Add relative current_url

[treyhunner]

Description

current_url returns the full URL:

>>> request.htmx.current_url
'https://www.pythonmorsel.com/some-page/

I'd like a way to access just the non-domain part of this URL (I'm using it for a ?next= login redirect).

>>> request.htmx.relative_current_url
'/some-page/

I imagine something like this might work.

    @cached_property
    def relative_current_url(self) -> str | None:
        url = self.current_url
        if url:
            urlunparse(urlparse(url)._r(scheme='', netloc=''))
        return url

[adamchainz]

I wouldn't feel comfortable just dropping the origin, as it's potentially security-relevant information. Some projects are deployed with multiple origins.

Instead, why not compare with absolute URL's? I don't quite know what you're trying to do, but perhaps you could use something like this:

if request.htmx.current_url == request.build_absolute_uri('/login/'):

[treyhunner]

@adamchainz I wasn't trying to compare, but was instead redirecting to a login page while specifying a next param. That next= param needs to be a relative URL (Django gets upset when it's an absolute URL).

Instead of this:

path = urlunparse(
    urlparse(self.request.htmx.current_url)._replace(netloc="", scheme="")
)
response = redirect_to_login(
    path,
    resolve_url(self.get_login_url()),
    self.get_redirect_field_name(),
)
return HttpResponseClientRedirect(response.url)

I wanted to be able to do this:

response = redirect_to_login(
    self.request.htmx.relative_current_url,
    resolve_url(self.get_login_url()),
    self.get_redirect_field_name(),
)
return HttpResponseClientRedirect(response.url)

In my case I plan to force redirects often with ?next=<CURRENT_RELATIVE_URL>. If you think this is too niche of a use case (or too challenging of a problem to solve without raising security concerns) I understand.

[adamchainz]

Yeah I think adding the helper is probably too niche/security sensitive to solve here. Calling Django’s redirect_to_login to generate a response that you then just use the URL from doesn't sit right.

(Django gets upset when it's an absolute URL).

How exactly? redirect_to_login() doesn't seem to mind. The later “allowed redirect” check should be done by the internal function url_has_allowed_host_and_scheme(), which checks for an allowed origin, so it should be fine with absolute URL's on the same origin?

(P.S. you probably want to use urlsplit() instead of urlparse(), as I learned recently. Django could probably do with that as well...)

[treyhunner]

Yeah I think adding the helper is probably too niche/security sensitive to solve here. Calling Django’s redirect_to_login to generate a response that you then just use the URL from doesn't sit right.

I do agree that the specific use case I showed is a hack that isn't wise. That was the first case that came up and I haven't yet refactored that code.

I just came across another case today though.

Something like this in a non-HTMX Django page:

{% url "users:login" %}?next={{ request.get_full_path }}

Doesn't have an equivalent in django-htmx land without a custom template filter/tag.

{% url "users:login" %}?next={{ request.htmx.relative_current_url }}

(Django gets upset when it's an absolute URL).

How exactly? redirect_to_login() doesn't seem to mind. The later “allowed redirect” check should be done by the internal function url_has_allowed_host_and_scheme(), which checks for an allowed origin, so it should be fine with absolute URL's on the same origin?

From my testing, passing an absolute URL to next= didn't seem to work.
Django seems to just ignore the next value entirely when it wasn't a relative URL (as of Django 3.2 at least).

(P.S. you probably want to use urlsplit() instead of urlparse(), as I learned recently. Django could probably do with that as well...)

Ah I do. Thanks for noting that!

[adamchainz]

Added in #287, as current_url_abs_path.

I did and a bit of research to check naming, and was reminded that relative URL’s mean “relative to a given base URL, like ./image.pngis relative to whichever URL. The right name is “absolute-path” as per [section 4.2 of this RFC](https://www.rfc-editor.org/rfc/rfc3986#page-26), which is also what Django (kinda) uses e.g. inModel.get_absolute_url(). So I picked the name with the _abs_path` suffix, which is kinda wordy but at least correct.