The Perl Toolchain Summit needs more sponsors. If your company depends on Perl, please support this very important event.

NAME

Getopt::Args2 - command-line argument and option processor

VERSION

0.0.11 (2018-08-18)

SYNOPSIS

    #!/usr/bin/env perl
    use Getopt::Args2;

    arg item => (
        isa      => 'Str',
        required => 1,
        comment  => 'the item to paint',
    );

    opt quiet => (
        isa     => 'Flag',
        alias   => 'q',
        comment => 'output nothing while working',
    );

    my $ref = optargs;

    print "Painting $ref->{item}\n" unless $ref->{quiet};

DESCRIPTION

Getopt::Args2 processes command line arguments, options, and subcommands according to the following definitions:

Command

A program run from the command line to perform a task.

Arguments

Arguments are positional parameters that pass information to a command. Arguments can be optional, but they should not be confused with Options below.

Options

Options are non-positional parameters that pass information to a command. They are generally not required to be present (hence the name Option) but that is configurable. All options have a long form prefixed by '--', and may have a single letter alias prefixed by '-'.

Subcommands

From the users point of view a subcommand is a special argument with its own set of arguments and options. However from a code authoring perspective subcommands are often implemented as stand-alone programs, called from the main script when the appropriate command arguments are given.

Getopt::Args2 is a re-write of the original OptArgs module with a cleaner code base and improved API. It should be preferred over OptArgs for new projects however OptArgs is not likely to disappear from CPAN anytime soon. Users converting to Getopt::Args2 from OptArgs need to be aware of the following:

Obvious API changes: cmd(), subcmd()

Commands and subcommands must now be explicitly defined using cmd() and subcmd().

class_optargs() no longer loads the class

Users must specifically require the class if they want to use it afterwards:

    my ($class, $opts) = class_optargs('App::demo');
    eval "require $class" or die $@; # new requirement
Bool options with no default display as "--[no-]bool"

A Bool option without a default is now displayed with the "[no-]" prefix. What this means in practise is that many of your existing Bool options should likely become Flag options instead.

Simple Commands

To demonstrate the simple use case (i.e. with no subcommands) lets put the code from the synopsis in a file called paint and observe the following interactions from the shell:

    $ ./paint
    usage: paint ITEM [OPTIONS...]

      arguments:
        ITEM          the item to paint

      options:
        --quiet, -q   output nothing while working

The optargs() function parses the command line according to the previous opt() and arg() declarations and returns a single HASH reference. If the command is not called correctly then an exception is thrown containing an automatically generated usage message as shown above. Because Getopt::Args2 fully knows the valid arguments and options it can detect a wide range of errors:

    $ ./paint wall message
    error: unexpected option or argument: red

So let's add that missing argument definition:

    arg message => (
        isa      => 'Str',
        comment  => 'the message to paint on the item',
        greedy   => 1,
    );

And then check the usage again:

    $ ./paint
    usage: paint ITEM [MESSAGE...] [OPTIONS...]

      arguments:
        ITEM          the item to paint
        MESSAGE       the message to paint on the item

      options:
        --quiet, -q   output nothing while working

Note that optional arguments are surrounded by square brackets, and that three dots (...) are postfixed to greedy arguments. A greedy argument will swallow whatever is left on the comand line:

    $ ./paint wall Perl is great
    Painting on wall: "Perl is great".

Note that it probably doesn't make sense to define any more arguments once you have a greedy argument. Let's imagine you now want the user to be able to choose the colour if they don't like the default. An option might make sense here:

    opt colour => (
        isa           => 'Str',
        default       => 'blue',
        show_default  => 1,
        comment       => 'the colour to use',
    );

This now produces the following usage output:

    usage: paint ITEM [MESSAGE...] [OPTIONS...]

      arguments:
        ITEM               the item to paint
        MESSAGE            the message to paint on the item
 
      options:
        --colour=STR, -c   the colour to use [default: blue]
        --quiet,      -q   output nothing while working

The command line is parsed first for arguments, then for options, in the same order in which they are defined. This probably only of interest if you are using trigger actions on your options (see FUNCTIONS below for details).

Multi-Level Commands

Commands with subcommands require a different coding model and syntax which we will describe over three phases:

Definitions

Your command structure is defined using calls to the cmd() and subcmd() functions. The first argument to both functions is the name of the Perl class that implements the (sub-)command.

    cmd 'App::demo' => (
        comment => 'the demo command',
        optargs => sub {
            arg command => (
                isa      => 'SubCmd',
                required => 1,
                comment  => 'command to run',
            );

            opt quiet => (
                isa     => 'Flag',
                alias   => 'q',
                comment => 'run quietly',
            );
        },
    );

    subcmd 'App::demo::foo' => (
        comment => 'demo foo',
        optargs => sub {
            arg action => (
                isa      => 'Str',
                required => 1,
                comment  => 'command to run',
            );
        },
    );

    subcmd 'App::demo::bar' => (
        comment => 'demo bar',
        optargs => sub {
            opt baz => (
                isa => 'Counter',
                comment => '+1',
            );
        },
    );

    # Command hierarchy for the above code:
    # demo COMMAND [OPTIONS...]
    #     demo foo ACTION [OPTIONS...]
    #     demo bar [OPTIONS...]

An argument of type 'SubCmd' is an explicit indication that subcommands can occur in that position. The command hierarchy is based upon the natural parent/child structure of the class names. This definition can be done in your main script, or in one or more separate packages or plugins, as you like.

Parsing

The class_optargs() function is called instead of optargs() to parse the @ARGV array and call the appropriate arg() and opt() definitions as needed. It's first argument is generally the top-level command name you used in your first cmd() call.

    my ($class, $opts) = class_optargs('App::demo');

    printf "Running %s with %s\n", $class, Dumper($opts)
      unless $opts->{quiet};

The additional return value $class is the name of the actual (sub-)command to which the $opts HASHref applies. Usage exceptions are raised just the same as with the optargs() function.

    error: unknown option "--invalid"

    usage: demo COMMAND [OPTIONS...]

        COMMAND       command to run
          bar           demo bar
          foo           demo foo

        --quiet, -q   run quietly

Note that options are inherited by subcommands.

Dispatch/Execution

Once you have the subcommand name and the option/argument hashref you can either execute the action or dispatch to the appropriate class/package as you like.

There are probably several ways to layout command classes when you have lots of subcommands. Here is one way that seems to work for this module's author.

lib/App/demo.pm, lib/App/demo/subcmd.pm

I typically put the actual (sub-)command implementations in lib/App/demo.pm and lib/App/demo/subcmd.pm. App::demo itself only needs to exists if the root command does something. However I tend to also make App::demo the base class for all subcommands so it is often a non-trivial piece of code.

lib/App/demo/OptArgs.pm

App::demo::OptArgs is where I put all of my command definitions with names that match the actual implementation modules.

    package App::demo::OptArgs;
    use Getopt::Args2;

    cmd 'App::demo' => (
        comment => 'the demo app',
        optargs => sub {
            #...
        },
    )

The reason for keeping this separate from lib/App/demo.pm is speed of loading. I don't want to have to load all of the modules that App::demo itself uses just to find out that I called the command incorrectly.

bin/demo

The command script itself is then usually fairly short:

    #!/usr/bin/env perl
    use Getopt::Args2 'class_optargs';
    use App::demo::OptArgs;

    my ($class, $opts) = class_optargs('App::demo');
    eval "require $class" or die $@;
    $class->new->run($opts);

The above does nothing more than load the definitions from App::demo::OptArgs, obtain the command name and options hashref, and then loads the appropriate package to run the command.

Formatting of Usage Messages

Usage messages attempt to present as much information as possible to the caller. Here is a brief overview of how the various types look and/or change depending on things like defaults.

The presentation of Bool options in usage messages is as follows:

    Name        Type        Default         Presentation
    ----        ----        -------         ------------
    option      Bool        undef           --[no-]option
    option      Bool        true            --no-option
    option      Bool        false           --option
    option      Counter     *               --option

The Flag option type is like a Bool that can only be set to true or left undefined. This makes sense for things such as --help or --version for which you never need to see a "--no" prefix.

    Name        Type        Default         Presentation
    ----        ----        -------         ------------
    option      Flag        always undef    --option

Note that Flags also makes sense for "negative" options which will only ever turn things off:

    Name        Type        Default         Presentation
    ----        ----        -------         ------------
    no_option   Flag        always undef    --no-option

    # In Perl
    opt no_foo => (
        isa     => 'Flag',
        comment => 'disable the foo feature',
    );

    # Then later do { } unless $opts->{no_foo}

The remaining types are presented as follows:

    Name        Type        isa_name        Presentation
    ----        ----        --------        ------------
    option      ArrayRef    -               --option=STR
    option      HashRef     -               --option=STR
    option      Int         -               --option=INT
    option      Num         -               --option=NUM
    option      Str         -               --option=STR
    option      *           XX              --option=XX

Defaults TO BE COMPLETED.

FUNCTIONS

The following functions are exported by default.

arg( $name, %parameters )

Define a command argument, for example:

    arg name => (
        comment  => 'the file to parse',
        default  => '-',
        greedy   => 0,
        isa      => 'Str',
        required => 1,
    );

The arg() function accepts the following parameters:

comment

Required. Used to generate the usage/help message.

default

The value set when the argument is not given. Conflicts with the 'required' parameter.

If this is a subroutine reference it will be called with a hashref containg all option/argument values after parsing the source has finished. The value to be set must be returned, and any changes to the hashref are ignored.

greedy

If true the argument swallows the rest of the command line.

fallback

A hashref containing an argument definition for the event that a subcommand match is not found. This parameter is only valid when isa is a SubCmd. The hashref must contain "isa", "name" and "comment" key/value pairs, and may contain a "greedy" key/value pair.

This is generally useful when you want to calculate a command alias from a configuration file at runtime, or otherwise run commands which don't easily fall into the Getopt::Args2 subcommand model.

isa

Required. Is mapped to a Getopt::Long type according to the following table:

     optargs         Getopt::Long
    ------------------------------
     'Str'           '=s'
     'Int'           '=i'
     'Num'           '=f'
     'ArrayRef'      's@'
     'HashRef'       's%'
     'SubCmd'        '=s'
required

Set to a true value when the caller must specify this argument. Conflicts with the 'default' parameter.

show_default

If set to a true value then usage messages will show the default value.

class_optargs( $class, [ @argv ] ) -> ($subclass, $opts)

Parse @ARGV by default (or @argv when given) for the arguments and options defined for command $class. @ARGV will first be decoded into UTF-8 (if necessary) from whatever I18N::Langinfo says your current locale codeset is.

Throws an error / usage exception object (typically Getopt::Args2::Usage) for missing or invalid arguments/options.

Returns the following two values:

$subclass

The actual subcommand name that was matched by parsing the arguments. This may be the same as $class.

$opts

a hashref containing key/value pairs for options and arguments combined.

As an aid for testing, if the passed in argument @argv (not @ARGV) contains a HASH reference, the key/value combinations of the hash will be added as options. An undefined value means a boolean option.

cmd( $class, %parameters ) -> Getopt::Args2::Cmd

Define a top-level command identified by $class which is typically a Perl package name. The following parameters are accepted:

abbrev

When set to true then subcommands can be abbreviated, up to their shortest, unique values.

comment

A description of the command. Required.

optargs

A subref containing calls to arg() and opt. Note that options are inherited by subcommands so you don't need to define them again in child subcommands.

opt( $name, %parameters )

Define a command option, for example:

    opt colour => (
        alias        => 'c',
        comment      => 'the colour to paint',
        default      => 'blue',
        show_default => 1,
        isa          => 'Str',
    );

Any underscores in $name are be replaced by dashes (-) for presentation and command-line parsing. The arg() function accepts the following parameters:

alias

A single character alias.

comment

Required. Used to generate the usage/help message.

default

The value set when the option is not used.

If this is a subroutine reference it will be called with a hashref containg all option/argument values after parsing the source has finished. The value to be set must be returned, and any changes to the hashref are ignored.

hidden

When true this option will not appear in usage messages unless the usage message is a help request.

This is handy if you have developer-only options, or options that are very rarely used that you don't want cluttering up your normal usage message.

isa

Required. Is mapped to a Getopt::Long type according to the following table:

    isa              Getopt::Long
    ---              ------------
     'ArrayRef'      's@'
     'Flag'          '!'
     'Bool'          '!'
     'Counter'       '+'
     'HashRef'       's%'
     'Int'           '=i'
     'Num'           '=f'
     'Str'           '=s'
isa_name

When provided this parameter will be presented instead of the generic presentation for the 'isa' parameter.

ishelp

When true creates a trigger parameter that generates a usage message exception. In other words it is just a shortcut for the following:

    opt help => (
        isa     => 'Flag',
        alias   => 'h',
        comment => 'print help message and exit',
        trigger => sub {
            my ( $cmd, $value ) = @_;
            die $cmd->usage(Getopt::Args2::STYLE_FULL);
        }
    );

Note that this option conflicts with the trigger parameter.

show_default

If set to a true value then usage messages will show the default value.

trigger

The trigger parameter lets you define a subroutine that is called immediately as soon as the option presence is detected. This is primarily to support --help or --version options which typically don't need the full command line to be processed before generating a response.

    opt version => (
        isa     => 'Flag',
        alias   => 'V',
        comment => 'print version string and exit',
        trigger => sub {
            my ( $cmd, $value ) = @_;
            die "$cmd version $VERSION\n";
        }
    );

The trigger subref is passed two parameters: a Getopt::Args2::Cmd object and the value (if any) of the option. The Getopt::Args2::Cmd object is an internal one, but one public interface is has (in addition to the usage() method described in 'ishelp' above) is the usage_tree() method which gives a usage overview of all subcommands in the command hierarchy.

    opt usage_tree => (
        isa     => 'Flag',
        alias   => 'U',
        comment => 'print usage tree and exit',
        trigger => sub {
            my ( $cmd, $value ) = @_;
            die $cmd->usage_tree;
        }
    );

    # demo COMMAND [OPTIONS...]
    #     demo foo ACTION [OPTIONS...]
    #     demo bar [OPTIONS...]
optargs( [@argv] ) -> HASHref

Parse @ARGV by default (or @argv when given) for the arguments and options defined for the default global command. Argument decoding and exceptions are the same as for class_optargs, but this function returns only the combined argument/option values HASHref.

subcmd( $class, %parameters ) -> Getopt::Args2::Cmd

Defines a subcommand identified by $class which must include the name of a previously defined (sub)command + '::'.

Accepts the same parameters as cmd() in addition to the following:

hidden

Hide the existence of this subcommand in usage messages created with Getopt::Args2::STYLE_NORMAL. This is handy if you have developer-only or rarely-used commands that you don't want cluttering up your normal usage message.

usage( $class, [STYLE] ) -> Str

Only exported on request, this function returns the usage string for the command $class.

SEE ALSO

Getopt::Long

This module is duplicated on CPAN as OptArgs2, to cover both its original name and yet still be found in the mess that is Getopt::*.

SUPPORT & DEVELOPMENT

This distribution is managed via github:

    https://github.com/mlawren/p5-OptArgs2/tree/devel

This distribution follows the semantic versioning model:

    http://semver.org/

Code is tidied up on Git commit using githook-perltidy:

    http://github.com/mlawren/githook-perltidy

AUTHOR

Mark Lawrence <nomad@null.net>

LICENSE

Copyright 2016 Mark Lawrence <nomad@null.net>

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 3 of the License, or (at your option) any later version.