NAME main.rb

SYNOPSIS a class factory and dsl for generating command line programs real quick

URI http://codeforpeople.com/lib/ruby/ http://rubyforge.org/projects/codeforpeople/ http://github.com/ahoward/main

INSTALL gem install main

DESCRIPTION main.rb features the following:

- unification of option, argument, keyword, and environment parameter
  parsing
- auto generation of usage and help messages
- support for mode/sub-commands
- io redirection support
- logging hooks using ruby's built-in logging mechanism
- intelligent error handling and exit codes 
- use as dsl or library for building Main objects
- parsing user defined ARGV and ENV
- zero requirements for understanding the obtuse apis of *any* command
  line option parsers
- leather pants

in short main.rb aims to drastically lower the barrier to writing uniform command line applications.

for instance, this program

require 'main'

Main {
  argument 'foo'
  option 'bar'

  def run
    p params['foo']
    p params['bar']
    exit_success!
  end
}

sets up a program which requires one argument, 'bar', and which may accept one command line switch, '--foo' in addition to the single option/mode which is always accepted and handled appropriately: 'help', '--help', '-h'. for the most part main.rb stays out of your command line namespace but insists that your application has at least a help mode/option.

main.rb supports sub-commands in a very simple way

require 'main'

Main {
  mode 'install' do
    def run() puts 'installing...' end
  end

  mode 'uninstall' do
    def run() puts 'uninstalling...' end
  end
}

which allows a program, called 'a.rb', to be invoked as

ruby a.rb install

and

ruby a.rb uninstall

for simple programs main.rb is a real time saver but it's for more complex applications where main.rb's unification of parameter parsing, class configuration dsl, and auto-generation of usage messages can really streamline command line application development. for example the following 'a.rb' program:

require 'main'

Main {
  argument('foo'){
    cast :int
  }
  keyword('bar'){
    arity 2
    cast :float
    defaults 0.0, 1.0
  }
  option('foobar'){
    argument :optional
    description 'the foobar option is very handy'
  }
  environment('BARFOO'){
    cast :list_of_bool
    synopsis 'export barfoo=value'
  }

  def run
    p params['foo'].value
    p params['bar'].values
    p params['foobar'].value
    p params['BARFOO'].value
  end
}

when run with a command line of

BARFOO=true,false,false ruby a.rb 42 bar=40 bar=2 --foobar=a

will produce

42
[40.0, 2.0]
"a"
[true, false, false]

while a command line of

ruby a.rb --help

will produce

NAME
  a.rb

SYNOPSIS
  a.rb foo [bar=bar] [options]+

PARAMETERS
  * foo [ 1 -> int(foo) ]

  * bar=bar [ 2 ~> float(bar=0.0,1.0) ]

  * --foobar=[foobar] [ 1 ~> foobar ]
      the foobar option is very handy

  * --help, -h

  * export barfoo=value

and this shows how all of argument, keyword, option, and environment parsing can be declartively dealt with in a unified fashion - the dsl for all parameter types is the same - and how auto synopsis and usage generation saves keystrokes. the parameter synopsis is compact and can be read as

  * foo [ 1 -> int(foo) ]

    'one argument will get processed via int(argument_name)'

      1        : one argument
      ->       : will get processed (the argument is required)
      int(foo) : the cast is int, the arg name is foo

  * bar=bar [ 2 ~> float(bar=0.0,1.0) ]

    'two keyword arguments might be processed via float(bar=0.0,1.0)'

      2                  : two arguments
      ~>                 : might be processed (the argument is optional)
      float(bar=0.0,1.0) : the cast will be float, the default values are
                           0.0 and 1.0

  * --foobar=[foobar] [ 1 ~> foobar ]

    'one option with optional argument may be given directly'

  * --help, -h

    no synopsis, simple switch takes no args and is not required

  * export barfoo=value

    a user defined synopsis

SAMPLES

<========< samples/a.rb >========>

~ > cat samples/a.rb

require 'main'

ARGV.replace %w( 42 ) if ARGV.empty?

Main {
  argument('foo'){
    required                    # this is the default
    cast :int                   # value cast to Fixnum
    validate{|foo| foo == 42}   # raises error in failure case 
    description 'the foo param' # shown in --help
  }

  def run
    p params['foo'].given?
    p params['foo'].value
  end
}

~ > ruby samples/a.rb

true
42

<========< samples/b.rb >========>

~ > cat samples/b.rb

require 'main'

ARGV.replace %w( 40 1 1 ) if ARGV.empty?

Main {
  argument('foo'){
    arity 3                             # foo will given three times
    cast :int                           # value cast to Fixnum
    validate{|foo| [40,1].include? foo} # raises error in failure case 
    description 'the foo param'         # shown in --help
  }

  def run
    p params['foo'].given?
    p params['foo'].values
  end
}

~ > ruby samples/b.rb

true
[40, 1, 1]

<========< samples/c.rb >========>

~ > cat samples/c.rb

require 'main'

ARGV.replace %w( foo=40 foo=2 bar=false ) if ARGV.empty?

Main {
  keyword('foo'){
    required  # by default keywords are not required
    arity 2
    cast :float
  }
  keyword('bar'){
    cast :bool
  }

  def run
    p params['foo'].given?
    p params['foo'].values
    p params['bar'].given?
    p params['bar'].value
  end
}

~ > ruby samples/c.rb

true
[40.0, 2.0]
true
false

<========< samples/d.rb >========>

~ > cat samples/d.rb

require 'main'

ARGV.replace %w( --foo=40 -f2 ) if ARGV.empty?

Main {
  option('foo', 'f'){
    required  # by default options are not required, we could use 'foo=foo'
              # above as a shortcut
    argument_required
    arity 2
    cast :float
  }

  option('bar=[bar]', 'b'){  # note shortcut syntax for optional args
    # argument_optional      # we could also use this method
    cast :bool
    default false
  }

  def run
    p params['foo'].given?
    p params['foo'].values
    p params['bar'].given?
    p params['bar'].value
  end
}

~ > ruby samples/d.rb

true
[40.0, 2.0]
nil
false

<========< samples/e.rb >========>

~ > cat samples/e.rb

require 'main'

ARGV.replace %w( x y argument )

Main {
  argument 'argument'
  option 'option'

  def run() puts 'run' end

  mode 'a' do
    option 'a-option'
    def run() puts 'a-run' end
  end

  mode 'x' do
    option 'x-option'

    def run() puts 'x-run' end

      mode 'y' do
        option 'y-option'

        def run() puts 'y-run' end
      end
  end
}

~ > ruby samples/e.rb

y-run

<========< samples/f.rb >========>

~ > cat samples/f.rb

require 'main'

ARGV.replace %W( compress /data )

Main {
  argument('directory'){ description 'the directory to operate on' }

  option('force'){ description 'use a bigger hammer' }

  def run
    puts 'this is how we run when no mode is specified'
  end

  mode 'compress' do
    option('bzip'){ description 'use bzip compression' }

    def run
      puts 'this is how we run in compress mode' 
    end
  end

  mode 'uncompress' do
    option('delete-after'){ description 'delete orginal file after uncompressing' }

    def run
      puts 'this is how we run in un-compress mode' 
    end
  end
}

~ > ruby samples/f.rb

this is how we run in compress mode

<========< samples/g.rb >========>

~ > cat samples/g.rb

require 'main'

ARGV.replace %w( 42 ) if ARGV.empty?

Main {
  argument( 'foo' )
  option( 'bar' )

  run { puts "This is what to_options produces: #{params.to_options.inspect}" }
}

~ > ruby samples/g.rb

This is what to_options produces: {"help"=>nil, "foo"=>"42", "bar"=>nil}

<========< samples/h.rb >========>

~ > cat samples/h.rb

require 'main'

# block-defaults are instance_eval'd in the main instance and be combined with
# mixins
#
# ./h.rb   #=> forty-two
# ./h.rb a #=> 42 
# ./h.rb b #=> 42.0 
#

Main {
  fattr :default_for_foobar => 'forty-two' 

  option(:foobar) do
    default{ default_for_foobar }
  end

  mixin :foo do
    fattr :default_for_foobar => 42
  end

  mixin :bar do
    fattr :default_for_foobar => 42.0
  end


  run{ p params[:foobar].value }

  mode :a do
    mixin :foo
  end

  mode :b do
    mixin :bar
  end
}

~ > ruby samples/h.rb

"forty-two"

<========< samples/j.rb >========>

~ > cat samples/j.rb

#!/usr/bin/env ruby

require 'open-uri'

require 'main'
require 'digest/sha2'

# you have access to a sequel/amalgalite/sqlite db for free
#

Main {
  name :i_can_haz_db

  db {
    create_table(:mp3s) do
      primary_key :id
      String :url
      String :sha
    end unless table_exists?(:mp3s)
  }

  def run
    url = 'http://s3.amazonaws.com/drawohara.com.mp3/ween-voodoo_lady.mp3'
    mp3 = open(url){|fd| fd.read}
    sha = Digest::SHA2.hexdigest(mp3)

    db[:mp3s].insert(:url => url, :sha => sha)
    p db[:mp3s].all
    p db
  end
}

~ > ruby samples/j.rb

[{:url=>"http://s3.amazonaws.com/drawohara.com.mp3/ween-voodoo_lady.mp3", :sha=>"54c99ac7588dcfce1e70540b734805e9c69ff98dcca001e6f2bdec140fb0f9dc", :id=>1}, {:url=>"http://s3.amazonaws.com/drawohara.com.mp3/ween-voodoo_lady.mp3", :sha=>"54c99ac7588dcfce1e70540b734805e9c69ff98dcca001e6f2bdec140fb0f9dc", :id=>2}]
#<Sequel::Amalgalite::Database: "amalgalite://Users/ahoward/.i_can_haz_db/db.sqlite">

DOCS test/main.rb vim -p lib/main.rb lib/main/*rb API section below

HISTORY 5.1.0 - support for STATE_DIRNAME and STATE_BASENAME environment variables 5.0.0 ? 4.4.0 - support for automatic sequel/sqlite/amalgalite dbs for persistent state across invocations

    Main {
      db {
        create_table :foo do
          String key
          String val
        end unless table_exists? :foo
      }

      def run
        db[:foo].create(:key => 'using', :val => 'amalgalite')
      end
    }

  - support for automatic config files with auto populated template data

      Main {
        config :email => '[email protected]', :password => 'pa$$word'

        def run
          email = config[:email]
        end
      }

  - new paramter types :pathname, :path, :slug, :input, and :output 

  - input/output parameters.  can be filenames or '-' to supply
    stdin/stdout respectively

      Main {
        input :i
        output :o

        def run
          i = params[:i].value
          o = params[:o].value

          line = i.gets
          o.puts line
        end
      }

   - clean up warnings running with 'ruby -w'

   - fix a failing test

   - ability to ignore parameters in sub modes

      Main {
        argument :foo
        argument :bar

        def run
          p param[:bar].value
        end

        mode :ignoring do
          params[:foo].ignore!
        end
      }

4.0.0 - avoid duping ios. new methods Main.push_ios! and Main.pop_ios! are utilized for testing. this was done to make it simple to wrap daemon/servolux programs around main, althought not strictly required. not the version bump - there is not reason to expect existing main programs to break, but it is and interface change which requires a major version bump.

3.0.0 - major refactor to support modes via module/extend vs. subclassing. MIGHT NOT be backward compatible, though no known issues thus far.

2.9.0 - support ruby 1.9

2.8.3 - support for block defaults

2.8.2 - fixes and tests for negative arity/attr arguments, options, eg

    argument(:foo){
      arity -1  
    }

    def run  # ARGV == %w( a b c )
      p foo  #=> %w( a b c )
    end

  thanks nathan

2.8.1 - move from attributes.rb to fattr.rb

2.8.0 - added 'to_options' method for Parameter::Table. this allows you to convert all the parameters to a simple hash.
for example

    Main {
      option 'foo'
      argument 'baz'

      run { puts params.to_options.inspect } 

    }

2.7.0 - removed bundled arrayfields and attributes. these are now dependancies mananged by rubygems. a.k.a. you must have rubygems installed for main to work.

2.6.0 - added 'mixin' feaature for storing, and later evaluating a block of code. the purpose of this is for use with modes where you want to keep your code dry, but may not want to define something in the base class for all to inherit. 'mixin' allows you to define the code to inherit once and the selectively drop it in child classes (modes) on demand. for example

    Main {
      mixin :foobar do
        option 'foo'
        option 'bar'
      end

      mode :install do
        mixin :foobar
      end

      mode :uninstall do
        mixin :foobar
      end

      mode :clean do
      end
    }

- mode definitions are now deferred to the end of the Main block, so you
  can do this

    Main {
      mode 'a' do
        mixin :foo
      end

      mode 'b' do
        mixin :foo
      end

      def inherited_method
        42
      end

      mixin 'foo' do
        def another_inherited_method
          'forty-two'
        end
      end
    }

- added sanity check at end of paramter contruction

- improved auto usage generation when arity is used with arguments

- removed 'p' shortcut in paramerter dsl because it collided with
  Kernel.p.  it's now called 'param'.  this method is availble *inside* a
  parameter definition

    option('foo', 'f'){
      synopsis "arity = #{ param.arity }"
    }
  
- fixed bug where '--' did not signal the end of parameter parsing in a
  getoptlong compliant way

- added (before/after)_parse_parameters, (before/after)_initialize, and
  (before/after)_run hooks

- fixed bug where adding to usage via

    usage['my_section'] = 'custom message'

  totally horked the default auto generated usage message

- updated dependancies in gemspec.rb for attributes (~> 5.0.0) and
  arrayfields (~> 4.3.0)

- check that client code defined run, iff not wrap_run! is called.  this is
  so mains with a mode, but no run defined, still function correctly when 
  passed a mode

- added new shortcut for creating accessors for parameters.  for example

    option('foo'){
      argument :required
      cast :int
      attr
    }

    def run
      p foo ### this attr will return the parameter's *value*
    end

  a block can be passed to specify how to extract the value from the
  parameter

    argument('foo'){
      optional
      default 21 
      cast :int
      attr{|param| param.value * 2}
    }

    def run
      p foo #=> 42 
    end

- fixed bug where 'abort("message")' would print "message" twice on exit
  if running under a nested mode (yes again - the fix in 2.4.0 wasn't
  complete)

- added a time cast, which uses Time.parse

    argument('login_time'){ cast :time }

- added a date cast, which uses Date.parse

    argument('login_date'){ cast :date }

2.5.0 - added 'examples', 'samples', and 'api' kewords to main dsl. each keyword takes a list of strings which will be included in the help message

    Main {
      examples "foobar example", "barfoo example"

      samples <<-txt
        do this

        don't do that
      txt

      api %(
        foobar string, hash

        barfoo hash, string
      ) 
    }

  results in a usage message with sections like

  ...

  EXAMPLES
    foobar example
    barfoo example

  SAMPLES
    do this

    don't do that

  API
    foobar string, hash

    barfoo hash, string

  ...

2.4.0 - fixed bug where 'abort("message")' would print "message" twice on exit if running under a nested mode.

- allowed parameters to be overridden completely in subclasses (modes)

2.3.0 - re-worked Main.new such that client code may define an #initialize methods and the class will continue to work. that is to say it's fine to do this

    Main {
      def initialize
        @a = 42
      end

      def run
        p @a
      end

      mode 'foo' do
        def run
          p @a
        end
      end
    }

  the client #initialize will be called *after* main has done it's normal
  initialization so things like @argv, @env, and @stdin will all be there
  in initialize.  of course you could have done this before but you'd have
  to both call super and call it with the correct arguments - now you can
  simply ignore it.

2.2.0 - added ability for parameter dsl error handlers to accept an argument, this will be passed the current error. for example

    argument(:x) do
      arity 42
    
      error do |e|
        case e
          when Parameter::Arity
        ...
      end
    end  

- refined the mode parsing a bit: modes can now be abbreviated to uniqness
  and, when the mode is ambiuous, a nice error message is printed, for
  example:

    ambiguous mode: in = (inflate or install)?

2.1.0 - added custom error handling dsl for parameters, this includes the ability to prepend, append, or replace the standard error handlers:

    require 'main'

    Main {
      argument 'x' do
        error :before do
          puts 'this fires *before* normal error handling using #instance_eval...'
        end

        error do
          puts 'this fires *instead of* normal error handling using #instance_eval...'
        end

        error :after do
          puts 'this fires *after* normal error handling using #instance_eval...'
        end
      end
      
      run(){ p param['x'].given? }
    }

- added ability to exit at any time bypassing *all* error handling using
  'throw :exit, 42' where 42 is the desired exit status.  throw without a
  status simply exits with 0.

- added 'help!' method which simply dumps out usage and exits

2.0.0 - removed need for proxy.rb via Main::Base.wrap_run! - added error handling hooks for parameter parsing - bundled arrayfields, attributes, and pervasives although gems are tried first - softened error messages for parameter parsing errors: certain classes of errors are now 'softspoken' and print only the message, not the entire stacktrace, to stderr. much nicer for users. this is configurable. - added subcommand/mode support - added support for user defined exception handling on top level exceptions/exits - added support for negative arity. this users ruby's own arity semantics, for example:

    lambda{|*a|}.arity     == -1
    lambda{|a,*b|}.arity   == -2
    lambda{|a,b,*c|}.arity == -3
    ...

  in otherwords parameters now support 'zero or more', 'one or more' ...
  'n or more' argument semantics

1.0.0 - some improved usage messages from jeremy hinegardner

0.0.2 - removed dependancy on attributes/arrayfields. main now has zero gem dependancies.

- added support for io redirection.  redirection of stdin, stdout, and
  stderr can be done to any io like object or object that can be
  inerpreted as a pathname (object.to_s)

- main objects can now easily be created and run on demand, which makes
  testing a breeze

    def test_unit_goodness!
      main = 
        Main.new{
          stdout StringIO.new 
          stderr '/dev/null'

          def run
            puts 42
          end
        }

      main.run
      main.stdout.rewind

      assert main.stdout.read == "42\n"
    end

- added API section to readme and called it 'docs'

- wrote a bunch more tests.  there are now 42 of them.

0.0.1

initial version.  this version extracts much of the functionality of alib's
(gen install alib) Alib.script main program generator and also some of jim's
freeze's excellent CommandLine::Aplication into what i hope is a simpler and
more unified interface 

API

Main {

###########################################################################

CLASS LEVEL API

###########################################################################

the name of the program, auto-set and used in usage

program 'foo.rb'

a short description of program functionality, auto-set and used in usage

synopsis "foo.rb arg [options]+"

long description of program functionality, used in usage iff set

description <<-hdoc
  this text will automatically be indented to the right level.

  it should describe how the program works in detail
hdoc

used in usage iff set

author '[email protected]'

used in usage

version '0.0.42'

stdin/out/err can be anthing which responds to read/write or a string

which will be opened as in the appropriate mode

stdin '/dev/null'
stdout '/dev/null'
stderr open('/dev/null', 'w')

the logger should be a Logger object, something 'write'-able, or a string

which will be used to open the logger. the logger_level specifies the

initalize verbosity setting, the default is Logger::INFO

logger(( program + '.log' ))
logger_level Logger::DEBUG

you can configure exit codes. the defaults are shown

exit_success # 0
exit_failure # 1
exit_warn    # 42

the usage object is rather complex. by default it's an object which can

be built up in sections using the

usage["BUGS"] = "something about bugs'

syntax to append sections onto the already pre-built usage message which

contains program, synopsis, parameter descriptions and the like

however, you always replace the usage object wholesale with one of your

chosing like so

usage <<-txt
  my own usage message
txt

###########################################################################

MODE API

###########################################################################

modes are class factories that inherit from their parent class. they can

be nested arbitrarily deep. usage messages are tailored for each mode.

modes are, for the most part, independant classes but parameters are

always a superset of the parent class - a mode accepts all of it's parents

paramters plus and additional ones

option 'inherited-option'
argument 'inherited-argument'

mode 'install' do
  option 'force' do
    description 'clobber existing installation'
  end

  def run
    inherited_method()
    puts 'installing...'
  end

  mode 'docs' do
    description 'installs the docs'

    def run
      puts 'installing docs...'
    end
  end
end

mode 'un-install' do
  option 'force' do
    description 'remove even if dependancies exist'
  end

  def run
    inherited_method()
    puts 'un-installing...'
  end
end

def run
  puts 'no mode yo?'
end

def inherited_method
  puts 'superclass_method...'
end

###########################################################################

PARAMETER API

###########################################################################

all the parameter types of argument|keyword|option|environment share this

api. you must specify the type when the parameter method is used.

alternatively used one of the shortcut methods

argument|keyword|option|environment. in otherwords

parameter('foo'){ type :option }

is synonymous with

option('foo'){ }

option 'foo' {
#
# required - whether this paramter must by supplied on the command line.
# note that you can create 'required' options with this keyword
#
  required # or required true
#
# argument_required - applies only to options.
#
  argument_required # argument :required
#
# argument_optional - applies only to options.
#
  argument_optional # argument :optional
#
# cast - should be either a lambda taking one argument, or a symbol
# designation one of the built in casts defined in Main::Cast.  supported
# types are :boolean|:integer|:float|:numeric|:string|:uri.  built-in
# casts can be abbreviated
#
  cast :int
#
# validate - should be a lambda taking one argument and returning
# true|false
#
  validate{|int| int == 42}
#
# synopsis - should be a concise characterization of the paramter.  a
# default synopsis is built automatically from the parameter.  this
# information is displayed in the usage message
#
  synopsis '--foo'
#
# description - a longer description of the paramter.  it appears in the
# usage also.
#
  description 'a long description of foo'
#
# arity - indicates how many times the parameter should appear on the
# command line.  the default is one.  negative arities are supported and
# follow the same rules as ruby methods/procs.
#
  arity 2
#
# default - you can provide a default value in case none is given.  the
# alias 'defaults' reads a bit nicer when you are giving a list of
# defaults for paramters of > 1 arity
#
  defaults 40, 2
#
# you can add custom per-parameter error handlers using the following
#
  error :before do
    puts 'this fires *before* normal error handling using #instance_eval...'
  end

  error do
    puts 'this fires *instead of* normal error handling using #instance_eval...'
  end

  error :after do
    puts 'this fires *after* normal error handling using #instance_eval...'
  end
}

###########################################################################

INSTANCE LEVEL API

###########################################################################

you must define a run method. it is the only method you must define.

def run
  #
  # all parameters are available in the 'params' hash and via the alias
  # 'param'.  it can be indexed via string or symbol.  the values are all
  # Main::Parameter objects
  #
    foo = params['foo']
  #
  # the given? method indicates whether or not the parameter was given on
  # the commandline/environment, etc.  in particular this will not be true
  # when a default value was specified but no parameter was given 
  #
    foo.given?
  #
  # the list of all values can be retrieved via 'values'.  note that this
  # is always an array.
  #
    p foo.values
  #
  # the __first__ value can be retrieved via 'value'.  note that this
  # never an array.
  #
    p foo.value
  #
  # the methods debug|info|warn|error|fatal are delegated to the logger
  # object
  #
    info{ "this goes to the log" }
  #
  # you can set the exit_status at anytime.  this status is used when
  # exiting the program.  exceptions cause this to be ext_failure if, and
  # only if, the current value was exit_success.  in otherwords an
  # un-caught exception always results in a failing exit_status
  #
    exit_status exit_failure
  #
  # a few shortcuts both set the exit_status and exit the program.
  #
    exit_success!
    exit_failure!
    exit_warn!
end

}