Questions about the nomination and use of data sources

[xdxTao]

Hello, I have built a service on my own server, and the imported data comes from https://download.geofabrik.de/asia/china.html

After testing, I found that the data is incomplete. The query results of the service I built are different from those from the Nominatim API. The results of Nominatim are more accurate, so I want to know where Nominatim's data source is downloaded from. If possible, I hope to tell me the download address of China's data directly, thank you very much

[mtmail]

Both use the same OpenStreetMap data. Can you give examples where your Nominatim installation returns different results than nominatim.openstreetmap.org?

[xdxTao]

This is the request and result of the self-built service

{ "type": "FeatureCollection", "geocoding": { "version": "0.1.0", "attribution": "Data © OpenStreetMap contributors, ODbL 1.0. https://osm.org/copyright", "licence": "ODbL", "query": "拉萨市" }, "features": [ { "type": "Feature", "properties": { "geocoding": { "place_id": 126888, "osm_type": "relation", "osm_id": 2745046, "osm_key": "boundary", "osm_value": "administrative", "type": "city", "label": "拉萨市, 850000, 中国", "name": "拉萨市", "country": "中国", "postcode": "850000", "admin": { "level5": "拉萨市" } } }, "geometry": { "type": "Point", "coordinates": [ 91.1173015, 29.6542054 ] } } ] }

This is the request and result of Nominatim

{ "type": "FeatureCollection", "geocoding": { "version": "0.1.0", "attribution": "Data © OpenStreetMap contributors, ODbL 1.0. http://osm.org/copyright", "licence": "ODbL", "query": "拉萨市" }, "features": [ { "type": "Feature", "properties": { "geocoding": { "place_id": 220815117, "osm_type": "relation", "osm_id": 2745046, "osm_key": "boundary", "osm_value": "administrative", "type": "city", "label": "拉萨市, 西藏自治区, 850000, 中国", "name": "拉萨市", "postcode": "850000", "state": "西藏自治区", "country": "中国", "country_code": "cn", "admin": { "level5": "拉萨市", "level4": "西藏自治区" } } }, "geometry": { "type": "Point", "coordinates": [ 91.1173015, 29.6542054 ] } } ] }

The self-built service is missing a level 4 layer. All the queries under this are indeed missing this level. I use "osm_id": 2745046 to query my self-built database and there is no data, so I suspect that the data downloaded from https://download.geofabrik.de/asia/china.html is incomplete.

[xdxTao]

Sorry, forgot to post the request link above

http://10.0.1.18:8090/search/search?q=拉萨市&format=geocodejson&addressdetails=1&accept-language=zh
https://nominatim.openstreetmap.org/search?q=拉萨市&format=geocodejson&addressdetails=1&accept-language=zh

Looking forward to your reply

[mtmail]

Indeed the state 'Tibet' is incomplete in the China file on download.geofabrik.de. It's not related to Nominatim software.

In OpenStreetMap data schema a relation (here https://www.openstreetmap.org/relation/153292) is made up of multiple ways and each way is made up of many nodes. https://wiki.openstreetmap.org/wiki/Elements

The China file you downloaded contains the relation, but it's missing 17 ways. So Nominatim didn't import the relation.

wget https://download.geofabrik.de/asia/china-latest.osm.pbf

# https://osmcode.org/osmium-tool/
# https://www.openstreetmap.org/relation/913012
osmium getid --add-referenced --output jiangsu.osm china-latest.osm.pbf r913012
# https://www.openstreetmap.org/relation/153292
osmium getid --add-referenced --output tibet.osm china-latest.osm.pbf r153292

osmium check-refs --check-relations jiangsu.osm
There are 248152 nodes, 9297 ways, and 3089 relations in this file.
Nodes     in ways      missing: 0
Nodes     in relations missing: 0
Ways      in relations missing: 0
Relations in relations missing: 0

osmium check-refs --check-relations tibet.osm
There are 98570 nodes, 2463 ways, and 778 relations in this file.
Nodes     in ways      missing: 0
Nodes     in relations missing: 0
Ways      in relations missing: 17 <===
Relations in relations missing: 0

You can contact Geofabrik asking to improve the extraction process. Basically once per night they take the global OpenStreetMap data (called the planet https://planet.openstreetmap.org/) and cut out countries into individual files. Sometimes that's not precise enough or borders changed. It's possible the borders changed only a couple of metres.

You could download all of Asia and extract the needed regions yourself.

You can manually add the relation with all ways to your Nominatim installation :

cd $nominatim-project-directory
nominatim add-data --relation 153292
nominatim index

If that didn't work try to reindex everything in that state.

nominatim refresh --data-area r153292
nominatim index