firedev.com

Bottomless Ruby Hash

The other day somebody asked if there is a way to blindly assign nested values to Ruby Hash without creating each key. Turns out there is, and it has an interesting side-effect. Welcome Bottomless Hash.

Let’s try assigning ‘through’ a key in plain old Ruby first.

params = {}
params[:world][:russia] = :moscow
=> NoMethodError: undefined method `[]=' for nil:NilClass

Okay, it doesn’t work like this. Fortunately Hashes in Ruby can be initialized with a default value. First thing to try seems quite obvious. Let’s put an empty hash as a default value.

params = Hash.new({})
params[:world][:russia] = :moscow

params[:world]
=> {:russia=>:moscow}

Seems legit. But let’s dig a little deeper.

params[:world][:thailand]=:bangkok

params
=> {}

Why is it empty? Let’s add more stuff to it.

params[:underworld]=:hell

params
=> {:underworld=>:hell}

What is going on? Some wicked magic? Not quite. First the :world key was initialized with the same empty Hash that we put as a default value. It is easy to access still, since the same Hash is returned when the key is empty. However now all our cities are available in both worlds.

params[:world][:thailand]
=> :bangkok

params[:underworlds][:russia]
=> :moscow

Okay, we got to fix it. To initialize new hash for values we need to pass a block. Block accepts two variables - the hash itself and the key it was accessed with. Lets just inject new empty hash as a value for the key.

params = Hash.new do |hash, key|
  hash[key] = Hash.new
end

params[:world][:thailand]=:phuket

params
=> {:world=>{:thailand=>:phuket}}

Amazing. Okay but what if we need to add another level?

params[:asia][:thailand][:bangkok] = :chao_praya
=> NoMethodError: undefined method `[]=' for nil:NilClass

Oh no, not again. What can we do? Let’s add another layer! So the nested Hash could in turn create more hashes:

params = Hash.new do |hash0, key0|
  hash0[key0] = Hash.new do |hash1, key1|
    hash1[key1] = Hash.new
  end
end

params[:asia][:thailand][:moscow] = :moscow_river

It works! But what if the rabbit hole is even deeper?

params[:asia][:thailand][:bangkok][:river] = :chao_praya
=> NoMethodError: undefined method `[]=' for nil:NilClass

Okay, have to solve this once and for all. Let’s throw some functional programming in the mix and see where it takes us. What we need is a kind of procedure that would return a new hash with the same procedure hidden inside waiting for an empty key to come in.

What would such procedure look like? Quite familiar in fact. We just need to pack it with & symbol-to-proc pretzel to shove into Hash initializer.

procedure = lambda do |hash, key|
  hash[key] = Hash.new(procedure)
end

params = Hash.new(&procedure)
params[:russia][:moscow] = :moscow_river
params
=> {:russia=>{:moscow=>:moscow_river}}

Okay, that part is solved, now lets tie it a little tighter so we don’t need to create lambda beforehand. Ruby Hash sports default_proc method that can be used to access the block hash was initialized with. Thank makes it super sweet, thanks Pavel for pointing me on that.

params = Hash.new { |h, k| h[k] = Hash.new(&h.default_proc) }

params[:world][:thailand][:bangkok][:bangna]
params
=> {:world=>{:thailand=>{:bangkok=>{:bangna=>{}}}}}

This is sweet, but what is the practical point of bottomless hash? Interesting side-effect is that it never fails you when reading values.

params[:i][:dont][:know]
=> {}

And the beauty of this is that you can merge any hash with it to produce a bottomless version. So you can blindly access the keys.

unknown = { key: :value }
bottomless = params.merge other

bottomless[:missing][:value]
=> {}

No matter how long the chain is, bottomless hash won’t raise an error. Okay, it returns and empty hash instead of nil, which is truthy. But this can be checked with empty? even in plain ruby.

As it was mentioned in comments we can encapsulated the behaviour into a Class that would return an empty Bottomless hash or convert a given hash into a bottomless version:

class BottomlessHash < Hash
  def initialize
    super &-> h, k { h[k] = self.class.new }
  end

  def self.from_hash(hash)
    new.merge(hash)
  end
end

And some tests for the good night sleep:

class BottomlessHash < Hash
  def initialize
    super &-> h, k { h[k] = self.class.new }
  end

  def self.from_hash(hash)
    new.merge(hash)
  end
end

class Hash
  def bottomless
    BottomlessHash.from_hash(self)
  end
end

describe BottomlessHash do
  subject { described_class.new }

  it 'does not raise on missing key' do
    expect do
      subject[:missing][:key]
    end.to_not raise_error
  end

  it 'returns an empty value on missing key' do
    expect(subject[:missing][:key]).to be_empty
  end

  it 'stores and returns keys' do
    subject[:existing][:key] = :value
    expect(subject[:existing][:key]).to eq :value
  end

  describe '#from_hash' do
    let (:hash) do
      { existing: { key: { value: :hello } } }
    end

    subject do
      described_class.from_hash(hash)
    end

    it 'returns old hash values' do
      expect(subject[:existing][:key][:value]).to eq :hello
    end

    it 'provides a bottomless version' do
      expect(subject[:missing][:key]).to be_empty
    end

    it 'stores and returns new values' do
      subject[:existing][:key] = :value
      expect(subject[:existing][:key]).to eq :value
    end

    it 'converts nested hashes as well' do
      expect do
        subject[:existing][:key][:missing]
      end.to_not raise_error
    end
  end
end

And an ActiveSupport-like extension for the Hash class so you can call .bottomless on any hash in the system. Thanks Patrick!

class Hash
  def bottomless
    BottomlessHash.from_hash(self)
  end
end

Hope this little experiment might get useful for data processing or when dealing with unknown nested structures from the outside world.

Bottomless Ruby Hash
Making Bottomless Hash Ruby Gem
Github firedev/bottomless_hash
Gist bottomless_hash.rb