Nested Before/Around Run In Weird Order

[benmmurphy]

Subject of the issue

If you have a describe block with an before(:each) and then nest a describe block with an around(:each) the around(:each) runs before the before(:each). I would have thought the around(:each) is just a nice way of writing before(:each) after(:each) and sharing some state between them. but because of the way around(:each) always runs before before(:each) even if it is nested lower it makes around(:each) much less useful.

there doesn't seem to be a test case for this behaviour in the documentation so I'm not sure if this was deliberate or not. though, i guess it would be hard to change this behaviour because probably a lot of people rely on it.

https://relishapp.com/rspec/rspec-core/v/2-0/docs/hooks/around-hooks#before/after(:each)-hooks-are-wrapped-by-the-around-hook

Your environment

• Ruby version: 2.4.0
• rspec-core version: 3.7

Steps to reproduce

describe "filter" do

  before(:each) do
    puts "before each outer"
  end

  describe "inner" do
    around(:each) do |example|
      puts "around each inner"
      example.run
    end

    it "should" do
      puts "example"
    end
  end

end

actual:

around each inner
before each outer
example

expected:

before each outer
around each inner
example

here is how it works if they are both before(:each) rather than using an around(:each) for the inner.

describe "filter" do

  before(:each) do
    puts "before each outer"
  end

  describe "inner" do
    before(:each) do
      puts "before each inner"
    end

    it "should" do
      puts "example"
    end
  end

end

before each outer
before each inner
example

[JonRowe]

Verified this is a bit odd, we should at least document the order that we expect this to run in, I think at least around hooks should run after before when defined in an inner context

[deivid-rodriguez]

So, I think this would be a good fix for RSpec 4? It's a bit odd, indeed!

[deivid-rodriguez]

I'd like to give this a try, would you be willing to accept a PR for RSpec 4?

[JonRowe]

Yes, for RSpec 4, I would recommend developing it as in an opt-in feature, and we will eventually need a way to warn about the change in behaviour.

[deivid-rodriguez]

Sounds like a good plan to me! I'll see if I can find some time to work on it!

[JonRowe]

Closing for now as there have been additional docs added, potential reorder can be looked at later.