Expand description
Rust SDK for Compute@Edge.
This Rustdoc page is a reference for how to use individual APIs from this SDK. For a guide-level introduction to using Compute@Edge with this SDK, see Rust on Compute@Edge at the Fastly Developer Hub.
Migrating from SDK version 0.5.*
and earlier
Version 0.6.0
introduced some substantial changes to the API to improve convenience,
capability, and discoverability. While there are many changed names and new types, the execution
model of a Compute@Edge program has not changed. This guide highlights the changes that most
existing Compute@Edge programs will need to make in order to work with the improved API.
New Request
and Response
; no more RequestExt
and ResponseExt
Replace Request<Body>
with Request
, and Response<Body>
with Response
, and then remove
references to RequestExt
and ResponseExt
.
In previous versions of the API, Request
and Response
were provided by the third-party
http
crate. While those types are flexible enough to represent Compute@Edge HTTP values, the
new types are purpose-built for Compute@Edge and include a number of conveniences, such as
automatic conversion of method argument types and built-in support for JSON and form
data. Methods like Request::send()
that were previously in the extension traits RequestExt
and ResponseExt
are now defined directly on the types, so those traits are no longer needed.
For example, a minimal program using #[fastly::main]
now looks like:
use fastly::{Error, Request, Response};
#[fastly::main]
fn main(ds_req: Request) -> Result<Response, Error> {
Ok(ds_req.send("example_backend")?)
}
“Client” instead of “downstream”
Call Request::from_client()
to get the client request, instead of the previous
fastly::downstream_request()
function.
Other functions and methods that used “downstream” to refer to the client that makes the initial
HTTP request now also use “client”. For example, Response::send_downstream()
is now
Response::send_to_client()
, and downstream_client_ip_addr()
is now
Request::get_client_ip_addr()
.
Header methods instead of HeaderMap
Access headers directly through Request
and Response
, rather than through
http::header::HeaderMap
. For example, code that previously
looked like:
let my_val = req.headers().get(header_name);
req.headers_mut().insert(other_header_name, header_value);
Should now look like:
let my_val = req.get_header(header_name);
req.set_header(other_header_name, header_value);
Automatic argument conversion
Pass trusted string literals and other similar types to functions like Request::get_header()
directly, rather than converting them explicitly.
In previous Compute@Edge programs, it was very common to see code like:
HeaderName::try_from("my-header").unwrap()
This has an explicit type conversion and error handling, despite the fact that we know the
string "my-header"
will always be a valid header name.
Many method arguments now have types like impl ToHeaderName
rather than HeaderName
, allowing
you to mix and match strings, byte vectors, and other types into your argument types with
conversions performed automatically. See the documentation of the convert
module’s traits
like ToHeaderName
for details about which types can be converted,
and when to watch out for panics.
Builder-style methods on Request
and Response
Remove uses of http::request::Builder
and http::response::Builder
, and use the methods of
Request
and Response
prefixed by with_
to use builder-style method chaining. For example:
Request::get("https://example.com")
.with_header("my-header", "hello!")
.with_header("my-other-header", "Здравствуйте!")
.send("example_backend")?;
There are still non-builder-style getter and setter methods on Request
and Response
, so
you can mix and match styles. For example:
let mut req = Request::get("https://example.com").with_header("my-header", "hello!");
if needs_translation {
req.set_header("my-other-header", "Здравствуйте!");
}
req.send("example_backend")?;
Methods to read and modify prefixes of HTTP bodies
Rather than buffering entire bodies with methods like Body::into_bytes()
or
Body::into_string()
, you can use a new set of methods to inspect and modify a certain number
of bytes from the beginning of the body:
These methods automatically write the modified prefix back to the front of the body, making them ideal for situations where you want to inspect and possibly change some metadata at the beginning of a file. See the methods’ documentation for examples.
Modules
Macros
Structs
Error
type, a wrapper around a dynamic error type.