fix(extensions): measure brushingRadius in true meters on MapView

[chrisgervang]

Follow up of #10661

Background

brushing_isPointInRange divides the distance by project.commonUnitsPerMeter. Below zoom 12 (PROJECTION_MODE_WEB_MERCATOR) that is the equatorial value, so brushingRadius on MapView is radius × cos(latitude) in real meters, while auto-offset mode and GlobeView already measure true meters. The same radius selects different objects across the zoom-12 boundary and between views.

	Before	After
brushing-source on MapView: brushingRadius: 3000 m, pointer at the canvas centre, San Francisco at zoom 11.5

The before image is the golden from #10661; the after image is the regenerated golden. The circle grows by 1/cos(37.75°) and now matches the extent of brushing-source-globe.

Change List

• brushing shader module: under WEB_MERCATOR + LNGLAT without pseudo-meters, divide the distance by project_size_at_latitude() at the mean latitude of the pointer and the object, mirroring project_size()
  • Behavior change on MapView below zoom 12: the circle grows by 1/cos(latitude), 1.26x at 37.75°, 2x at 60°
  • Makes brushingRadius mean meters as documented; there is no post-9.4 upgrade guide section to note it in yet
• brushing-source.png regenerated: the map circle now matches its globe twin (about 140 px for 3 km at zoom 11.5); other brushing goldens unchanged
• Documentation: brushingRadius property text states the measurement rule per view; docs/whats-new.md notes the MapView change

Ran the render spec headless on WebGL, tsc and lint.

---

Note

Medium Risk
Changes GPU distance logic for brushing on MapView below zoom 12, so apps may see a larger selection circle at higher latitudes; behavior is intentional to match documented meters.

Overview
BrushingExtension now treats brushingRadius as real meters on MapView at all zoom levels, matching GlobeView and the docs.

In the brushing shader, when Web Mercator + lng/lat is active without pseudo-meters, the pointer-to-object distance is scaled by project_size_at_latitude() at the mean latitude of the pointer and the feature—same idea as project_size(). Below zoom 12 the effective brush circle grows by 1/cos(latitude) (e.g. ~1.26× at San Francisco), so radius no longer jumps at the zoom-12 boundary or diverges from globe behavior.

Docs clarify that radius is in meters in every view; what’s new calls out the MapView fix. Render goldens for the brushing example were regenerated so map and globe circles align.

^{Reviewed by Cursor Bugbot for commit 2f2f1ac. Bugbot is set up for automated code reviews on this repo. Configure here.}

[coveralls]

Coverage is 82.555% — claude/fix-brushing-radius-meters-GveWO into claude/fix-brushing-globe-GveWO. No base build found for claude/fix-brushing-globe-GveWO.

[greptile-apps]

Confidence Score: 5/5

[Medium risk] Changes how brushing radius is measured on map views.

The PR appears safe to merge; no outstanding finding or new actionable issue was established.

Summary

The PR adjusts MapView brushing distances to account for latitude below the auto-offset threshold, updates the brushing documentation and golden image, and adds globe path-dash coverage.

• The three previous findings are manually resolved; none is carried forward as outstanding.
• No new actionable issue was established.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[Pointer and object positions] --> B[Projected common-space distance]
  B --> C{Web Mercator lng/lat without pseudo-meters?}
  C -- Yes --> D[Adjust distance at mean latitude]
  C -- No --> E[Keep projected distance]
  D --> F[Compare with brushingRadius]
  E --> F

Loading

_{Reviews (7) · Last reviewed commit: "docs(extensions): qualify the brushingRa..."}