DataFusion Proto

DataFusion is an extensible query execution framework, written in Rust, that uses Apache Arrow as its in-memory format.

This crate is a submodule of DataFusion that provides a protocol buffer format for representing query plans and expressions.

Serializing Expressions

Based on examples/expr_serde.rs

use datafusion_common::Result;
use datafusion_expr::{col, lit, Expr};
use datafusion_proto::bytes::Serializeable;

fn main() -> Result<()> {
    // Create a new `Expr` a < 32
    let expr = col("a").lt(lit(5i32));

    // Convert it to an opaque form
    let bytes = expr.to_bytes()?;

    // Decode bytes from somewhere (over network, etc.)
    let decoded_expr = Expr::from_bytes(&bytes)?;
    assert_eq!(expr, decoded_expr);
    Ok(())
}

Serializing Logical Plans

Based on examples/logical_plan_serde.rs

use datafusion::prelude::*;
use datafusion_common::Result;
use datafusion_proto::bytes::{logical_plan_from_bytes, logical_plan_to_bytes};

#[tokio::main]
async fn main() -> Result<()> {
    let ctx = SessionContext::new();
    ctx.register_csv("t1", "tests/testdata/test.csv", CsvReadOptions::default())
        .await
        ?;
    let plan = ctx.table("t1").await?.into_optimized_plan()?;
    let bytes = logical_plan_to_bytes(&plan)?;
    let logical_round_trip = logical_plan_from_bytes(&bytes, &ctx)?;
    assert_eq!(format!("{:?}", plan), format!("{:?}", logical_round_trip));
    Ok(())
}

Serializing Physical Plans

Based on examples/physical_plan_serde.rs

use datafusion::prelude::*;
use datafusion_common::Result;
use datafusion_proto::bytes::{physical_plan_from_bytes,physical_plan_to_bytes};

#[tokio::main]
async fn main() -> Result<()> {
    let ctx = SessionContext::new();
    ctx.register_csv("t1", "tests/testdata/test.csv", CsvReadOptions::default())
        .await
        ?;
    let logical_plan = ctx.table("t1").await?.into_optimized_plan()?;
    let physical_plan = ctx.create_physical_plan(&logical_plan).await?;
    let bytes = physical_plan_to_bytes(physical_plan.clone())?;
    let physical_round_trip = physical_plan_from_bytes(&bytes, &ctx)?;
    assert_eq!(format!("{:?}", physical_plan), format!("{:?}", physical_round_trip));
    Ok(())
}

Generated Code

The prost/tonic code can be generated by running, which in turn invokes the Rust binary located in gen

This is necessary after modifying the protobuf definitions or altering the dependencies of gen, and requires a valid installation of protoc.

./regen.sh