Ignore too long shortcuts

[XioNoX]

See for example this screenshot. LTN says "This shortcut is 2.9x the length of the shortest route using all roads, not just this neighbourhood".

Would it be possible to add an option to ignore those ? Maybe customizable (eg. hide all shortcuts longer than X times longer than the "regular" route) ?

[dabreegster]

Going further... what if there's also an option to exclude shortcuts if they enter and exit on the same logical "road" (going by name)? Originally I tried something like that, but it's an imperfect heuristic. It's easier now in v2 to expose it as a setting and just let the user try things out, to improve the signal/noise ratio.

[XioNoX]

In the case of a bendy street (or L shaped) a shortcut can still start and end on the same street. That's why I think the shortcut length is what should matter the most (like how it is now).
As there are many possible use-cases, it looks quite complex to have a single solution to improve the signal/noise ratio. I guess the difficulty is to not add a 1000 control knobs in the settings :)
For example I could imagine some use-cases (eg. when the main roads are really congested) where "long" shortcuts could still be problematic and needed to be displayed.

[dabreegster]

Few more questions, then I'll implement this...

• Should this length filter apply just to the "browse shortcuts to understand them" bit, or also to the overall heatmap? I think both...
• Once speed limits are working and routing takes that into account, should we consider both ratios of distance and expected time?

[XioNoX]

Should this length filter apply just to the "browse shortcuts to understand them" bit, or also to the overall heatmap? I think both...

I'd say both too.

Once speed limits are working and routing takes that into account, should we consider both ratios of distance and expected time?

I think it would be nice to have. I'd need to test/compare both to really see if one is better than the other for my usecase, but I expect the result to be about the same.
Except in some places where the signed speed is not respected, reminds me of #86

[XioNoX]

New neighborhood, new challenges !

Here the roundabout is used in the wrong direction (in France it's counter-clockwise). But that's a detail.

The main point here is that even though the A to B shortcut (through https://www.openstreetmap.org/way/128909043 ) is indeed shorter than the main roads, it's a pointless shortcut as it on a one way street going back to A.

I have no suggestion/idea though on how to programmatically ignore those shortcuts, but maybe it could be useful to be able to manually "flag" a shortcut as "ignored" ?

A bit similar there :

Cars that come from the right of "Boulevard de l'Europe" can't make a left turn there, so with a one way modal filter on https://www.openstreetmap.org/way/25439590 there are no real shortcuts anymore.

[dabreegster]

Here the roundabout is used in the wrong direction (in France it's counter-clockwise). But that's a detail.

https://www.openstreetmap.org/way/789313242#map=19/48.409491/-4.483283
Ah good catch, I've forgotten that oneways are implied sometimes. https://wiki.openstreetmap.org/wiki/Key:oneway#Implied_oneway_restriction Need to handle junction=roundabout

it's a pointless shortcut as it on a one way street going back to A.

Ahhh... an amusing case. I'll have to think about how to handle this kind of thing. Thanks!

[dabreegster]

it's a pointless shortcut as it on a one way street going back to A

We can determine this when initially picking the pairs of entry and exit intersections. This has nothing to do with the shortcut through the neighbourhood at all. We could check what's allowed at the exit intersection. From that exit, if the only reachable point is the entry intersection, then it's this simple redundant case. If there's anything in between, then maybe it's still a shortcut somebody would take, but it could be weighted less, based on how "useful" exiting there is.

[TFCx]

There are slightly more complex cases like in this example (there are twice the same problem on top-left and bottom-left) :

https://www.openstreetmap.org/#map=17/43.600230/3.902979

Both of those shortcuts aren't because they would they can't be use to reach anything you wouldn't already by able to reach before taking them.

We could check what's allowed at the exit intersection. From that exit, if the only reachable point is the entry intersection, then it's this simple redundant case. If there's anything in between, then maybe it's still a shortcut somebody would take, but it could be weighted less, based on how "useful" exiting there is.

So I would suggest that you should not only take the first-downstream node of the exit, you should also take the first upstream node of the entry.