Redundancy of oneway=yes for junction=roundabout?

In the OpenStreetMap changeset (Changeset: 51763240 | OpenStreetMap), I noticed that the removal of oneway=yes for junction=roundabout was reversed.

Could someone provide some context for this change? Was it due to concerns about mass removal without proper discussion, or were there other reasons behind it?

I’m curious because according to the wiki (Tag:junction=roundabout - OpenStreetMap Wiki), using oneway=yes with junction=roundabout is considered redundant. Can anyone confirm if the wiki is accurate in this regard?

Two weeks ago, I actually asked the person involved in this change about it, but I haven’t received a response from them yet. So, I decided to ask here.

1 Like

It’s right there in the changeset comment:

Revert undiscussed mass removal of ‘oneway=yes’ on roundabouts performed by user Grauer after seeing this flagged on Osmose; after complaints from the Dutch community the user asked DWG to help with revert.

For the avoidance of doubt, the changeset that you were referring to was 6 years ago. It’s pretty common to back out an undiscussed mechanical edit in this way in cases like this. Data that is “redundant” isn’t an error; it’s just unnecessary. It can be seen now that there is significant usage of oneway=yes with junction=roundabout, but it is not ubiquitous.

There is actually fairly significant usage of way["oneway"="no"]["junction"="roundabout"]. I suspect that quite a few of those are in error, but a mechanical edit that removes oneway=yes would make those errors significantly harder to fix, and hence a bad idea.


OSM knows that junction=roundabout and junction=circular are oneway and knows which way to point the oneway depending on the country where mapped. In the editor the little triangles on the way would point anti-clockwise e.g in Continental Europe, over in e.g. Great Britain and Ireland they’d point clockwise. Thusly oneway=yes is superfluous and vrs QA progs warn for it, think JOSM even tells you before saving.

Some add oneway=yes or forget to remove it when a junction gets converted, saw it even few days ago.

1 Like

Yes, it’s pretty old, but I just noticed it now. I did read the changeset comment, but I wasn’t sure if the issue was that it wasn’t discussed or if it was something specific to the Netherlands, since the complaint came from the Dutch community. Thanks for clearing that up.

I have converted many regular roundabouts to turbo roundabouts, removing lots of oneway=yes tags, because junction=roundabout on the ways by definition implies oneway=yes. At the same time, I tagged lanes=1 on all these same pieces of roadway.
In hindsight, lanes=1 is completely redundant because all the ways on a turbo roundabout are single lane ways. Only a few superturbo roundabouts sport multi-lane ways, so I could easily have tagged only the exceptions.

junction=roundabout, to me, is a slighty different matter. It is supposed to mark a junction as an object, at first mainly to assign a name and applied to the junction node, later applied to the closed way circle replacing the node (junction=roundabout, junction=circular), which still represents the whole junction.
Next stage, the circle was cut up into pieces, but together these pieces still formed the junction circle.
With turbo roundabouts however, al the lanes on the roundabout are mapped a separate ways, turning the roundabout into a system of interconnecting ways and junctions.
And now I’m not convinced the junction=roundabout tag is appropriate for all these separate pieces of way. Some roundabouts now have 20 ways with the tag, and now it seems better (simpler, more direct, not as impication but explicit) to simply tag oneway=yes on them, and reserve junction=roundabout to an element that actually represents the whole junction.

In practice this means I understand that mappers would like to preserve the explicit oneway=yes tag, and do not want it mass removed, and I will no longer remove them.

1 Like

When you’re doing that you’re adding a lot more extra information, and looking at each roundabout in detail to see how it is best mapped.

If as part of that process you reckon a way is better without a oneway=yes tag (and the local community agrees), I’d have thought that it would be perfectly fine to remove it.

What would be a problem would be mechanically removing tags from a whole bunch of roundabouts that you haven’t even looked at, because you’ll miss the edge cases.


It is, but in hindsight I regret having removed the tag, especially since I did explicitly tag lanes=1 (replacing lanes=2) which is completely redundant (but the local community still agreed to tag it).

OSM “knows” nothing, the direction is just defaulting to the mapped way-direction.

If you want to split hairs, fine, it goes ‘default’ to the mapped direction only because of the junction tag i.e. a rule. I’ve seen oneway=-1 too BTW and then JOSM protesting that the mapped direction and oneway direction are not the same and then when you hit the Reverse key you get that odd -1 notation with before a complicated screen what you want to change.

I am not aware of any renderers which consider the country when rendering oneway arrows. This doesn’t make any sense, as renderers will handle each way in isolation, and the way may not form a closed loop so there is notion of clockwise or anticlockwise, so all you have is a line going from A to B. If it has a tag that indicates it’s oneway, you render arrows in the direction of the line. If it’s a tag that indicates it’s oneway reversed, you render arrows in the opposite direction or flip the line.

junction=circular is not generally considered a oneway tag. None of the featured layers on do, and I wasn’t able to find any other map layers that do. If you want a circular junction to be considered oneway, put a oneway tag on it.

Have to admit I’m wrong in my statement after revisiting a roundabout and a circular which I had converted from roundabout because round was the only true part about it, no right of way, traffic signals and stop spots on the ring, in fact a section was bus only, not part of the main traffic stream.

  1. roundabout without oneway tags, the arrows make it explicit how it flows.


  1. The noted circular converted from roundabout and reading back now remember to have been prompted to add the oneway=yes tags (as the wiki says is required on circulars)


Will test the reverse direction of a roundabout to see what happens on (which I did not know but redirects to the standard carto view at site). Starting to think it will be rendered in the wrong direction, then add a oneway=-1 to force the hand which will give an alert in JOSM during validation, can’t remember what ID does. There’s a new one not yet mapped though at least half year in operation with dense Strava heatmap data already so good testing ground.


I think it should. junction=circular is not just a shape indication, it is used for a roundabout, just the priority is (partly) different.
From the wiki: “Where traffic on a circular junction itself does have complete right of way, this is a roundabout.” "A road junction where the traffic goes around a non-traversable island but does not always have right of way. "

If rotaries are not considered oneway by default, that strengthens the case for tagging oneway=yes on junction=roundabout as well. Because a roundabout can be made into a rotary by placing or removing a single priority sign, which I know actually happens in Nederland.
(Here, roundabouts are not priority by definition; priority has to be explicitly signed on all approaches).

As promised I tested on a new roundabout, actually this is the second new discovered today but mapped first as very very simple, just 3 bus routes traversing this one. The used tag for the test is highway=tertiary + junction=circular NO oneway=yes tag though the wiki says to use on this junction type. The result is below. Note the little black triangles pointing in the right direction. Saved and no protests from the JOSM validator,

And that’s why I stated what I stated. Yes, mapped in the right direction, NO oneway tag on the circular and treated as a oneway by the editor.

Yes, mapped in the right direction, NO oneway tag on the circular and treated as a oneway by the editor.

not sure if this is a bug, but if you expected otherwise, try reporting it to the josm developers

Maybe it should - but that’s not how it’s treated. For myself, I’d rather see explicit oneway in all cases.

Verification is everything, I had not reversed the temp tagging of circular on the new roundabout and low and behold, ID editor does the same thing

The QA site i shall not mention has not flagged this yet either 24 hours later, but I’m going to leave it up a little longer as maybe they don’t run this particular ‘missing oneway tag check’ everyday.

Could it be this one gets attention when there’s actually a stop spot or a traffic light on the ring, which is what I remember from the Breccia circular.