escape colon in path segment

[yarafan]

Description

Considering example

f =  Faraday.new(url: 'https://service.com/') do |faraday|
    faraday.adapter(Faraday.default_adapter)
  end
f.post('service:search')

URI.parse parses relative path service:search as opaque_part (according to https://tools.ietf.org/html/rfc2396#section-3) (eg mailto:foo@example.com), so Faraday::Connection#build_exclusive_url produces wrong URI as result and passing query params leads to InvalidURIError

irb(main):004:0> URI.parse('mailto:foo@example.com').opaque
=> "foo@example.com"
irb(main):005:0> URI.parse('service:search').opaque
=> "search"

Fixes #1214

Todos

List any remaining work that needs to be done, i.e:

• Rubocop (not sure what to do with ClassLength)

[yarafan]

Not sure whether final uri should be unescaped

[iMacTia]

That's a good point, this may cause issues in case the URL contains escaped characters like spaces? Escaped columns may also be part of the URL, so I'm not sure unescaping here is a good idea at all.
E.g. Faraday.get('https://service.com/path%20with%20spaces')

[iMacTia]

I like the approach and even more the test coverage! But unfortunately the escaping solution might not be the best one 🤔. I'm curious to see if other tests are passing (I'd expect at least one to fail, if we're testing spaces!), but Rubocop is in the way...

The ClassLength cop is a tricky one, so feel free to increase the limit in the .rubocop_todo.yml file for this time. Since you added 2 lines that should be 236 now, I guess?

[iMacTia]

That's a good point, this may cause issues in case the URL contains escaped characters like spaces? Escaped columns may also be part of the URL, so I'm not sure unescaping here is a good idea at all.
E.g. Faraday.get('https://service.com/path%20with%20spaces')

[iMacTia]

@yarafan this looks good now, but is your service call working as expected even with the : escaped?
@luizkowalski maybe you want to have a quick look as well?

[yarafan]

@iMacTia I don't know any service with such uri. I tried to access uri in issue and got 302.
Would be great if @luizkowalski also check this case

[luizkowalski]

@yarafan @iMacTia wow I disabled email from comments from GitHub (cause I get too many) so I missed this one, guys, so sorry! I will check tomorrow morning first thing and will come back to you!

Thanks for the PR!

[luizkowalski]

Hey guys!

I can confirm it works and this is how the URL appeared on the logs

INFO  Rails : request: POST https://subdomain.service.com/api/loans%3Asearch?limit=50&offset=2900