The router currently restricts all further redirects after an absolute
redirect. Because there's no documented reason for _why_ this
restriction is in place, I'm now deeming this unnecessary. Developers
should not be restricted in this manner. Instead, configs that may
have caused infinite redirects in the past should be updated to not be
infinite. It is confusing to ignore configs with redirects after an
absolute redirect occurred because it creates different matching rules
depending on the whether an absolute redirect has happened or not.
For additional context on why I believe removing this restriction is
necessary, #13373 asks for allowing `redirectTo` to be a function. It
would make sense to allow this function to return a `UrlTree` like other
guards in the Router. When guards in the `Router` return `UrlTree`, they
cancel the current navigation and start a new one to re-do the route
matching. Since we're already in the router matching part, we don't need
to cancel the navigation. However, the restriction on absolute redirects
here then creates a weird situation where developers wouldn't see any
other redirects if they returned a `UrlTree` as an absolute redirect
from `redirectTo`.
resolves#39770
BREAKING CHANGE: Absolute redirects no longer prevent further redirects.
Route configurations may need to be adjusted to prevent infinite
redirects where additional redirects were previously ignored after an
absolute redirect occurred.
PR Close#51731
The `UrlTree` is a fundamental concept in the router's API design.
Generally speaking, a `UrlTree` would be constructed via things like
`Router.createUrlTree` or `UrlSerializer#parse`. The second example here
is the core reason to make the constructor public, regardless of how
develpers generally create a `UrlTree` in an application.
Applications can provide their own `UrlSerializer` and may want to implement their
own `parse` method. This means they would also need to create a
`UrlTree` but would not be able to do so effectively because the
constructor is marked `@internal`.
In addition, the `UrlTree` constructor only has 3 parameters, all of
which are already public. There's nothing "special" about it other than
the potential desire to push developers to use other methods for
constructing a `UrlTree` (i.e. `Router.createUrlTree`).
Also see http://b/234604699#comment9 for additional context.
PR Close#47186
This custom error message for invalid guards requires a fair bit of
extra logic in several places. This change reduces special logic for an error
case that would fail anyways. Additionally, there were already a couple
places that _did not_ have this special error and we have not seen any
indication that there is more confusion for those (resolvers and canMatch).
PR Close#46678
