feat(udf): support inlined Rust UDF #14903

wangrunji0408 · 2024-01-31T11:31:42Z

I hereby agree to the terms of the RisingWave Labs, Inc. Contributor License Agreement.

What's changed and what's your intention?

This PR allows users to inline their Rust UDF code in the create-function statement.

create function gcd(int, int) returns int language rust as $$
    fn gcd(mut a: i32, mut b: i32) -> i32 {
        while b != 0 {
            let t = b;
            b = a % b;
            a = t;
        }
        a
    }
$$;

The function will be built into a wasm binary in the frontend. This requires the Rust toolchain (with the wasm32-wasi target) installed in frontend's environment. The build process will run for at least 10s, during which time the command is blocking from user's perspective. It may also require network access to download dependencies. Therefore, in the future we should consider how to:

improve the user experience while waiting for a build to complete
limit resources for build to prevent DoS attack

Checklist

I have written necessary rustdoc comments
I have added necessary unit tests and integration tests
I have added test labels as necessary. See details.
I have added fuzzing tests or opened an issue to track them. (Optional, recommended for new SQL features Sqlsmith: Sql feature generation #7934).
My PR contains breaking changes. (If it deprecates some features, please create a tracking issue to remove them in the future).
All checks passed in ./risedev check (or alias, ./risedev c)
My PR changes performance-critical code. (Please run macro/micro-benchmarks and show the results.)

My PR contains critical fixes that are necessary to be merged into the latest release. (Please check out the details)

Documentation

My PR needs documentation updates. (Please use the Release note section below to summarize the impact on users)

Release note

Add support for inline Rust UDFs.

Signed-off-by: Runji Wang <[email protected]>

BugenZhao · 2024-02-01T05:14:32Z

This is amazing!

This requires the Rust toolchain (with the wasm32-wasi target) installed in frontend's environment

Should we consider including it into the Docker image?

BugenZhao · 2024-02-01T05:19:42Z

src/frontend/src/handler/create_function.rs

+            body = Some(function_body.clone());
+
+            let wasm_binary =
+                tokio::task::spawn_blocking(move || arrow_udf_wasm::build("", &script))


IIUC, this means that we don't support specifying external crate dependencies now. This sounds good as the build.rs could run arbitrary code and is quite unsafe in a managed environment.

I'm considering whether it's possible to support a whitelist of 3rd-party crates in the future, just like how Rust's online playground does: https://arc.net/l/quote/qtopwevk

Or if wasm can serve as a with-host-tools toolchain in the distant future... 🤣

https://doc.rust-lang.org/nightly/rustc/platform-support.html#tier-2-with-host-tools

I'm considering whether it's possible to support a whitelist of 3rd-party crates in the future, just like how Rust's online playground does: https://arc.net/l/quote/qtopwevk

Good idea! We can follow their way: pre-download all the crates that Rust playground supports and add --offline to cargo build.

BugenZhao · 2024-02-01T05:22:50Z

src/frontend/src/handler/create_function.rs

+            body = Some(function_body.clone());
+
+            let wasm_binary =
+                tokio::task::spawn_blocking(move || arrow_udf_wasm::build("", &script))


The experimental feature of cargo script would also be helpful: rust-lang/cargo#12207

Yes I have investigated it. But cargo script seems to only support executables, not libraries. 😕

BugenZhao · 2024-02-01T05:24:57Z

src/frontend/src/handler/create_function.rs

+            let system_params = session.env().meta_client().get_system_params().await?;
+            let object_name = format!("{:?}.wasm", md5::compute(&wasm_binary));
+            upload_wasm_binary(
+                system_params.wasm_storage_url(),


Are we still going to deprecate the external storage for WASM binaries? I'm not sure if a Rust WASM binary fits the metadata store. 😕

What do you mean by "yes"? 😄

We will soon move WASM binaries from object store to meta store. I think meta store is capable given that we will move to SQL backend soon and binaries usually won't be larger than 10MB. 😄

stdrc · 2024-02-02T08:40:07Z

This requires the Rust toolchain (with the wasm32-wasi target) installed in frontend's environment.

Can we detect the existance of the toolchain before building? Seems in this PR the query will run for a while before reporting missing wasm32-wasi target.

The build process will run for at least 10s, during which time the command is blocking from user's perspective.

Is it possible to print some notice before building or make it run in background?

stdrc

Looks so successful to me!

Signed-off-by: Runji Wang <[email protected]>

wangrunji0408 · 2024-02-04T06:46:06Z

Can we detect the existance of the toolchain before building? Seems in this PR the query will run for a while before reporting missing wasm32-wasi target.

Sure. I've added a feature to automatically install the target on build if --offline is not given. Besides, I also added this target to our docker image.

wangrunji0408 · 2024-02-04T07:00:22Z

Is it possible to print some notice before building or make it run in background?

I did a quick investigation and found NoticeResponse in pgwire protocol maybe useful.

wangrunji0408 added 2 commits January 31, 2024 16:45

bump rquickjs

bb0745a

Signed-off-by: Runji Wang <[email protected]>

support inlined rust udf

fab4572

Signed-off-by: Runji Wang <[email protected]>

wangrunji0408 requested review from TennyZhuang, BugenZhao and xxchan January 31, 2024 11:31

wangrunji0408 requested a review from a team as a code owner January 31, 2024 11:31

github-actions bot added the type/feature label Jan 31, 2024

BugenZhao approved these changes Feb 1, 2024

View reviewed changes

BugenZhao requested a review from fuyufjh February 1, 2024 05:24

BugenZhao reviewed Feb 1, 2024

View reviewed changes

stdrc approved these changes Feb 2, 2024

View reviewed changes

wangrunji0408 added 2 commits February 4, 2024 13:42

auto install wasm32-wasi target

71e3d5c

Signed-off-by: Runji Wang <[email protected]>

add wasm32-wasi target to dockerfile

2414228

Signed-off-by: Runji Wang <[email protected]>

wangrunji0408 added this pull request to the merge queue Feb 4, 2024

Merged via the queue into main with commit f5386ca Feb 4, 2024
28 of 29 checks passed

wangrunji0408 deleted the wrj/inlined-rust-udf branch February 4, 2024 07:16

wangrunji0408 mentioned this pull request Feb 29, 2024

feat(udf): introduce #[derive(StructType)] for struct types in Rust UDF #15372

Merged

9 tasks

BugenZhao mentioned this pull request Apr 5, 2024

risingwave 1.8.0 risingwavelabs/homebrew-risingwave#34

Merged

1 task

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(udf): support inlined Rust UDF #14903

feat(udf): support inlined Rust UDF #14903

wangrunji0408 commented Jan 31, 2024

BugenZhao commented Feb 1, 2024

BugenZhao Feb 1, 2024

BugenZhao Feb 1, 2024

wangrunji0408 Feb 1, 2024

BugenZhao Feb 1, 2024

wangrunji0408 Feb 1, 2024

BugenZhao Feb 1, 2024

wangrunji0408 Feb 1, 2024

BugenZhao Feb 2, 2024

wangrunji0408 Feb 4, 2024

stdrc commented Feb 2, 2024

stdrc left a comment

wangrunji0408 commented Feb 4, 2024 •

edited

Loading

wangrunji0408 commented Feb 4, 2024

feat(udf): support inlined Rust UDF #14903

feat(udf): support inlined Rust UDF #14903

Conversation

wangrunji0408 commented Jan 31, 2024

What's changed and what's your intention?

Checklist

Documentation

Release note

BugenZhao commented Feb 1, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

stdrc commented Feb 2, 2024

stdrc left a comment

Choose a reason for hiding this comment

wangrunji0408 commented Feb 4, 2024 • edited Loading

wangrunji0408 commented Feb 4, 2024

wangrunji0408 commented Feb 4, 2024 •

edited

Loading