tonic-error 0.1.2

A helper trait to assist with passing error types through return tonic::Status messages.

Usage

This works with the thiserror crates, but using that is not required. If you are not using thiserror then at the moment you will need to manually implement std::fmt::Display for your type. Your error type will also need to derive serde::{Serialize, Deserialize}.

In order to use this, you will need to #[derive(TonicError)] on your error type.

#[derive(Debug, Error, TonicError, Serialize, Deserialize)]
pub enum MathsError {
    #[error("division by zero for inputs: a={0} b={1}")]
    DivByZero(i32, i32),
}

The TonicError trait provides two functions that will be able to convert your type to and from a tonic::Status. fn to_status(&self) -> Status and fn from_status(s: &'de Status) -> Result<Self, TonicErrorError> - the TonicErrorError is used to wrap serde and tonic errors, as well as if the incorrect Status code is set, or the key is not a part of the Status metadata map.

These examples are taken from the included examples.

Server Side

async fn div(&self, req: Request<DivRequest>) -> Result<Response<DivResponse>, Status> {
    let req = req.into_inner();
    if req.b == 0 {
        // We invoke `to_status` on our error type.
        return Err(MathsError::DivByZero(req.a, req.b).to_status());
    }
    let result = req.a as f64 / req.b as f64;
    Ok(Response::new(DivResponse { result }))
}

Client Side

pub async fn div(&mut self, a: i32, b: i32) -> Result<f64, MathsError> {
    let req = Request::new(DivRequest { a, b });
    let resp = self
        .client
        .div(req)
        .await
        // We map the `tonic::Status` to our custom error type.
        .map_err(|s| TonicError::from_status(&s).unwrap())?;

    Ok(resp.into_inner().result)
}

Example

See the tonic-error-example subdirectory in this repo for a working client/server example.

License

This is released under the Apache 2.0 license.