Routing modes #66
Routing modes #66
Conversation
|
@fibonacci1729 I went for now with your constructor idea from #51. It would be easy to change it to make the decision at handle time instead of construction time if you're leaning that way. |
itowlson
force-pushed
the
route-on-trailing-wildcard
branch
from
0b3bfc1 to
67e175b
Compare
| @@ -154,9 +154,38 @@ pub type Params = Captures<'static, 'static>; | |||
| /// # fn handle_single_segment(req: Request, params: Params) -> anyhow::Result<Response> { todo!() } | |||
| /// # fn handle_exact(req: Request, params: Params) -> anyhow::Result<Response> { todo!() } | |||
| /// ``` | |||
| /// | |||
| /// Route based on the trailing segment of a Spin wildcard route, instead of on the full path | |||
There was a problem hiding this comment.
I know this follows an existing pattern, but it somehow feels weird that the example code here is on the struct itself instead of on the method. But maybe I would feel the same way if it were the other way around?
There was a problem hiding this comment.
@rylev Good point. My rough thinking is that these don't document handle() or suffix() in isolation, but rather the interaction between new()/suffix(), get/post/put/etc, and handle() - that is, they are examples of "how to use the Router struct", rather than an individual method in isolation. It does lead to a rather large and hard to navigate docs block though.
And there's something to be said for also having examples on individual methods, because a reader might have an anchored link that takes them directly to a method and bypasses the struct docs.
For now, I'm going to punt on this for this PR, and we can re-assess and/or try different things as part of a separate examples pass.
There was a problem hiding this comment.
This looks great @itowlson ! Thanks for knocking it out. I can work on getting a release out once we land this.
Signed-off-by: itowlson <[email protected]>
itowlson
force-pushed
the
route-on-trailing-wildcard
branch
from
67e175b to
5bd986f
Compare
Fixes #51
The unit test is rather synthetic because it relies on the same assumption that the functional code uses, but I couldn't think of a way to force a request through Real Actual Spin. I did test with an actual Spin binary and it behaved as desired. Better ideas welcome.