feat!: Use a radix trie for route matching

[ElijahAhianyo]

Description

The current router implementation uses a Vec to store routers, which has some real limitations. For example, using a vec meant route conflict detection was cumbersome and hacky to get right. Using a Radix trie is the right data structure for this problem. This PR delegates the core Trie logic to the matchit crate. We still keep our business logic in a light wrapper over the Matchit Router.

Breaking Changes

1. Registering a route with a duplicate parameter at the same segment will now fail

Router::with_urls([
    Route::with_handler_and_name("/foo/{bar}", index, "index"),
        Route::with_handler_and_name("/foo/{baz}", add_example_form, "add"),
])

2. Registering 2 views with the same url_prefix will fail

struct TestProject;
impl Project for TestProject {
   ...
    fn register_apps(&self, apps: &mut AppBuilder, _context: &RegisterAppsContext) {
        apps.register_with_views(App1, "");
        apps.register_with_views(App2, "");
    }
}

3. When a handler and router are registered for the same route, the handler will take precedence at lookup time

let nested_router = Router::with_urls([Route::with_handler_and_name(
        "/inner/{id}",
        nested,
       "nested",
)]);

Router::with_urls([
    Route::with_handler_and_name("/foo", index, "index"),
    Route::with_router("/foo", nested_router),
])

4. A route is treated as different from its trailing-slash counterpart:

Router::with_urls([
    Route::with_handler_and_name("/foo", index, "index"),
    Route::with_handler_and_name("/foo/", another_handler, "another" ),
])

5. This does not apply to routers. The example below is invalid and will fail:

Router::with_urls([
    Route::with_router("/foo", nested_router_a),
    Route::with_router("/foo/", nested_router_b ),
])

Type of change

• Bug fix
• New feature
• Documentation
• Refactor / cleanup
• Performance improvement
• Other (describe above)

[github-actions]

Bencher Report

Project	cot
Branch	elijah/router-trie
Testbed	github-ubuntu-latest

Click to view all benchmark results

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
empty_router/empty_router	📈 view plot 🚷 view threshold	14.14 ms (+62.95%) Baseline: 8.68 ms	16.57 ms (85.32%)
json_api/json_api	📈 view plot 🚷 view threshold	1.14 ms (+8.23%) Baseline: 1.05 ms	1.37 ms (82.96%)
nested_routers/nested_routers	📈 view plot 🚷 view threshold	1.09 ms (+11.01%) Baseline: 0.98 ms	1.26 ms (86.86%)
single_root_route/single_root_route	📈 view plot 🚷 view threshold	1.03 ms (+8.99%) Baseline: 0.95 ms	1.22 ms (84.64%)
single_root_route_burst/single_root_route_burst	📈 view plot 🚷 view threshold	18.29 ms (+7.00%) Baseline: 17.09 ms	21.56 ms (84.83%)

🐰 View full continuous benchmarking report in Bencher

[ElijahAhianyo]

Should be rebased on and merged after #586

[m4tx]

Do we have any way of registering multiple routers at the same URL? This is an important feature - people might want to merge different routers at the same base URL.

If matchit doesn't let us do that because it detects a conflict, we should merge the routers. I'm not sure if we should have separate methods to do that explicitly, or keep the implicit behavior - I'll let you decide.

[m4tx]

This is not really "conflict" in most cases, is it? For instance, a numeral-only parameter name doesn't really conflict with anything. I think this error message is misleading.

[m4tx]

Why not just use expect here?

[m4tx]

Where is the change coming from? If there's a change in trailing slash handling coming from the migration, I think we should consistently use trailing slashes for URLs.

[m4tx]

Same here, we should probably modify the URL routes to include the trailing slashes.