Symptom (reported 2026-05-27)
While interacting with the explorer (panning, zooming, applying a material filter), the map dots appeared to stop refreshing — the visible dots remained at a state inconsistent with the current facet counts and active filter.
URL captured from the stale tab:
https://isamples.org/explorer.html?material=https%3A%2F%2Fw3id.org%2Fisample%2Fvocabulary%2Fmaterial%2F1.0%2Forganicmaterial#v=1&lat=34.9540&lng=33.3063&alt=258920&heading=360.0
When the same URL was opened in a fresh tab, the map showed a different (likely correct) cold-load state. So this is not a problem with the URL itself; it's a divergence between interactive state and URL-encoded state on the same page.
Why this is interesting
Adjacent to #234's "filter coherence" thesis. Adjacent to #235 (which fixed #facetNote not appearing on URL-load — a sibling instance of interactive-vs-URL divergence). Possibly an instance of the same underlying class of bug: some recompute paths (here, the map-dot rendering) are bound to camera/filter event handlers that don't fire (or fire and short-circuit) the same way they do on a fresh load.
Cluster-mode dot rendering specifically depends on loadRes() being called when the camera-changed or source-filter-changed event fires. If a material/feature/specimen filter change doesn't trigger a re-render of cluster dots (which honors only the source filter today, per the C2 behavior documented in #234), the visible dots could persist past a state where they no longer represent "what's in view."
What's needed
- Reproduction recipe: exact sequence of clicks / pans / filter toggles that produces the stale visual state. Hard to investigate without this.
- Once reproducible, instrument the cluster re-render path (
loadRes, camera.changed listener, source-filter handler) to identify which event was expected to fire and didn't.
Diagnostic findings so far
A Playwright probe of the captured URL at cold-load (2026-05-27):
#facetNote correctly visible: "Material / feature / specimen filters apply at sample zoom level — zoom in or click a cluster." (so the C2 disclosure is working on cold load)- Source counts at cold load were global, not viewport-scoped — possibly a separate finding (B1 may not be firing on URL-load with no
moveEnd event ever firing); needs follow-up probe.
Cross-refs
Acceptance
- Reproducible repro recipe documented
- Root-cause traced to a specific event/handler path
Symptom (reported 2026-05-27)
While interacting with the explorer (panning, zooming, applying a material filter), the map dots appeared to stop refreshing — the visible dots remained at a state inconsistent with the current facet counts and active filter.
URL captured from the stale tab:
When the same URL was opened in a fresh tab, the map showed a different (likely correct) cold-load state. So this is not a problem with the URL itself; it's a divergence between interactive state and URL-encoded state on the same page.
Why this is interesting
Adjacent to #234's "filter coherence" thesis. Adjacent to #235 (which fixed
#facetNotenot appearing on URL-load — a sibling instance of interactive-vs-URL divergence). Possibly an instance of the same underlying class of bug: some recompute paths (here, the map-dot rendering) are bound to camera/filter event handlers that don't fire (or fire and short-circuit) the same way they do on a fresh load.Cluster-mode dot rendering specifically depends on
loadRes()being called when the camera-changed or source-filter-changed event fires. If a material/feature/specimen filter change doesn't trigger a re-render of cluster dots (which honors only the source filter today, per the C2 behavior documented in #234), the visible dots could persist past a state where they no longer represent "what's in view."What's needed
loadRes,camera.changedlistener, source-filter handler) to identify which event was expected to fire and didn't.Diagnostic findings so far
A Playwright probe of the captured URL at cold-load (2026-05-27):
#facetNotecorrectly visible: "Material / feature / specimen filters apply at sample zoom level — zoom in or click a cluster." (so the C2 disclosure is working on cold load)moveEndevent ever firing); needs follow-up probe.Cross-refs
#facetNoteURL-load fix (a precedent for this class of bug)Acceptance