Return Result<{Self,&Self,&mut Self}, _> instead of Result<(), _> to allow method chaining

[StaloneLab]

I'm working on the issue #299 right now and I saw some methods return Result<(), ()>, the issue I mentionned is about making them return a ParseResult<()>, but wouldn't it be better if they returned ParseResult<Self>, in order to allow method chaining with ? or (old) unwrap()?

This would actually be useful mainly for setters, I set up an actual list of them:

1. #set_href
2. HostInternal#serialize<S> (part of serde crate, but I think can be replaced)
3. (IDNA) #domain_to_unicode
4. Url#set_port
5. Url#set_ip_host
6. Url#set_password
7. Url#set_username
8. Url#set_scheme

Could work on this, be I need to be validated by someone before.

[SimonSapin]

Is this different from #299 ?

I’d prefer not to make breaking changes until there’s a stronger need to do so.

[StaloneLab]

Not for Errors here, but for Ok().
That's what I was saying just over, issue #299 suggest to return Result<(), SomeError> instead of Result<(), ()>, I'm stating Result<Self, SomeError> would be better.

I’d prefer not to make breaking changes until there’s a stronger need to do so.

Any reason for that?

[SimonSapin]

() as error is a big nono, though, isn't it?

Thank you for your well-argumented constructive input to this discussion.

[SimonSapin]

I’d prefer not to make breaking changes until there’s a stronger need to do so.

Any reason for that?

To be able to url 2.0 in Servo we’d first need to upgrade Hyper. That would be a breaking change in Hyper since Url is part of its public API. Then we’ll also need to do the same for each crate that depends on Hyper, then for crates that depend on those… There’s a cascading effect.

It’d involve coordinating pull requests to dozens of projects, with almost as many maintainers. It’d be a big pain.

[bstrie]

Considering that Hyper (along with everyone else in the web domain) is at this moment preparing for a breaking change due to async/await stabilization, if ever there was a time to consider such a breaking change it would be now.

[jdm]

Hyper no longer depends on rust-url. That being said, there was recently a breaking major release of rust-url, and it was, indeed, a big pain.