feat: update state representation of seg layer for skeletons by seankmartin · Pull Request #145 · MetaCell/neuroglancer

seankmartin · 2026-05-01T11:31:10Z

The primary goal of this is to remove a decent portion of what was stored in the JSON state previously for spatial skeletons. The first part of this is to remove the inferred values that were serialized in the JSON state. The user can only directly change the 2D and 3D grid level targets from the UI. All other values around this, such as the LOD and actual chosen grid level are inferred, and should not be stored in the JSON state. Because the user can now only change these targets (they could previously change the other values in the state, even if not in the UI) we can simplify the restoration and handling of these values as there is a clearer and easier to handle flow of information.

The second part of removing what was stored in the state is to remove the relative/non-relative checkbox state. This is instead going to be inferred from the zoom-relative setting on panels by using information about the near-far plane distance. Fixing this has not been included in this PR, just removing the relative handling. New relative handling will come later, likely alongside changes to what the target value units are (likely px).

Removing this state lead to some knock on changes for parts of the code that appear to have been setup to support these settings. I've commented a few places in github about what was removed. But in summary:

The render scale widget subclass for spatial skeletons can be simplified. We don't need a lot of the custom functionality or styling. This also means we don't need the chunk tracking that was in place previously for informing the render scale widget (at least as far as I can tell - we can restore some of this if needed for a different purpose). We can track this via the regular method of the render scale histogram and the chunk stats.
The serializing and restoration of state for spatial skeletons is much more minimal and can be kept inline instead of needing custom functions and module for this.
With these changes, the draw method on some of the new layer types becomes a stub. Before this PR had already been exploring consolidating these in claude generated: experiment with consolidating layers #147 but this might make that more tenable.

will need to update the logic after this, but this moves to the version of the state where the user controlled value is the only thing stored in the state, and everything else needed for the computation is derived from this single control this will require updates in segmentation/index.ts further and then in skeleton/frontend.ts as some of the old things there won't be used

the old was adding custom css and behaviour but we can rely on the base instead for simplicity

on simplifing the render scale widget, I believe there is a good chunk of dead code. I don't think we need this as it should be handled separately.

seankmartin

I wanted to make some pointers about this. I also wanted to note that the reason this touches so much in the frontend code is that when the chunk stats are not passed to the render scale widget, they appear to be fully unusued.

So removing the state, meant the render scale widget didn't need customisation. And I didn't want to leave around the now seemingly dead code for the stats after that (though it's possible there is still some or I'm missing a different way it is used).

seankmartin · 2026-05-04T16:15:19Z

  }));
 }

-function findClosestSpatialSkeletonGridLevel(


removed in favor of always finding by spacing

seankmartin · 2026-05-04T16:22:22Z

      this.spatialSkeletonGridResolutionTarget3dExplicit = true;
-      this.applySpatialSkeletonGridResolutionTarget("3d");
-    });
-    this.spatialSkeletonGridResolutionRelative2d.changed.add(() => {


Removed these listeners in general as relative removed, both 2d and 3d

seankmartin · 2026-05-04T16:22:55Z

-    this.spatialSkeletonGridPixelSize2d.changed.add(() => {
-      if (this.spatialSkeletonGridResolutionRelative2d.value) {
-        this.applySpatialSkeletonGridResolutionTarget("2d");
-      }
-    });
-    this.spatialSkeletonGridPixelSize3d.changed.add(() => {
-      if (this.spatialSkeletonGridResolutionRelative3d.value) {
-        this.applySpatialSkeletonGridResolutionTarget("3d");
-      }
-    });
-    this.spatialSkeletonGridLevel2d.changed.add(() => {
-      if (this.suppressSpatialSkeletonGridLevel2d) return;
-      if (this.spatialSkeletonGridLevels.value.length === 0) return;
-      this.setSpatialSkeletonGridLevel(
-        "2d",
-        this.spatialSkeletonGridLevel2d.value,
-        true,
-      );
-    });
-    this.spatialSkeletonGridLevel3d.changed.add(() => {
-      if (this.suppressSpatialSkeletonGridLevel3d) return;
-      if (this.spatialSkeletonGridLevels.value.length === 0) return;
-      this.setSpatialSkeletonGridLevel(
-        "3d",
-        this.spatialSkeletonGridLevel3d.value,
-        true,
-      );


removed as no longer changable by user, only 2d and 3d targets are

seankmartin · 2026-05-04T16:24:07Z

-  spatialSkeletonGridChunkStats2d = new WatchableValue({
-    presentCount: 0,
-    totalCount: 0,
-  });
-  spatialSkeletonGridChunkStats3d = new WatchableValue({
-    presentCount: 0,
-    totalCount: 0,
-  });


Updating how these chunk stats work, I think we can be more integrated with neuroglancer. Still ensuring I've done this right to be honest.

seankmartin · 2026-05-04T16:29:57Z

-  private spatialSkeletonGridLevel2dExplicit = false;
-  private spatialSkeletonGridLevel3dExplicit = false;
-  private suppressSpatialSkeletonGridLevel2d = false;
-  private suppressSpatialSkeletonGridLevel3d = false;


our logic can be simpler now that we only have two possible user controls.

seankmartin · 2026-05-04T16:42:10Z

+    x[json_keys.HIDDEN_OPACITY_3D_JSON_KEY] =
+      this.displayState.hiddenObjectAlpha.toJSON();
+    x[json_keys.SKELETON_CROSS_SECTION_RENDER_SCALE_JSON_KEY] =
+      this.displayState.spatialSkeletonGridResolutionTarget2d.toJSON();
+    x[json_keys.SKELETON_PERSPECTIVE_RENDER_SCALE_JSON_KEY] =
+      this.displayState.spatialSkeletonGridResolutionTarget3d.toJSON();


since we serialize a lot less now, I don't think it needs its own module or tests or custom fns

seankmartin · 2026-05-04T16:48:10Z

-      | undefined;
+    const spatialDisplayState = displayState as SegmentationDisplayState &
+      SpatialSkeletonDisplayState;
+    const anyDisplayState = displayState as any;


later aim to fix this, but for now at least the cast happens once

seankmartin · 2026-05-04T16:48:50Z

    );
  }

-  private updateChunkStatsWatchable() {


Tried to remove a lot of code related to the histogram updates, I think this is more broadly handled. Related to previous comment, still verifying

seankmartin · 2026-05-04T16:57:21Z

-      ),
-    );
-  }
+  draw(_renderContext: SliceViewRenderContext) {}


Without updating chunk stats, these draw methods become stubs. It's possible some level of consolidation between layers could happen. That was briefly explored in #147 to see if it caused any major bugs but hit the token limit. can try to continue that experiment another time.

seankmartin · 2026-05-04T16:58:46Z

    this.logScaleBinSize = this.histogram.logScaleBinSize;
-  }
-
-  constructor(


We don't need a lot of this custom functionality after removing the relative checkbox and trying to use the histogram more directly instead of the chunk stats.

seankmartin · 2026-05-04T17:09:38Z

The one part of this which is a bit awkward is that you can now have something off the scale. However, I believe the real fix for that will be us changing the target value to something in pixels, and then won't need

  setSpatialSkeletonGridSizes(gridSizes: SpatialSkeletonGridSize[]) {
    const sortedSizes = [...gridSizes].sort(
      (a, b) => Math.min(b.x, b.y, b.z) - Math.min(a.x, a.y, a.z),
    );
    const levels = buildSpatialSkeletonGridLevels(sortedSizes);
    const { origin: histogramOrigin, binSize: histogramBinSize } =
      getSpatialSkeletonGridHistogramConfig(levels);
    if (
      this.spatialSkeletonGridRenderScaleHistogram2d.logScaleOrigin !==
        histogramOrigin ||
      this.spatialSkeletonGridRenderScaleHistogram2d.logScaleBinSize !==
        histogramBinSize
    ) {
      this.spatialSkeletonGridRenderScaleHistogram2d.logScaleOrigin =
        histogramOrigin;
      this.spatialSkeletonGridRenderScaleHistogram2d.logScaleBinSize =
        histogramBinSize;
      this.spatialSkeletonGridRenderScaleHistogram2d.changed.dispatch();
    }
    if (
      this.spatialSkeletonGridRenderScaleHistogram3d.logScaleOrigin !==
        histogramOrigin ||
      this.spatialSkeletonGridRenderScaleHistogram3d.logScaleBinSize !==
        histogramBinSize
    ) {
      this.spatialSkeletonGridRenderScaleHistogram3d.logScaleOrigin =
        histogramOrigin;
      this.spatialSkeletonGridRenderScaleHistogram3d.logScaleBinSize =
        histogramBinSize;
      this.spatialSkeletonGridRenderScaleHistogram3d.changed.dispatch();
    }
    this.spatialSkeletonGridLevels.value = levels;

because our histogram will be in an expected range

seankmartin added 9 commits May 1, 2026 12:44

refactor: remove custom functions for spatial skeleton json management

d4a25f6

refactor: remove code related to old state management

68b2194

refactor: remove customisation from skeleton grid render scale

be339cb

the old was adding custom css and behaviour but we can rely on the base instead for simplicity

refactor: remove chunk stats code

659008e

on simplifing the render scale widget, I believe there is a good chunk of dead code. I don't think we need this as it should be handled separately.

merge feature/edit-mode

012b15a

refactor: stub draw methods

c36911d

refactor: remove uneeded guards

6a08e2f

refactor: remove custom state fn

915315a

seankmartin commented May 4, 2026

View reviewed changes

fix: remove resolving to nearest grid

6b1e341

seankmartin added 4 commits May 4, 2026 20:12

fix: update tests for removed code

17d92f3

fix: only show current / current as lod in render scale widget

a0fe01a

fix: don't triple count histogram for 3 panels

3aa6395

fix: remove some counting issues on panels

9dad654

seankmartin requested a review from afonsobspinto May 5, 2026 08:07

seankmartin marked this pull request as ready for review May 5, 2026 08:07

afonsobspinto approved these changes May 5, 2026

View reviewed changes

afonsobspinto merged commit b662b52 into feature/edit-mode May 5, 2026
1 check passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: update state representation of seg layer for skeletons#145

feat: update state representation of seg layer for skeletons#145
afonsobspinto merged 14 commits intofeature/edit-modefrom
feat/NA-641

seankmartin commented May 1, 2026 •

edited

Loading

Uh oh!

seankmartin left a comment

Uh oh!

seankmartin May 4, 2026

Uh oh!

seankmartin May 4, 2026 •

edited

Loading

Uh oh!

seankmartin May 4, 2026

Uh oh!

seankmartin May 4, 2026

Uh oh!

seankmartin May 4, 2026

Uh oh!

seankmartin May 4, 2026

Uh oh!

seankmartin May 4, 2026

Uh oh!

seankmartin May 4, 2026

Uh oh!

seankmartin May 4, 2026

Uh oh!

seankmartin May 4, 2026

Uh oh!

seankmartin commented May 4, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

seankmartin commented May 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

seankmartin left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

seankmartin May 4, 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!

seankmartin commented May 4, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

seankmartin commented May 1, 2026 •

edited

Loading

seankmartin May 4, 2026 •

edited

Loading