module ShiftMeansAppend
=A0def <<(x) =A0#1
=A0end
end
class String
=A0include ShiftMeansAppend
end
class Array
=A0include ShiftMeansAppend
end
def append_to_self(x)
=A0unless x.is_a? ShiftMeansAppend =A0#2
=A0 =A0raise ArgumentError, "I can't trust this object's left-shift opera= tor."
=A0end
=A0x << x
end
Hello,
For the code above, I can't understand that:
#1 why redefine "<<" and set its action to nothing?
#2 why x.is_a? ShiftMeansAppend can be true when x is an array or a strin=
g?
Because, it seems to me, that the author of that code sees Ruby
through Java colored classes and doesn't understand Ruby all that
well.
I see that the code comes from a Ruby Cookbook book, and is discussing
how to not use duck-typing.
He's trying to do the equivalent of the technique in Java of using an
interface as a marker.
#1 is totally unnecessary, Since String and Array both have #<<
methods, and classes are searched for methods before any modules they
include, the method in the module will never be invoked. It seems
that the author thinks that like Java a method needs to be defined in
an interface or a superclass to be considered the same method. Not
so in Ruby which doesn't have interfaces. Modules just add methods to
the repertoire of an objects instances.
#2 is_a? tests whether the argument is anywhere on the chain of
classes and modules which are on chain searched for an objects
methods. So
x.is_a? y
is true if y is the class of x, or on the chain of superclasses of
that class, or a module included by one of those classes
ruby-1.9.1-p376 > a =3D Object.new
=3D> #<Object:0x00000101176068>
ruby-1.9.1-p376 > a.kind_of? Object
=3D> true
ruby-1.9.1-p376 > a.kind_of? Kernel
=3D> true
ruby-1.9.1-p376 > a.kind_of? Comparable
=3D> false
It also returns true for a module if the singleton class of the
object includes the module, which can be accomplshed with the
Object#extend method
ruby-1.9.1-p376 > a.extend Comparable
=3D> #<Object:0x00000101176068>
ruby-1.9.1-p376 > a.kind_of? Comparable
=3D> true
The motivating example a function which takes an object and appends it
to itself is rather strained, IMHO. Personally I'd never be tempted to
write such a method. Rather I'd implement it on String and Array
module SelfAppendable
def append_to_self
self << self
end
end
class String
include SelfAppendable
end
class Array
include SelfAppendable
end
And I don't know if the Author really expects the result of appending
an Array to itself, inserting a element into an Array with << inserts
a reference to the element, so this will generate a recursive array,
which can cause problems.
ruby-1.8.6-p383 > [1, 2, 3].append_to_self
=3D> [1, 2, 3, [...]]
ruby-1.8.6-p383 > [1, 2, 3] << [1, 2, 3]
=3D> [1, 2, 3, [1, 2, 3]]
ruby-1.8.6-p383 > ([1, 2, 3] << [1, 2, 3]).flatten
=3D> [1, 2, 3, 1, 2, 3]
ruby-1.8.6-p383 > [1, 2, 3].append_to_self
=3D> [1, 2, 3, [...]]
ruby-1.8.6-p383 > [1, 2, 3].append_to_self.flatten
ArgumentError: tried to flatten recursive array
--=20
Rick DeNatale
Blog:
http://talklikeaduck.denhaven2.com/
Twitter:
http://twitter.com/RickDeNatale
WWR:
http://www.workingwithrails.com/person/9021-rick-denatale
LinkedIn:
http://www.linkedin.com/in/rickdenatale