cluster-distance inaccurate below zoom level 14

[hactar]

Thanks for tippecanoe, a marvelous piece of software.

I'm running it with the following arguments: "--force", "-r1", "-z22", "-Z0", "-aC", "--cluster-distance=26", "--layer=result", "--accumulate-attribute=meldungId:comma"

My assumption is that if I then draw circles on the clusters where their radius is cluster-distance / 2, these circles should never overlap. This does appear to work for zoom levels like 16, 15, and 14, but the more the user zooms out, the more the circles start overlapping. Below are some screenshots with these circles, they have a radius 10, so a diameter of 20. Yet around zoom levels of 13, 12, 11 etc they start overlapping more and more.

Zoom level 15, everything fine.

Zoom level 13, things start going wrong.

Is my configuration wrong for what I'm trying to achieve? Any help would be appreciated.

[e-n-f]

Thanks for the report. As you have discovered, tippecanoe's cluster distance calculation only takes into account the index distance between the first point of one cluster and the first point of the subsequent cluster, rather than the actual physical distance between the representative points of the clusters, and doesn't compare at all the locations of clusters that are not index-adjacent.

#359 adds a check to prevent starting a new cluster if the location of the feature under consideration is still physically close to the first feature of the current cluster, even if it would be far enough away in index distance to start a new cluster. Could you try that and see if it improves your too-close clusters?

[hactar]

Thanks for the explanation and the attempt of a quick fix. I gave that branch a quick go with the same parameters, but unfortunately the problem remains - I think it helps for one overlap on the right side of the map, but the other ones remain.