Why does Swift URL's init?(string: String, relativeTo: URL?) add only the protocol?

Question

When I do

let baseUrl = URL(string: "ftp://www.foobar.com/foo/bar/")
let finalUrl = URL(string: "//barfoo.com/bar/foo", relativeTo: baseUrl)
print(finalUrl?.absoluteString ?? "ooops")

it prints ftp://barfoo.com/bar/foo

Is this expected behaviour? Is this documented somewhere?

[EDIT] I was just wondering, because I'd expect that all the stuff, including server and path would be taken from the base url, like here:

let baseUrl = URL(string: "ftp://www.foobar.com/foo/bar/")
let finalUrl = URL(string: "boo/far", relativeTo: baseUrl)
print(finalUrl?.absoluteString ?? "ooops")

which prints ftp://www.foobar.com/foo/bar/boo/far.

Why is the server and everything else from the base url being ignored in the first example? Is that due to the // at the beginning? And is this part of some RFC or something or documented somewhere else?

This looks correct. What were you expecting this to return? What do you mean by "replace only the protocol?" By "protocol" I assume you mean scheme, and that's the one thing that **isn't** replaced. — Rob Napier, Aug 08 '23 at 14:11
Thanks @RobNapier. Yes I mean the scheme (which will be interpreted in a way that will end up in using a certain protocol). And yes, it's a bit confusing ... With "replace" actually I mean "added". What I found confusing is, that the only thing taken from the baseUrl `ftp://www.foobar.com/foo/bar` is the `ftp:`, while "relativeTo" suggests that the full server and path will be used from the baseUrl. I will clarify that with an edit. — Kai Huppmann, Aug 08 '23 at 15:05

score 3 · Accepted Answer · answered Aug 08 '23 at 15:31

As you note, this is due to the // at the beginning. From the URI RFC, the format of a URI is:

  URI         = scheme ":" hier-part [ "?" query ] [ "#" fragment ]
  hier-part   = "//" authority path-abempty
              / path-absolute
              / path-rootless
              / path-empty

When you begin your URI with //, you're indicating that it should replace the authority and path, which it does. As described in Section 3.3, a path-absolute cannot begin with //:

                / path-absolute   ; begins with "/" but not "//"

If your intention is to replace the path without modifying the authority (userinfo, host, port), then you can pass an absolute path (beginning with /) rather than an authority and path (beginning with //):

let finalUrl = URL(string: "/barfoo.com/bar/foo", relativeTo: baseUrl)
// ftp://www.foobar.com/barfoo.com/bar/foo

And if you want to append a path, you can pass a rootless path (not beginning with a /):

let finalUrl = URL(string: "barfoo.com/bar/foo", relativeTo: baseUrl)
// ftp://www.foobar.com/foo/barfoo.com/bar/foo

For operations more complex that this, you should definitely look at URLComponents instead, which allows you to query and modify parts more explicitly and safely.

Thanks very much. As I found out he URI RFC, the. Section 5.2.2 describes an algorithm (https://datatracker.ietf.org/doc/html/rfc3986#section-5.2.2), to construct a target URI from base uri and reference uri, which exactly matches, what we see here, in your and my examples. — Kai Huppmann, Aug 08 '23 at 21:00

Why does Swift URL's init?(string: String, relativeTo: URL?) add only the protocol?

1 Answers1