Why does array.each behavior depend on Array.new syntax?

Question

I\'m using Ruby 1.9.2-p290 and found:

a = Array.new(2, []).each {|i| i.push(\"a\")}    
=> [[\"a\", \"a\"], [\"a\", \"a\"]]

Which is not

Answer 1

In the first case you're using a single instance of an Array as a default for the elements of the main Array:

a = Array.new(2, []).each {|i| i.push("a")}

The second argument is simply recycled, so the push is applied to the same instance twice. You've only created one instance here, the one being supplied as an argument, so it gets used over and over.

The second method is the correct way to do this:

b = Array.new(2) {Array.new}.each {|i| i.push("b")

This deliberately creates a new instance of an Array for each position in the main Array. The important difference here is the use of the block { ... } which executes once for each position in the new Array. A short-form version of this would be:

b = Array.new(2) { [ ] }.each {|i| i.push("b")