Suggest adding an entrance=* node to fix 'disconnected highway' errors

[k-yle]

Closes #6341

If an isolated path or road connects to a building outline, the validator has a new autofix suggestion, which adds entrance=yes or amenity=parking_entrance to the node:

[tordans]

Thanks for this, I think it is a good addition.

Does https://github.com/openstreetmap/iD/pull/10729/files#diff-aa67428c964bd0d0ac6939c0545ffa404cbc3237b1dffec3b2a22491fbac8988R232-R234 mean that this suggestion will either suggest entrance=yes for path like highways and otherwise amenity=parking_entrance?

I wonder if this is the best second option in most cases.

In Berlin, we commonly use barrier=gate for those hw=service ways that connect to a building

• https://www.openstreetmap.org/node/6320148742
• https://www.openstreetmap.org/node/6896812829

Parking entrance is also present sometimes, though https://www.openstreetmap.org/node/8003765396

[k-yle]

Does https://github.com/openstreetmap/iD/pull/10729/files#diff-aa67428c964bd0d0ac6939c0545ffa404cbc3237b1dffec3b2a22491fbac8988R232-R234 mean that this suggestion will either suggest entrance=yes for path like highways and otherwise amenity=parking_entrance?

Yes. Maybe this autofix should only apply to path-like ways? I suspect that isolated vehiclular ways are quite rare, compared to isolated pathways

[tordans]

Yes. Maybe this autofix should only apply to path-like ways?

Maybe to make this PR easier to merge because I consider this case very clear.

I suspect that isolated vehiclular ways are quite rare, compared to isolated pathways

Actually, we have many such cases in Berlin where a barrier=gate (and maybe barrier=entrance – no gate, just no one added the building passage way, yet) would be welcome additions. See those lines in https://www.openstreetmap.org/#map=19/52.471610/13.444500 for example.

FYI: Those service ways are important data that we add in Berlin to have data that we can use during parking pane processing (https://parkraum.osm-verkehrswende.org/). But those tags at the building are left out often because its already a lot of clicks to add the driveway.

[k-yle]

I've removed the amenity=parking_entrance logic, now this PR only considers path-like ways

[tyrasd]

Hm… I'm not convinced this is a good idea for the disconnected_way validator: Even after adding the entrance information, the way itself is still not connected to the rest of the road/path network. As far as I can see, the warning is suppressed in the case of the presence of an entrance node because there are some edge cases of false positives (like a path connecting two buildings, see #3906) that could not be detected in a better way. Maybe it also helps with false positives of mapped indoor corridors, and similar situations (although I would also argue that even in those cases there should be a connection to the outside routable network).

So, wouldn't it be better to have this as a dedicated validation rule that checks for missing entrances on highway<->building intersections? As suggested in #6341 (comment) by @1ec5

[1ec5]

Either seems reasonable to me, but now that you mention it, the suggestion is incongruent with the warning message, which implies that a road or path connection is required. I guess the explanatory text would need to mention building entrances as an exception.

Another case to consider is a footpath that you can only access from within a building, such as in a courtyard or backyard.