Relative URLs with unknown base

[rictic]

I'm writing some code that transforms source code, specifically URL strings in HTML, CSS, and JS. These URLs may be relative (e.g. ./foo.js, /bar.css, ../search?query=cats) and my code won't have the context of the base URL that these URLs would be interpreted relative to at runtime.

I didn't see any public API that supports parsing or providing structured handling of this sort of URL. Is it a use case you'd be interested in supporting?

[SimonSapin]

I think I would call this outside of the scope of this crate. The Url type always represents an absolute URL.

What transformation do you need to do on those URL strings?

[rictic]

The main operation I need right now is to resolve a url from a base. e.g. if we encountered the url ./foo.js inside ./public/docs.html I'd want to produce the URL ./public/foo.js. This is specifically for prototyping build-time tooling around the WIP import maps spec.

Correctly handling absolute URLs, absolute paths, query params and hashes, not to mention encoding makes for a decent amount of code and a lot of overlap with other url handling.

[SimonSapin]

In terms of implementation I expect that dealing with .. path segments without a base URL is just different enough from with one that I expect that supporting either in the same code base would be non-trivial.

In terms of public API, I expect this new feature would be entirely separate from what’s already in the url crate. (The Url type would not be involved.)

For these reasons I think this would be better in a separate crate. Feel free to copy some of the code according to the license, though.

[KiChjang]

To me, this issue sounds like a confusion between URLs and URIs. My understanding is that the use case that @rictic described pertains to URIs, because they are relative paths without a base at compile-time. IIRC, URLs are explicitly required to have absolute paths.

Given the above, I'd suggest the author to look into using the http crate for URI parsing. See hyperium/http#73 for more information about the history behind http not including the url crate as one of its components, in particular this comment from @SimonSapin

[SimonSapin]

I don’t believe that’s what the difference between URL and URI is. But I also don’t believe that distinction is useful at all, or relevant here.

[rictic]

Seems reasonable, thanks!