Crates.io | gotham_restful |
lib.rs | gotham_restful |
version | 0.9.0 |
source | src |
created_at | 2019-10-27 21:47:42.845494 |
updated_at | 2024-05-14 15:33:40.826039 |
description | RESTful additions for the gotham web framework |
homepage | |
repository | https://github.com/msrd0/gotham_restful |
max_upload_size | |
id | 176243 |
size | 144,229 |
This crate is an extension to the popular gotham web framework for Rust. It allows you to create resources with assigned endpoints that aim to be a more convenient way of creating handlers for requests.
This crate is just as safe as you’d expect from anything written in safe Rust - and
#![forbid(unsafe_code)]
ensures that no unsafe was used.
There are a set of pre-defined endpoints that should cover the majority of REST APIs. However, it is also possible to define your own endpoints.
Assuming you assign /foobar
to your resource, the following pre-defined endpoints exist:
Endpoint Name | Required Arguments | HTTP Verb | HTTP Path |
---|---|---|---|
read_all | GET | /foobar | |
read | id | GET | /foobar/:id |
search | query | GET | /foobar/search |
create | body | POST | /foobar |
update_all | body | PUT | /foobar |
update | id, body | PUT | /foobar/:id |
delete_all | DELETE | /foobar | |
delete | id | DELETE | /foobar/:id |
Each of those endpoints has a macro that creates the neccessary boilerplate for the Resource. A simple example looks like this:
/// Our RESTful resource.
#[derive(Resource)]
#[resource(read)]
struct FooResource;
/// The return type of the foo read endpoint.
#[derive(Serialize)]
struct Foo {
id: u64
}
/// The foo read endpoint.
#[read]
fn read(id: u64) -> Success<Foo> {
Foo { id }.into()
}
Defining custom endpoints is done with the #[endpoint]
macro. The syntax is similar to that
of the pre-defined endpoints, but you need to give it more context:
use gotham_restful::gotham::hyper::Method;
#[derive(Resource)]
#[resource(custom_endpoint)]
struct CustomResource;
/// This type is used to parse path parameters.
#[derive(Clone, Deserialize, StateData, StaticResponseExtender)]
struct CustomPath {
name: String
}
#[endpoint(
uri = "custom/:name/read",
method = "Method::GET",
params = false,
body = false
)]
fn custom_endpoint(path: CustomPath) -> Success<String> {
path.name.into()
}
Some endpoints require arguments. Those should be
i64
or String
.RequestBody
.search?id=1
. The
type needs to implement QueryStringExtractor
.Additionally, all handlers may take a reference to gotham’s State
. Please note that for async
handlers, it needs to be a mutable reference until rustc’s lifetime checks across await bounds
improve.
By default, every request body is parsed from json, and every respone is converted to json using serde_json. However, you may also use raw bodies. This is an example where the request body is simply returned as the response again, no json parsing involved:
#[derive(Resource)]
#[resource(create)]
struct ImageResource;
#[derive(FromBody, RequestBody)]
#[supported_types(mime::IMAGE_GIF, mime::IMAGE_JPEG, mime::IMAGE_PNG)]
struct RawImage {
content: Vec<u8>,
content_type: Mime
}
#[create]
fn create(body: RawImage) -> Raw<Vec<u8>> {
Raw::new(body.content, body.content_type)
}
You can read request headers from the state as you would in any other gotham handler, and specify custom response headers using Response::header.
#[derive(Resource)]
#[resource(read_all)]
struct FooResource;
#[read_all]
async fn read_all(state: &mut State) -> NoContent {
let headers: &HeaderMap = state.borrow();
let accept = &headers[ACCEPT];
let mut res = NoContent::default();
res.header(VARY, "accept".parse().unwrap());
res
}
To make life easier for common use-cases, this create offers a few features that might be helpful when you implement your web server. The complete feature list is
auth
Advanced JWT middlewarecors
CORS handling for all endpoint handlersdatabase
diesel middleware supporterrorlog
log errors returned from endpoint handlersfull
enables all features except without-openapi
openapi
router additions to generate an openapi specwithout-openapi
(default) disables openapi
support.In order to enable authentication support, enable the auth
feature gate. This allows you to
register a middleware that can automatically check for the existence of an JWT authentication
token. Besides being supported by the endpoint macros, it supports to lookup the required JWT secret
with the JWT data, hence you can use several JWT secrets and decide on the fly which secret to use.
None of this is currently supported by gotham’s own JWT middleware.
A simple example that uses only a single secret looks like this:
#[derive(Resource)]
#[resource(read)]
struct SecretResource;
#[derive(Serialize)]
struct Secret {
id: u64,
intended_for: String
}
#[derive(Deserialize, Clone)]
struct AuthData {
sub: String,
exp: u64
}
#[read]
fn read(auth: AuthStatus<AuthData>, id: u64) -> AuthSuccess<Secret> {
let intended_for = auth.ok()?.sub;
Ok(Secret { id, intended_for })
}
fn main() {
let auth: AuthMiddleware<AuthData, _> = AuthMiddleware::new(
AuthSource::AuthorizationHeader,
AuthValidation::default(),
StaticAuthHandler::from_array(b"zlBsA2QXnkmpe0QTh8uCvtAEa4j33YAc")
);
let (chain, pipelines) = single_pipeline(new_pipeline().add(auth).build());
gotham::start(
"127.0.0.1:8080",
build_router(chain, pipelines, |route| {
route.resource::<SecretResource>("secret");
})
)
.expect("Failed to start gotham");
}
The cors feature allows an easy usage of this web server from other origins. By default, only
the Access-Control-Allow-Methods
header is touched. To change the behaviour, add your desired
configuration as a middleware.
A simple example that allows authentication from every origin (note that *
always disallows
authentication), and every content type, looks like this:
#[derive(Resource)]
#[resource(read_all)]
struct FooResource;
#[read_all]
fn read_all() {
// your handler
}
fn main() {
let cors = CorsConfig {
origin: Origin::Copy,
headers: Headers::List(vec![CONTENT_TYPE]),
max_age: 0,
credentials: true
};
let (chain, pipelines) = single_pipeline(new_pipeline().add(cors).build());
gotham::start(
"127.0.0.1:8080",
build_router(chain, pipelines, |route| {
route.resource::<FooResource>("foo");
})
)
.expect("Failed to start gotham");
}
The cors feature can also be used for non-resource handlers. Take a look at CorsRoute
for an example.
The database feature allows an easy integration of diesel into your handler functions. Please
note however that due to the way gotham’s diesel middleware implementation, it is not possible
to run async code while holding a database connection. If you need to combine async and database,
you’ll need to borrow the connection from the State
yourself and return a boxed future.
A simple non-async example looks like this:
#[derive(Resource)]
#[resource(read_all)]
struct FooResource;
#[derive(Queryable, Serialize)]
struct Foo {
id: i64,
value: String
}
#[read_all]
fn read_all(conn: &mut PgConnection) -> QueryResult<Vec<Foo>> {
foo::table.load(conn)
}
type Repo = gotham_middleware_diesel::Repo<PgConnection>;
fn main() {
let repo = Repo::new(&env::var("DATABASE_URL").unwrap());
let diesel = DieselMiddleware::new(repo);
let (chain, pipelines) = single_pipeline(new_pipeline().add(diesel).build());
gotham::start(
"127.0.0.1:8080",
build_router(chain, pipelines, |route| {
route.resource::<FooResource>("foo");
})
)
.expect("Failed to start gotham");
}
The OpenAPI feature is probably the most powerful one of this crate. Definitely read this section carefully both as a binary as well as a library author to avoid unwanted suprises.
In order to automatically create an openapi specification, gotham-restful needs knowledge over
all routes and the types returned. serde
does a great job at serialization but doesn’t give
enough type information, so all types used in the router need to implement
OpenapiType
. This can be derived for almoust any type and there
should be no need to implement it manually. A simple example looks like this:
#[derive(Resource)]
#[resource(read_all)]
struct FooResource;
#[derive(OpenapiType, Serialize)]
struct Foo {
bar: String
}
#[read_all]
fn read_all() -> Success<Foo> {
Foo {
bar: "Hello World".to_owned()
}
.into()
}
fn main() {
gotham::start(
"127.0.0.1:8080",
build_simple_router(|route| {
let info = OpenapiInfo {
title: "My Foo API".to_owned(),
version: "0.1.0".to_owned(),
urls: vec!["https://example.org/foo/api/v1".to_owned()]
};
route.with_openapi(info, |mut route| {
route.resource::<FooResource>("foo");
route.openapi_spec("openapi");
route.openapi_doc("/");
});
})
)
.expect("Failed to start gotham");
}
Above example adds the resource as before, but adds two other endpoints as well: /openapi
and /
.
The first one will return the generated openapi specification in JSON format, allowing you to easily
generate clients in different languages without worying to exactly replicate your api in each of those
languages. The second one will return documentation in HTML format, so you can easily view your
api and share it with other people.
The openapi feature has some gotchas you should be aware of.
The name of a struct is used as a “link” in the openapi specification. Therefore, if you have two structs with the same name in your project, the openapi specification will be invalid as only one of the two will make it into the spec.
By default, the without-openapi
feature of this crate is enabled. Disabling it in favour of the
openapi
feature will add additional type bounds and method requirements to some of the traits and
types in this crate, for example instead of Endpoint
you now have to implement
EndpointWithSchema
. This means that some code might only compile on either feature, but not
on both. If you are writing a library that uses gotham-restful, it is strongly recommended to pass
both features through and conditionally enable the openapi code, like this:
#[derive(Deserialize, Serialize)]
#[cfg_attr(feature = "openapi", derive(openapi_type::OpenapiType))]
struct Foo;
Like all rust crates, this crate will follow semantic versioning guidelines. However, changing the MSRV (minimum supported rust version) is not considered a breaking change.
Copyright (C) 2019-2024 Dominic Meiser and contributors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
https://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.