minzoom is too high

[andrewharvey]

I'm trying to understand the minzoom calculation logic at

mapnik-omnivore/lib/utils.js

Line 6 in 805d115

module.exports.zoomsBySize = function(filepath, extent, callback) {

.

It uses the original file size to number of tiles ratio to find the minzoom. Is this the best heuristic, since formats which compress worse (geojson vs shp) will have a higher minzoom for the exact same data.

Secondly the automated minzoom I'm seeing is too high. Here is a sample file to demonstrate, if that link doesn't work, it's the last file here

As a SHP file the minzoom is set as 5, the exact same data as a GEOJSON file the minzoom is set as 6.

In practice, that sample file needs to show up at least at minzoom 3 but ideally all the way out to minzoom 0.

I ran tile-live-copy with minzoom=1 and the tile created for zoom 1 is only 1.4KB so it's not that huge is it?

EDIT: I assume this code is what's being used when I upload a new Tileset in Mapbox Studio, as that's where I'm running into this issue.

[rclark]

Is this the best heuristic, since formats which compress worse (geojson vs shp) will have a higher minzoom for the exact same data.

Also, datasets with varying density have issues (see #140). This is definitely not the best heuristic, but it is also incredibly efficient compared to other algorithms that we've tried out -- like trying to get a sense of vertex density or rendering tiles optimistically until you hit a threshold. Definitely open to other ideas for other ways to assess this though!