Contact the Swish-e Discussion list
for help in configuring the script.
DEBUGGING¶
Most problems with using this script have been a result of improper
configuration. Please get the script working with default settings before
adjusting the configuration settings.
The key to debugging CGI scripts is to run them from the command line, not with
a browser.
First, make sure the program compiles correctly:
$ perl -c swish.cgi
swish.cgi syntax OK
Next, simply try running the program:
$ ./swish.cgi | head
Content-Type: text/html; charset=ISO-8859-1
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<title>
Search our site
</title>
</head>
<body>
Under Windows you will need to run the script as:
C:\wwwroot\swishtest> perl swish.cgi
Now, you know that the program compiles and will run from the command line.
Next, try accessing the script from a web browser.
If you see the contents of the CGI script instead of its output then your web
server is not configured to run the script. With Apache look at settings like
ScriptAlias, SetHandler, and Options.
If an error is reported (such as Internal Server Error or Forbidden) you need to
locate your web server's error_log file and carefully read what the problem
is. Contact your web administrator for help locating the web server's error
log.
If you don't have access to the web server's error_log file, you can modify the
script to report errors to the browser screen. Open the script and search for
"CGI::Carp". (Author's suggestion is to debug from the command line
-- adding the browser and web server into the equation only complicates
debugging.)
The script does offer some basic debugging options that allow debugging from the
command line. The debugging options are enabled by setting an environment
variable "SWISH_DEBUG". How that is set depends on your operating
system and the shell you are using. These examples are using the
"bash" shell syntax.
Note: You can also use the "debug_options" configuration setting, but
the recommended method is to set the environment variable.
You can list the available debugging options like this:
$ SWISH_DEBUG=help ./swish.cgi >outfile
Unknown debug option 'help'. Must be one of:
basic: Basic debugging
command: Show command used to run swish
headers: Show headers returned from swish
output: Show output from swish
summary: Show summary of results
dump: Show all data available to templates
Debugging options may be combined:
$ SWISH_DEBUG=command,headers,summary ./swish.cgi >outfile
You will be asked for an input query and the max number of results to return.
You can use the defaults in most cases. It's a good idea to redirect output to
a file. Any error messages are sent to stderr, so those will still be
displayed (unless you redirect stderr, too).
Here are some examples:
~/swishtest$ SWISH_DEBUG=basic ./swish.cgi >outfile
Debug level set to: 1
Enter a query [all]:
Using 'not asdfghjklzxcv' to match all records
Enter max results to display [1]:
------ Can't use DateRanges feature ------------
Script will run, but you can't use the date range feature
Can't locate Date/Calc.pm in @INC (@INC contains: modules /usr/local/lib/perl5/5.6.0/i586-linux /usr/local/lib/perl5/5.6.0 /usr/local/lib/perl5/site_perl/5.6.0/i586-linux /usr/local/lib/perl5/site_perl/5.6.0 /usr/local/lib/perl5/site_perl/5.005/i586-linux /usr/local/lib/perl5/site_perl/5.005 /usr/local/lib/perl5/site_perl .) at modules/DateRanges.pm line 107, <STDIN> line 2.
BEGIN failed--compilation aborted at modules/DateRanges.pm line 107, <STDIN> line 2.
Compilation failed in require at ./swish.cgi line 971, <STDIN> line 2.
--------------
Can't exec "./swish-e": No such file or directory at ./swish.cgi line 1245, <STDIN> line 2.
Child process Failed to exec './swish-e' Error: No such file or directory at ./swish.cgi line 1246, <STDIN> line 2.
Failed to find any results
The above indicates two problems. First problem is that the Date::Calc module is
not installed. The Date::Calc module is needed to use the date limiting
feature of the script.
The second problem is a bit more serious. It's saying that the script can't find
the swish-e binary file. In this example it's specified as being in the
current directory. Either correct the path to the swish-e binary, or make a
local copy or symlink to the swish-e binary.
~/swishtest$ cat .swishcgi.conf
return {
title => 'Search the Apache Documentation',
swish_binary => '/usr/local/bin/swish-e',
date_ranges => 0,
};
Now, let's try again:
~/swishtest$ SWISH_DEBUG=basic ./swish.cgi >outfile
Debug level set to: 1
---------- Read config parameters from '.swishcgi.conf' ------
$VAR1 = {
'date_ranges' => 0,
'title' => 'Search the Apache Documentation'
};
-------------------------
Enter a query [all]:
Using 'not asdfghjklzxcv' to match all records
Enter max results to display [1]:
Found 1 results
Can't locate SWISH::TemplateDefault.pm in @INC (@INC contains: modules /usr/local/lib/perl5/5.6.0/i586-linux /usr/local/lib/perl5/5.6.0 /usr/local/lib/perl5/site_perl/5.6.0/i586-linux /usr/local/lib/perl5/site_perl/5.6.0 /usr/local/lib/perl5/site_perl/5.005/i586-linux /usr/local/lib/perl5/site_perl/5.005 /usr/local/lib/perl5/site_perl .) at ./swish.cgi line 608.
This means that the swish.cgi script could not locate a required module. To
correct this locate where the SWISH::Template module is installed and add a
"use lib" line to your configuration file (or to the swish.cgi
script):
~/swishtest$ cat .swishcgi.conf
use lib '/home/bill/local/lib/perl';
return {
title => 'Search the Apache Documentation',
date_ranges => 0,
};
~/swishtest$ SWISH_DEBUG=basic ./swish.cgi >outfile
Debug level set to: 1
---------- Read config parameters from '.swishcgi.conf' ------
$VAR1 = {
'date_ranges' => 0,
'title' => 'Search the Apache Documentation'
};
-------------------------
Enter a query [all]:
Using 'not asdfghjklzxcv' to match all records
Enter max results to display [1]:
Found 1 results
That is much better!
The "use lib" statement tells Perl where to look for modules by adding
the path supplied to an array called @INC.
Note that most modules are in the SWISH namespace. For example, the default
output module is called SWISH::TemplateDefault. When Perl is looking for that
module it is looking for the file
SWISH/TemplateDefault.pm. If the
"use lib" statement is set as:
use lib '/home/bill/local/lib/perl';
then Perl will look (among other places) for the file
/home/bill/local/lib/perl/SWISH/TemplateDefault.pm
when attempting to load the SWISH::TemplateDefault module. Relative paths may
also be used.
use lib 'modules';
will cause Perl to look for the file:
./modules/SWISH/TemplateDefault.pm
relative to where the swish.cgi script is running. (This is not true when
running under mod_perl).
Here's another common problem. Everything checks out, but when you run the
script you see the message:
Swish returned unknown output
Ok, let's find out what output it is returning:
~/swishtest$ SWISH_DEBUG=headers,output ./swish.cgi >outfile
Debug level set to: 13
---------- Read config parameters from '.swishcgi.conf' ------
$VAR1 = {
'swish_binary' => '/usr/local/bin/swish-e',
'date_ranges' => 0,
'title' => 'Search the Apache Documentation'
};
-------------------------
Enter a query [all]:
Using 'not asdfghjklzxcv' to match all records
Enter max results to display [1]:
usage: swish [-i dir file ... ] [-S system] [-c file] [-f file] [-l] [-v (num)]
...
version: 2.0
docs: http://sunsite.berkeley.edu/SWISH-E/
*** 9872 Failed to run swish: 'Swish returned unknown output' ***
Failed to find any results
Oh, looks like /usr/local/bin/swish-e is version 2.0 of swish. We need 2.1-dev
and above!
Frequently Asked Questions¶
Here's some common questions and answers.
How do I change the way the output looks?¶
The script uses a module to generate output. By default it uses the
SWISH::TemplateDefault.pm module. The module used is selected in the swish.cgi
configuration file. Modules are located in the example/modules/SWISH directory
in the distribution, but are installed in the $prefix/lib/swish-e/perl/SWISH/
directory.
To make simple changes you can edit the installed SWISH::TemplatDefault module
directly, otherwise make a copy of the module and modify its package name. For
example, change directories to the location of the installed module and copy
the module to a new name:
$ cp TemplateDefault.pm MyTemplateDefault.pm
Then at the top of the module adjust the "package" line to:
package SWISH::MyTemplateDefault;
To use this modules you need to adjust the configuration settings (either at the
top of
swish.cgi or in a configuration file:
template => {
package => 'SWISH::MyTemplateDefault',
},
The module does not need to be in the SWISH namespace, and can be stored in any
location as long as the module can be found via the @INC array (i.e. modify
the "use lib" statement in swish.cgi if needed).
How do I use a templating system with swish.cgi?¶
In addition to the TemplateDefault.pm module, the swish-e distribution includes
two other Perl modules for generating output using the templating systems
HTML::Template and Template-Toolkit.
Templating systems use template files to generate the HTML, and make maintaining
the look of a large (or small) site much easier. HTML::Template and
Template-Toolkit are separate packages and can be downloaded from the CPAN.
See
http://search.cpan.org.
Two basic templates are provided as examples for generating output using these
templating systems. The example templates are located in the
example
directory. The module
SWISH::TemplateHTMLTemplate uses the file
swish.tmpl to generate its output, while the module
SWISH::TemplateToolkit uses the
swish.tt file. (Note: swish.tt
was renamed from search.tt Jun 03, 2004.)
To use either of these modules you will need to adjust the "template"
configuration setting. Examples for both templating systems are provided in
the configuration settings near the top of the
swish.cgi program.
Use of these modules is an advanced usage of
swish.cgi and are provided
as examples only.
All of the output generation modules are passed a hash with the results from the
search, plus other data use to create the output page. You can see this hash
by using the debugging option "dump" or by using the included
SWISH::TemplateDumper module:
~/swishtest >cat .swishcgi.conf
return {
title => 'Search the Apache Documentation',
template => {
package => 'SWISH::TemplateDumper',
},
};
And run a query. For example:
http://localhost/swishtest/swish.cgi?query=install
Why are there three different highlighting modules?¶
Three are three highlighting modules included with the swish-e distribution.
Each is a trade-off of speed vs. accuracy:
SWISH::DefaultHighlight - reasonably fast, but does not highlight phrases
SWISH::PhraseHighlight - reasonably slow, but is reasonably accurate
SWISH::SimpleHighlight - fast, some phrases, but least accurate
Eh, the default is actually "PhraseHighlight". Oh well.
All of the highlighting modules slow down the script. Optimizations to these
modules are welcome!
My ISP doesn't provide access to the web server logs¶
There are a number of options. One way it to use the CGI::Carp module. Search in
the swish.cgi script for:
use Carp;
# Or use this instead -- PLEASE see perldoc CGI::Carp for details
# use CGI::Carp qw(fatalsToBrowser warningsToBrowser);
And change it to look like:
#use Carp;
# Or use this instead -- PLEASE see perldoc CGI::Carp for details
use CGI::Carp qw(fatalsToBrowser warningsToBrowser);
This should be only for debugging purposes, as if used in production you may end
up sending quite ugly and confusing messages to your browsers.
Why does the output show (NULL)?¶
Swish-e displays (NULL) when attempting to display a property that does not
exist in the index.
The most common reason for this message is that you did not use StoreDescription
in your config file while indexing.
StoreDescription HTML* <body> 200000
That tells swish to store the first 200,000 characters of text extracted from
the body of each document parsed by the HTML parser. The text is stored as
property "swishdescription".
The index must be recreated after changing the swish-e configuration.
Running:
~/swishtest > ./swish-e -T index_metanames
will display the properties defined in your index file.
This can happen with other properties, too. For example, this will happen when
you are asking for a property to display that is not defined in swish.
~/swishtest > ./swish-e -w install -m 1 -p foo
# SWISH format: 2.1-dev-25
# Search words: install
err: Unknown Display property name "foo"
.
~/swishtest > ./swish-e -w install -m 1 -x 'Property foo=<foo>\n'
# SWISH format: 2.1-dev-25
# Search words: install
# Number of hits: 14
# Search time: 0.000 seconds
# Run time: 0.038 seconds
Property foo=(NULL)
.
To check that a property exists in your index you can run:
~/swishtest > ./swish-e -w not dkdk -T index_metanames | grep foo
foo : id=10 type=70 META_PROP:STRING(case:ignore) *presorted*
Ok, in this case we see that "foo" is really defined as a property.
Now let's make sure
swish.cgi is asking for "foo" (sorry for
the long lines):
~/swishtest > SWISH_DEBUG=command ./swish.cgi > /dev/null
Debug level set to: 3
Enter a query [all]:
Using 'not asdfghjklzxcv' to match all records
Enter max results to display [1]:
---- Running swish with the following command and parameters ----
./swish-e \
-w \
'swishdefault=(not asdfghjklzxcv)' \
-b \
1 \
-m \
1 \
-f \
index.swish-e \
-s \
swishrank \
desc \
swishlastmodified \
desc \
-x \
'<swishreccount>\t<swishtitle>\t<swishdescription>\t<swishlastmodified>\t<swishdocsize>\t<swishdocpath>\t<fos>\t<swishrank>\t<swishdocpath>\n' \
-H \
9
If you look carefully you will see that the -x parameter has "fos"
instead of "foo", so there's our problem.
How do I use the SWISH::API perl module with swish.cgi?¶
Use the "use_library" configuration directive:
use_library => 1,
This will only provide improved performance when running under mod_perl or other
persistent environments.
Why does the "Run time" differ when using the SWISH::API module¶
When using the SWISH::API module the run (and search) times are calculated
within the script. When using the swish-e binary the swish-e program reports
the times. The "Run time" may include the time required to load and
compile the SWISH::API module.
MOD_PERL¶
This script can be run under mod_perl (see
http://perl.apache.org). This will
improve the response time of the script compared to running under CGI by
loading the swish.cgi script into the Apache web server.
You must have a mod_perl enabled Apache server to run this script under
mod_perl.
Configuration is simple. In your httpd.conf or your startup.pl file you need to
load the script. For example, in httpd.conf you can use a perl section:
<perl>
use lib '/usr/local/apache/cgi-bin'; # location of the swish.cgi file
use lib '/home/yourname/swish-e/example/modules'; # modules required by swish.cgi
require "swish.cgi";
</perl>
Again, note that the paths used will depend on where you installed the script
and the modules. When running under mod_perl the swish.cgi script becomes a
perl module, and therefore the script does not need to be installed in the
cgi-bin directory. (But, you can actually use the same script as both a CGI
script and a mod_perl module at the same time, read from the same location.)
The above loads the script into mod_perl. Then to configure the script to run
add this to your httpd.conf configuration file:
<location /search>
PerlSetVar Swish_Conf_File /home/yourname/swish-e/myconfig.pl
allow from all
SetHandler perl-script
PerlHandler SwishSearch
</location>
Note that you use the "Swish_Conf_File" setting in httpd.conf to tell
the script which config file to use. This means you can use the same script
(and loaded modules) for different search sites (running on the same Apache
server). You can just specify differnt config files for each Location and they
can search different indexes and have a completely different look for each
site, but all share the same code.
Note that the config files are cached in the swish.cgi script. Changes to
the config file will require restarting the Apache server before they will be
reloaded into the swish.cgi script. This avoids calling
stat() for
every request.
Unlike CGI, mod_perl does not change the current directory to the location of
the script, so your settings for the swish binary and the path to your index
files must be absolute paths (or relative to the server root).
Using the SWISH::API module with mod_perl will provide the most performance
improvements. Use of the SWISH::API module can be enabled by the configuration
setting "use_library":
use_library => 1,
Without highlighting code enabled, using the SWISH::API module resulted in about
20 requests per second, where running the swish-e binary slowed the script
down to about 8 requests per second.
Note that the highlighting code is slow. For the best search performance turn
off highlighting. In your config file you can add:
highlighting => 0, # disable highlighting
and the script will show the first 500 chars of the description (or whatever you
set for "max_chars"). Without highlight one test was processing
about 20 request per second. With The "PhraseHighlight" module that
dropped to a little better than two requests per second,
"DefaultHighlight" was about 2.3 request per second, and
"SimpleHighlight" was about 6 request per second.
Experiement with different highlighting options when testing performance.
Please post to the swish-e discussion list if you have any questions about
running this script under mod_perl.
Here's some general request/second on an Athlon XP 1800+ with 1/2GB RAM, Linux
2.4.20.
Highlighting Mode
None Phrase Default Simple
Using SWISH::API 45 1.5 2 12
----------------------------------------------------------------------------
Using swish-e 12 1.3 1.8 7.5
binary
As you can see the highlighting code is a limiting factor.
SpeedyCGI¶
SpeedyCGI (also called PersistentPerl) is another way to run Perl scripts
persistently. SpeedyCGI is good if you do not have mod_perl available or do
not have root access. SpeedyCGI works on Unix systems by loading the script
into a "back end" process and keeping it in memory between requests.
New requests are passed to the back end processes which avoids the startup
time required by a Perl CGI script.
Install SpeedyCGI from
http://daemoninc.com/ (your OS may provide a packaged
version of SpeedyCGI) and then change the first line of swish.cgi. For
example, if the speedy binary is installed in /usr/bin/speedy, use the line:
#! /usr/bin/speedy -w -- -t60
The -w option is passed to Perl, and all options following the double-dash are
SpeedyCGI options.
Note that when using SpeedyCGI configuration data is cached in memory. If you
change the swish.cgi configuration file (.swishcgi.conf) then touch the main
swish.cgi script to force reloading of configuration data.
Spidering¶
There are two ways to spider with swish-e. One uses the "http" input
method that uses code that's part of swish. The other way is to use the new
"prog" method along with a perl helper program called
"spider.pl".
Here's an example of a configuration file for spidering with the
"http" input method. You can see that the configuration is not much
different than the file system input method. (But, don't use the http input
method -- use the -S prog method shown below.)
# Define what to index
IndexDir http://www.myserver.name/index.html
IndexOnly .html .htm
IndexContents HTML* .html .htm
DefaultContents HTML*
StoreDescription HTML* <body> 200000
MetaNames swishdocpath swishtitle
# Define http method specific settings -- see swish-e documentation
SpiderDirectory ../swish-e/src/
Delay 0
You index with the command:
swish-e -S http -c spider.conf
Note that this does take longer. For example, spidering the Apache documentation
on a local web server with this method took over a minute, where indexing with
the file system took less than two seconds. Using the "prog" method
can speed this up.
Here's an example configuration file for using the "prog" input
method:
# Define the location of the spider helper program
IndexDir ../swish-e/prog-bin/spider.pl
# Tell the spider what to index.
SwishProgParameters default http://www.myserver.name/index.html
IndexContents HTML* .html .htm
DefaultContents HTML*
StoreDescription HTML* <body> 200000
MetaNames swishdocpath swishtitle
Then to index you use the command:
swish-e -c prog.conf -S prog -v 0
Spidering with this method took nine seconds.
Stemmed Indexes¶
Many people enable a feature of swish called word stemming to provide
"fuzzy" search options to their users. The stemming code does not
actually find the "stem" of word, rather removes and/or replaces
common endings on words. Stemming is far from perfect, and many words do not
stem as you might expect. Plus, currently only English is supported. But, it
can be a helpful tool for searching your site. You may wish to create both a
stemmed and non-stemmed index, and provide a checkbox for selecting the index
file.
To enable a stemmed index you simply add to your configuration file:
UseStemming yes
If you want to use a stemmed index with this program and continue to highlight
search terms you will need to install a perl module that will stem words. This
section explains how to do this.
The perl module is included with the swish-e distribution. It can be found in
the examples directory (where you found this file) and called something like:
SWISH-Stemmer-0.05.tar.gz
The module should also be available on CPAN (
http://search.cpan.org/).
Here's an example session for installing the module. (There will be quite a bit
of output when running make.)
% gzip -dc SWISH-Stemmer-0.05.tar.gz |tar xof -
% cd SWISH-Stemmer-0.05
% perl Makefile.PL
or
% perl Makefile.PL PREFIX=$HOME/perl_lib
% make
% make test
(perhaps su root at this point if you did not use a PREFIX)
% make install
% cd ..
Use the
PREFIX if you do not have root access or you want to install the
modules in a local library. If you do use a PREFIX setting, add a "use
lib" statement to the top of this swish.cgi program.
For example:
use lib qw(
/home/bmoseley/perl_lib/lib/site_perl/5.6.0
/home/bmoseley/perl_lib/lib/site_perl/5.6.0/i386-linux/
);
Once the stemmer module is installed, and you are using a stemmed index, the
"swish.cgi" script will automatically detect this and use the
stemmer module.
DISCLAIMER¶
Please use this CGI script at your own risk.
This script has been tested and used without problem, but you should still be
aware that any code running on your server represents a risk. If you have any
concerns please carefully review the code.
See
http://www.w3.org/Security/Faq/www-security-faq.html
Security on Windows questionable.
SUPPORT¶
The SWISH-E discussion list is the place to ask for any help regarding SWISH-E
or this example script. See
http://swish-e.org.
Before posting please review:
http://swish-e.org/2.2/docs/INSTALL.html#When_posting_please_provide_the_
Please do not contact the author or any of the swish-e developers directly.
LICENSE¶
swish.cgi $Revision: 1830 $ Copyright (C) 2001 Bill Moseley search@hank.org
Example CGI program for searching with SWISH-E
This program is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free Software
Foundation; either version 2 of the License, or (at your option) any later
version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
A PARTICULAR PURPOSE. See the GNU General Public License for more details.
AUTHOR¶
Bill Moseley