Why use Ruby's attr_accessor, attr_reader and attr_writer?

You may use the different accessors to communicate your intent to someone reading your code, and make it easier to write classes which will work correctly no matter how their public API is called.

class Person
  attr_accessor :age
  ...
end

Here, I can see that I may both read and write the age.

class Person
  attr_reader :age
  ...
end

Here, I can see that I may only read the age. Imagine that it is set by the constructor of this class and after that remains constant. If there were a mutator (writer) for age and the class were written assuming that age, once set, does not change, then a bug could result from code calling that mutator.

But what is happening behind the scenes?

If you write:

attr_writer :age

That gets translated into:

def age=(value)
  @age = value
end

If you write:

attr_reader :age

That gets translated into:

def age
  @age
end

If you write:

attr_accessor :age

That gets translated into:

def age=(value)
  @age = value
end

def age
  @age
end

Knowing that, here's another way to think about it: If you did not have the attr_... helpers, and had to write the accessors yourself, would you write any more accessors than your class needed? For example, if age only needed to be read, would you also write a method allowing it to be written?

All of the answers above are correct; attr_reader and attr_writer are more convenient to write than manually typing the methods they are shorthands for. Apart from that they offer much better performance than writing the method definition yourself. For more info see slide 152 onwards from this talk (PDF) by Aaron Patterson.

It is important to understand that accessors restrict access to variables, but not their content. In ruby, like in some other OO languages, every variable is a pointer to an instance. So if you have an attribute to an Hash, for example, and you set it to be "read only" you always could change its content, but not the content of pointer. Look at this:

> class A
>   attr_reader :a
>   def initialize
>     @a = {a:1, b:2}
>   end
> end
=> :initialize
> a = A.new
=> #<A:0x007ffc5a10fe88 @a={:a=>1, :b=>2}>
> a.a
=> {:a=>1, :b=>2}
> a.a.delete(:b)
=> 2
> a.a
=> {:a=>1}
> a.a = {}
NoMethodError: undefined method `a=' for #<A:0x007ffc5a10fe88 @a={:a=>1}>
        from (irb):34
        from /usr/local/bin/irb:11:in `<main>'

As you can see is possible delete a key/value pair from the Hash @a, as add new keys, change values, eccetera. But you can't point to a new object because is a read only instance variable.

Not all attributes of an object are meant to be directly set from outside the class. Having writers for all your instance variables is generally a sign of weak encapsulation and a warning that you're introducing too much coupling between your classes.

As a practical example: I wrote a design program where you put items inside containers. The item had attr_reader :container, but it didn't make sense to offer a writer, since the only time the item's container should change is when it's placed in a new one, which also requires positioning information.

You don't always want your instance variables to be fully accessible from outside of the class. There are plenty of cases where allowing read access to an instance variable makes sense, but writing to it might not (e.g. a model that retrieves data from a read-only source). There are cases where you want the opposite, but I can't think of any that aren't contrived off the top of my head.

Why use Ruby's attr_accessor, attr_reader and attr_writer?

Related

Recent Posts