Mass remove `gnis:created` and similar tags? [final version presented]

Writing things down is fun and good. Thanks to everyone who writes things down.

1 Like

:+1:

Same.

1 Like

Thanks for cleaning this stuff up! I saw the changesets roll through my area, and I was very happy thinking about all those tags going away.

2 Likes

gnis:edited is similar to gnis:created: it says when the record was last modified in GNIS prior to the import into OSM. The only potential use of this tag would be to compare it to edited dates in the current GNIS database to see which ones need to be updated in OSM. But as with tiger:reviewed, someone could’ve already updated the feature in OSM without touching that tag. Any “three-way merge” of a given feature could instead compare version 1’s tags and coordinates with the current version and current GNIS. I think that would be more reliable than merely comparing timestamps anyways.

1 Like

@Kai_Johnson is working on various tooling to try and sort out “this got updated in GNIS but not yet on OSM” for various classes of things.

Note that this proposal remains active - not all such tags were removed. See for example Node: ‪Grace United Methodist Church‬ (‪357813137‬) | OpenStreetMap and gnis:created | Keys | OpenStreetMap Taginfo

1 Like

Anyone finding this thread later, you can know if it’s completed by looking at the counts of various tags here: Search results | OpenStreetMap Taginfo

Maybe we will succeed this time. Maybe we will just dent it some more. Either way, it is appreciated.

I created Key:gnis:County - OpenStreetMap Wiki and plan to create similar for other keys.

Even after they will be gone such documentation (rather than redirect to very large article) may be useful to people looking at history of objects. And to people who will notice bot edit removing this tags.

2 Likes

Is anyone interested in list of suspicious combinations like gnis:feature_type = Mine on power=tower ?

See

  • gnis:feature_type = Mine
  • landuse = quarry
  • material = wood
  • name = Rattlesnake Park Mine
  • power = tower

on Node: ‪Rattlesnake Park Mine‬ (‪2308446206‬) | OpenStreetMap

Or should gnis:feature_type be mass removed from all objects, not only where it clearly duplicates existing main keys?

1 Like

Also: is it welcome to remove gnis:name where it matches already present name or official_name tag values?

Should we mass remove also gnis:reviewed - see gnis:reviewed | Keys | OpenStreetMap Taginfo

I expect that gnis:edited should be removed.

NHD:GNIS_Name probably should be handled with other NHD:* tags.

(this above is based on writing draft code of cleanup bot and test dry run)

I generally think that with these import tags, it’s usually just the unique IDs that we care about, which might allow us to link back to original data sets. So I’m all for purging things that add no value.

2 Likes

This is obviously a case of someone incorrectly connecting a power line to an unrelated GNIS POI. The gnis:feature_type mismatch does help us find the mistake, but so could the combination of landuse=quarry with power=tower.

2 Likes

I think it’s a waste of effort to create documentation for content we’re actively removing. Also, it all says “gnis” so anyone who wonders what it’s all about can just look up “gnis” on the OSM Wiki.

I’m fine with wasted effort, as long as it’s just somebody else’s voluntary wasted effort :smiley:

3 Likes

There was another comment on US Slack how this tag is pointless and I remember similar comments in previous, now gone, Slack comments - so I plan to just mass remove all instances of it.

Comments about other potential changes are appreciated - see Mass remove `gnis:created` and similar tags? - #30 by Mateusz_Konieczny

I hesitate especially about gnis:reviewed and NHD:GNIS_Name and gnis:name

1 Like

There’s also the curiosity of ~8000 objects with gnis:feature_id but no name

I have done absolutely no investigation of this bucket of work.

I think this cleanup is a great idea.

In that same vein though, there’s about 1400 instances of objects with a gnis:* tag (other than fcode or ftype since most of those don’t have names anyway) that are missing gnis:feature_id. Meaning, they probably had a feature_id but it got deleted? By mistake? On purpose?

“nodes”: “475”,
“ways”: “845”,
“relations”: “82”,
“total”: “1402”

7MB, takes like 10 seconds to run as long as I don’t limit it by area for some reason, idk:

Example: way/426905026

Another great class of cleanup. The example way seems to have lost the feature_id when it was redrawn as an area (they were all imported as nodes). I have updated it after crawling the GNIS national file and selecting the one with the correct lat/lon.

I guess that these could be skipped from gnis: removal… Or maybe keep just single gnis: tag on them? Or open notes?

Maybe you were thinking of a similar NHD key?