I recently start to read the book "Programming Ruby 1.9&2.0". It shows a trick for explicit enumerator
triangular_numbers = Enumerator.new do |yielder|
number = 0
count = 1
loop do
number += count
count += 1
yielder.yield number
end
end
5.times { print triangular_numbers.next, " " }
puts
I wonder why this yielder.yield will leave the loop temporarily and also return the value of number until the next enumerator object created. It seems different than the usual case when a yield inside the loop block. I check the APIdock and find that the source code of Proc.yield() is the same as Proc.call(). For the Yielder object in the Enumerator class, the Yielder has override the yield(). But why the yielder.yield will temporarily leave the loop block?
Reference: APIdock Yielder yield(), Ruby MRI rb_proc_call
I also stumbled on that example in the book. After some thinking on how the example might work and browsing Ruby doc I found the Fiber class which I think is what is used by Enumerator behind the scene:
http://www.ruby-doc.org/core-2.0/Fiber.html
Fiber concept implements "light weight cooperative concurrency" which is quite interesting, not that difficult to understand and what's more important it's different from other "yields" for calling blocks or dealing with thread controls.
I think Enumerator has a Fiber object inside which it passes to the block. Then looks like every time when you call "next" on your Enumerator, it calls "resume" on the Fiber object to allow it calculate the next number and when the block calls "yield" on the Fiber, the control goes back to the "next" method. And so on.
Here's my version of possible implementation of Enumerator (of course, only that part that is discussed in the book's example):
class MyExplicitEnumerator
def initialize (&block)
@yielder = Fiber.new { block.call Fiber }
end
def next
@yielder.resume
end
end
e = MyExplicitEnumerator.new do |yielder|
number = 1
loop do
yielder.yield number
number += 1
end
end
p e.next
p e.next
# output
# 1
# 2