Junction restrictions on unnecessary and unreachable pedestrian crossings

[Elesbaan70]

This resolves #1302.

• When a segment has no pedestrian lanes and is not part of the shortest path between two other segments with pedestrian lanes on the same node, crossings for that segment are disallowed by default.
• When none of the segments on a node have pedestrian lanes, crossings are disallowed and are not configurable.

The "shortest path" is determined by the number of segments a pedestrian must cross, without regard for geometry; so it may not reflect true distances.

graph TD

    Allowed[Allowed=true<br/>Configurable=true]
    Configurable[Allowed=false<br/>Configurable=true]
    Disallowed[Allowed=false<br/>Configurable=false]

    HasPedestrians{Does this segment<br/>have pedestrian lanes?}
    HasPedestrians -->|Yes| Allowed
    HasPedestrians -->|No| OtherSegments

    OtherSegments{How many other segments<br/>have pedestrian lanes?}
    OtherSegments -->|More than 1| ShortestPath
    OtherSegments -->|1| Configurable
    OtherSegments -->|0| Disallowed

    ShortestPath{Does any shortest path go<br/>through this segment?}
    ShortestPath -->|Yes| Allowed
    ShortestPath -->|No| Configurable

[kianzarrin]

see comments above

[kianzarrin]

I like the idea of checking distance from both sides. it did not occur to me.

[kianzarrin]

Whats happening here?

I see zebra crossing flag at that segment end.

[Elesbaan70]

Whats happening here?

I missed that! The existing code is hardcoded to bypass the computation when it's not configurable. (Great example of an unsafe optimization.) Node Controller wasn't affected because of the way the patch works.

I'm on mobile, but I'll fix it when I'm back at my computer.

[Elesbaan70]

@kianzarrin Have your concerns been addressed?

@krzychu124 Have you had a chance to look at this?

[kianzarrin]

at bend nodes pedestrian crossing is configurable. Is this intentional?

[Elesbaan70]

This is not a change. The code was simply moved from another location in this commit: Elesbaan70@8e732e6

[kianzarrin]

@Elesbaan70 are bend nodes within the scope of this PR?

[Elesbaan70]

I would say no. That code was only touched to address an issue with the way default values were handled when not configurable. This PR is about segments without pedestrian lanes, and not really related to the type of node. And this is how the code worked before.

In my opinion, it should be configurable anyway.

[kianzarrin]

works good in game.

[krzychu124]

@krzychu124 Have you had a chance to look at this?

hehe, not yet but I'll try to find some time today or tomorrow. Changes make sense and should work as described, I just need to test it in game.
Unfortunately I already noticed (current Test and Stable) that sometimes pedestrians don't care about disabled crossing at highway segments or may even stuck if they finally notice it so there must be a bug hidden somewhere in the pathfinding code 😕

[Elesbaan70]

Unfortunately I already noticed (current Test and Stable) that sometimes pedestrians don't care about disabled crossing at highway segments or may even stuck if they finally notice it so there must be a bug hidden somewhere in the pathfinding code 😕

The original idea that led to this change was just disabling all crossings when no segment on the node has pedestrian lanes at all. Should I put some conditional compiling in place to limit this to just that one change, until the issue you're describing is resolved?

[kianzarrin]

Unfortunately I already noticed (current Test and Stable) that sometimes pedestrians don't care about disabled crossing at highway segments or may even stuck if they finally notice it so there must be a bug hidden somewhere in the pathfinding code

is it with NCR?

[krzychu124]

No, I posted screenshot yesterday on our discord. I've checked networks with Network detective and highway road doesn't have pedestrian lanes, only empty lanes (who knows what for). It's might be just some kind of asset issues but I didn't find anything obvious. I'll try to find the source of that problem at the weekend.

Should I put some conditional compiling in place to limit this to just that one change

I didn't test this PR yet so it's not related (not caused by your changes). Like I said I'll investigate what is going on and try to fix before we release next Test version

[Elesbaan70]

My concern was that if disabling crossings is what triggers this problem, then until it's resolved, we might want to avoid introducing a disabled default state on nodes where it might matter.

[krzychu124]

Yes, I understand, but it's still configurable (assuming there is at least one segment with pedestrian lanes) so there is IMO big chance that user will try to configure it the same way - will spot the same problem before I fix it, plus the thing it is already broken in some cases 😉