Stumbled on a junction in Keepright which has evident turn restrictions missing. Similar encountered last week. The roads arriving at a T junction, classic Italian model, are one ways for right, left, on and off are uncut miles long and in the previous case so long to extend to the next village, puzzling over where that only_straight_on ‘to’ was coming from.
The long roads from this junction before planting the first turn restriction was without cuts, the bottom way northbound a single mapped way extending past the junction to the eastbound end.
The question is: How long does one reasonably let a road run in a from or to position before one applies the scissors? Some of these ways have side driveways/service ways used as a reason to set the cut there, some just absolutely nothing the question also arising when navigators decide it is enough with the limitation.
In instances, when the traffic sign says only a straight on is allowed, a no left is coded to avoid the arbitrary cutting of such ways. For navigators is has the same effect, not entering into ponderings on navigators not being fast enough to catch up on the driver having chosen the correct exit way i.e. too late to compute that the driver took the straight ahead exit way for the exampled junction southbound but really needed to go eastbound.
(As an aside when adding destination:forward/backward, the same happens, the destination(s) continue to remain in the navigator display till that next cut happens)
Technically, the restriction only applies going from one edge of the routing graph, via a vertex or series of edges, to another edge. Any intersection with another navigable path will inherently result in two edges in the routing graph, even if OSM represents those two edges as a single way. This goes back to the definition of a graph in mathematics.
Even so, iD automatically splits a way at the nearest intersection when you use it as a to or from member in a turn restriction relation. This isn’t airtight, but it mitigates issues such as discontiguous to ways that often arise when mappers are changing something non-routing-related along a very long way.
Now that is an excellent piece to know for the dismissing of the ‘feel too long’ sensation. Hoping it will do the same thing to the destination tags on the same way. Found more than a few that have shot past a junction where a turn has to be taken.
JOSM behaves no different than ID, cut a road that has a turn restriction associated and the ‘to’ or ‘from’ part stays with the section which is attached to the ‘via’ node/way. Not seen these go broken in either editor.
destination or any other tag on a way will behave differently. All the edges associated with the way will have the same destination. A destination_sign relation would behave more like a restriction relation, but few if any routers support destination_sign relations.
