Use UDFs in Rust

This article provides a step-by-step guide for defining Rust functions in RisingWave.

Rust functions are compiled into WebAssembly modules and then run on the embedded WebAssembly virtual machine in RisingWave. Rust UDFs provide higher performance (near native) compared to Python and Java. The RisingWave kernel handles the management of Rust UDFs, eliminating the need for additional maintenance. However, for security reasons, Rust UDFs currently do not support access to external networks and are limited to computational tasks only, with restricted CPU and memory resources. Therefore, we recommend using Rust UDFs for computationally intensive tasks like packet parsing and format conversion.

Declare your functions in RisingWave

You can utilize the CREATE FUNCTION command to develop UDFs in Rust. The syntax is outlined below:

CREATE FUNCTION function_name ( arg_name arg_type [, ...] )
    [ RETURNS return_type | RETURNS TABLE ( column_name column_type [, ...] ) ]
    LANGUAGE rust
    AS [ $$ function_body $$ | 'function_body' ];

For instance, the scalar function gcd can be defined as follows:

CREATE FUNCTION gcd(int, int) RETURNS int LANGUAGE rust AS $$
    fn gcd(mut x: i32, mut y: i32) -> i32 {
        while y != 0 {
            (x, y) = (y, x % y);
        }
        return x;
    }
$$;

The Rust code must start with a fn function, and the function's name, parameters, and return type must match those declared in the CREATE FUNCTION statement. Refer to the Data type mapping for details on the correspondence between SQL types and Rust types.

For table functions, your function must return an impl Iterator<Item = T> type, where T is the type of the returned elements. For example, to generate a sequence from 0 to n-1:

CREATE FUNCTION series(n int) RETURNS TABLE (x int) LANGUAGE rust AS $$
    fn series(n: i32) -> impl Iterator<Item = i32> {
        (0..n).into_iter()
    }
$$;

If your function returns a structured type, the syntax is a bit more complex. For instance, the following function parses a key-value pair from a string:

CREATE FUNCTION key_value(varchar) RETURNS STRUCT<key varchar, value varchar> LANGUAGE rust AS $$
    #[derive(StructType)]
    struct KeyValue<'a> {
        key: &'a str,
        value: &'a str,
    }
    #[function("key_value(varchar) -> struct KeyValue")]
    fn key_value(kv: &str) -> Option<KeyValue<'_>> {
        let (key, value) = kv.split_once('=')?;
        Some(KeyValue { key, value })
    }
$$;

First, define a structure using struct and annotate it with #[derive(StructType)]. Its fields must match the struct type declared in the CREATE FUNCTION statement. Then, define the function and annotate it with the #[function("...")] macro. The string in the macro represents the SQL signature of the function, with the tail return type being struct StructName. For the specific syntax, see arrow-udf.

Currently, in CREATE FUNCTION statement, Rust code can only use libraries from the standard library, chrono, rust_decimal, serde_json, and does not support other third-party libraries. If you wish to use other libraries, you may consider compiling WebAssembly modules manually.

Use your functions in RisingWave

After creating UDFs in RisingWave, you can employ them in SQL queries just as you would with any built-in functions. For example:

SELECT gcd(25, 15);
SELECT * FROM series(5);

Alternative: Manually build your functions into a WebAssembly module

If you want to use other libraries in your Rust functions, you can manually build your functions into a WebAssembly module and then load it into RisingWave.

Prerequisites

• Ensure that you have Rust toolchain (stable channel) installed on your computer.

• Ensure that the Rust standard library for wasm32-wasi target is installed:

  rustup target add wasm32-wasi

1. Create a project

Create a Rust project named udf:

cargo new --lib udf
cd udf

Add the following lines to Cargo.toml:

[lib]
crate-type = ["cdylib"]

[dependencies]
arrow-udf = "0.2"

2. Define your functions

In src/lib.rs, define your functions using the function macro:

use arrow_udf::function;

// define a scalar function
#[function("gcd(int, int) -> int")]
fn gcd(mut x: i32, mut y: i32) -> i32 {
    while y != 0 {
        (x, y) = (y, x % y);
    }
    x
}

// define a table function
#[function("series(int) -> setof int")]
fn series(n: i32) -> impl Iterator<Item = i32> {
    0..n
}

You can find more usages in these functions and more examples in these tests.

See the correspondence between SQL types and Rust types in the Data type mapping.

3. Build the project

Build your functions into a WebAssembly module:

cargo build --release --target wasm32-wasi

You can find the generated WASM module at target/wasm32-wasi/release/udf.wasm.

Optional: It is recommended to strip the binary to reduce its size:

# Install wasm-tools
cargo install wasm-tools

# Strip the binary
wasm-tools strip ./target/wasm32-wasi/release/udf.wasm > udf.wasm

4. Declare your functions in RisingWave

In RisingWave, use the CREATE FUNCTION command to declare the functions you defined.

There are two ways to load the WASM module:

1. Embed the WASM binary into SQL with base64 encoding. You can use the following command in psql:

   \set wasm_binary `base64 -i path/to/udf.wasm`
   CREATE FUNCTION gcd(int, int) RETURNS int LANGUAGE wasm USING BASE64 :'wasm_binary';

2. Load the WASM binary from the local file system of the frontend. Note the fs:// is URI schema, and /path/to/udf.wasm is the real path.

   CREATE FUNCTION gcd(int, int) RETURNS int LANGUAGE wasm USING LINK 'fs:///path/to/udf.wasm';

5. Use your functions in RisingWave

Once the UDFs are created in RisingWave, you can use them in SQL queries just like any built-in functions. For example:

SELECT gcd(25, 15);
SELECT series(5);

Data type mapping

The following table shows the data type mapping between SQL and Rust:

SQL type	Rust type as argument	Rust type as return value
boolean	bool	bool
smallint	i16	i16
integer	i32	i32
bigint	i64	i64
real	f32	f32
double precision	f64	f64
decimal	rust_decimal::Decimal	rust_decimal::Decimal
date	chrono::NaiveDate	chrono::NaiveDate
time	chrono::NaiveTime	chrono::NaiveTime
timestamp	chrono::NaiveDateTime	chrono::NaiveDateTime
timestamptz	not supported yet	not supported yet
interval	arrow_udf::types::Interval	arrow_udf::types::Interval
jsonb	serde_json::Value	serde_json::Value
varchar	&str	impl AsRef<str>, e.g. String, Box<str>, &str
bytea	&[u8]	impl AsRef<[u8]>, e.g. Vec<u8>, Box<[u8]>, &[u8]
smallint[]	&[i16]	impl Iterator<Item = i16>
integer[]	&[i32]	impl Iterator<Item = i32>
bigint[]	&[i64]	impl Iterator<Item = i64>
real[]	&[f32]	impl Iterator<Item = f32>
double precision[]	&[f64]	impl Iterator<Item = f64>
varchar[]	&arrow::array::StringArray	impl Iterator<Item = &str>
bytea[]	&arrow::array::BinaryArray	impl Iterator<Item = &[u8]>
others[]	not supported yet	not supported yet
struct<..>	not supported yet	user defined struct