Safely Call Superclass Methods from Ruby Modules
source link: https://avdi.codes/safely-call-superclass-methods-from-ruby-modules/
Go to the source link to view the article. You can view the picture content, updated content and better typesetting reading experience. If the link is broken, please click the button below to view the snapshot at that time.
In today’s RubyTapas throwback, we tackle a metaprogramming topic. How do you reliably call superclass methods from included modules… when you can’t know for sure if those methods exist, or if they have been overridden out from under you? This video and article will show you some techniques you can use.
Director’s commentary: This was originally published on RubyTapas October 29th, 2012. The metaprogramming techniques shown here are still valid, if rarely needed. The video quality, on the other hand, makes me cringe… I wasn’t nearly as good at pacing back then, and I had a lot more tolerance for “whoopsies” in a three-minute video.
By the way, if you want to learn more about this kind of deep metaprogramming for libraries, you might like Much Ado About Null, a bonus ebook included in the Gold Edition of Confident Ruby.
Here’s the original episode video, and if you scroll down you’ll find the video script and code.
When authoring a reusable module, we may find we want to call a superclass method, but only if it exists. For instance, here’s a module which defines its own #hello
method. We want to be able to include it in many different classes, some of which may inherit from other classes that define #hello
.
If the class it is included in also inherits a #hello
method from somewhere else, the module will simply embellish its output a bit. But since the including class might not define a #hello
method, the module also includes its own full implementation. The question is, how do we tell if an ancestor implements #hello
?
module YeOlde
def hello(subject="World")
if ???
super
else
"Good morrow, #{subject}!"
end
puts "Well met indeed!"
end
end
The answer is to use Ruby’s defined?
operator, with super
as its argument.
When we include this module in a class whose parent defines #hello
, it uses the parent greeting. When we include it in a class with no #hello
method, it uses its own greeting.
module YeOlde
def hello(subject="World")
if defined?(super)
super
else
puts "Good morrow, #{subject}!"
end
puts "Well met indeed!"
end
end
class Greeter
def hello(subject)
puts "Hello, #{subject}"
end
end
class GreeterChild < Greeter
include YeOlde
end
class NonGreeter
include YeOlde
end
GreeterChild.new.hello("Bob")
NonGreeter.new.hello("Sally")
Hello, Bob Well met indeed! Good morrow, Sally! Well met indeed!
Let’s look at another situation involving the use of super
in a module. Let’s say we have a module which defines a #logged_send
method. #logged_send
acts just like a call to Ruby’s #send
, except it also logs the method call and arguments.
module Logged
def logged_send(name, *args, &block)
puts "Sending #{name}(#{args.map(&:inspect).join(', ')})"
send(name, *args, &block)
end
end
When we include this module in most classes it works just fine.
class Greeter
include Logged
def hello(subject)
puts "Hello, #{subject}"
end
end
Greeter.new.logged_send(:hello, "Major Tom")
But one day we add in another module which overrides #send
to do something completely different. Suddenly, #logged_send
doesn’t work so well.
module PigeonPost
def send(*messages)
# ...
puts "Your message is winging its way to its recipient!"
end
end
class Greeter
include PigeonPost
include Logged
def hello(subject)
puts "Hello, #{subject}"
end
end
Greeter.new.logged_send(:hello, "Major Tom")
The problem here is that when the Logged
module called #send
, expecting the default Object
implementation of #send
, it got the PigeonPost
implementation instead.
How can we ensure that Logged
always gets the original definition of #send
? Let’s take it step by step. Inside the #logged_send
method, we first need to get a method object referring to the original definition of #send
from the Object
class.
original_send = Object.instance_method(:send)
#<UnboundMethod: Object(Kernel)#send>
This yields an UnboundMethod
object. This object then needs to be bound to a specific object instance, in this case self
.
bound_send = original_send.bind(self)
This results in a callable Method
object.
The last step is to call the Method
object.
bound_send.call(name, *args, &block)
When we put it all together and try again, things work as intended!
module Logged
def logged_send(name, *args, &block)
puts "Sending #{name}(#{args.map(&:inspect).join(', ')})"
original_send = Object.instance_method(:send)
bound_send = original_send.bind(self)
bound_send.call(name, *args, &block)
end
end
This isn’t the most straightforward technique in the world, and I don’t need it very often. But every now and then it’s a real lifesaver.
Now, a few viewers are probably yelling at their screens right now, saying “you should have just used #__send__
instead!”. To which I say: you’re absolutely right, at least for this example. But that’s a topic for another day.
Happy hacking!
Let's correspond!
May I write to you? Every week or so you'll get a SIGAVDI newsletter with a few handpicked links to interesting articles, talks, or resources; an update on what I've published lately; and some reflections at the intersection of software and life. Replies go directly to me, and I try to respond to every single one (eventually!)
Recommend
About Joyk
Aggregate valuable and interesting links.
Joyk means Joy of geeK