The tiles produced by PostGIS and published via
pg_tileserv are “Mapbox vector tiles”, a widely used de facto standard encoding of vector tiles.
The purpose of vector tiles is to efficiently transfer map features over the network, so they optimize for size, using a variety of techniques while retaining enough context to be useful to the client mapping environment.
Coordinates in tiles are quantized to integer values, and the default resolution of vector tiles is 4096 by 4096. The default resolution can be altered using the
DefaultResolution configuration parameter.
Tiles are rendered independently. For features with wide styles near borders, a copy of the feature needs to appear in both neighboring tiles, or a rendering failure will occur.
The default tile buffer is 256 pixels, which is enough for most rendering cases. You can make your tiles smaller if you have narrow rendering styles, by reducing the
DefaultBuffer configuration parameter.
A single feature can end up in multiple tiles, and the unique identifier allows the client side renderer to do things like roll-overs and highlights on features that span tile boundaries.
The tile server can automatically populate the “id” element, but only in cases where:
- PostGIS version is >= 3.0, as the ST_AsMVT() function did not support feature id until then.
- The table being published has a integer primary key defined. This key will be used as the “id” automatically.
For function layers, the “id” can be populated, but that task is left to the function author, who will be calling the
ST_AsMVT() function in their code, and must remember to populate the feature id name field. The column that is chosen to populate the “id” element must be unique per feature.