This is the first in a series of entries which I’m pulling over articles from an old blog, revising, etc…

The Problem

One of the neat things about Ruby is it’s ability to create accessor methods for you, by simply declaring attr_reader, attr_writer, or attr_accessor.  However, there’s not an easy way to define a default value for an attribute.  The original version of this code was based off of Create getter and setter on a valorized variable and ideas from Ruby Quiz #67: Metakoans.  The main difference between it and Sandro Paganotti’s version is that you can pass in a block which will be evaluated each time it is invoked (in case the value might depend on something else).  Additionally, it will work with boolean types.

The Solution

class Object
  def self.attribute(*arg,&block)
    (name, default) = arg
    short_name = name.to_s.sub(/\?/,"")
    self.send(:define_method, name) {
      if instance_variables.include? "@#{short_name}"
           self.instance_eval "@#{short_name}"
      else
        if block_given?
          instance_eval &block
        else
          default
        end
      end
    }
    self.send(:define_method, "#{short_name}="){ |value|
      self.instance_eval "@#{short_name} = value"
    }
  end
end

In order to use it, you could do something like this:

class Foo
  attribute :bar
  attribute(:fud) {instance_variables.include?("@bar") ? @bar : ' '}
  attribute :fi, 10
  attribute :flag?, true
end

And, in action, it would look like this:

>> f=Foo.new
=> #<Foo:0x2ad8c6af7728>
>> f.bar # defined, but not set to anything (nil)
=> nil
>> f.fud # since bar is not defined, return an empty space
=> " "
>> f.fi # just return the default
=> 10
>> f.flag? # and here's a boolean
=> true
>> f.bar = 5 # now we set bar
=> 5
>> f.fud # Since bar is set, then fud will return its value
=> 5
>> f.fud = 20 # now we explicitly set fud, and the block is no longer used
=> 20
>> f.fud # fud returns 20 as expected
=> 20
>> f.bar # and, as expected, changing fud has no effect on bar
=> 5

There might be a better way to do this; but one of the things I love about Ruby is that you can add new features to the language very easily.

At various points in my life, I’ve used many methods of parsing command line options, from rolling my own, to using getopt or other libraries.  Recently I came across a new one which has its own view of how to parse commandline arguments.

Trollop is, as they put it, “Yet Another Fine CommandLine Argument Parsing Library” (YAFCLAP).  It’s goal is to be as easy as possible.  You can install it as a single file in lib, or use it as a gem.  Trollop doesn’t care.  To install, you can download it from rubyforge, or use the ever popular:

gem install trollop

(you may need to prepend sudo, or run it as root).

Let’s take a look at a simple example:

require '<a class="zem_slink" title="RubyGems" rel="homepage" href="http://docs.rubygems.org">rubygems</a>'
require 'trollop'
opts = Trollop::options do
  opt :host, "Host to connect to", :default => "localhost"
  opt :port, "jmx remote port", :default => 4444
  opt :interval, "How often do we poll (in seconds)", :default => 10
  opt :count, "How many intervals, 0 == infinite, use ^C (SIGINT) to exit and write output", :default => 0
end
p opts

The first argument to opt is the name of the option.  We don’t have to specify a switch; it’s smart enough to figure out a mapping for them.  The second is a description of the argument.  You can specify a default value and/or a type.  In this example, we’ve specified defaults.  And here it is in use:

$ ruby trollop_simple.rb
{:interval=>10, :count=>0, :help=>false, :host=>"localhost", :port=>4444}
$ ruby trollop_simple.rb -h
Error: option '-h' needs a parameter.
Try --help for help.
$ ruby trollop_simple.rb --help
Options:
      --host, -h <s>:   Host to connect to (default: localhost)
      --port, -p <i>:   jmx remote port (default: 4444)
  --interval, -i <i>:   How often do we poll (in seconds) (default: 10)
     --count, -c <i>:   How many intervals, 0 == infinite, use ^C (SIGINT) to
                        exit and write output (default: 0)
          --help, -e:   Show this message
 
 

Notice that it has set up mappings for us as well as created a nicely formatted help — though it’s a default, without a lot of information.   Let’s add a version and usage:

require 'rubygems'
require 'trollop'
opts = Trollop::options do
  version <<EOF
jmx_mon -- we be jammin' with version zz9-za aka Arthur Dent
Don't forget your towel!
EOF
 
  banner <<EOF
jmx_mon is a program to monitor a jmx bean\'s value over time.
It can output the data as a csv, a graph in a pdf, or both.
 
Usage jmx_mon [options]
where [options] are:
EOF
  opt :host, "Host to connect to", :default => "localhost"
  opt :port, "jmx remote port", :default => 4444
  opt :interval, "How often do we poll (in seconds)", :default => 10
  opt :count, "How many intervals, 0 == infinite, use ^C (SIGINT) to exit and write output", :default => 0
end
p opts

Now let’s see what we have:

$ ruby trollop_simple.rb --help
jmx_mon is a program to monitor a jmx bean's value over time.
It can output the data as a csv, a graph in a pdf, or both.
 
Usage jmx_mon [options]
where [options] are:
      --host, -h <s>:   Host to connect to (default: localhost)
      --port, -p <i>:   jmx remote port (default: 4444)
  --interval, -i <i>:   How often do we poll (in seconds) (default: 10)
     --count, -c <i>:   How many intervals, 0 == infinite, use ^C (SIGINT) to
                        exit and write output (default: 0)
       --version, -v:   Print version and exit
          --help, -e:   Show this message
 
$ ruby trollop_simple.rb -v
jmx_mon -- we be jammin' with version zz9-za aka Arthur Dent
Don't forget your towel!
 

We now have a nicer usage, as well as version.  Also, it will stop automagickally if we ask for help or the version.

All of the options are placed in the opts hash; let’s see what we can do with it:

require 'rubygems'
require 'trollop'
 
aliases = {
      :MEMORY_HEAP => "Heap memory usage, multiple values",
      :MEMORY_HEAP_USED => "Used heap memory"
}
 
opts = Trollop::options do
  version <<EOF
jmx_mon -- we be jammin' with version zz9-za aka Arthur Dent
Don't forget your towel!
EOF
 
  banner <<EOF
jmx_mon is a program to monitor a jmx bean\'s value over time.
It can output the data as a csv, a graph in a pdf, or both.
 
Usage jmx_mon [options]
where [options] are:
EOF
  opt :host, "Host to connect to", :default => "localhost"
  opt :port, "jmx remote port", :default => 4444
  opt :interval, "How often do we poll (in seconds)", :default => 10
  opt :count, "How many intervals, 0 == infinite, use ^C (SIGINT) to exit and write output", :default => 0
  opt :aliases, "Print defined bean aliases & exit.", :default => false
  opt :towel, "Don't forget this!", :default => false
end
 
if opts[:aliases_given]
  aliases.each_pair do |key, value|
    puts "#{key}: #{value}"
  end
  exit
end
 
Trollop::die :towel, "You forgot your towel.  You've been eaten by a Ravenous Bugblatter Beast of Traal" unless opts[:towel]
 
p opts

We’ve added two more options, one for printing out aliases, and another indicating whether or not you have your towel — with deadly results should you forget.

$ ruby trollop_simple.rb -a    
MEMORY_HEAP: Heap memory usage, multiple values
MEMORY_HEAP_USED: Used heap memory
 
$ ruby trollop_simple.rb   
Error: argument --towel You forgot your towel.  You've been eaten by a Ravenous Bugblatter Beast of Traal.
Try --help for help.
 

There’s more that you can do with trollop; the website and documentation gives more examples.  However, I’d suggest considering trollop the next time you need to parse the command line.

On a particular, nameless, heterogenous project using ruby on rails, ant, and shell scripts, there was an instance where data was not being properly cleaned before going into the database (primarily due to user errors in property files which were used by ant to feed the database).   As a result, the database was filled with extraneous spaces which were causing many issues.  A perfect illustration of GIGO.

One problem was that there were multiple databases, each with their own data.   By the time the issue was discovered, a fix was needed soonest, so the solution was less than elegant, but it does illustrate the use of callbacks in ActiveRecord:

module DbParanoia
# This is a really stupid thing to have to back-fit; it is making
# sure, however, that the users enter good data.  We ensure that
# strings are stripped on read and write.
 
  def trim_space
    self.attributes.each do |attr|
      attr.strip! if attr.instance_of? String
    end
  end
 
  alias_method :after_find, :trim_space
  alias_method :before_save, :trim_space
end

The above module is using the after_find and before_save callbacks to make sure that:

1. Any spaces coming back from the database are trimmed.
2. Any new records and/or updates will have the spaces trimmed.

You can use this method to perform pretty much any sort of validation or data manipulation of the data of an ActiveRecord class.  Other callbacks include:

• after_create
• after_destroy
• after_save
• after_update
• after_validation
• after_validation_on_create
• after_validation_on_update
• before_create
• before_destroy
• before_save
• before_update
• before_validation
• before_validation_on_create
• before_validation_on_update

You can read more about ActiveRecord::Callbacks at http://api.rubyonrails.org/classes/ActiveRecord/Callbacks.html.

One of the features of Ruby’s Kernel module (which is imported into Object, so it’s available in every class) is the binding.  The binding allows you to keep a pointer to the scope at a particular location in your application and then later evaluate code from that location.

For example (and this came up as the result of a question on the ruby talk  mailing list), one of the things you can do is obtain a list of the local variables of a method:

def my_vars(a)
  b=a
  c=b+a
  puts local_variables
end

returns the three variables, a, b, and c.  That’s fine, we’re still within the scope of the method.  But how about outside of our method?  This is where #binding and the Binding class come into play:

def outside(b)
  puts eval("local_variables", b)
end
 
def my_vars(a)
  b=a
  c=b+a
  outside(binding)
end

One thing to note: in order to work with the binding, we have to #eval code within its scope.  Thus the call to #eval with the method we are invoking and the binding.  This leads to the question from the mailing list:

we already have #has_block? to see if a block was passed. So how about
a #has_arguments? to query if _any_ arguments have been passed — Intransition

So how would we go about writing a #has_arguments? — here is a first stab:

class Object
  def has_arguments?(b)
    vars = eval("local_variables",b)
    return false if vars.length == 0
    vars.each do |v|
      return false if eval("#{v}.nil?",b) 
    end
    return true
  end
end
 
class Foo
  def initialize(first=nil)
    if (has_arguments? binding)
      puts "We're defined"
    else
      puts "Not!"
    end
  end
end

The only real drawback here is that we’ve got to pass in the binding itself.  Let’s try again:

class Object
  def has_arguments?(&b)
    vars = eval("local_variables",b.binding)
    return false if vars.length == 0
    vars.each do |v|
      return false if eval("#{v}.nil?",b) 
    end
    return true
  end
end
 
class Foo
  def initialize(first=nil)
    if (has_arguments? {}) # note the empty block
      puts "We're defined"
    else
      puts "Not!"
    end
  end
end

In this case, instead of explicitly calling #binding, we’re passing in an empty block — the block has it’s own scope, but it happens to contain the scope of then enclosing method.  It’s still not perfect, however:

1. You need to pass in an empty block.
2. It fails if you have local variables other than those passed in as arguments. (Pointed out by Joel VanderWerf)

Ok, it’s got some serious flaws (Charles Nutter posted a different way of solving the problem, which does work for all cases, and doesn’t use #binding).  But it does give an idea of what you can do with #binding.

#method_missing is a very useful part of the ruby language.  However, there’s one common mistake that developers make when using it.  They forget to add a call to #super.

Let’s take a look:

$ irb
>> class Foo
>>   end
=> nil
>> f=Foo.new
=> #<Foo:0xb75122fc>
>> f.i_do_not_exist
NoMethodError: undefined method `i_do_not_exist' for #<Foo:0xb75122fc>
 from (irb):5
>> class Foo
>>   def method_missing(symbol,*args)
>>     if (symbol.to_s === "i_do_not_exist")
>>       puts "Yes, actually you do"
>>       end
>>     end
>>   end
=> nil
>> f.i_do_not_exist
Yes, actually you do
=> nil
>> f.i_exist
=> nil

Unfortunately, we no longer get the error when we try to invoke a method which does not exist.  When we add an invocation to super, it’ll behave as expected:

>> class Foo
  def method_missing(symbol,*args)
    if (symbol.to_s === "i_do_not_exist")
      puts "Yes, actually you do"
    else   
      super
    end
  end
end
>>   >>     >>       >>       ?>         >>       >>     >>   => nil
>> f.i_exist
NoMethodError: undefined method `i_exist' for #<Foo:0xb75122fc>
 from (irb):31:in `method_missing'
 from (irb):35
>> f.i_do_not_exist
Yes, actually you do
=> nil
>> 

One of the neat things about ruby is that, in general, operators are methods, so they can be overridden.

When working with multidimensional arrays, some languages allow you to do the following:

a[0,0]

whereas ruby uses:

a[0][0]

Which makes sense, since Ruby implements it as an array of arrays and you’re chaining methods to get to the element. But, since we can override the [] method, we can then say a[0,0].

class Grid
  attr_reader :cols, :rows
 
  # initialize our grid and optionally prepopulate it.
  # klass can behave in the following fashions:
  # nil -- nil
  # a value (like "", 0, true) -- that value
  # a class which responds to new -- an instance of that class
   def initialize(rows=1, cols=1, klass = nil)
    @rows = rows
    @cols = cols
    @grid = (0 ... @rows).collect do |y|
      (0 ... @cols).collect do |x|
        if klass.nil? then
          nil
        elsif klass.respond_to? :new
          klass.new
        else
          klass
        end
      end
    end
  end
 
# This method allows us to say +grid[x, y]+ instead of +grid[y][x]+
  def [](x, y=nil)
    (x,y) = x if x.instance_of? Array
    @grid[y][x]
  end
 
  # This method allows us to set the value of a grid grid entry via
  # +grid[x, y]+ instead of +grid[y][x]+
  def []=(*args)
    if (args[0].instance_of? Array)
      ((x, y),value) = args
    else
      (x,y,value) = args
    end
    @grid[y][x] = value
  end
 
end

So, it can be used like this:

$ irb
>> require 'grid'
=> true
>> g=Grid.new
=> #< Grid:0xb747d760 @cols=1, @rows=1, @grid=[[nil]] >
>> g=Grid.new(2,2)
=> #< Grid:0xb747bb04 @cols=2, @rows=2, @grid=[[nil, nil], [nil, nil]] >
>> g=Grid.new(2,2,0)
=> #< Grid:0xb7479aac @cols=2, @rows=2, @grid=[[0, 0], [0, 0]] >
>> g=Grid.new(2,2,String)
=> #< Grid:0xb7477810 @cols=2, @rows=2, @grid=[["", ""], ["", ""]] >
>> g=Grid.new(2,2,false)
=> #< Grid:0xb74755d8 @cols=2, @rows=2, @grid=[[false, false], [false, false]] >
>> g[0,0]
=> false
>> g[0,0]=true
=> true
>> g
=> #< Grid:0xb74755d8 @cols=2, @rows=2, @grid=[[true, false], [false, false]] >

It’s an interesting technique — and given that most people think of arrays/coordinates in pairs, I think that it can make the code more “readable”.

To paren, or not to paren: that is the question:

Whether ’tis nobler in the mind to suffer

The bugs and errors of maintenance programs,

Or to take arms against a host of typos,

And by debugging end them? To hack: to slash;

No more; and by a slash to say we end

The heart-ache and the thousand natural shocks

Of cut and paste code, ’tis a maturation

Devoutly to be wished. To hack, not slash;

Not slash: perchance to code: ay, there’s the rub;

For in that new programme what code may come

When we have shuffled off this unix box,

Must give us pause: there’s the respect

That makes wuffy of so long life;

– Not William Shakespeare

There’s been a good deal of conversation of late on comp.lang.ruby as to whether or not to use parentheses. And when to use them. Ruby is such that parentheses are optional, except, of course, when they’re not. Here’s some examples of where they are not:

irb(main):002:0> "parenthesis".length()
=> 11
irb(main):003:0> "parenthesis".length
=> 11
 
...
 
irb(main):009:0&gt; class <a class="zem_slink" title="Foo" rel="wikipedia" href="http://en.wikipedia.org/wiki/Foo">Foo</a>
irb(main):010:1&gt; attr_accessor :bar, :groo
irb(main):011:1&gt; attr_reader(:gruff)
irb(main):012:1&gt; end
=&gt; nil
irb(main):013:0&gt; f=Foo.new
=&gt; #<foo:0x2a98ad0bc0>
irb(main):014:0&gt; f.gruff
=&gt; nil
irb(main):015:0&gt; f.bar
=&gt; nil
 
</foo:0x2a98ad0bc0>

Notice that Ruby also doesn’t care if you use parenthesis or not for the arguments; in this case it doesn’t matter whether or not the parenthesis are there, because the meaning is clear, at least to the compiler. And now, for something completely different, an instance where meaning is not clear to the compiler.

irb(main):016:0> class Foo
irb(main):017:1> def fud(thing)
irb(main):018:2> thing.reverse
irb(main):019:2> end
irb(main):020:1> end
=> nil
irb(main):021:0> f.fud "hi"
=> "ih"
irb(main):022:0> f.fud "hi" ? true : false
(irb):22: warning: string literal in condition
NoMethodError: undefined method `reverse' for true:TrueClass
        from (irb):18:in `fud'
        from (irb):22
        from /usr2/jest/tools//lib/ruby/site_ruby/1.8/rubygems/dependency.rb:19
irb(main):023:0> f.fud("hi") ? true : false
=> true

In the first instance, it is trying to evaluate

"hi" ? true : false

f.fud("hi") ? true : false

In terms of background, I learned C 20 years ago. I cut my teeth on basic back in 1980. So I’ve been around for a little while. I mention this because several posters mentioned that one’s preference for using parenthesis may depend on one’s background, with those who come from C potentially more likely to use parenthesis whether they were needed or not.

I am of the opinion that parenthesis can make the code harder to read.

attr_accessor :first, :last
has_many :pets

attr_accessor(:first, :last)
has_many(:pets)

So, what do you think? Is there a compelling reason to use parentheses?

In irbrc goodness, I touched briefly on my_methods, a method of discovering the methods of an object which are distinct from those inherited from Object. The method looks like this:

class Object
  def my_methods
    (self.methods - Object.methods).sort
  end
end

Here’s the promised code:

class Object
  def my_methods(_super=false)
    _methods = (_super) ? self.class.superclass.new.methods : Object.methods
    (self.methods - _methods).sort
  end
end

Here’s my (current) .irbrc, with comments

# I am including gems in my irb session.  Therefore, the next require
require 'rubygems'
 
# wirble is a wonderful gem which add nice features to irb
require 'wirble'
 
# Dr. Nic's useful helper gem.  It makes it so you can say:
#   foo.map_by_bar
# instead of having to say:
#   foo.map{|f| f.bar}
# more to the point you can do
#   people.map_by_first_and_last
# instead of
#   people.map{|p| [p.first, p.last]}
require 'map_by_method'
 
# what_methods is a way to see which methods, when performed on an object, return a particular value:
# >> "hi".what? 'h'
# "hi".chop! == "h"
# "hi".chop == "h"
# => ["chop!", "chop"]
require 'what_methods'
 
# pretty print --> provides a "pretty" view of an object
require 'pp'
 
# auto indent irb; it's useful for when you're just experimenting
IRB.conf[:AUTO_INDENT]=true
 
# start wirble (with color)
Wirble.init
#Wirble.colorize
 
# What are the methods which belong only to an object, and not those inherited from Object
class Object
  def my_methods
    (self.methods - Object.methods).sort
  end
end
 
# strip html tags from a string
class String
  def strip_tags
    self.gsub(/<\S[^><]*>/,"")
  end
end

I’ve shown mine. Show me yours?

Here are links to the referenced rubygems:

• what_methods
• map_by_method
• wirble