Very impressive!

Btw: 7 decimal digits already correspond to about 1.1cm.

Aha! Here we probably have our bad boy: osm.org/way/512314784 (a building with one node in Vienna and the rest of its nodes in Zambia).

@maxerickson: If that was the case, could it also explain that this glitch is only visible on zoom 19 and not on any other zoom level?

wow, that’s odd.

The brown area has an apparently horizontal edge (at a latitude of approximately 40.13558) when you follow it from here eastwards (up until a longitude of ~27.47677), and a regular zig-zag structure in southwards direction. Towards the north-west it looks “randomly glitchy”.

Very cool, but a main killer feature from the original walking papers is missing here (as far as I can see): the ability to automatically georeference scanned-in walking papers to quickly have them in an OSM editor as a background layer.

@Jedrzej Pelka: You mean the short flash of a black screen when switching layers, right? Does it look better now? I’ve now actually added in a bit of fading to make the layer switching smoother.

@imagico: Right. But I’m not sure if Joost (or anyone else) has already had time to actually look into it, though.

@Zverik yeah, I think so.

@CloCkWeRX: There’s already a ticket for that on github: #219.

JOSM’s ContourMerge plugin can also be quite helpful in such situations (especially if you want to avoid creating “unnecessary” multipolygon relations for areas without actual holes): First, split the forest very roughly along the river, unglue the middle section and use ContourMerge to join the two forest halves with the respective riverbank outlines.

PS: For those interested in the technical implementation details: Recently I’ve switched out Vladimir Agafonkin’s rbush library with his kdbush implementation, which offers even faster indexing, resulting shaving off another few seconds of startup times. Awesome stuff!

Jennings, I think I just run the commands as explained here (I’m not sure why it reads zoom=16 here – this should also be 15 or lower). That generates a 1 gigapixel (32768x32768) density map, which cut into tiles corresponds to max zoom level 7 on the slippy map: http://tyrasd.github.io/osm-node-density/#7/-7.961/-248.373/latest,places

I tried to use my tms tile source https://s3-eu-west-1.amazonaws.com/osm-node-density/2016/{z}/{x}/{y}.png with mapboxgl as well, and see similarly blurry results. Not sure what’s going on. Maybe it’s related to https://github.com/mapbox/mapbox-gl-js/issues/4552?

Hey. Great work and thanks for providing this as a service!

PS: are the cached (raw) augmented diffs also publicly available?

Biggest problem is probably finding unambigous syntax for all the sorts of mentions (node/way/relation/changeset/user) that people want.

What about just using URIs as identifiers? Github shortens typed URLs to other tickets: http://github.com/project/issues/123 becomes #123. That should work for us as well (e.g. http://www.openstreetmap.org/<type>/<id> becomes type/id or something similar). For user mentions the @username pattern is already a well established.

Wikipedia editors created a lot of stubs from Geonames which went to generate new Wikidata items having only the label in the sh iso code

oh… that sounds like a lot of fun to clean up after this import.

In fact, I already found quite a lot of duplicate wikidata entry pairs where one stems from a “real” wikipedia article and one from a geonames-imported stub article. (e.g. Q1526768 and Q18473363.) Fortunately, one can quite easily fix them by merging the items with this wikidata tool: https://www.wikidata.org/wiki/Special:MergeItems. (@sabas: I’ve signed those points on your tool as “non mappabile” because the respective item is now a redirect, is that the correct way to do it?).

You can also see which OSM tags are already linked from wikidata on the following (daily updated) list on taginfo: https://taginfo.openstreetmap.org/projects/wikidata_org#tags. Strangely, it only reports 872 items.

Here’s a similar example: osm.org/way/397462262#map=19/39.92535/32.83750&layers=D 🙈

Tools like keepright often indirectly show such cases by detecting the non-intersecting overlapping highways.

@joost: nope. here you go: https://gist.github.com/tyrasd/80b2cbd9a3991e9b295124c1a6165bc2. Code and freshly processed raw data is included. osmnodes-eu.geostatgrid.csv uses keys that are compatible with the grid ids from the pop dataset, but note that their “reference grid” doesn’t include cells with 0 population.

@goclem: In this visualization (as well as the updated one) from 2016, there is no minimum number of nodes in a pixel. The maximum depends very much on the zoom level you’re looking at. Frederic’s analysis from 2013 may give you some more insight into what absolute numbers one might have to expect.

@dekstop: Thanks for the additional explanation! It’s much clearer now.