Regression / Breaking change in v3.1.0? about js_regex

[jaynetics]

Hi @tagliala,

The test should be updated. The new source (?:something) is more correct. It retains the "non-capturingness" of the input.

In case you're wondering why that matters despite you doing nothing with group captures: even pure matching as in client_side_validations can be affected by the incorrect "capturingness" of substitutions.

A contrived example:

/(?-mix:a)(b)\1/ # matches 'abb'
/(a)(b)\1/       # js_regex 3.0.0 output, matches 'aba'
/(?:a)(b)\1/     # js_regex 3.1.0 output, matches 'abb' as it should

[tagliala]

@janosch-x thanks

So I should replace

  def test_regexp_remove_group_options
    expected_regexp = { source: '(something)', options: 'g' }
    assert_equal expected_regexp, /(?-mix:something)/.to_json(nil)
  end

with

  def test_regexp_convert_group_options
    expected_regexp = { source: '(?:something)', options: 'g' }
    assert_equal expected_regexp, /(?-mix:something)/.to_json(nil)
  end

?

[jaynetics]

@tagliala correct!