The Script

  #!/usr/bin/perl
  use YourApp;
  YourApp->run;

The Application Class

  package YourApp;
  use App::Cmd::Setup -app;
  1;

When a new application instance is created, it loads all of the command classes it can find, looking for modules under the Command namespace under its own name. In the above snippet, for example, YourApp will look for any module with a name starting with "YourApp::Command::".

The Command Classes

  # ABSTRACT: set up YourApp
  package YourApp::Command::initialize;
  use YourApp -command;
  1;

Now, a user can run this command, but he'll get an error:

  $ yourcmd initialize
  YourApp::Command::initialize does not implement mandatory method 'execute'

Oops! This dies because we haven't told the command class what it should do when executed. This is easy, we just add some code:

  sub execute {
    my ($self, $opt, $args) = @_;

    print "Everything has been initialized.  (Not really.)\n";
  }

Now it works:

  $ yourcmd initialize
  Everything has been initialized.  (Not really.)

Default Commands

commands

lists available commands.

  $yourcmd commands
  Available commands:

    commands: list the application's commands
        help: display a command's help screen

        init: set up YourApp
    

Note that by default the commands receive a description from the "# ABSTRACT" comment in the respective command's module, or from the "=head1 NAME" Pod section.

help

allows one to query for details on command's specifics.

  $yourcmd help initialize
   yourcmd initialize [-z] [long options...]

          -z --zero        ignore zeros
    

Of course, it's possible to disable or change the default commands, see App::Cmd.

Arguments and Options

  $ yourcmd reset -zB --new-seed xyzzy foo.db bar.db

"-zB" and "--new-seed xyzzy" are "options" and "foo.db" and "bar.db" are "arguments."

With a properly configured command class, the above invocation results in nicely formatted data:

  $opt = {
    zero      => 1,
    no_backup => 1, #default value
    new_seed  => 'xyzzy',
  };

  $args = [ qw(foo.db bar.db) ];

Arguments are processed by Getopt::Long::Descriptive (GLD). To customize its argument processing, a command class can implement a few methods: "usage_desc" provides the usage format string; "opt_spec" provides the option specification list; "validate_args" is run after Getopt::Long::Descriptive, and is meant to validate the $args, which GLD ignores. See Getopt::Long for format specifications.

The first two methods provide configuration passed to GLD's "describe_options" routine. To improve our command class, we might add the following code:

  sub usage_desc { "yourcmd %o [dbfile ...]" }

  sub opt_spec {
    return (
      [ "skip-refs|R",  "skip reference checks during init", ],
      [ "values|v=s@",  "starting values", { default => [ 0, 1, 3 ] } ],
    );
  }

  sub validate_args {
    my ($self, $opt, $args) = @_;

    # we need at least one argument beyond the options; die with that message
    # and the complete "usage" text describing switches, etc
    $self->usage_error("too few arguments") unless @$args;
  }

Global Options

To add a "--help" option to all your commands create a base class like:

  package MyApp::Command;
  use App::Cmd::Setup -command;

  sub opt_spec {
    my ( $class, $app ) = @_;
    return (
      [ 'help' => "this usage screen" ],
      $class->options($app),
    )
  }

  sub validate_args {
    my ( $self, $opt, $args ) = @_;
    if ( $opt->{help} ) {
      my ($command) = $self->command_names;
      $self->app->execute_command(
        $self->app->prepare_command("help", $command)
      );
      exit;
    }
    $self->validate( $opt, $args );
  }

Where "options" and "validate" are "inner" methods which your command subclasses implement to provide command-specific options and validation.

Note: this is a new file, previously not mentioned in this tutorial and this tip does not recommend the use of global_opt_spec which offers an alternative way of specifying global options.

Ignoring a Single Module.

  package YourApp::Command::foo::NotACommand;

  use YourApp -ignore;

  <whatever you want here>

This will register this package's namespace with YourApp to be excluded from its plugin validation magic. It otherwise makes no changes to "::NotACommand"'s namespace, does nothing magical with @ISA, and doesn't bolt any hidden functions on.

Its also probably good to notice that it is ignored only by "YourApp". If for whatever reason you have two different "App::Cmd" systems under which "::NotACommand" is visible, you'll need to set it ignored to both.

This is probably a big big warning NOT to do that.

Ignoring Multiple modules from the App level.

  sub should_ignore {
    my ( $self, $command_class ) = @_;
    return 1 if not $command_class->isa( 'App::Cmd::Command' );
    return;
  }

This will prematurely mark for ignoring all packages that don't subclass "App::Cmd::Command", which causes non-commands ( or perhaps commands that are coded wrongly / broken ) to be silently skipped.

Note that by overriding this method, you will lose the effect of any of the other ignore mechanisms completely. If you want to combine the original "should_ignore" method with your own logic, you'll want to steal "Moose"'s "around" method modifier.

  use Moose::Util;

  Moose::Util::add_method_modifier( __PACKAGE__, 'around', [
    should_ignore => sub {
      my $orig = shift;
      my $self = shift;
      return 1 if not $command_class->isa( 'App::Cmd::Command' );
      return $self->$orig( @_ );
  }]);