Testing a clone fails, unless the clone is invoked first

[forforf]

When doing a customized deep clone, I ran across a situation where the test will fail unless the variable assigned by let is called first (see the spec code below).

The cloning does work, but the normal test case fails.

Code to reproduce:

File: cloner.rb

class Cloner
  def clone_array_of_hashes(a)
    a.map{|h| h.clone}
  end
end

File: cloner_spec.rb

require_relative "./cloner"

describe Cloner do
  let(:cloner) {Cloner.new}
  let(:orig_ah) { [{a: "A"},{b: "B"}] }
  let(:cloned_ah) { cloner.clone_array_of_hashes(orig_ah) }

  it "passes" do
    cloned_ah[0] #this is the only difference between the two tests
    orig_ah[0][:a] = "AA"
    expect(orig_ah[0]).to_not eq(cloned_ah[0])
  end

  it "fails but should pass" do
    orig_ah[0][:a] = "AA"
    expect(orig_ah[0]).to_not eq(cloned_ah[0])
  end
end

Notes:

• Invoking cloned_ah instaed of cloned_ah[0] works to pass the test as well.
• The issue exists whether clone or dup is used

The issue occurs with RSpec 2.14 and 2.12, and Ruby 1.9.3 and 2.0.0

I had a similar issue a while ago with regexes (#820) with the variable needing to be invoked first, but I'm not sure if the similarities extend beyond that.

[soulcutter]

This isn't a bug - let variables are lazily instantiated, so in your failing test cloned_ah is instantiated after you modify orig_ah. If you need to control the order of instantiation you should probably be explicit about the ordering using a before block.

Alternately you could probably get away with using let! for cloned_ah, which is functionally equivalent to initializing it in a before block.

https://www.relishapp.com/rspec/rspec-core/v/2-14/docs/helper-methods/let-and-let

[forforf]

Got it, obvious in hindsight. Thanks!