Add lasso selection to node graph #3333

wade-cheng · 2025-11-03T02:32:37Z

Closes #2532.

wade-cheng · 2025-11-08T22:48:37Z

I should say while this is on my mind right now, performance should probably be looked at for this PR. We convert the entire vec of polygon points from one unit to another once every frame.

0HyperCube

Thanks for your work on this. It seems to work nicely.

It is good to hear you thinking about the performance. However current architectural decisions unfortunately means an excessive number of allocations are necessary.

0HyperCube · 2025-11-23T15:50:30Z

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

+	/// plus a flag indicating if it has been dragged since the mousedown began.
+	lasso_selection_curr: Option<(Vec<DVec2>, bool)>,


nit: It is probably clearer to make a struct { lasoo_points: Vec<DVec2>, has_dragged: bool } rather than relying on the doc comment to name the fields. However I suppose the current implementation is consistent with the box selection.

I wonder if it is even necessary to store the has_dragged bool. Can't you just check if the number of points is ≥ 2?

Re: paragraph 2. Good point, pretty sure that's all the functionality that the dragged bool is used for in the box selection logic anyways.

Hm. Perhaps the has_dragged field helps make it explicit that we do something related to whether we have moved from the initial mousedown or not. But I agree with removing it.

Also, I took a look, and the only reason box selection can't do without the field either is because we want to update_node_graph_hints only after moving from the initial mousedown. That is, if the initial mousedown was to update the hints, we could remove the field. But for the analogous lasso logic, we just check if the length is at least two, again.

implemented in 4f86201 below

0HyperCube · 2025-11-23T15:53:05Z

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

+					let lasso_selection_viewport: Vec<DVec2> = lasso_selection_curr
+						.iter()
+						.map(|selection_point| network_metadata.persistent_metadata.navigation_metadata.node_graph_to_viewport.transform_point2(*selection_point))
+						.collect();


There is no need to call .collect() since you use into_iter() later. You might have to move the frontend message to here though if the borrow checker is unhappy.

The borrow checker indeed was. And yes, moving the frontend message there fixes it.

implemented in 49efe27 below

0HyperCube · 2025-11-23T15:54:48Z

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

+					// WARNING WARNING WARNING: this commented-out code is copy pasted from UpdateBoxSelection above and has not been edited for lasso
+					// The mouse button was released but we missed the pointer up event
+					// if ((e.buttons & 1) === 0) {
+					// 	completeBoxSelection();
+					// 	boxSelection = undefined;
+					// } else if ((e.buttons & 2) !== 0) {
+					// 	editor.handle.selectNodes(new BigUint64Array(previousSelection));
+					// 	boxSelection = undefined;
+					// }
+


What is this relating to? It looks like javascript.

Oh neat, that is javascript for some reason. Yeah, the identical comment appears in UpdateBoxSelection, and got copied over when I copy-pasted the entire UpdateBoxSelection block. Didn't know if it was important so I left it.

To answer the question more explicitly, don't really know what this relates to, but your guess would be better than mine anyhow.

The box selection seems to be working fine for many months, so it is probably safe to delete this comment (in both places).

implemented in 4f86201 below

0HyperCube · 2025-11-23T15:55:48Z

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

+						let node_graph_point = network_metadata
+							.persistent_metadata
+							.navigation_metadata
+							.node_graph_to_viewport


nit: Probably less verbose to let node_graph_to_viewport = network_metadata.persistent_metadata.navigation_metadata.node_graph_to_viewport;

implemented in 0aa3860 below

0HyperCube · 2025-11-23T16:00:10Z

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

+					if nodes != previous_selection {
+						responses.add(NodeGraphMessage::SelectedNodesSet {
+							nodes: nodes.into_iter().collect::<Vec<_>>(),
+						});
+					}


It is probably easiest to just always send this message (rather than allocating the previous_selection HashMap each frame).

I can't intuit which is better, but I'll take your word for it. When I or whoever makes the change, I assume they should also make the identical analogous change to the box selection code?

I guess it doesn't matter very much either way.

Leaving it for now, then.

0HyperCube · 2025-11-23T16:05:05Z

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

+						fn points_to_polygon(points: &[DVec2]) -> Vec<PathSeg> {
+							points
+								.windows(2)
+								.map(|w| PathSeg::Line(Line::new(dvec2_to_point(w[0]), dvec2_to_point(w[1]))))
+								.chain(std::iter::once(PathSeg::Line(Line::new(
+									dvec2_to_point(*points.last().unwrap()),
+									dvec2_to_point(*points.first().unwrap()),
+								))))
+								.collect()
+						}
+						points_to_polygon(lasso_selection_curr)


Rather than allocating, consider a simplye point in polygon algorithm https://wrfranklin.org/Research/Short_Notes/pnpoly.html.

Hmm, I feel the disappeal of needing to allocate, but I'm not sure how point-in-polygon helps. I stole click_targets.node_click_target.intersect_path from... uh, probably the box selection code, and I didn't look further because I knew it worked. I'd need to think about it lots and dig deeper into the code, but first impression: aren't nodes not points? We rather need nodes (rounded rectangles) in polygon?

That is, without some digging, I currently don't know how to get the bounding boxes of all the nodes.

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

wade-cheng · 2025-11-24T23:37:55Z

Made some initial tweaks, discussion above. Feel free to mark whatever resolved if you agree with changes.

I suppose the last holdup is the "point in polygon" note that I'd be stuck on.

wade-cheng mentioned this pull request Nov 3, 2025

Add lasso selection mode to the node graph as well #2532

Open

wade-cheng force-pushed the node-selection-parity branch from 4059fde to 9e255fa Compare November 3, 2025 02:54

wade-cheng changed the title ~~Improve node graph and artboard selection parity~~ Add lasso selection to node graph Nov 11, 2025

0HyperCube reviewed Nov 23, 2025

View reviewed changes

wade-cheng force-pushed the node-selection-parity branch 2 times, most recently from 0203b77 to 7e4eb0f Compare November 24, 2025 20:16

wade-cheng added 4 commits November 24, 2025 15:23

transcribe 0hypercube's frontend help code

9907a38

change lasoo to lasso

451f762

finish adding node graph lasso select (also with alt-subtract)

0ec28b9

add hint for lasso select

03aa7f9

wade-cheng force-pushed the node-selection-parity branch from 7e4eb0f to 03aa7f9 Compare November 24, 2025 20:27

wade-cheng added 4 commits November 24, 2025 17:05

remove lasso_selection_curr drag flag and extraneous comments

4f86201

refer to correct scope in error messages

ea97704

remove one collect after appeasing the borrow checker

49efe27

make node_graph_to_viewport less verbose (dedup dot chains)

0aa3860

		/// plus a flag indicating if it has been dragged since the mousedown began.
		lasso_selection_curr: Option<(Vec<DVec2>, bool)>,

Uh oh!

Add lasso selection to node graph #3333

Are you sure you want to change the base?

Add lasso selection to node graph #3333

Uh oh!

Conversation

wade-cheng commented Nov 3, 2025 • edited by Keavon Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

wade-cheng commented Nov 8, 2025

Uh oh!

0HyperCube left a comment

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!

wade-cheng Nov 23, 2025 • 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!

Choose a reason for hiding this comment

Uh oh!

wade-cheng Nov 24, 2025 • 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!

Uh oh!

wade-cheng commented Nov 24, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

wade-cheng commented Nov 3, 2025 •

edited by Keavon

Loading

wade-cheng Nov 23, 2025 •

edited

Loading

wade-cheng Nov 24, 2025 •

edited

Loading