Background and context 🗺️

I’ve been using JOSM to map missing public transport routes in Chennai. It’s a project I’ve been doing for personal reasons ^{which I intend to elaborate in another post} and I began with iD only to find later that the order of the route segments matter and iD doesn’t make it easy to do reorder segments.

So I started using JOSM with a helper plugin I found. This has made my workflow much easier and faster. One small annoyance though is that I could only download a small section of a fairly large city to edit the map in JOSM. I discovered I could download larger areas if I carefully picked specific ways or nodes since the limitation (to downloading data) is the total size of data that OSM can return.

So I went and searched the web for examples of OverpassQL and their explanations, learned what I could about how the QL works in the time I had and strung together a quick and dirty query that downloads ONLY roads and bus stops within the bounding-box (that I get to easily draw in JOSM).

This probably could be improved for concision. This is the OverpassQL I use to download roads and existing relations in order for me to map bus routes using JOSM’s overpass-based data import dialogue window.

[out:xml][timeout:90][bbox:{{bbox}}];
(
  wr["highway"="primary"];
  wr["highway"="primary_link"];
  wr["highway"="secondary"];
  wr["highway"="secondary_link"];
  wr["highway"="tertiary"];
  wr["highway"="tertiary_link"];
  wr["highway"="trunk"];
  wr["highway"="trunk_link"];
  node["highway"="bus_stop"];
  node["bus"="yes"];
  node["public_transport"="platform"];
  node["public_transport"="stop_position"];
);
(._;>;);
out meta;

This has some nice conveniences for me:

1. It does not download residential roads and therefore does not clutter the viewport
2. Does not download railway lines
3. I can download a fairly large area - usually the size of an entire city.
4. I can do batch / scripted editing of bus stops to add tags (put down your pitchforks - I realized my mistakes early on and have stopped clobbering others’ work unless absolutely necessary)
5. I can save my JOSM session and not have it eat up a lot of my disk space.

Explanation of the QL

• nwr is short form for a combination of nodes, ways and relations. It’s basically an OR clause.
  • Other available combos are nw, nr, and wr.
  • Then of course, there are standalone keywords for node, way, and relation.
• Wildcards are done by specifying just the key without the value: node["public_transport"]; matches all values for the key public_transport
• Maybe I should add the PTv2 stop_area relation: relation["type"="public_transport"]["public_transport"="stop_area"]; to this code??
• In the areas I edit, there are hardly any motorways so I don’t bother adding them unless I find a missing section.
  • I could add a wildcard for highways, but I don’t want the viewport cluttered, so..
• ⚠️ The parentheses groups the queries, if you remove it, only results matching the last statement will be returned.

Notes:

The size on disk for an empty roads-and-bus-stops-only session is ~7 MiB for a city the size of Chennai which I’m editing currently. But there are not many routes or other relations mapped within the city. Even for a better mapped city like Bangalore or Delhi, I think the size on disk will not exceed ~10 MiB. Don’t quote me on that though.

OverpassQL documentation is not very friendly. Maybe there should be a page with a zillion examples of all combinations that can be written using this QL with explanations for each of them. I dream.

⚠️ I am probably missing deprecated tags for bus stop platforms🚉. But so far, I’m unaware of any.

Addendum1 (05 Feb 2022)

I realized I’m going to have to download relation data more carefully, meaning something like this:

[out:xml][timeout:90][bbox:{{bbox}}];
(
  ... snip ...
  relation["public_transport"="stop_area"];
  relation["route"="bus"];
  relation["route_master"="bus"];
);
(._;>;);
out meta;

I also realized that (quite late 😖) that I could take advantage of layers in JOSM. Downloading data as a new layer - and renaming it to my convenience - was the way to go. I could hide the layer if it ever gets in the way.

Original post made - 02 Feb 2022