Which tags are indexed?

[karussell]

Using this query to find railway stations I get only this node because the other railway station is only tagged with public_transport=station. But using a query with this tag won't give any results. Is the problem that public_transport is not indexed? Is this related to #462 ?

And is this also the reason why this toilet isn't found? Here is the photon query.

[lonvia]

Photon imports all data from the Nominatim database that have either a name or a housenumber, see

photon/src/main/java/de/komoot/photon/PhotonDoc.java

Lines 200 to 206 in ce3424b

	public boolean isUsefulForIndex() {
	if ("place".equals(tagKey) && "houses".equals(tagValue)) return false;
	if (linkedPlaceId > 0) return false;
	return houseNumber != null || !name.isEmpty();
	}

The Nominatim instance used to produce the pre-built Photon indexes uses a special style that drops all the OSM data that doesn't have a name.

Please note once again that Photon is mainly meant to be a forward geocoder. It is supposed to find places by their name and address. It is not meant to be a POI database. It would need to have a different design, if it was supposed to be a POI database. If you are using a query of the form reverse?lat=..&lon=..&osm_tag=...&limit=... then you are very close to using Photon out-of-spec. Please have a look at Overpass API.

[karussell]

Thanks for these details.

When I use osm_tag=historic it shows me only historic=memorial although e.g. this tower (building=yes):

https://www.openstreetmap.org/way/244850905

has a name and an address with housenumber and it does not show up for the following requests:

https://photon.komoot.io/reverse?lat=50.850814&lon=4.349807&radius=50&limit=10&osm_tag=historic

https://photon.komoot.io/reverse?lat=50.850814&lon=4.349807&radius=50&limit=10&osm_tag=historic:tower

Are there other requirements?

Do I need to add certain "extra-tags" in the command line?

[karussell]

Fetching the ID=244850905 in ElasticSearch directly gives two documents:

• one with "osm_key" : "man_made" and "osm_value" : "tower" and
• the other with "osm_key" : "tourism" and "osm_value" : "attraction".

Both with default.name = Tour Noire - Zwarte Toren.

Where is the code which tags trigger creating multiple documents? Or is this limitation coming from nominatim? (But it seems the needed tag is stored in nominatim.)

Could we improve the situation and index all tags like:

{
  "tags": [
    { "historic": "tower" },
    { "tourism": "attraction" },
    ...
  ]}

instead of creating multiple documents? Assuming the following mapping (or a dynamic mapping):

{
  "mappings": {
    "properties": {
      "tags": {
        "type": "nested",
        "properties": {
          "historic": { "type": "keyword" },
          "tourism": { "type": "keyword" },
          ...
        }}}}}

the query would look like this { "term": { "tags.historic": "tower" } }. Or would this explode the index?

[lonvia]

The duplication comes from Nominatim. We should get rid of it eventually but I'd prefer to reconsider this in a larger context of overhauling Nominatim's simplistic class/type classification.

[lonvia]

historic is an odd tag because it is sometimes a main tag, sometimes just an attribute. In Nominatim it is only used as classification when there are no other useful tags. Photon does not save the full tagging of places. It just gets a crude classification from Nominatim which happens to correspond to one of the OSM tags. You cannot use Photon to filter by arbitrary tags.