Don't use obvious directions at ramp bifurcations

[oxidase]

Issue

Fixes #4895 as

"before-after" comparison in SF-area http://bl.ocks.org/d/1f6b51ce1cafff2463d913cd05d40f73

Tasklist

• CHANGELOG.md entry (How to write a changelog entry)
• update relevant Wiki pages
• add tests (see testing documentation
• review
• adjust for comments
• cherry pick to release branch

Requirements / Relations

Link any requirements here. Other pull requests this PR is based on?

[chaupow]

Codewise it looks good to me! 👍

I just wonder about this scenario:

Given the node map
    """
         /-----------c
    a---b------------d
    """

And the ways
    | nodes | highway       | name | destination |
    | ab    | motorway      |      |             |
    | bc    | motorway_link |      | City 17     |
    | bd    | motorway_link |      | Ravenholm   |

If ab is not a ramp:

1. Does this scenario exist? 🤔
2. Should it also be handled the same?

I am thinking about this scenario:
https://www.mapillary.com/app/?lat=49.86228055555556&lng=8.598833333333333&z=17&pKey=hb7H7YDWiRMYbnzio1dv6A&x=0.6825900002246823&y=0.5478995910837546&zoom=1&focus=photo

The signs show a straight vs a exit right, but there are 4 lanes, two go straight, two go to the right - so it is somehow a weird mix of a fork / exit?

Routing this way A5 -> A67:

http://map.project-osrm.org/?z=15&center=49.857215%2C8.602273&loc=49.871102%2C8.600750&loc=49.848803%2C8.596029&hl=en&alt=0

does not have an instruction (even though there is a name change)

Routing this way A5 -> A5:

http://map.project-osrm.org/?z=15&center=49.857215%2C8.602273&loc=49.871102%2C8.600750&loc=49.850325%2C8.608990&hl=en&alt=0

has an instruction (even though there is no name change).

Maybe this should another issue - and this is not osrm code related but more a guidance question?

[oxidase]

Does this scenario exist? thinking

No, i have added an intersection from a motorway to two ramps to the feature test.

Should it also be handled the same?

Yes, atm such intersections are handled in motorway_handler as forks (related issue #3106). PR makes both cases consistent: in motorway_handler and turn_handler will be used assignFork for intersections with motorway links.

A5 -> A5 or A67 case is a usual motorway intersection and handled in motorway_handler. An "obvious" instruction for A5 -> A67 is {type = NewName (1), direction_modifier = Straight (4)}, so it is suppressed later in the guidance post-processing.

[dgearhart]

@oxidase for a fork - typically, I would see ab as a motorway_link too

@chaupow For the A5 -> A67 case: I believe this is a bug, it should call out a Continue on A67 since it is a motorway to motorway transition and the user needs to know to continue straight onto A67.

@chaupow For the A5 -> A67 case: Since the edge is a motorway_link - it should be called out as a right exit so the user knows to take the exit even though the name remains the same. This scenario happens in the US too.

[dgearhart]

@oxidase 👍 LGTM thanks

[chaupow]

I'll be giving a 👍 here - @oxidase you might wanna figure out whether there should be an issue for the case I highlighted or not :)

[chaupow]

LGTM