4 releases (breaking)

new 0.4.0 Sep 25, 2023
0.3.0 May 3, 2023
0.2.0 Sep 11, 2022
0.1.1 May 13, 2022
0.1.0 May 13, 2022

#36 in Date and time

Download history 529/week @ 2023-06-08 294/week @ 2023-06-15 266/week @ 2023-06-22 256/week @ 2023-06-29 196/week @ 2023-07-06 193/week @ 2023-07-13 251/week @ 2023-07-20 181/week @ 2023-07-27 154/week @ 2023-08-03 178/week @ 2023-08-10 240/week @ 2023-08-17 263/week @ 2023-08-24 174/week @ 2023-08-31 255/week @ 2023-09-07 232/week @ 2023-09-14 271/week @ 2023-09-21

953 downloads per month
Used in 6 crates (4 directly)

MPL-2.0 license

337 lines


Progenitor is a Rust crate for generating opinionated clients from API descriptions in the OpenAPI 3.0.x specification. It makes use of Rust futures for async API calls and Streams for paginated interfaces.

It generates a type called Client with methods that correspond to the operations specified in the OpenAPI document.

The primary target is OpenAPI documents emitted by Dropshot-generated APIs, but it can be used for many OpenAPI documents. As OpenAPI covers a wide range of APIs, Progenitor may fail for some OpenAPI documents. If you encounter a problem, you can help the project by filing an issue that includes the OpenAPI document that produced the problem.

Using Progenitor

There are three different ways of using the progenitor crate. The one you choose will depend on your use case and preferences.


The simplest way to use Progenitor is via its generate_api! macro.

In a source file (often main.rs, lib.rs, or mod.rs) simply invoke the macro:


You'll need to add the following to Cargo.toml:

+futures = "0.3"
+progenitor = { git = "https://github.com/oxidecomputer/progenitor" }
+reqwest = { version = "0.11", features = ["json", "stream"] }
+serde = { version = "1.0", features = ["derive"] }

In addition, if the OpenAPI document contains string types with the format field set to date or date-time, include

+chrono = { version = "0.4", features = ["serde"] }

Similarly, if there is a format field set to uuid:

+uuid = { version = "1.0.0", features = ["serde", "v4"] }

And if there are any websocket channel endpoints:

+base64 = "0.21"
+rand = "0.8"

If types include regular expression validation:

+regress = "0.4.1"

The macro has some additional fancy options to control the generated code:

    spec = "path/to/openapi_document.json",      // The OpenAPI document
    interface = Builder,                         // Choose positional (default) or builder style
    tags = Separate,                             // Tags may be Merged or Separate (default)
    inner_type = my_client::InnerType,           // Client inner type available to pre and post hooks
    pre_hook = closure::or::path::to::function,  // Hook invoked before issuing the HTTP request
    post_hook = closure::or::path::to::function, // Hook invoked prior to receiving the HTTP response
    derives = [ schemars::JsonSchema ],          // Additional derive macros applied to generated types

Note that the macro will be re-evaluated when the spec OpenAPI document changes (when its mtime is updated).


Progenitor includes an interface appropriate for use in a build.rs file. While slightly more onerous than the macro, a builder has the advantage of making the generated code visible.

The build.rs file should look something like this:

fn main() {
    let src = "../sample_openapi/keeper.json";
    println!("cargo:rerun-if-changed={}", src);
    let file = std::fs::File::open(src).unwrap();
    let spec = serde_json::from_reader(file).unwrap();
    let mut generator = progenitor::Generator::default();

    let tokens = generator.generate_tokens(&spec).unwrap();
    let ast = syn::parse2(tokens).unwrap();
    let content = prettyplease::unparse(&ast);

    let mut out_file = std::path::Path::new(&std::env::var("OUT_DIR").unwrap()).to_path_buf();

    std::fs::write(out_file, content).unwrap();

In a source file (often main.rs, lib.rs, or mod.rs) include the generated code:

include!(concat!(env!("OUT_DIR"), "/codegen.rs"));

You'll need to add the following to Cargo.toml:

+futures = "0.3"
+progenitor-client = { git = "https://github.com/oxidecomputer/progenitor" }
+reqwest = { version = "0.11", features = ["json", "stream"] }
+serde = { version = "1.0", features = ["derive"] }

+prettyplease = "0.1.25"
+progenitor = { git = "https://github.com/oxidecomputer/progenitor" }
+serde_json = "1.0"
+syn = "1.0"

(chrono, uuid, base64, and rand as above)

Note that progenitor is used by build.rs, but the generated code required progenitor-client.

Static Crate

Progenitor can be run to emit a stand-alone crate for the generated client. This ensures no unexpected changes (e.g. from updates to progenitor). It is however, the most manual way to use Progenitor.


cargo progenitor

    -i INPUT            OpenAPI definition document (JSON or YAML)
    -o OUTPUT           Generated Rust crate directory
    -n CRATE            Target Rust crate name
    -v VERSION          Target Rust crate version

For example:

cargo install cargo-progenitor
cargo progenitor -i sample_openapi/keeper.json -o keeper -n keeper -v 0.1.0

... or within the repo:

cargo run --bin cargo-progenitor -- progenitor -i sample_openapi/keeper.json -o keeper -n keeper -v 0.1.0

This will produce a package in the specified directory.

Options --license and --registry-name may also be used to improve metadata before publishing the static crate.

The output will use the published progenitor-client crate by default if progenitor was built from a released version. However, when using progenitor built from the repository, the progenitor-client will be inlined into the static crate by default. The command line flag --include-client can be used to override the default behaviour.

To ensure the output has no persistent dependency on Progenitor, enable --include-client.

Here is an excerpt from the emitted Cargo.toml:

bytes = "1.3.0"
chrono = { version = "0.4.23", default-features=false, features = ["serde"] }
futures-core = "0.3.25"
percent-encoding = "2.2.0"
reqwest = { version = "0.11.13", default-features=false, features = ["json", "stream"] }
serde = { version = "1.0.152", features = ["derive"] }
serde_urlencoded = "0.7.1"

The dependency versions in the generated Cargo.toml are the same as the versions that were used when progenitor was built.

Note that there is a dependency on percent-encoding which macro- and build.rs-generated clients is included from progenitor-client.

Generation Styles

Progenitor can generate two distinct interface styles: positional and builder (described below). The choice is simply a matter of preference that many vary by API and taste.

Positional (current default)

The "positional" style generates Client methods that accept parameters in order, for example:

impl Client {
    pub async fn instance_create<'a>(
        &'a self,
        organization_name: &'a types::Name,
        project_name: &'a types::Name,
        body: &'a types::InstanceCreate,
    ) -> Result<ResponseValue<types::Instance>, Error<types::Error>> {
        // ...

A caller invokes this interface by specifying parameters by position:

let result = client.instance_create(org, proj, body).await?;

Note that the type of each parameter must match precisely--no conversion is done implicitly.


The "builder" style generates Client methods that produce a builder struct. API parameters are applied to that builder, and then the builder is executed (via a send method). The code is more extensive and more complex to enable simpler and more legible consumers:

impl Client
    pub fn instance_create(&self) -> builder::InstanceCreate {

mod builder {
    pub struct InstanceCreate<'a> {
        client: &'a super::Client,
        organization_name: Result<types::Name, String>,
        project_name: Result<types::Name, String>,
        body: Result<types::InstanceCreate, String>,

    impl<'a> InstanceCreate<'a> {
        pub fn new(client: &'a super::Client) -> Self {
            // ...

        pub fn organization_name<V>(mut self, value: V) -> Self
            V: TryInto<types::Name>,
            // ...

        pub fn project_name<V>(mut self, value: V) -> Self
            V: TryInto<types::Name>,
            // ...

        pub fn body<V>(mut self, value: V) -> Self
            V: TryInto<types::InstanceCreate>,
            // ...

        pub async fn send(self) ->
            Result<ResponseValue<types::Instance>, Error<types::Error>>
            // ...

Note that, unlike positional generation, consumers can supply compatible (rather than invariant) parameters:

let result = client

The string parameters will implicitly have TryFrom::try_from() invoked on them. Failed conversions or missing required parameters will result in an Error result from the send() call.

Generated struct types also have builders so that the body parameter can be constructed inline:

let result = client
        .memory(types::ByteCount(1024 * 1024 * 1024)),

Consumers do not need to specify parameters and struct properties that are not required or for which the API specifies defaults. Neat!


~240K SLoC