shoulda-matchers RSpec expect syntax

Question

What is the correct format for using shoulda-matchers and RSpec's new expect syntax?

Erect

Sorry I meant shoulda – Erect

Regarding the vote-to-close based on this being primarily opinion based, I would argue that this particular instance of asking if something is a "good idea" is an exception because there is a fact-based explanation that the concern is a non-issue. – Irriguous

edited the quesrion detail to make it be appropriate for SO. – Kob

Answer 1

While one could certainly use the shoulda-matchers with the new expect syntax as follows:

it 'should validate presence of :email' do
  expect(subject).to validate_presence_of :email
end

or the more concise but less readable:

it { expect(subject).to validate_presence_of :email }

the one-liner should format these matchers are typically used with is explicitly supported in 2.14 even when config.syntax == :expect. When should is being used with an implicit subject as in:

describe User
  it { should validate_presence_of :email }
end

it does not rely on the monkey patching of Kernel that should otherwise depends on.

This is covered in https://github.com/rspec/rspec-expectations/blob/master/Should.md. In fact, that documentation even uses the above shoulda matcher example to illustrate this exception.

See also Using implicit `subject` with `expect` in RSpec-2.11, which discusses a configuration option which lets you use as an alternative to it.

expect_it { to validate_presence_of :email }

Update: As of RSpec 3.0 (beta2), you will also be able to use:

it { is_expected.to validate_presence_of :email }

Irriguous

The first block would also be surrounded by the desribe User, right? As it it looks like a omparison of 3 lines to 3 different lines, but isn't quite that. If so might want to add that for clarity in first code block. – Kob

@MichaelDurrant Yes, it would. Agreed. Thanks for pointing this out. Answer updated. – Irriguous

Anyone have any idea what the downvote is for? Is it related to the close votes for this being "primarily opinion based"? – Irriguous

It's too bad that you need the it block. I always liked shoulda matches because of their concise syntax. – Erect

Actually, you don't need the block. That was what I intended as the essence of my answer. The one-liner should form is completely acceptable and fully accepted/supported by the latest convention. But you did give me an idea. What if RSpec introduced expect_it as an alias for expect(subject)? – Irriguous

Answer 2

I'll suplement the answer of @peter-alfvin. In case you test the model and its migration themselves with shoulda-matchers you can't use :expect outside of it block, so can't write:

RSpec.describe ModelName, type: :model do
   expect(subject).to belong_to(:user)
end

And you will get the expection:

`expect` is not available on an example group (e.g. a `describe` or `context` block).

but correct version is:

RSpec.describe ModelName, type: :model do
   it { expect(subject).to belong_to(:user) }
end

Committal