scuffle-batching

Crates.ioscuffle-batching
lib.rsscuffle-batching
version
sourcesrc
created_at2024-11-28 16:45:48.200393
updated_at2024-12-08 09:23:52.707072
descriptionOptimized batching and dataloading for external services.
homepage
repositoryhttps://github.com/scufflecloud/scuffle
max_upload_size
id1464767
Cargo.toml error:TOML parse error at line 18, column 1 | 18 | autolib = false | ^^^^^^^ unknown field `autolib`, expected one of `name`, `version`, `edition`, `authors`, `description`, `readme`, `license`, `repository`, `homepage`, `documentation`, `build`, `resolver`, `links`, `default-run`, `default_dash_run`, `rust-version`, `rust_dash_version`, `rust_version`, `license-file`, `license_dash_file`, `license_file`, `licenseFile`, `license_capital_file`, `forced-target`, `forced_dash_target`, `autobins`, `autotests`, `autoexamples`, `autobenches`, `publish`, `metadata`, `keywords`, `categories`, `exclude`, `include`
size0
Troy (TroyKomodo)

documentation

https://docs.rs/scuffle-batching

README

scuffle-batching

[!WARNING]
This crate is under active development and may not be stable.

crates.io docs.rs


A crate designed to batch multiple requests into a single request.

Why do we need this?

Often when we are building applications we need to load multiple items from a database or some other external resource. It is often expensive to load each item individually, and this is typically why most drivers have some form of multi-item loading or executing. This crate provides an improved version of this functionality by combining multiple calls from different scopes into a single batched request.

Tradeoffs

Because we are buffering requests for a short period of time we do see higher latencies when there are not many requests. This is because the overhead from just processing the requests is lower then the time we spend buffering.

However, this is often negated when we have a large number of requests as we see on average lower latencies due to more efficient use of resources. Latency is also more consistent as we are doing fewer requests to the external resource.

Usage

Here is an example of how to use the DataLoader interface to batch multiple reads from a database.

struct MyUserLoader(SomeDatabase);

impl DataLoaderFetcher for MyUserLoader {
    type Key = i64;
    type Value = User;

    async fn load(&self, keys: HashSet<Self::Key>) -> Option<HashMap<Self::Key, Self::Value>> {
        let users = self.0.fetch("SELECT * FROM users WHERE id IN ($1)").bind(keys).await.map_err(|e| {
            error!("Failed to fetch users: {}", e);
        })?;

        Some(users.into_iter().map(|user| (user.id, user)).collect())
    }
}

let loader = DataLoader::new(MyUserLoader(database));

// Will only make a single request to the database and load both users
// You can also use `loader.load_many` if you have more then one item to load.
let (user1, user2) = join!(loader.load(1), loader.load(2));

Another use case might be to batch multiple writes to a database.

struct MyUserUpdater(SomeDatabase);

impl BatchExecutor for MyUserUpdater {
    type Request = User;
    type Response = bool;

    async fn execute(&self, requests: Vec<(Self::Request, BatchResponse<Self::Response>)>) {
        let (users, responses) = requests.into_iter().unzip();

        // You would need to build the query somehow, this is just an example
        if let Err(e) = self.0.update("INSERT INTO users (id, name) VALUES ($1, $2), ($3, $4)").bind(users).await {
            error!("Failed to insert users: {}", e);

            for response in responses {
                // Reply back saying we failed
                response.send(false);
            }

            return;
        }

        // Reply back to the client that we successfully inserted the users
        for response in responses {
            response.send(true);
        }
    }
}

let batcher = Batcher::new(MyUserUpdater(database));

// Will only make a single request to the database and insert both users
// You can also use `batcher.execute_many` if you have more then one item to insert.
let (success1, success2) = join!(batcher.execute(user1), batcher.execute(user2));
if !success1 {
    error!("Failed to insert user 1");
}

if !success2 {
    error!("Failed to insert user 2");
}

License

This project is licensed under the MIT or Apache-2.0 license. You can choose between one of them if you use this work.

SPDX-License-Identifier: MIT OR Apache-2.0

Commit count: 73

cargo fmt