NAME Moops - Moops Object-Oriented Programming Sugar SYNOPSIS use Moops; role NamedThing { has name => (is => "ro", isa => Str); } class Person with NamedThing; class Company with NamedThing; class Employee extends Person { has job_title => (is => "rwp", isa => Str); has employer => (is => "rwp", isa => InstanceOf["Company"]); method change_job ( Object $employer, Str $title ) { $self->_set_job_title($title); $self->_set_employer($employer); } method promote ( Str $title ) { $self->_set_job_title($title); } } STATUS Unstable. Will probably never be stable. A lot of the modules that Moops is built on have problems. In particular, Devel::CallParser is broken on a lot of Perl versions, and Parse::Keyword has *fundamental errors in the way it handles closures* (which Moops works around using PadWalker). Moops will remain on CPAN for the foreseeable future and I'll continue to accept patches that fix bugs, but don't expect any new features to be added. For a replacement, consider Zydeco. It's not a drop-in replacement but it has a similar syntax to Moops, and provides many of the same features. DESCRIPTION Moops is sugar for declaring and using roles and classes in Perl. The syntax is inspired by MooseX::Declare, and Stevan Little's p5-mop-redux project (which is in turn partly inspired by Perl 6). Moops has fewer than half of the dependencies as MooseX::Declare, loads in about 25% of the time, and the classes built with it run significantly faster. Moops does not use Devel::Declare, instead using Perl's pluggable keyword API; *this requires Perl 5.14 or above*. Moops uses Moo to build classes and roles by default, but allows you to use Moose if you desire. (And Mouse experimentally.) Classes The `class` keyword declares a class: class Foo { # ... } A version number can be provided: class Foo 1.2 { # ... } If no version is provided, your class' $VERSION variable is set to the empty string; this helps the package be seen by Class::Load. If your class extends an existing class through inheritance, or consumes one or more roles, these can also be provided when declaring the class. class Foo::Bar 1.2 extends Foo 1.1 with Magic::Monkeys { # ... } If you use Moops within a package other than `main`, then package names used within the declaration are "qualified" by that outer package, unless they contain "::". So for example: package Quux; use Moops; class Foo { } # declares Quux::Foo class Xyzzy::Foo # declares Xyzzy::Foo extends Foo { } # ... extending Quux::Foo class ::Baz { } # declares Baz If you wish to use Moose or Mouse instead of Moo; include that in the declaration: class Foo using Moose { # ... } It's also possible to create classes `using Tiny` (Class::Tiny), but there's probably little point in it, because Moops uses Moo internally, so the more capable Moo is already loaded and in memory. (The `using` option is exempt from the package qualification rules mentioned earlier.) Moops uses MooseX::MungeHas in your classes so that the `has` keyword supports some Moo-specific features, even when you're using Moose or Mouse. Specifically, it supports `is => 'rwp'`, `is => 'lazy'`, `builder => 1`, `clearer => 1`, `predicate => 1`, and `trigger => 1`. If you're using Moo, the MooX::late extension is enabled too, which allows Moose-isms in Moo too. With the combination of these features, there should be very little difference between Moo, Mouse and Moose `has` keywords. Moops uses Lexical::Accessor to provide you with private (lexical) attributes - that is, attributes accessed via a coderef method in a lexical variable. class Foo { lexical_has foo => ( isa => Int, accessor => \(my $_foo), default => 0, ); method increment_foo () { $self->$_foo( 1 + $self->$_foo ); } method get_foo () { return $self->$_foo; } } my $x = Foo->new; $x->increment_foo(); # ok say $x->get_foo(); # says "1" $x->$_foo(42); # dies; $_foo does not exist in this scope Moose classes are automatically accelerated using MooseX::XSAccessor if it's installed. Note that it is possible to declare a class with an empty body; use a trailing semicolon. class Employee extends Person with Employment; If using Moose or Mouse, classes are automatically made immutable. namespace::autoclean is automatically used in all classes. Between the class declaration and its body, Attribute::Handlers-style attributes may be provided: class Person :mutable { # ... } class Employee extends Person with Employment :mutable; The following attributes are defined for classes: * `:assertions` - enables assertion checking (see below) * `:dirty` - suppresses namespace::autoclean * `:fp` - use Function::Parameters instead of Kavorka * `:mutable` - suppresses making Moose classes immutable * `:ro` - make attributes declared with `has` default to 'ro' * `:rw` - make attributes declared with `has` default to 'rw' * `:rwp` - make attributes declared with `has` default to 'rwp' Roles Roles can be declared similarly to classes, but using the `role` keyword. role Stringable using Moose # we know you meant Moose::Role { # ... } Roles do not support the `extends` option. Roles can be declared to be `using` Moo, Moose, Mouse or Tiny. (Note that if you're mixing and matching role frameworks, there are limitations to which class builders can consume which roles. Mouse is generally the least compatible; Moo and Moose classes should be able to consume each others' roles; Moo can also consume Role::Tiny roles.) If roles use Moo, the MooX::late extension is enabled. namespace::autoclean is automatically used in all roles. Roles take similar Attribute::Handlers-style attributes to classes, but don't support `:mutable`. A note on consuming roles In a standard: class MyClass with MyRole { ...; } You should note that role composition is delayed to happen at the *end* of the class declaration. This is usually what you want. However the interaction between method modifiers and roles is complex, and *sometimes* you'll want the role to be applied to the class part-way through the declaration. In this case you can use a `with` statement *inside* the class declaration: class MyClass { ...; with "MyRole"; ...; } Namespaces The `namespace` keyword works as above, but declares a package without any class-specific or role-specific semantics. namespace Utils { # ... } namespace::autoclean is not automatically used in namespaces. Attribute::Handlers-style attributes are supported for namespaces, but most of the built-in attributes make any sense without class/role semantics. (`:assertions` does.) Traits written as Moops extensions may support namespaces. Functions and Methods Moops uses Kavorka to declare functions and methods within classes and roles. Kavorka provides the `fun` and `method` keywords. class Person { use Scalar::Util 'refaddr'; has name => (is => 'rwp'); # Moo attribute method change_name ( Str $newname ) { $self->_set_name( $newname ) unless $newname eq 'Princess Consuela Banana-Hammock'; } fun is_same_as ( Object $x, Object $y ) { refaddr($x) == refaddr($y) } } my $phoebe = Person->new(name => 'Phoebe'); my $ursula = Person->new(name => 'Ursula'); Person::is_same_as($phoebe, $ursula); # false Note function signatures use type constraints from Types::Standard; MooseX::Types and MouseX::Types type constraints should also work, *provided you use their full names, including their package*. The `is_same_as` function above could have been written as a class method like this: class Person { # ... method is_same_as ( $class: Object $x, Object $y ) { refaddr($x) == refaddr($y) } } # ... Person->is_same_as($phoebe, $ursula); # false The `method` keyword is not provided within packages declared using `namespace`; only within classes and roles. See also Kavorka::Manual::Methods and Kavorka::Manual::Functions. Within Moose classes and roles, the MooseX::KavorkaInfo module is loaded, to allow access to method signatures via the meta object protocol. (This is currently broken for `around` method modifiers.) In Moops prior to 0.025, Function::Parameters was used instead of Kavorka. If you wish to continue to use Function::Parameters in a class you can use the `:fp` attribute: class Person :fp { ...; } Or to do so for all classes in a lexical scope: use Moops function_parameters_everywhere => 1; class Person { ...; } Or the environment variable `MOOPS_FUNCTION_PARAMETERS_EVERYWHERE` can be set to true to enable it globally, but this feature is likely to be removed eventually. Method Modifiers Within classes and roles, `before`, `after` and `around` keywords are provided for declaring method modifiers. These use the same syntax as `method`. If your class or role is using Moose or Mouse, then you also get `augment` and `override` keywords. See also Kavorka::Manual::MethodModifiers. Multi Methods Moops uses Kavorka to implement multi subs and multi methods. See also Kavorka::Manual::MultiSubs. Type Constraints The Types::Standard type constraints are exported to each package declared using Moops. This allows the standard type constraints to be used as barewords. Type constraints can be used in attribute definitions (`isa`) and method signatures. Because Types::Standard is based on Type::Tiny, the same type constraints may be used whether you build your classes and roles with Moo, Moose our Mouse. Alternative libraries can be imported using the `types` option; a la: class Document types Types::XSD::Lite { has title => (is => 'rw', isa => NormalizedString); } Note that if an alternative type constraint library is imported, then Types::Standard is *not* automatically loaded, and needs to be listed explicitly: class Document types Types::Standard, Types::XSD::Lite { # ... } Type libraries built with Type::Library, MooseX::Types and MouseX::Types should all work. Bear in mind that type constraints from, say, a MooseX::Types library won't be usable in, say, Moo attribute definitions. However, it's possible to wrap them with Type::Tiny, and make them usable: class Foo types MooseX::Types::Common::Numeric using Moo { use Types::TypeTiny qw( to_TypeTiny ); has favourite_number => ( is => 'rwp', isa => to_TypeTiny(PositiveInt) ); } Type Libraries You can use the `library` keyword to declare a new type library: library MyTypes extends Types::Standard declares EmptyString, NonEmptyString { declare EmptyString, as Str, where { length($_) == 0 }; declare NonEmptyString, as Str, where { length($_) > 0 }; } class StringChecker types MyTypes { method check ( Str $foo ) { return "empty" if EmptyString->check($foo); return "non-empty" if NonEmptyString->check($foo); return "impossible?!"; } } Libraries declared this way can extend existing type libraries written with Type::Library, MooseX::Types or MouseX::Types. Note that this also provides a solution to the previously mentioned problem of using MooseX::Types type libraries in Moo classes: library MyWrapper extends MooseX::Types::Common::Numeric; class Foo types MyWrapper using Moo { has favourite_number => ( is => 'rwp', isa => PositiveInt, ); } Constants The useful constants `true` and `false` are imported into all declared packages. (Within classes and roles, namespace::autoclean will later remove them from the symbol table, so they don't form part of your package's API.) These constants can help make attribute declarations more readable. has name => (is => 'ro', isa => Str, required => true); Further constants can be declared using the `define` keyword (see PerlX::Define): namespace Maths { define PI = 3.2; } Constants declared this way will *not* be swept away by namespace::autoclean, and are considered part of your package's API. Assertions Declared packages can contain assertions (see PerlX::Assert). These are normally optimized away at compile time, but you can force them to be checked using the `:assertions` attribute. class Foo { assert(false); # not checked; optimized away } class Bar :assertions { assert(false); # checked; fails; throws exception } More Sugar strict and FATAL warnings are imported into all declared packages. However the `uninitialized`, `void`, `once` and `numeric` warning categories are explicitly excluded, as are any warnings categories added to Perl after version 5.14. Perl 5.14 features, including the `state` and `say` keywords, and sane Unicode string handling are imported into all declared packages. Try::Tiny is imported into all declared packages. Scalar::Util's `blessed` and Carp's `confess` are imported into all declared packages. Outer Sugar The "outer" package, where the `use Moops` statement appears also gets a little sugar: strict, the same warnings as "inner" packages, and Perl 5.14 features are all switched on. true is loaded, so you don't need to do this at the end of your file: 1; Custom Sugar It is possible to inject other functions into all inner packages using: use Moops imports => [ 'List::Util' => [qw( first reduce )], 'List::MoreUtils' => [qw( any all none )], ]; This is by far the easiest way to extend Moops with project-specific extras. There is a shortcut for injecting strictures into all inner packages: use Moops -strict; EXTENDING Moops is written to hopefully be fairly extensible. Extending Moops via imports The easiest way to extend Moops is to inject additional imports into the inner packages using the technique outlined in "Custom Sugar" above. You can wrap all that up in a module: package MoopsX::Lists; use base 'Moops'; use List::Util (); use List::MoreUtils (); sub import { my ($class, %opts) = @_; push @{ $opts{imports} ||= [] }, ( 'List::Util' => [qw( first reduce )], 'List::MoreUtils' => [qw( any all none )], ); $class->SUPER::import(%opts); } 1; Now people can do `use MoopsX::Lists` instead of `use Moops`. Extending Moops via keyword traits Roles in the `Moops::TraitFor::Keyword` namespace are automatically loaded and applied to keyword objects when a corresponding Attribute::Handlers-style attribute is seen. For examples extending Moops this way, see the Moops::TraitFor::Keyword::dirty, Moops::TraitFor::Keyword::mutable, Moops::TraitFor::Keyword::ro, Moops::TraitFor::Keyword::rw and Moops::TraitFor::Keyword::rwp traits. Extending Moops via parser traits For more complex needs, you can create a trait which will be applied to Moops::Parser. Parser traits might want to override: * The `keywords` class method, which returns the list of keywords the parser can handle. * The `class_for_keyword` object method, which returns the name of a subclass of Moops::Keyword which will be used for translating the result of parsing the keyword into a string using Perl's built-in syntax. Hopefully you'll be able to avoid overriding the `parse` method itself, as it has a slightly messy API. Your `class_for_keyword` subclass can either be a direct subclass of Moops::Keyword, or of Moops::Keyword::Class or Moops::Keyword::Role. The keyword subclass might want to override: * The `known_relationships` class method, which returns a list of valid inter-package relationships such as `extends` and `using` for the current keyword. * The `qualify_relationship` class method, which, when given the name of an inter-package relationship, indicates whether it should be subjected to package qualification rules (like `extends` and `with` are, but `using` is not). * The `version_relationship` class method, which, when given the name of an inter-package relationship, indicates whether it should accept a version number. * The `generate_package_setup` object method which returns a list of strings to inject into the package. * The `arguments_for_function_parameters` object method which is used by the default `generate_package_setup` method to set up the arguments to be passed to Function::Parameters. * The `check_prerequisites` method which performs certain pre-flight checks and may throw an exception. Hopefully you'll be able to avoid overriding the `generate_code` method. You can apply your trait using: use Moops traits => [ 'Moops::TraitFor::Parser::FooKeyword', 'Moops::TraitFor::Parser::BarKeyword', ]; BUGS If seeing test failures on threaded Perl 5.21+, it may be a bug in Devel::CallParser 0.002. Try installing Alt::Devel::CallParser::ButWorking. Please report any other bugs to <http://rt.cpan.org/Dist/Display.html?Queue=Moops>. GOTCHAS Certain software (like the CPAN indexer!) greps Perl source code looking for `package` statements to determine which packages a module provides. Moops uses `class` and `role` keywords to declare packages, so it may be necessary to provide some dummy `package` statements at the end of your module if you need to support such software. SUPPORT IRC: support is available through in the *#moops* channel on irc.perl.org <http://www.irc.perl.org/channels.html>. For general Moose/Moo queries which don't seem to be related to Moops' syntactic sugar, your question may be answered more quickly in the *#moose* channel. Web: if you ask a question on PerlMonks in Seekers of Perl Wisdom <http://www.perlmonks.org/?node_id=479> with "Moops" in the subject line, it should be answered pretty quickly. There is a moops tag <http://stackoverflow.com/questions/tagged/moops> on StackOverflow. SEE ALSO Similar: MooseX::Declare, <https://github.com/stevan/p5-mop-redux>. Main functionality exposed by this module: Moo/MooX::late, Kavorka, Try::Tiny, Types::Standard, namespace::autoclean, true, PerlX::Assert. Internals fueled by: Keyword::Simple, Module::Runtime, Import::Into, Attribute::Handlers. <http://en.wikipedia.org/wiki/The_Bubble_Boy_(Seinfeld)>. AUTHOR Toby Inkster <tobyink@cpan.org>. COPYRIGHT AND LICENCE This software is copyright (c) 2013-2014 by Toby Inkster. This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself. DISCLAIMER OF WARRANTIES THIS PACKAGE IS PROVIDED "AS IS" AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.