De-duplicating OSM admins

[nilsnolde]

this was recently raised in valhalla/web-app#406 and followed up with manual de-duplication at valhalla/web-app#407.

consider just a "Paris" query:

there's 3 duplicate entries which have the exact same geometries but different admin_levels:

• 8: https://www.openstreetmap.org/relation/7444
• 7: https://www.openstreetmap.org/relation/1641193
• 6: https://www.openstreetmap.org/relation/71525#map=12/48.8589/2.3469&layers=V

I know OSM admins are messy, but I wonder if all 3 are really legit relations or if there's sense in de-duplication at/post geocoder data ingestion?

sorry for the OSM admin & geocoding noob question;)

[mtmail]

Was this written by an AI?

[mtmail]

I'm going to hide this answer. It's not low quality but it's not clear somebody (a human) has reviewed it. Anybody can ask a LLM and it will produce pages of text with high confidence. It's also not clear who "I" in "I'm not an OSM admin-boundary regular" is.

[1ec5]

I know OSM admins are messy, but I wonder if all 3 are really legit relations or if there's sense in de-duplication at/post geocoder data ingestion?

I wonder if the French community felt the need to duplicate the boundary because Paris is a peer of administrative areas at multiple levels. In 2024, the U.S. community decided against modeling consolidated city-cities as redundant boundaries. We handled every case individually based on the exact nature of the consolidation. For example:

• The City and County of San Francisco is modeled as a single boundary relation, because the county was abolished when the city took over and took on a dual identity as both a city and a county. The admin_level=6 tag reflects the greater status as a county, at the cost of being able to find every California city by admin_level=8.
• Orleans Parish remains coextensive with New Orleans because they have distinct names and are legally distinct, as reflected on signage.
• We kept the District of Columbia but discovered that Washington, the capital, isn’t even an administrative boundary anymore, other than an obscure legal remnant that we mapped in OpenHistoricalMap instead.

I’m unfamiliar with the administrative structure in France, but maybe the French community would be interested in taking a closer look. I agree that OSM can be messy, but for something as famous as Paris, I’m sure you’d get some opinions by contacting the French community on their forum.