Allow http:// when guessing URL pattern

[eksperimental]

I coudln't figure out what was wrong with a test, and it was because source_ref was set to "http://github.com/elixir-lang/ex_doc/" and not "https://..."
https://github.com/elixir-lang/ex_doc/blob/8269ca8/lib/ex_doc.ex#L145-L159

GitHub, GitLab and BitBucket will redirect you to https if you try to fetch a project via "http://"

My proposal is to accept http:// and return the guessed address with http://

WDYGT?

[josevalim]

My proposal is to accept http:// and return the guessed address with http://

Isn't this what we do? If you are proposing to rewrite http to https I don't think we should rewrite it. Users can fix on their side.

[milmazz]

I'm with @josevalim here, I don't think we should rewrite from http to https. @eksperimental Is that what are you suggesting?

[eksperimental]

@milmazz no! that's not 😮

As of now, if you pass source_url: "http://github.com/elixir-lang/ex_doc"
the guessed url pattern will be "http://github.com/elixir-lang/ex_doc" and not "http://github.com/elixir-lang/ex_doc/blob/master/%{path}#L%{line}"

What I meant by:

My proposal is to accept http:// and return the guessed address with http://

...was that if you provide source_url: "http://github.com/elixir-lang/ex_doc", it should set source_url_pattern: "http://github.com/elixir-lang/ex_doc/blob/master/%{path}#L%{line}"

Sorry If i was not clear enough

[eksperimental]

@josevalim nope. that's not precisely what we are doing now. it only accepts "https://"

[milmazz]

@eksperimental Thanks for your explanation! Now I understand what's your goal here, but, in order to keep things simple I would say that we leave the current behavior as is, because otherwise we also should support URLs like this:

• source_url: "http://www.github.com/elixir-lang/ex_doc" -> source_url_pattern: "http://www.github.com/elixir-lang/ex_doc/blob/master/%{path}#L%{line}"
• source_url: "https://www.github.com/elixir-lang/ex_doc" -> source_url_pattern: "https://www.github.com/elixir-lang/ex_doc/blob/master/%{path}#L%{line}"
• ... and so on.

[eksperimental]

but https is really easy to mistake for http, more than www.
We had a test including http as source_ref in ExDoc itself, and it made me waste some of my time trying to figure out that one. I just corrected it in my last commit.

[josevalim]

In this case let's do a previous pass and rewrite to https for the cases we attempt to guess the pattern. Basically let's do the thing I originally said we should not do :P -- *José Valimwww.plataformatec.com.br <http://www.plataformatec.com.br/>Founder and Director of R&D*

[milmazz]

@josevalim @eksperimental We could cover both transformations in just one step:

url = "http://www.github.com/elixir-lang/ex_doc"
String.replace(url, ~r{https?://(www\.)?(github.com|gitlab.com|bitbucket.org)/(.*)}, "https://\\2/\\3")

And then guess the URL.

[josevalim]

Let's go with only http -> https for now.

[josevalim]

We will add more complexity if we need it later on.

[josevalim]

Closing in favor of the PR.