Forward given filter block to matcher

[cbliard]

Fixes #2616

[twalpole]

Sorry for the delay, this will be looked at/considered this coming weekend

[twalpole]

I don't think the tests added are actually validating the filter blocks are run since the have_selector calls would pass with or without the filter blocks.

[cbliard]

You're right. When writing the tests, I knew that some of them were not actually accurate as they would pass with or without the block.

This one for instance will happily pass with or without the block:

expect(page).to have_selector(:css, 'input#test_field') { |input| input.value == 'monkey' }

Contrary to this one which would not pass without the block:

expect(page).to have_no_selector(:css, 'input#test_field') { |input| input.value == 'not a monkey' }

How should the tests be formulated then? I saw in lib/capybara/spec/session/element/match_css_spec.rb that some tests are written like expect { expect(page).to match_something }.to raise_exception(RSpec::Expectations::ExpectationNotMetError). Following it, the test can be written like this:

expect { expect(page).to have_selector(:css, 'input#test_field') { |input| input.value == 'not a monkey' } }
  .to raise_exception(RSpec::Expectations::ExpectationNotMetError)

Would it be better? Any other suggestions?

[twalpole]

Either that or changing the tests to consider number of matched elements like

expect(page).to have_selector(:css, 'input', count: 1) { |input| input.value == 'monkey' }

which would be easier to read/understand

[twalpole]

Looking at this further, this breaks the default binding of the do block to to which is defined to allow a proc for failure messages. Not really sure what the best solution here actually is. This PR is overriding Rubys default binding precedence and could break existing code... Ignore, I mistread the RSpec code

[twalpole]

Merged via cacd032

[cbliard]

Thanks @twalpole for merging.

I'm still feeling bad about the tests. Do you mind if I open another PR to provide better tests with count: 1 as you suggested?

[twalpole]

@cbliard I had already fixed the tests