Many Enumerable Returns

As threatened then, here’s the followup to my last post on the #Enumerables section from Ruby Monk, how I felt like a bit of an idiot for a few hours, and what I learned from that.

tl;dr - enumerable blocks aren’t magic; yield is magic.

This question is a little further along from the last, and was framed so:

Try implementing a method called occurrences that accepts a string argument and uses inject to build a Hash. The keys of this hash should be unique words from that string. The value of those keys should be the number of times this word appears in that string.

So far so, so good. So I wrote this:

def occurrences(str)
  str.scan(/\w+/).inject(Hash.new(0)) do |hashy, i|
    hashy[i.downcase] += 1
  end
end

Which spat out:

TypeError can't convert String into Integer

And left me confused for a good few minutes. OK, getting on for a quarter of an hour. What was going on? - what I’d written was very similar to the example above:

[4, 8, 15, 16, 23, 42].inject({}) { |a, i| a.update(i => i) }

So I caved and looked at the answer:

def occurrences(str)
	str.scan(/\w+/).inject(Hash.new(0)) do |build, word|
    	build[word.downcase] +=1
    	build
	end
end

Which left me none the wiser. Why was the block re-iterating the accumulator function at the end? To test this I played around with p-ing the lines of the block… and discovered something interesting. Namely,

a.update(i => i) # => a

But…

build[word.downcase] +=1 # => build[word.downcase], the new value of that key

The block needs to return the accumulator - the first example is just lucky that it does so already!

The only reason the accumulator in an Enumerable#inject accumulates is that it’s returned from the block on each iteration. In other words, somewhere in the definition of #inject for each class that can be made enumerable, the method yields to the block, and then keeps the value returned to be passed in again as the new accumulator argument.

I’d previously thought of #inject as working by magic, whereas in fact it was working by a method I could probably write myself given enough time. Something like this…

bob = [1,2,3,4,5,6]

def bob.inject(default = nil)
  accumulator = default || self[0]
  if default
    self.each do |element|
      accumulator = yield(accumulator, element)
    end
  else
    self.drop(1).each do |element|
      accumulator = yield(accumulator, element)
    end
  end
  puts "all adds up to: "   # just to prove it's this method being
                            # called, not the superclasses...
  p accumulator
end

Which gives us such fun as:

bob.inject() {|a,e| a += e}
# => all adds up to: 21
bob.inject(10) {|a,e| a += e}
# => all adds up to: 31
bob.inject([]) {|a,e| a << e**2}
# => all adds up to: [1, 4, 9, 16, 25, 36]
bob.inject({}) {|a,e| a[e] = "x"*e; a}
# => {5=>"xxxxx", 6=>"xxxxxx", 1=>"x", 2=>"xx", 3=>"xxx", 4=>"xxxx"}

I relied on #each here, but we could easily write an each method using a for... in... loop or similar. The genius is in yield, which is the real magic that’s going on here.

Ruby Monk has more about the magic of yield, and why it’s weird in a language that professes that everything is an object. Like a lot in Ruby, I discovered a small thing didn’t work, patiently played with it until I found out why, and then ‘worked’ that small new piece of knowledge to give me greater insight into what was going on. I’m finding this to be the most satisfying method to learn by, both because it makes me feel like I’m learning to a deeper degree than I would by just reading the answers out of a book, and in addition, when the books do cover the subject, I can better apply what’s written there to what I’ve seen in action.

postscipt - ²⁷⁄₇₁₄

Of course, David Black covers the same ground, but better (gets each off the ground using a for loop), in chapter 6 of The Well Grounded Rubyist. Love that book.