include matcher is not diffable enough

[ojab]

Subject of the issue

Consider:

RSpec.describe 'foo' do
  RSpec::Matchers.define :my_include do |expected|
    match do |actual|
      @expected_as_array = [[expected]]
      expect(actual).to include(expected)
    end

    diffable
  end

  let(:expected) { {a: 'a' * 100, b: 2} }
  let(:actual) { [{a: 'a' * 100  , b: 3}] }

  it { expect(actual).to include(expected) }
  it { expect(actual).to my_include(expected) }
end

diff for my_includes shows exact lines where the issue occurred, because include matcher tries to compare single object with the whole array.
It's especially useful when working with real/more complicated data structures, for me it's unfortunately spend 3 minutes reading output or try to diff by hand usually on similar failures.

Your environment

• Ruby version: 3.0.2
• rspec-expectations version: current mastermain

[pirj]

A PR is welcome. You can start with a pending spec.
If you're up to improve this, check rspec/rspec-support#365 (and listed issues) and rspec/rspec-support#516 for details.

[ojab]

I don't quite understand the expected solution here, while the single expected hash is quite easy, for multiple hashes we probably should find the closest objects in the actual enumerable and show diffs for them, but I guess it will lead to #1161.

And even if we knew what should be done, not sure if I'd have the capacity to do it.
So only see something say something issue for now and workaround in a private matcher 😅

[JonRowe]

Our current differ is limited in that it has to diff everything as strings, so it "prints" objects to strings then diffs those strings. As the include matcher is diffing an array vs a single element, it's always going to be "wrong" when it comes to the diff, and we don't currently have an api to provide the differ with a more intelligent solution.

Mutating the expected is a work around, but it might cause problems for re-use (which we support) and other downstream complications, for now I'm going to close this as a "can't fix" as its essentially a duplicate of out "we'd like to improve the differ" issues.

[ghost]

@JonRowe What Issue should I subscribe to and read to help with the Differ limitations?

[JonRowe]

Theres a couple over on rspec/rspec-support, contributions for improving the differ or making it configurable with say super_diff would be welcome