Experimental | urlbuilder

[garrettladley]

Hello everyone, there's a new experimental urlbuilder feature in templ-go/x.

Checkout the PR here.

We're looking to collect community feedback on the feature! Does it meet your needs? If not, what else would be necessary?

Thanks again, G

[mac-hel]

Is there a way to build relative url?

[joerdav]

Currently no, only URLs that contain a scheme and a host work at the moment. But good suggestion, I think that would be useful!

[acaloiaro]

I came to suggest the same.

This PR (templ-go/x#5) is the API that I prefer.

[acaloiaro]

@joerdav I'm curious if you have any feedback on the API design of templ-go/x#5

[joerdav]

I understand the design choices, but I think the fact we are circumventing "net/url" for most of the construction of the URL we are missing some of the assurances that the std library gives us.

I think that potentially the safest way to design this builder is to wrap a *url.URL:

type URLBuilder struct {
	url *url.URL
	err      error // allows error handling to be deferred.
}

Then we can provide a urlbuilder.New func, as you already have done. But the builder functions modify the url directly:

func (u *URLBuilder) Fragment(fragment string) *URLBuilder {
	u.url.Fragment = fragment
	return u
}

// urlbuilder.New().Path("user", userid)
func (u *URLBuilder) Path(base string, elements ...string) *URLBuilder {
	path, err := url.JoinPath(base, elements...)
	u.err = err
	u.url.Path = path
	return u
}

It could also be beneficial to provide a "Parse" function that allows passing of an initial URL. It could set the err value that would be returned in the final Build step.

[acaloiaro]

I agree, *url.URL should be a core implementation detail.

As an API, my goal of commenting here is to remove the current scheme/host requirement for those who have a need to build relative URLs.

Just to clarify @garrettladley came up with this nice little API. I'm simply advocating for its design, if not its current implementation details.

[Nigel2392]

Maybe I'm missing something, and I hate to be that guy, but what is the use case for this?

Would a custom format function that returns a valid url object not suffice?

To me, at first glance, it looks shiny and cool, but not very practical.