Config::Auto - Magical config file parser

  use Config::Auto;

  # Not very magical at all.
  my $config = Config::Auto::parse("myprogram.conf", format => "colon");

  # Considerably more magical.
  my $config = Config::Auto::parse("myprogram.conf");

  # Highly magical.
  my $config = Config::Auto::parse();

This module was written after having to write Yet Another Config File Parser for some variety of colon-separated config. I decided never again.

When you call CWConfig::Auto::parse with no arguments, we first look at CW$0 to determine the program's name. Let's assume that's CWsnerk. We look for the following files:

    snerkconfig
    ~/snerkconfig
    /etc/snerkconfig
    /usr/local/etc/snerkconfig
    snerk.config
    ~/snerk.config
    /etc/snerk.config
    /usr/local/etc/snerk.config
    snerkrc
    ~/snerkrc
    /etc/snerkrc
    /usr/local/etc/snerkrc
    .snerkrc
    ~/.snerkrc
    /etc/.snerkrc
    /usr/local/etc/.snerkrc

Additional search paths can be specified with the CWpaths option.

We take the first one we find, and examine it to determine what format it's in. The algorithm used is a heuristic which is a fancy way of saying that it doesn't work. (Mark Dominus.) We know about colon separated, space separated, equals separated, XML, Perl code, Windows INI, BIND9 and irssi style config files. If it chooses the wrong one, you can force it with the CWformat option.

If you don't want it ever to detect and execute config files which are made up of Perl code, set CW$Config::Auto::DisablePerl = 1.

When using the perl format, your configuration file will be eval'd. This will cause taint errors. To avoid these warnings, set CW$Config::Auto::Untaint = 1.

When using the perl format, your configuration file will be eval'd using do(file). This will cause taint errors if the filename is not untainted. To avoid these warnings, set CW$Config::Auto::Untaint = 1. This setting will not untaint the data in your configuration file and should only be used if you trust the source of the filename.

Then the file is parsed and a data structure is returned. Since we're working magic, we have to do the best we can under the circumstances - You rush a miracle man, you get rotten miracles. (Miracle Max) So there are no guarantees about the structure that's returned. If you have a fairly regular config file format, you'll get a regular data structure back. If your config file is confusing, so will the return structure be. Isn't life tragic?

Here's what we make of some common Unix config files:

/etc/resolv.conf:

    $VAR1 = {
          'nameserver' => [ '163.1.2.1', '129.67.1.1', '129.67.1.180' ],
          'search' => [ 'oucs.ox.ac.uk', 'ox.ac.uk' ]
        };

/etc/passwd:

    $VAR1 = {
          'root' => [ 'x', '0', '0', 'root', '/root', '/bin/bash' ],
          ...
        };

/etc/gpm.conf:

    $VAR1 = {
          'append' => '""',
          'responsiveness' => '',
          'device' => '/dev/psaux',
          'type' => 'ps2',
          'repeat_type' => 'ms3'
        };

/etc/nsswitch.conf:

    $VAR1 = {
          'netgroup' => 'nis',
          'passwd' => 'compat',
          'hosts' => [ 'files', 'dns' ],
          ...
    };

Although CWConfig::Auto is at its most magical when called with no parameters, its behavior can be reined in by use of one or two arguments.

If a filename is passed as the first argument to CWparse, the same paths are checked, but CWConfig::Auto will look for a file with the passed name instead of the CW$0-based names.

 use Config::Auto;

 my $config = Config::Auto::parse("obscure.conf");

The above call will cause CWConfig::Auto to look for:

 obscure.conf
 ~/obscure.conf
 /etc/obscure.conf

Parameters after the first are named. forces CWConfig::Auto to interpret the contents of the configuration file in the given format without trying to guess. add additional directories to the search paths. The current directory is searched first, then the paths specified with the path parameter. CWpath can either be a scalar or a reference to an array of paths to check.

CWConfig::Auto recognizes the following formats:

* perl => perl code

* colon => colon separated (e.g., key:value)

* space => space separated (e.g., key value)

* equal => equal separated (e.g., key=value)

* bind => bind style (not available)

* irssi => irssi style (not available)

* xml => xml (via XML::Simple)

* ini => .ini format (via Config::IniFiles)

* list => list (e.g., ??)

When using a Perl config file, the configuration is borked

Give CWConfig::Auto more hints (e.g., add #!/usr/bin/perl to beginning of file) or indicate the format in the parse() command.

BIND9 and irssi file format parsers currently don't exist. It would be good to add support for CWmutt and CWvim style CWset-based RCs.

This module by Jos Boumans, CWkane@cpan.org.

This module is copyright (c) 2003 Jos Boumans <kane@cpan.org>. All rights reserved.

This library is free software; you may redistribute and/or modify it under the same terms as Perl itself.