What problem does this solve?
Some real, distinct nearby places (e.g. a bookshop, library, and internet
cafe in the same block) render as several separate, overlapping individual
pins instead of one cluster badge, which reads as visual clutter ("places
placed together"). Checked the data directly — there are no literal
duplicate entries in data/places/*.json; this is a clustering-tuning
issue, not a data-quality one.
src/components/map/map-view.tsx's Supercluster config (radius: 60, maxZoom: 18) intentionally clusters across categories (the pie-chart
cluster icon exists to show a mixed-category count), but the radius is loose
enough that nearby-but-distinct places sometimes don't merge.
Proposed solution
Raise the Supercluster radius (try 80-100) and/or lower maxZoom so
nearby cross-category places consolidate into one cluster badge more often,
at the zoom levels where overlap is currently worst. Needs a visual pass
across a few zoom levels to dial in, not a one-shot formula.
Alternatives considered (optional)
Could partition clustering per-category instead of allowing cross-category
merges, but that would lose the existing pie-chart "mix of categories here"
signal, which is a deliberate design choice.
Checklist
What problem does this solve?
Some real, distinct nearby places (e.g. a bookshop, library, and internet
cafe in the same block) render as several separate, overlapping individual
pins instead of one cluster badge, which reads as visual clutter ("places
placed together"). Checked the data directly — there are no literal
duplicate entries in
data/places/*.json; this is a clustering-tuningissue, not a data-quality one.
src/components/map/map-view.tsx's Supercluster config (radius: 60, maxZoom: 18) intentionally clusters across categories (the pie-chartcluster icon exists to show a mixed-category count), but the radius is loose
enough that nearby-but-distinct places sometimes don't merge.
Proposed solution
Raise the Supercluster
radius(try 80-100) and/or lowermaxZoomsonearby cross-category places consolidate into one cluster badge more often,
at the zoom levels where overlap is currently worst. Needs a visual pass
across a few zoom levels to dial in, not a one-shot formula.
Alternatives considered (optional)
Could partition clustering per-category instead of allowing cross-category
merges, but that would lose the existing pie-chart "mix of categories here"
signal, which is a deliberate design choice.
Checklist