[OSM-talk] QA tool for finding nameless highways that are armchair-fixable

Lukas Toggenburger ltoggenburger at swissonline.ch
Mon Nov 28 20:31:28 UTC 2022


I am a bit unsure on how to proceed from here...

Maybe let's start with some technical facts:

> IMHO a highway=service, track etc typically do not have names.
> Hopefully you account for that.

> And i am not shure the assumption that other road classes
> always have names holds up.

In the current incarnation, in order to show up, a nameless way must be connected to two other ways that have exactly the same tags and also all three ways must be of the same higway=*.

> Higher class roads may have ref= but not name=*

Currently I am only considering interpolating missing name=*, but looking at gaps of ref=* might be a good idea for a future version.

> I use Geofabrik's excellent OSM Inspector

Yeah, nice! 😎 I wrote the "Addresses" view of it.
(code is at https://github.com/ltog/osmi-addresses )

What I do not like so much about it: it shows you a lot of errors, but there is not much you can do about it after you fixed your reachable neighbourhood. It's the same for the "Highways" view (options "Highway name/ref") that looks at the same issue as the "ohni" software: https://tools.geofabrik.de/osmi/?view=highways&lon=8.00885&lat=47.99930&zoom=9&baselayer=Geofabrik%20Standard&overlays=name_missing_major%2Cname_missing_minor

I specifically wanted the software to give the users the possibility to find spots where they can not only watch, but contribute with a low number of false-positives. I was hoping for feedback especially on that aspect (trade-off between false-positives and high number of results).

> if you have a road in 3 segments A B C and A+C have the same name,
> then not only does it seem reasonable to me to add the name on B
> but also the reply "do a survey" is a dogmatic answer: the ground
> does not contain a sign every time osm cuts a way because
> of a change in the number of strips for example.
> So by survey, you will be reduced to deducing that a segment between
> 2 others with the same name, probably also has the same name

That is also my point of view.

Best regards

Lukas



More information about the talk mailing list