Support URL canonicalization/normalization

[sanmai-NL]

This is a feature request. For a project I need to iterate over the path segments that make up the path component of a URL, relative to a base URL.

It appears this library currently (at 1.2.1) does not help with this. To do it manually, I think I must canonicalize the URLs in the pair so that I can directly compare as strings their path components and determine whether/where they diverge or extend.

See these tests to get a taste of what I mean.

[SimonSapin]

With this library, parsed_base_url.join(str) parses the given string relatively to the given base URL. Parsing does do some normalization along the way. For example (from the first test case you linked), the protocol is ASCII-lowercased and . and .. path segments are resolved.

https://tools.ietf.org/html/rfc3986#section-6 discusses how there are many possible ways to define equivalence and normalization, with a "comparison ladder" although this is not strictly one-dimensional. https://url.spec.whatwg.org/#url-equivalence picks one way, and this library implements it with PartialEq for Url (that is, overloading the == operator).

Does this do what you want? If not, how precisely?

[sanmai-NL]

I think my original issue had to do with the special behavior of the join method based on the presence of a trailing slash in the parsed URL.

I've made an example that shows how to do what I was looking for back then. @SimonSapin: Do you find it educational enough to add as an example?

But the usefulness of my solution hinges on the quality of the normalization. See also iron/staticfile#90 (comment) . @SimonSapin: Are you interested in improving the quality/degree of normalization further?

[SimonSapin]

I’ve filed #461 to have a look at iron/staticfile#90 (comment) specifically. Closing as it is unlikely that this crate will provide normalization beyond what is specified in https://url.spec.whatwg.org/, if only because there is no single answer to what the behavior should be.