SC wants to create duplicate kerb and tactile tags

[BenWiederhake]

How to Reproduce
In the following scenario, screenshot from iD editor:

There is a single way (highway=footway, footway=crossing) from node A through B, C, D, to node E. Nodes B and D are already correctly tagged as kerbs, as the iD editor also shows in the screenshot above. And yet, SC shows quests for the nodes A and E, which are (correctly) not tagged with kerb= at all, because it's connected to a highway=footway, footway=sidewalk way, and why would there be a kerb.
This quest caused me to make some "Determine the heights of kerbs" and "Specify whether kerbs have tactile paving" changesets, because I didn't see that a kerb was already tagged, but at a slightly different node.

This changeset (not in the screenshot) resulted in the footpath seemingly having four kerbs, the two original (and correct) kerbs, plus the two kerbs I just (accidentally) tagged.

I can DM you the actual coordinates, but would prefer to not make them public.

Expected Behavior
The result should be that no such changeset is made. I can see two approaches for that:

• Algorithmic: Detecting this scenario algorithmically and not even prompting for this quest would be ideal, but I don't know how that could possibly be done. Maybe something like this? "Walk the path towards the crossing, and search for other nodes with kerb= set, and if any such exist, do not raise a quest marker"
• Visual: I think the simplest option would be to display all kerbs on the map, just like some other quests show nodes/ways that are relevant for orientation. This should enable the user to raise their eyebrows and notice that there already is a kerb here.
  (Maybe there's a way to tag that the kerb is elsewhere? I have a feeling that kerb=no would cause other issues?)

Versions affected
SC 51.1, but it should also affect the latest master branch.

[arrival-spring]

SC assumes that when highway=footway meets highway=crossing there is a kerb of some sort.

The question is whether this is correct mapping to have AB and DE mapped as crossing when they're behind the kerb

If it is correct then an algorithmic approach would be best, so this quest is not asked in such situations. It should be possible to check if the next node on a crossing way is already tagged as a kerb and exclude such situations.

[westnordost]

The question is whether this is correct mapping to have AB and DE mapped as crossing when they're behind the kerb

I'd say, yes.

Even though the part that branches off from the sidewalk is still on the sidewalk (area), semantically it is part of the crossing, it is not a sidewalk.

But I am pretty sure I had this case in mind when I implemented this quest, so maybe there is a bug or some edge case in which it doesn't work as expected.

[westnordost]

I did the research. In a nutshell: The data shown is incorrect.

See #1305 (comment) for a longer explanation.

Code is in https://github.com/streetcomplete/StreetComplete/blob/master/app/src/main/java/de/westnordost/streetcomplete/osm/kerb/KerbUtil.kt#L71

[westnordost]

Also scroll down for further illustrations, in particular the last comments should be helpful.

[BenWiederhake]

Hmm. Something is wrong though, in #1305 (comment) you explicitly mention the case I'm looking at, and yet SC asks for this quest:

So SC still wants to create duplicate kerb and tactile tags. I'm not sure why you closed this issue, because it still exists.

[matkoniecz]

Without knowing specific location it will be hard to answer (it is possible to recover location from geometry shown here or by looking at edit log, but it would take some time and presumably you are trying to keep it less accessible)

But you need to link some location or prepare synthetic test case as an unit test, otherwise investigating such case is not viable.

[matkoniecz]

Also, #1305 (comment) is a different case and what you pictures matches rather #1305 (comment)

is road having proper sidewalk tags tagged? what is shown by sidewalk overlay?

[BenWiederhake]

I'd be happy to share it with you, I just don't want it posted publicly on the internet, that's why I'm trying to keep it vague.

We could chat at #that-one-sc-issue about it, if you'd like?

[westnordost]

Hmm. Something is wrong though, in #1305 (comment) you explicitly mention the case I'm looking at

Uh no, your second picture still shows a situation alike

[BenWiederhake]

Yes, that's exactly the situation, I agree on that.

As far as I understand, SC should detect that the crossing way already contains kerb nodes, and thus not generate a quest.

Or, in case this is not possible, it should at least display existing other kerbs, so that the user can see that this was a false detection.

[westnordost]

As far as I understand, SC should detect that the crossing way already contains kerb nodes, and thus not generate a quest.

No, where did you read that?