Scheme-relative URLs aren't handled correctly

[michaeltelford]

Description

Protocol-relative URLs aren't handled correctly. They are currently processed as relative when they are designed to be used for external resources. This has a knock on effect where these types of links are returned by Document#internal_links instead of external_links.

Additional reading: https://www.paulirish.com/2010/the-protocol-relative-url/

Reproduce

Steps to reproduce the behavior:

url = Wgit::Url.new '//fonts.google.com'
url.relative? # true - which isn't right, it should be false.

Expected Behavior

1. Url#relative? should return false for a protocol relative URL.
2. Document#external_links should return any protocol relative URLs, not internal_links. The returned URLs should have the correct @url.scheme prefixed so they are valid?.

Possible Solutions

Point 1 above can be achieved by adding the following line (not tested):

# Add this to Wgit::Url#relative? on line 164
return false if starts_with?('//)

Point 2 above can be achieved with something like (not tested):

# Replace this method in the Wgit::Document class
def external_links
    return [] if @links.empty?

    links = @links
        .map do |link|
            link = Wgit::Url.new("#{@url.scheme}:#{link}") if link.start_with?('//')
            link.omit_trailing_slash
        end
        .reject { |link| link.relative?(host: @url.to_origin) }

    Wgit::Utils.sanitize(links)
end

Tests

What tests would prove this bug is fixed?

• See the expected behavior and write tests to assert it.

[michaeltelford]

Fixed in v0.10.0.