Precedence of style lookup is inverted

[mojavelinux]

When resolving the style for a token using the style_for method (which eventually delegates to get_own_style), the precedence is backwards. What's expected is a more specific (nested) style to win over a more general (ancestor) style. However, the inverse is true.

I think that the following logic:

def get_own_style(token)
  token.token_chain.each do |anc|
    return styles[anc] if styles[anc]
  end    
  nil
end

should in fact be:

def get_own_style(token)
  token.token_chain.reverse_each do |anc|
    return styles[anc] if styles[anc]
  end    
  nil
end

The test case is as follows:

style Keyword::Type, fg: :gray, bold: true
style Keyword, fg: :green, bold: true

When I invoke:

style_for Token['Keyword.Type']

I expect to get:

{:fg => :gray, :bold => true}

Instead, I get:

{:fg => :green, :bold => true}

As a result, this makes the behavior of Rouge inconsistent with the behavior of Pygments.

[mojavelinux]

I think the reason this wasn't detected by the test suite is because the test is ambiguous. It will assert true regardless of order of precedence in the implementation. When I send a pull request, I'll make sure that the test catches this distinction.