9 releases (breaking)
0.7.0 | Nov 12, 2024 |
---|---|
0.5.0 | Sep 11, 2024 |
0.0.0 | Jul 31, 2024 |
#281 in Network programming
167 downloads per month
145KB
3K
SLoC
wayfind
A speedy, flexible router for Rust.
NOTE: wayfind
is still a work in progress.
Why another router?
wayfind
attempts to bridge the gap between existing Rust router options:
- fast routers, lacking in flexibility
- flexible routers, lacking in speed
Real-world projects often need fancy routing capabilities, such as projects ported from frameworks like Ruby on Rails, or those adhering to specifications like the Open Container Initiative (OCI) Distribution Specification.
The goal of wayfind
is to remain competitive with the fastest libraries, while offering advanced routing features when needed. Unused features shouldn't impact performance - you only pay for what you use.
Features
Dynamic Routing
Dynamic parameters can match any byte, excluding the path delimiter /
.
We support both:
- whole segment parameters:
/{name}/
- inline parameters:
/{year}-{month}-{day}/
Dynamic parameters are greedy in nature, similar to a regex .*
, and will attempt to match as many bytes as possible.
Example
use std::error::Error;
use wayfind::{Path, Router};
fn main() -> Result<(), Box<dyn Error>> {
let mut router = Router::new();
router.insert("/users/{id}", 1)?;
router.insert("/users/{id}/files/{filename}.{extension}", 2)?;
let path = Path::new("/users/123")?;
let search = router.search(&path)?.unwrap();
assert_eq!(*search.data, 1);
assert_eq!(search.route, "/users/{id}");
assert_eq!(search.parameters[0], ("id", "123"));
let path = Path::new("/users/123/files/my.document.pdf")?;
let search = router.search(&path)?.unwrap();
assert_eq!(*search.data, 2);
assert_eq!(search.route, "/users/{id}/files/{filename}.{extension}");
assert_eq!(search.parameters[0], ("id", "123"));
assert_eq!(search.parameters[1], ("filename", "my.document"));
assert_eq!(search.parameters[2], ("extension", "pdf"));
Ok(())
}
Wildcard Routing
Wildcard parameters can match any byte, including the path delimiter /
.
We support both:
- inline wildcards:
/{*path}.html
- mid-route wildcards:
/api/{*path}/help
- end-route catch-all:
/{*catch_all}
Like dynamic parameters, wildcard parameters are also greedy in nature.
Example
use std::error::Error;
use wayfind::{Path, Router};
fn main() -> Result<(), Box<dyn Error>> {
let mut router = Router::new();
router.insert("/files/{*slug}/delete", 1)?;
router.insert("/{*catch_all}", 2)?;
let path = Path::new("/files/documents/reports/annual.pdf/delete")?;
let search = router.search(&path)?.unwrap();
assert_eq!(*search.data, 1);
assert_eq!(search.route, "/files/{*slug}/delete");
assert_eq!(search.parameters[0], ("slug", "documents/reports/annual.pdf"));
let path = Path::new("/any/other/path")?;
let search = router.search(&path)?.unwrap();
assert_eq!(*search.data, 2);
assert_eq!(search.route, "/{*catch_all}");
assert_eq!(search.parameters[0], ("catch_all", "any/other/path"));
Ok(())
}
Optional Groups
Optional groups allow for parts of a route to be absent.
They are commonly used for:
- optional IDs:
/users(/{id})
- optional trailing slashes:
/users(/)
- optional file extensions:
/images/{name}(.{extension})
They work via 'expanding' the route into equivilant, simplified routes.
/release/v{major}(.{minor}(.{patch}))
/release/v{major}.{minor}.{patch}
/release/v{major}.{minor}
/release/v{major}
There is a small overhead to using optional groups, due to Arc
usage internally for data storage.
Example
use std::error::Error;
use wayfind::{Path, Router};
fn main() -> Result<(), Box<dyn Error>> {
let mut router = Router::new();
router.insert("/users(/{id})", 1)?;
router.insert("/files/{*slug}/{file}(.{extension})", 2)?;
let path = Path::new("/users")?;
let search = router.search(&path)?.unwrap();
assert_eq!(*search.data, 1);
assert_eq!(search.route, "/users(/{id})");
assert_eq!(search.expanded, Some("/users"));
let path = Path::new("/users/123")?;
let search = router.search(&path)?.unwrap();
assert_eq!(*search.data, 1);
assert_eq!(search.route, "/users(/{id})");
assert_eq!(search.expanded, Some("/users/{id}"));
assert_eq!(search.parameters[0], ("id", "123"));
let path = Path::new("/files/documents/folder/report.pdf")?;
let search = router.search(&path)?.unwrap();
assert_eq!(*search.data, 2);
assert_eq!(search.route, "/files/{*slug}/{file}(.{extension})");
assert_eq!(search.expanded, Some("/files/{*slug}/{file}.{extension}"));
assert_eq!(search.parameters[0], ("slug", "documents/folder"));
assert_eq!(search.parameters[1], ("file", "report"));
assert_eq!(search.parameters[2], ("extension", "pdf"));
let path = Path::new("/files/documents/folder/readme")?;
let search = router.search(&path)?.unwrap();
assert_eq!(*search.data, 2);
assert_eq!(search.route, "/files/{*slug}/{file}(.{extension})");
assert_eq!(search.expanded, Some("/files/{*slug}/{file}"));
assert_eq!(search.parameters[0], ("slug", "documents/folder"));
assert_eq!(search.parameters[1], ("file", "readme"));
Ok(())
}
Constraints
Constraints allow for custom logic to be injected into the routing process.
We support constraints for all types of parameters:
- Dynamic constraint:
/{name:constraint}
- Wildcard constraint:
/{*name:constraint}
The typical use-case for constraints would be to run a regex, or a simple FromStr
implementation, against a path segment.
A common mistake would be to use these for validation of parameters. This should be avoided.
If a constraint fails to match, and no other suitable match exists, it results in a Not Found
response, rather than any sort of Bad Request
.
They act as an escape-hatch for when you need to disambiguate routes.
The current constraint implementation has a number of limitations:
- constraints cannot take parameters
- checks cannot make use of any prior state
- checks cannot store data after a successful check
Default Constraints
wayfind
ships with a number of default constraints.
Curently, these can't be disabled.
Name | Method |
---|---|
u8 |
u8::from_str |
u16 |
u16::from_str |
u32 |
u32::from_str |
u64 |
u64::from_str |
u128 |
u128::from_str |
usize |
usize::from_str |
i8 |
i8::from_str |
i16 |
i16::from_str |
i32 |
i32::from_str |
i64 |
i64::from_str |
i128 |
i128::from_str |
isize |
isize::from_str |
f32 |
f32::from_str |
f64 |
f64::from_str |
bool |
bool::from_str |
ipv4 |
Ipv4Addr::from_str |
ipv6 |
Ipv6Addr::from_str |
Example
use std::error::Error;
use wayfind::{Constraint, Path, Router};
struct NamespaceConstraint;
impl Constraint for NamespaceConstraint {
const NAME: &'static str = "namespace";
fn check(segment: &str) -> bool {
segment
.split('/')
.all(|part| {
!part.is_empty() && part.chars().all(|c| c.is_ascii_alphanumeric() || c == '.' || c == '_' || c == '-')
})
}
}
fn main() -> Result<(), Box<dyn Error>> {
let mut router = Router::new();
router.constraint::<NamespaceConstraint>()?;
router.insert("/v2", 1)?;
router.insert("/v2/{*name:namespace}/blobs/{type}:{digest}", 2)?;
let path = Path::new("/v2")?;
let search = router.search(&path)?.unwrap();
assert_eq!(*search.data, 1);
assert_eq!(search.route, "/v2");
let path = Path::new("/v2/my-org/my-repo/blobs/sha256:1234567890")?;
let search = router.search(&path)?.unwrap();
assert_eq!(*search.data, 2);
assert_eq!(search.route, "/v2/{*name:namespace}/blobs/{type}:{digest}");
assert_eq!(search.parameters[0], ("name", "my-org/my-repo"));
assert_eq!(search.parameters[1], ("type", "sha256"));
assert_eq!(search.parameters[2], ("digest", "1234567890"));
let path = Path::new("/v2/invalid repo/blobs/uploads")?;
assert!(router.search(&path)?.is_none());
Ok(())
}
User-Friendly Error Messages
Where possible, we try to provide user-friendly error messages.
Example
use std::error::Error;
use wayfind::{Constraint, Router, errors::ConstraintError};
const ERROR_DISPLAY: &str = "
duplicate constraint name
The constraint name 'my_constraint' is already in use:
- existing constraint type: 'rust_out::ConstraintA'
- new constraint type: 'rust_out::ConstraintB'
help: each constraint must have a unique name
try:
- Check if you have accidentally added the same constraint twice
- Ensure different constraints have different names
";
struct ConstraintA;
impl Constraint for ConstraintA {
const NAME: &'static str = "my_constraint";
fn check(segment: &str) -> bool {
segment == "a"
}
}
struct ConstraintB;
impl Constraint for ConstraintB {
const NAME: &'static str = "my_constraint";
fn check(segment: &str) -> bool {
segment == "b"
}
}
fn main() -> Result<(), Box<dyn Error>> {
let mut router: Router<usize> = Router::new();
router.constraint::<ConstraintA>()?;
let error = router.constraint::<ConstraintB>().unwrap_err();
assert_eq!(error.to_string(), ERROR_DISPLAY.trim());
Ok(())
}
Router Display
Routers can print their routes as an tree diagram.
[*]
denotes nodes within the tree that can be matched against.
Currenty, this doesn't handle split multi-byte characters well.
Example
use std::error::Error;
use wayfind::Router;
const ROUTER_DISPLAY: &str = "
/
├─ user [*]
│ ╰─ /
│ ├─ createWithList [*]
│ ├─ log
│ │ ├─ out [*]
│ │ ╰─ in [*]
│ ╰─ {username} [*]
├─ pet [*]
│ ╰─ /
│ ├─ findBy
│ │ ├─ Status [*]
│ │ ╰─ Tags [*]
│ ╰─ {petId} [*]
│ ╰─ /uploadImage [*]
├─ store/
│ ├─ inventory [*]
│ ╰─ order [*]
│ ╰─ /
│ ╰─ {orderId} [*]
╰─ {*catch_all} [*]
";
fn main() -> Result<(), Box<dyn Error>> {
let mut router = Router::new();
router.insert("/pet", 1)?;
router.insert("/pet/findByStatus", 2)?;
router.insert("/pet/findByTags", 3)?;
router.insert("/pet/{petId}", 4)?;
router.insert("/pet/{petId}/uploadImage", 5)?;
router.insert("/store/inventory", 6)?;
router.insert("/store/order", 7)?;
router.insert("/store/order/{orderId}", 8)?;
router.insert("/user", 9)?;
router.insert("/user/createWithList", 10)?;
router.insert("/user/login", 11)?;
router.insert("/user/logout", 12)?;
router.insert("/user/{username}", 13)?;
router.insert("/{*catch_all}", 14)?;
assert_eq!(router.to_string(), ROUTER_DISPLAY.trim());
Ok(())
}
Performance
wayfind
is fast, and appears to be competitive against other top performers in all benchmarks we currently run.
However, as is often the case, your mileage may vary (YMMV). Benchmarks, especially micro-benchmarks, should be taken with a grain of salt.
Benchmarks
All benchmarks ran on a M1 Pro laptop.
Check out our codspeed results for a more accurate set of timings.
Context
For all benchmarks, we percent-decode the path before matching. After matching, we convert any extracted parameters to strings.
Some routers perform these operations automatically, while others require them to be done manually.
We do this to try and match behaviour as best as possible. This is as close to an "apples-to-apples" comparison as we can get.
matchit
inspired benches
In a router of 130 routes, benchmark matching 4 paths.
Library | Time | Alloc Count | Alloc Size | Dealloc Count | Dealloc Size |
---|---|---|---|---|---|
wayfind | 394.86 ns | 4 | 265 B | 4 | 265 B |
matchit | 493.32 ns | 4 | 416 B | 4 | 448 B |
xitca-router | 574.70 ns | 7 | 800 B | 7 | 832 B |
path-tree | 609.10 ns | 4 | 416 B | 4 | 448 B |
ntex-router | 1.7254 µs | 18 | 1.248 KB | 18 | 1.28 KB |
route-recognizer | 4.6767 µs | 160 | 8.505 KB | 160 | 8.537 KB |
routefinder | 6.4543 µs | 67 | 5.024 KB | 67 | 5.056 KB |
actix-router | 20.905 µs | 214 | 13.93 KB | 214 | 13.96 KB |
path-tree
inspired benches
In a router of 320 routes, benchmark matching 80 paths.
Library | Time | Alloc Count | Alloc Size | Dealloc Count | Dealloc Size |
---|---|---|---|---|---|
wayfind | 5.8725 µs | 59 | 2.567 KB | 59 | 2.567 KB |
path-tree | 9.0383 µs | 59 | 7.447 KB | 59 | 7.47 KB |
matchit | 9.1017 µs | 140 | 17.81 KB | 140 | 17.83 KB |
xitca-router | 10.762 µs | 209 | 25.51 KB | 209 | 25.53 KB |
ntex-router | 30.119 µs | 201 | 19.54 KB | 201 | 19.56 KB |
route-recognizer | 92.110 µs | 2872 | 191.7 KB | 2872 | 204.8 KB |
routefinder | 101.03 µs | 525 | 48.4 KB | 525 | 48.43 KB |
actix-router | 178.83 µs | 2201 | 128.8 KB | 2201 | 128.8 KB |
License
wayfind
is licensed under the terms of both the MIT License and the Apache License (Version 2.0).
Inspirations
- poem: Initial experimentations started out as a Poem router fork
- matchit: Performance leader among pre-existing routers
- path-tree: Extensive testing and router display feature
- ASP.NET Core: Constraints-based approach to routing