thbz's diary

Recent diary entries

Subway stations in Seoul (2)

Posted by thbz on 5 April 2015 in English (English)

So I did what I had planned in the previous diary and checked with some Overpass queries. For example, this query should probably produce no result.


Now, the following Overpass requests work:

  • Get the subway network in Seoul (lines 1 to 9) (see on a map):

  • Get all stops on line 2 (see on a map or get the data)



I have 3 conclusions.

  1. I used the old public transport scheme, aka version 1, except where I had already added public_transport=stop_position nodes. Therefore everything is not harmonized: in subway route relations, the stop role is associated sometimes with a railway=station node, sometimes with a public_transport=stop_position node (but only to one node in every subway station for each line).

    Maybe I should have removed all public_transport=stop_position tags and kept only railway=station nodes. The public_station=stop_position tag on railways does not seem very useful unless you are ready to really map the station according to the new Public Transportation scheme, i.e. add the public_transport:version=2 tag on the route relation in JOSM. Version 2 might let client software design very precise pedestrian routing (such as what Naver Maps does very well in Korea), but that requires very high quality data. It does not seem realistic in Seoul which has a very small OSM contributor community compared to other large cities (which makes contributing more fun than in other cities like Paris where you have almost nothing to map any more, except shops).

    Therefore it is probably better to be modest and keep with the old schema (public_transport:version=1): add railway=station nodes to the route relations with role stop and that's all, except maybe for a couple of subway stations in my neighborhood.

    However, using several relations for branches (e.g. on line 5) is a good thing, but that can also be done with version 1.

  2. JOSM is a good piece of software.

  3. JOSM is a very good piece of software. The more I am exploring its advanced features (layers and filters, this time), the more I am amazed how well it is adapted to the (experienced) user.

Subway stations in Seoul

Posted by thbz on 3 April 2015 in French (Français)

I have been working on the bus and subway routes in Seoul lately, but there are still many discrepancies and missing tags. So I thought I might try to do some planning instead of always improvising, une fois n'est pas coutume. And I'll practice my English, too, and learn more about Markdown.

The references:

The rules :

  1. Work on subway lines 1 to 9. Whether other lines should be dealt the same way might be discussed later.

  2. Every existing railway=station node that corresponds to a subway station should also be tagged as public_transport=station, subway=yes and (if it's only for lines 1 to 9) station=subway. Reciprocally, any public_transport=station should be tagged the same way if it corresponds to a subway station.

  3. Existing nodes should not be moved unless really necessary (i.e. do not apply the "This node should not be part of a railway track" rule for an existing railway=station node, unless I am ready to really map a public_transport=stop_area relation for that station).

  4. Only one railway=station node should exist for every subway station, unless it is used for very different networks (lines 1 to 9 will be considered as one single network, I may let several railway=station nodes if they exist in multi-network stations such as 서울역, but I'm not sure about that).

  5. Every existing public_transport=stop_position node in the subway network should be tagged as subway=yes and be located on a railway=rail or railway=subway way. It should not be tagged as railway=station or station=subway.

  6. All the public_transport=station nodes that belong to the same line should be members, with no role, of the relation(s) associated with that line. These relations may follow the new Public transport proposal (i.e one relation for each direction or branch) or not.

The planning:

Subway stations

  1. Select all railway=station and public_transport nodes in the Seoul Metropolitan area using the following Overpass request:

    out meta;
  2. Load these nodes into a JOSM calque (ça se dit comment, en anglais?). Also load in another calque the railway=subway and railway=rail ways in the same area:

    out meta;

    and/or maybe load in another calque the relation that corresponds to each subway line:

    out meta;
  3. For each subway station:

    • add the required tags for the corresponding station node;
    • remove the duplicates if any;
    • add the node to the relation associated to the subway lines this station belongs to (do it for each subway line separately):

      out meta;

Stop positions

  1. Load in JOSM the public_transport=stop_position nodes (maybe only if they are not too far from subway stations):

    out meta;
  2. Add subway=yes for those which are located in a subway station, and move them onto the railway=rail or railway=subway way if necessary.


  • set railway=subway for all railway=rail ways associated to a subway line?

  • create public_transport=stop_area relations for every station? At least for those with at least two public_transport=* nodes ?

  • apply the following railway_station rule:

    Using the word "station" in the name=* if it is not part of the name printed on signs at the station, i.e. avoid name=Station Newton

  • make the subway map look pretty for (I know, don't map for the renderer. Anyways, if I can apply the rules and make the renderer happy, there's no harm.)

Bon, personne n'a lu jusqu'ici, mais 괜찮아, c'est d'abord pour moi que j'ai pris ces notes.

Mieux que le prolongement jusqu'à la mer du boulevard Saint-Michel

Posted by thbz on 29 September 2012 in French (Français)

Tout à l'heure j'ai remarqué sur une modeste rue du 13e arrondissement de Paris qui paraissait traverser des quartiers où elle n'avait en principe rien à faire. Impression fugitive, car rien n'apparaissait lorsque je passais en mode édition (Potlatch). Sans doute était-ce une erreur du moteur d'affichage.

Un peu plus tard j'ai découvert, sous JOSM, que cette rue s'étendait bien jusqu'aux limites de l'arrondissement, et même au-delà ; en fait, un zoom arrière montrait qu'elle s'incurvait peu à peu vers le nord, aussi bien à l'ouest qu'à l'est, jusqu'à former un cercle dont le sommet septentrional se trouvait... quelque part entre le Royaume-Uni et le Danemark.

Je ne maîtrise pas encore parfaitement OSM, mais il semble bien que ce soit une fausse manipulation de ma part qui a ainsi projeté dans les eaux glaciales de la mer du Nord une petite rue résidentielle. Erreur corrigée à présent. Du moins je l'espère...

Location: Butte aux Cailles, Quartier de la Maison-Blanche, Paris 13e Arrondissement, Paris, Île-de-France, 75013, France
Older Entries | Newer Entries