
I look after a map style that’s designed to show rural pedestrian-focused information (hiking routes etc.). It natively supports zoom levels up to raster 24 and vector 26 (which is equivalent to raster 27), which means that small things like fire hydrants and manholes can be shown at high zooms without getting in the way of everything else. The raster and vector versions differ slightly, partly for technological restrictions (finding out “which routes this guidepost is part of” is easier on vector than raster).
A big question, though, is how do I find out what is missing? The styles evolved from an overlay for the the OSM Carto raster style in 2014 and have been added to ever since. However, the things that people map in OSM has increased a lot in that time. Sometimes I spot something obvious that’s missing, but how do I make sure that I haven’t ignored some relatively high-usage and important tag?
The answer is taginfo and especially the instance that Geofabrik run for Britain and Ireland. If I click “keys” at the top left I get a list of all the keys used in OSM in Britain and Ireland, in order of usage. For each of those (let’s take barrier as an example) I can click through to see all of the values used for that key in Britain and Ireland, and can then check whether I actually process them.
A list of “all tag/value combinations used by each project” is maintained within the project at github. The initial lists were created by a shell script; since then I’ve mostly updated the raster, vector (and mkgmap) versions manually.
Looking few the barrier list a few days ago I found that https://wiki.openstreetmap.org/wiki/Tag:barrier=height_restrictor was commonly used (about 3,000 uses in Britain and Ireland) but it was ignored by my mmap styles, so it made sense to add it. See here and here for raster and here and here for vector. Scripts automate the process of copying the new icon from raster, creating the vector sprite files and creating the multiple vector .json files for the different languages that the vector tiles support. An example is this, which can be seen in raster and vector.
Some tags and values are commonly used but I don’t currently show them. For example source is still the third-most used object key but it really would not make sense to show it. Some keys just aren’t used consistently (for example https://wiki.openstreetmap.org/wiki/Tag:natural=landform).
How far have I got? I’ve got on to the third page of keys, which is those “only” used 400,000 times in OSM in Britain and Ireland. To get to 100,000 uses I’d need to go to page 7, to 10,000 page 22, 1,000 page 57, 100 page 135, 10 page 329 and 1 page 647 - clearly it’s not worth looking at everything. The process so far has found about 25 combinations that I didn’t realise were quite so popular.
Discussion