URI.parse can blow up with non-standard URIs

[matthewrudy]

Here's an example tweet_spec that fails with URI

it "can handle strange urls" do
  urls_array = [
    {
      :url => "http://with_underscore.example.com/t.co",
      :expanded_url => "http://with_underscore.example.com/expanded",
      :display_url => "with_underscore.example.com/expanded…",
      :indices => [10, 33],
    }
  ]
  tweet = Twitter::Tweet.new(:id => 28669546014, :entities => {:urls => urls_array})
  uri = tweet.uris.first
  expect{ uri.url }.to_not raise_error
  expect{ uri.expanded_url }.to_not raise_error
  expect{ uri.display_url }.to_not raise_error
end

It fails with;

Failure/Error: expect{ uri.url }.to_not raise_error
       expected no Exception, got #<URI::InvalidURIError: the scheme http does not accept registry part: with_underscore.example.com (or bad hostname?)> with backtrace:
         # ./lib/twitter/base.rb:77:in `block in define_uri_method'
         # ./spec/twitter/tweet_spec.rb:325:in `block (4 levels) in <top (required)>'
         # ./spec/twitter/tweet_spec.rb:325:in `block (3 levels) in <top (required)>'

[sferik]

Technically, hostnames that contain underscores are not valid, so Ruby’s built-in URI library refuses to parse them. The addressable gem does a better job of handling some of these invalid-but-exist-in-the-real-world edge cases.

[matthewrudy]

that's great, thanks.

[sferik]

Try upgrading to version 5.0.1. That ought to fix it.