NAME

Config::MVP - multivalue-property package-oriented configuration

VERSION

version 2.200012

SYNOPSIS

If you want a useful synopsis, consider this code which actually comes from Config::MVP::Assembler:

  my $assembler = Config::MVP::Assembler->new;

  # Maybe you want a starting section:
  my $section = $assembler->section_class->new({ name => '_' });
  $assembler->sequence->add_section($section);

  # We'll add some values, which will go to the starting section:
  $assembler->add_value(x => 10);
  $assembler->add_value(y => 20);

  # Change to a new section...
  $assembler->change_section($moniker);

  # ...and add values to that section.
  $assembler->add_value(x => 100);
  $assembler->add_value(y => 200);

This doesn't make sense? Well, read on.

(You can also read the 2009 RJBS Advent Calendar article <http://advent.rjbs.manxome.org/2009/2009-12-20.html> on Config::MVP!)

DESCRIPTION

MVP is a mechanism for loading configuration (or other information) for libraries. It doesn't read a file or a database. It's a helper for things that do.

The idea is that you end up with a Config::MVP::Sequence object, and that you can use that object to fully configure your library or application. The sequence will contain a bunch of Config::MVP::Section objects, each of which is meant to provide configuration for a part of your program. Most of these sections will be directly related to a Perl library that you'll use as a plugin or helper. Each section will have a name, and every name in the sequence will be unique.

This is a pretty abstract set of behaviors, so we'll provide some more concrete examples that should help explain how things work.

EXAMPLE

Imagine that we've got a program called DeliveryBoy that accepts mail and does stuff with it. The "stuff" is entirely up to the user's configuration. He can set up plugins that will be used on the message. He writes a config file that's read by Config::MVP::Reader::INI, which is a thin wrapper around Config::MVP used to load MVP-style config from INI files.

Here's the user's configuration:

  [Whitelist]
  require_pgp = 1

  file = whitelist-family
  file = whitelist-friends
  file = whitelist-work

  [SpamFilter]
  filterset = standard
  max_score = 5
  action    = bounce

  [SpamFilter / SpamFilter_2]
  filterset = aggressive
  max_score = 5
  action    = tag

  [VerifyPGP]

  [Deliver]
  dest = Maildir

The user will end up with a sequence with five sections, which we can represent something like this:

  { name    => 'Whitelist',
    package => 'DeliveryBoy::Plugin::Whitelist',
    payload => {
      require_pgp => 1,
      files   => [ qw(whitelist-family whitelist-friends whitelist-work) ]
    },
  },
  { name    => 'SpamFilter',
    package => 'DeliveryBoy::Plugin::SpamFilter',
    payload => {
      filterset => 'standard',
      max_score => 5,
      action    => 'bounce',
    }
  },
  { name    => 'SpamFilter_2',
    package => 'DeliveryBoy::Plugin::SpamFilter',
    payload => {
      filterset => 'aggressive',
      max_score => 5,
      action    => 'tag',
    },
  },
  { name    => 'VerifyPGP',
    package => 'DeliveryBoy::Plugin::VerifyPGP',
    payload => { },
  },
  { name    => 'Deliver',
    package => 'DeliveryBoy::Plugin::Deliver',
    payload => { dest => 'Maildir' },
  },

The INI reader uses Config::MVP::Assembler to build up configuration section by section as it goes, so that's how we'll talk about what's going on.

Every section of the config file was converted into a section in the MVP sequence. Each section has a unique name, which defaults to the name of the INI section. Each section is also associated with a package, which was expanded from the INI section name. The way that names are expanded can be customized by subclassing the assembler.

Every section also has a payload -- a hashref of settings. Note that every entry in every payload is a simple scalar except for one. The "files" entry for the Whitelist section is an arrayref. Also, note that while it appears as "files" in the final output, it was given as "file" in the input.

Config::MVP provides a mechanism by which packages can define aliases for configuration names and an indication of what names correspond to "multi-value parameters." (That's part of the meaning of the name "MVP.") When the MVP assembler is told to start a section for "Whitelist" it expands the section name, loads the package, and inspects it for aliases and multivalue parameters. Then if multiple entries for a non-multivalue parameter are given, an exception can be raised. Multivalue parameters are always pushed onto arrayrefs and non-multivalue parameters are left as found.

...so what now?

So, once our DeliveryBoy program has loaded its configuration, it needs to initialize its plugins. It can do something like the following:

  my $sequence = $deliveryboy->load_config;

  for my $section ($sequence->sections) {
    my $plugin = $section->package->new( $section->payload );
    $deliveryboy->add_plugin( $section->name, $plugin );
  }

That's it! In fact, allowing this very, very block of code to load configuration and initialize plugins is the goal of Config::MVP.

The one thing not depicted is the notion of a "root section" that you might expect to see in an INI file. This can be easily handled by starting your assembler off with a pre-built section where root settings will end up. For more information on this, look at the docs for the specific components.

WHAT NEXT?

Making Packages work with MVP

Any package can be used as part of an MVP section. Packages can provide some methods to help MVP work with them. It isn't a problem if they are not defined

mvp_aliases

This method should return a hashref of name remappings. For example, if it returned this hashref:

  {
    file => 'files',
    path => 'files',
  }

Then attempting to set either the "file" or "path" setting for the section would actually set the "files" setting.

mvp_multivalue_args

This method should return a list of setting names that may have multiple values and that will always be stored in an arrayref.

The Assembler

Config::MVP::Assembler is a state machine that makes it easy to build up your MVP-style configuration by firing off a series of events: new section, new setting, etc. You might want to subclass it to change the class of sequence or section that's used or to change how section names are expanded into packages.

Sequences and Sections

Config::MVP::Sequence and Config::MVP::Section are the two most important classes in MVP. They represent the overall configuration and each section of the configuration, respectively. They're both fairly simple classes, and you probably won't need to subclass them, but it's easy.

Examples in the World

For examples of Config::MVP in use, you can look at Dist::Zilla or App::Addex.

AUTHOR

Ricardo Signes <rjbs@cpan.org>

CONTRIBUTORS

Alexandr Ciornii <alexchorny@gmail.com>
George Hartzell <hartzell@alerce.com>
Karen Etheridge <ether@cpan.org>
Kent Fredric <kentfredric@gmail.com>
Philippe Bruhat (BooK) <book@cpan.org>
Sven Kirmess <sven.kirmess@kzone.ch>

COPYRIGHT AND LICENSE

This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.