Software Developer

Synthesizing Composition With Delegation

Ruby Software Design Concert Series 🔗

  1. Dependency Injection: Plug In
  2. Shedding a Light on Duck Typing
  3. Synthesizing Composition With Delegation
  4. Inheritance: Derivative Songwriting
  5. Using Sonic Pi To Play Music With Ruby
  6. Stringing Code Together To Play Music

Setting the Stage 🔗

Any application will be comprised of multiple components - in Object-Oriented languages, typically classes. Sometimes these classes even work together! External users of one of these classes may not know that behind the scenes there are more classes working together, nor do they care. The public API does what they need it to, and anything else is an implementation detail. However, keeping the specialization of these different classes apart, but using them together, is beneficial.

To demonstrate using composition to model a complex system and using delegation in that composition, we will explore how a synthesizer can handle memory management to store presets of sounds. This example comes from my RubyConf 2020 talk about Ruby’s Coverage module.

Noise Reduction 🔗

The synthesizer is an instrument capable of producing a wide array of sounds. A collection of sounds and effects are known as a patch.

class Patch
  attr_reader :sound
  attr_reader :effect
  attr_reader :filter
  attr_reader :oscillator
end

You can save these patches on the synthesizer’s memory and recall them later for easy access.

patch = Patch.new
synth = Synthesizer.new

synth.save_patch(location: :b1, patch: patch)
synth.set_patch(patch)
synth.play_key(note: :a, duration: 1)

Save You the Trouble 🔗

Much like the actual instrument is comprised of various subcomponents, our Synthesizer is made up of various classes that specialize in its area of expertise.

For example, our synthesizer above doesn’t know how to save a patch to its onboard memory. It relies on its patch memory to handle that.

class Synthesizer
  def save_patch(location:, patch:)
    @patch_memory.write(location: location, patch: patch)
  end
end

All the synthesizer itself knows is what message to send to the memory to have it do that. The synthesizer is delegating the responsibility of storing these patches to the patch memory instance.

Anyone playing the synthesizer does not need to be concerned with how it’s storing these patches, just that it’s doing it. Anyone using our synthesizer class isn’t aware that there is a separate patch memory class that the synthesizer is using.

At the same time, our synthesizer doesn’t know directly how to access its memory. It relies on the PatchMemory class for that, and delegates any responsibility related to memory management to that class. As Sandi Metz describes in Practical Object-Oriented Design In Ruby, a synthesizer has a patch memory, as it has a series of other parts, and those are composed together to deliver all the functionality that a synthesizer provides.

Key Benefits to Delegation 🔗

Delegation provides a few important drivers that make it easier to wrangle complex systems.

Specialization 🔗

Our patch memory component is solely focused on interfacing with the onboard memory of the instrument, which is where it saves and recalls stored sounds. Its tests can dig into all of the edge cases and minutiae that need to be accounted for. The implementation can make very specific decisions so that it is extremely performant without other areas of the system needing to worry about that.

A synthesizer itself is a complex system. The memory management is only one small part of it. The strength and value-add of our Synthesizer class is in organizing all of these components together, knowing the right messages to pass to them, with a public API that doesn’t require intimate knowledge of all those details. If the internals of our Synthesizer class handled all of this responsibility itself, it would quickly become unwieldy, difficult to navigate, hard to read, a challenge to troubleshoot, a burden to test, and feared when changes are required.

Flexibility & Reuse 🔗

In reality, there are many different kinds of synthesizer, all of which have different capabilities. Some may be able to store 1,000 different patches on board. Others may only have capacity for four. Still more may have expandable memory, where you can plug in a USB device for nearly infinite storage.

Rather than needing to create entirely different synthesizer classes to handle any of these scenarios, instead we only need to model those differences in patch memory classes. Our synthesizer can then use any of those and still maintain the rest of its functionality, without needing to duplicate it across different classes.

In this example, our synthesizer changes its memory capabilities based on the brand that it is.

def initialize_memory
  if @brand == :moog
    @patch_memory = MoogPatchMemory.new
  elsif @brand == :nord
    @patch_memory = NordPatchMemory.new
  end
end

Thanks to duck typing, as long as these patch memory classes respond to the same messages, our Synthesizer class can use either of them interchangeably.

Rock On 🔗

Composing classes together allows us to create a fully-functional system. A class that uses another class to handle a request or responsibility is delegating that duty to the helper class. Delegation can encapsulate the knowledge of different specialties for code organization without external consumers needing to know or care about that implementation detail. Delegating responsibility to different classes can also make it easier for the system to change, making it more likely to promote code reuse.

Next we’re going to play one of the greatest hits in software design: inheritance.

This post originally published on The Gnar Company blog.