Filtering pmtiles file at low zoom level while accounting for dropped features?

[glaroc]

Hi, we have a point file containing several million points and created a pmtiles file with Tippecanoe from it. In our Maplibre React dashboard, we now want to be able to filter the points on the map using an id stored as a feature property. We can do this using an opacity filter in Maplibre based on a feature property stored. However, if a given property only has a few points, at low zoom levels, the map will generally be empty because the filter only applies to features shown at the current zoom level and the filtered features are dropped due to the Tippecanoe algorithm. We can use the --accumulate-attribute parameter in Tippecanoe to add the ids of the features that are dropped to the features kept at the lower zoom levels and then filter based on that. Then the file becomes very big and the navigation is much slower, and sometimes we have a long list of ids to filter on.

So, to make a long story short, has anyone found an efficient way of filtering point features from a pmtiles file while accounting for the dropped features at all zoom levels ?

[HarelM]

To be honest, sounds like you need a database and not a PMTiles file...
I would look into server side filtering and returning a geojson that can be presented on the map.

[glaroc]

Geojsons with tens of millions of points won't work so we'll need aggregation/selection of points by zoom level. We were hoping to not have to do this server-side, but will look into approaches possible with PostGIS. which we already use extensively. I managed to do the filtering efficiently with only pmtiles+Maplibre (no server-side) on this site: https://quebio.ca/arbresmtl , but for the issue described above, the number of points is an order of magnitude larger, so it is pushing the limits of the possibilities for cloud-optimized filtering. Thanks for your input on this!

[HarelM]

You might be able to leverage webgl capabilities somehow using a custom layer, maybe deck.gl, but these solutions won't be easier I think...