Implement node output destructuring by JustJ01 · Pull Request #3806 · GraphiteEditor/Graphite

JustJ01 · 2026-02-22T20:59:36Z

From the discord discussions: https://discord.com/channels/731730685944922173/1210129484255076354/1475229902788628573

gemini-code-assist · 2026-02-22T20:59:52Z

Summary of Changes

Hello @JustJ01, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request finalizes the implementation of node output destructuring, a feature that significantly enhances the granularity and flexibility of node interactions. By allowing complex struct outputs to be broken down into their constituent fields, it simplifies data flow and improves the usability of nodes within the graph, making it easier to connect specific data points without intermediate processing steps.

Highlights

Node Output Destructuring: Implemented the ability to deconstruct node outputs, allowing individual fields of a struct returned by a node to be exposed as separate outputs in the node graph.
New Destruct Trait: Introduced a Destruct trait in graphene-core to define how a struct's fields can be extracted and registered as node outputs.
Procedural Macro Support: Developed a new #[derive(Destruct)] procedural macro to automatically implement the Destruct trait for structs and generate corresponding field extractor nodes.
Node Metadata Extension: Extended the NodeMetadata struct to include a list of output_fields, providing registry information for destructured outputs.

Changelog

node-graph/gcore/src/registry.rs
- Added output_fields to NodeMetadata struct to store information about destructured outputs.
- Introduced StructField struct to represent metadata for individual fields.
- Defined the Destruct trait, which types can implement to specify their destructurable fields.
node-graph/node-macro/src/codegen.rs
- Modified generate_node_code to conditionally include output_fields in NodeMetadata based on the new deconstruct_output attribute.
- Updated the protonode_identifier function to store &'static str directly in OnceLock for improved efficiency.
node-graph/node-macro/src/destruct.rs
- Added a new module containing the implementation for the derive_destruct procedural macro.
- Implemented logic to generate Destruct trait implementations for structs.
- Included functionality to create individual extractor nodes for each field of a destructured struct.
- Provided parsing for the #[output(name = "...")] attribute to allow custom naming of destructured outputs.
node-graph/node-macro/src/lib.rs
- Integrated the new destruct module.
- Registered the #[proc_macro_derive(Destruct)] attribute macro.
node-graph/node-macro/src/parsing.rs
- Added a deconstruct_output field to the NodeFnAttributes struct.
- Updated the parsing logic to recognize and process the deconstruct_output attribute for node functions.

Activity

The implementation of this feature was discussed on Discord, as referenced in the PR description: https://discord.com/channels/731730685944922173/1210129484255076354/1475229902788628573

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	`/gemini review`	Performs a code review for the current pull request in its current state.
Pull Request Summary	`/gemini summary`	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	`/gemini help`	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

Keavon · 2026-02-22T21:00:26Z

Please rebase on the latest upstream master.

gemini-code-assist

Code Review

The pull request introduces functionality for node output destructuring, allowing for the extraction of individual fields from a node's output. This involves adding output_fields to NodeMetadata, implementing a Destruct trait, and generating extractor nodes via a new destruct.rs module. The changes appear well-structured and integrate smoothly with the existing macro system. The OnceLock usage for protonode_identifier has been updated to correctly handle &'static str by leaking the boxed string, which is a common pattern for static string storage in Rust. The new deconstruct_output attribute in NodeFnAttributes provides a clear way to enable this feature for specific nodes.

node-graph/nodes/gcore/src/extract_xy.rs

editor/src/messages/portfolio/document/node_graph/document_node_definitions.rs

TrueDoctor · 2026-02-23T20:23:34Z

Also, please configure your ai / editor / git to automatically run cargo fmt on save or pre commit

JustJ01 · 2026-02-23T23:03:33Z

@TrueDoctor I have made the requested changes which u asked for

Ayush2k02 · 2026-03-02T22:55:09Z

node-graph/nodes/raster/src/adjustments.rs


+#[cfg(feature = "std")]
+#[node_macro::node(name("Split Channels"), category("Raster: Channels"), deconstruct_output)]
+fn split_channels(_: impl Ctx, input: Table<Raster<CPU>>) -> SplitChannelsOutput {


Each call to extract_channel iterates over the entire raster. This means the image is traversed 4 times. Maybe do a single-pass implementation ?

Maybe do something like this

fn split_channels(_: impl Ctx, input: Table<Raster<CPU>>) -> SplitChannelsOutput { let (red, green, blue, alpha) = input.into_four_channels(); // hypothetical single-pass function SplitChannelsOutput { red, green, blue, alpha } }

I can commit this function and implimentation if you'd like

I think in this case we should not touch the split channels node at all since it makes the code less efficient, this was done as per Keavons request.

TrueDoctor · 2026-03-08T09:06:20Z

node-graph/node-macro/src/destruct.rs

+				static FIELDS: std::sync::OnceLock<Vec<#core_types::registry::StructField>> = std::sync::OnceLock::new();
+				FIELDS.get_or_init(|| vec![
+					#(#output_fields,)*
+				]).as_slice()


Why do you use heap memory + syncronization here instead of constructing a slice in place?

TrueDoctor · 2026-03-08T09:09:33Z

node-graph/node-macro/src/destruct.rs

+
+fn generate_extractor_node(core_types: &TokenStream2, fn_name: &syn::Ident, struct_name: &syn::Ident, field_name: &syn::Ident, ty: &Type, output_name: &LitStr) -> TokenStream2 {
+	quote! {
+		#[node_macro::node(category(""), name(#output_name))]


Why do we add the name() annotation here?

TrueDoctor · 2026-03-08T09:16:03Z

node-graph/node-macro/src/destruct.rs

+fn parse_output_name(attrs: &[syn::Attribute]) -> syn::Result<Option<String>> {
+	let mut output_name = None;
+
+	for attr in attrs {
+		if !attr.path().is_ident("output") {
+			continue;
+		}
+
+		let mut this_output_name = None;
+		match &attr.meta {
+			Meta::Path(_) => {
+				return Err(Error::new_spanned(attr, "Expected output metadata like #[output(name = \"Result\")]"));
+			}
+			Meta::NameValue(_) => {
+				return Err(Error::new_spanned(attr, "Expected output metadata like #[output(name = \"Result\")]"));
+			}
+			Meta::List(_) => {
+				attr.parse_nested_meta(|meta| {
+					if meta.path.is_ident("name") {
+						if this_output_name.is_some() {
+							return Err(meta.error("Multiple output names provided for one field"));
+						}
+						let value = meta.value()?;
+						let lit: LitStr = value.parse()?;
+						this_output_name = Some(lit.value());
+						Ok(())
+					} else {
+						Err(meta.error("Unsupported output metadata. Supported syntax is #[output(name = \"...\")]"))
+					}
+				})?;
+			}
+		}
+
+		let this_output_name = this_output_name.ok_or_else(|| Error::new_spanned(attr, "Missing output name. Use #[output(name = \"...\")]"))?;
+		if output_name.is_some() {
+			return Err(Error::new_spanned(attr, "Multiple #[output(...)] attributes are not allowed on one field"));
+		}
+		output_name = Some(this_output_name);
+	}
+
+	Ok(output_name)
+}


Is this function only used for overriding the field name? I don't think this code is currently used anywhere right? If it is indeed not used, it should be removed to reduce the code complexity

TrueDoctor · 2026-03-08T09:21:39Z

node-graph/preprocessor/src/lib.rs

+		let use_memo = !available_output_fields.is_empty()
+			&& node_registry.get(&memo_node).is_some_and(|memo_implementations| {
+				memo_implementations
+					.iter()
+					.any(|(_, node_io)| node_io.call_argument == *input_type && node_io.return_value == first_node_io.return_value)
+			});


use_memo will currently always be false since you did not the new structs to the memo node implementations

TrueDoctor · 2026-03-08T09:22:57Z

node-graph/preprocessor/src/lib.rs

+					inputs: vec![NodeInput::node(source_node_id, 0)],
+					call_argument: input_type.clone(),
+					implementation: DocumentNodeImplementation::ProtoNode(memo_node.clone()),
+					visible: false,


there is no need to set visible to false since the entire network is hidden anyway

TrueDoctor · 2026-03-08T09:26:52Z

node-graph/preprocessor/src/lib.rs

+				DocumentNode {
+					inputs: vec![NodeInput::node(source_node_id, 0)],
+					call_argument: input_type.clone(),
+					implementation: DocumentNodeImplementation::ProtoNode(memo_node.clone()),


shouldn't we get a static string for the identifier? why do we need heap allocations here?

TrueDoctor · 2026-03-08T09:27:30Z

@cubic-dev-ai

cubic-dev-ai · 2026-03-08T09:27:35Z

@cubic-dev-ai

@TrueDoctor I have started the AI code review. It will take a few minutes to complete.

cubic-dev-ai

No issues found across 10 files

Since this is your first cubic review, here's how it works:

cubic automatically reviews your code and comments on bugs and improvements
Teach cubic by replying to its comments. cubic learns from your replies and gets better over time
Add one-off context when rerunning by tagging @cubic-dev-ai with guidance or docs links (including llms.txt)
Ask questions if you need clarification on any suggestion

cubic-dev-ai

1 issue found across 2 files (changes from recent commits).

Prompt for AI agents (unresolved issues)


Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="node-graph/node-macro/src/destruct.rs">

<violation number="1" location="node-graph/node-macro/src/destruct.rs:5">
P1: Custom agent: **PR title enforcement**

PR title must be imperative (e.g., "Complete node output destructuring implementation"), but it uses past tense "Completed", which violates the PR title enforcement rules.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

node-graph/node-macro/src/destruct.rs

TrueDoctor · 2026-04-04T10:08:38Z

editor/src/node_graph_executor/runtime.rs

+				// Exclude nodes with destructured outputs since their definition networks already contain the complete structure
+				let mut substitutions = preprocessor::generate_node_substitutions();
+				let node_metadata = graphene_std::registry::NODE_METADATA.lock().unwrap();
+				substitutions.retain(|id, _| node_metadata.get(id).is_none_or(|meta| meta.output_fields.is_empty()));
+				drop(node_metadata);
+				substitutions


why is this done here instea of in the generate_node_substitutions method?

This is all out of my wheelhouse. If you're able to take it from here, I'd appreciate that. I basically just attempted to get it to work from a QA perspective without really understanding what's going on internally, since the alternative would have been to close the PR outright. Sorry I don't know the answer to the question, although it's probably worth fixing if it doesn't match your expectations.

I mean you made the changes you should be able to address comments regarding them. the generate substitutions functions is used in other places as well, and it looks to me like the changes should probably apply to all callers so it would likely make sense to integrate the changes in the function (or even better instead of first generating the substitutions and then filtering them out, we should just not generate them in the first place) I won't have time to take over the pr myself

Add Destruct macro and trait for automatically acessing struct fields

2e97e81

gemini-code-assist bot reviewed Feb 22, 2026

View reviewed changes

Complete node output destructuring implementation

e215927

JustJ01 force-pushed the finish-deconstruct-outputs branch from e27f090 to e215927 Compare February 22, 2026 22:10

JustJ01 added 3 commits February 23, 2026 06:12

implemenet preprocessing expansion

f05970c

formatted files

f81c4da

convert Split Vec2 and Split Channels to proto nodes

9db03a5

TrueDoctor reviewed Feb 23, 2026

View reviewed changes

node-graph/nodes/gcore/src/extract_xy.rs Outdated Show resolved Hide resolved

TrueDoctor reviewed Feb 23, 2026

View reviewed changes

editor/src/messages/portfolio/document/node_graph/document_node_definitions.rs Outdated Show resolved Hide resolved

removed dead code

051b690

Ayush2k02 suggested changes Mar 2, 2026

View reviewed changes

TrueDoctor requested changes Mar 8, 2026

View reviewed changes

cubic-dev-ai bot reviewed Mar 8, 2026

View reviewed changes

Keavon force-pushed the master branch 6 times, most recently from d6228da to e58c1de Compare March 16, 2026 23:03

Keavon force-pushed the master branch 5 times, most recently from 9b97ab7 to 2e842cb Compare March 19, 2026 11:00

cubic-dev-ai bot reviewed Apr 3, 2026

View reviewed changes

node-graph/node-macro/src/destruct.rs Show resolved Hide resolved

Keavon changed the title ~~Completed node output destructuring implementation~~ Implement node output destructuring Apr 3, 2026

Keavon added 2 commits April 3, 2026 03:21

Merge branch 'master' into finish-deconstruct-outputs

e5e4936

Potentially solve outstanding issues

958810d

Keavon force-pushed the finish-deconstruct-outputs branch from 3adba15 to 958810d Compare April 3, 2026 10:52

TrueDoctor reviewed Apr 4, 2026

View reviewed changes

Uh oh!

Conversation

JustJ01 commented Feb 22, 2026

Uh oh!

gemini-code-assist bot commented Feb 22, 2026

Summary of Changes

Highlights

Footnotes

Uh oh!

Keavon commented Feb 22, 2026

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

Uh oh!

Uh oh!

TrueDoctor commented Feb 23, 2026

Uh oh!

JustJ01 commented Feb 23, 2026

Uh oh!

Ayush2k02 Mar 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

TrueDoctor commented Mar 8, 2026

Uh oh!

cubic-dev-ai bot commented Mar 8, 2026

Uh oh!

cubic-dev-ai bot left a comment

Choose a reason for hiding this comment

Uh oh!

cubic-dev-ai bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Ayush2k02 Mar 2, 2026 •

edited

Loading