Just wondering if anyone has done any performance tests on efficient production of PMtiles (preferred) or MVT.
Currently I am ready ESRI shapefile data in to SQLite (SpatiaLite) file; multiple layers so multiple tables produced in this file. I am then converting it to PMtiles using ogr2ogr as this functionality was added in v3.8 of GDAL.
The alternative is to use tippecanoe and read in geoJSON data but these would be much bigger files than my SQLite file hence the question on performance before I go this route.
For WaterwayMap.org I use tippecanoe to generate the mvt pmtiles. It works pretty well. My input is GeoJSONSeq, and lots of small features. The big WWM maps are ~20 GiB files and converted to ~1.4 GiB pmtiles in ~30 min. Tippecanoe is a little customizable, but not that much. There’s a few features / improvements I’d like to have. #1, #2.
Some stats from rough experimentation. 10 layers of linear data producing vector tiles to zoom 16.
ogr2ogr reading all layers from a single SQLite file = 2 minutes 5 seconds
tippecanoe reading from fgb files = 1 minute 15 seconds
ogr2ogr produced a larger file (73 vs 53MB) and had retained more accuracy in the original data ; overzooming beyond Z17 with ogr2ogr had better quality