exit_to vs destination (USA)

Posted by k1wi on 11 September 2015 in English (English). Last updated on 19 September 2015.

The usage of destination tags in the United States on the OSM database is currently nearly equal to exit_to tags. Here is a graph of the evolution from January 2013 to September 2015.

You can read more about this on the wiki article:

EDIT: I originally used this Overpass Turbo query to create the graph: using the dev overpass server, as the normal one was issuing weird errors. However the dev server is not reliable for older data (as pointed out by mmd). I was able to use the normal server again and return correct older data by using simpler queries:,,

Comment from mvexel on 15 September 2015 at 22:32

Really interesting. Thanks for pulling those metrics!

For reference, here’s my diary entry on this topic. The timing happens to coincide with destination taking off as a tag in the U.S. :) Note that some of my colleagues at Telenav spent a bunch of time adding destination tags to exit ramps in the US following that post, affecting ~9000 exits.

I am wondering if we can design a good transition away from exit_to - perhaps a combination of a scripted update for the simple cases, and a MapRoulette challenge for the remaining ones?

Comment from pnorman on 15 September 2015 at 23:07

What would be interesting is counts of exit_to without destination and destination without exit_to.

But harder to query.

Comment from mvexel on 17 September 2015 at 15:15

I put together a MapRoulette challenge idea to fix these:


What do you think? Feedback on instructions?

Comment from mvexel on 17 September 2015 at 15:16

I should really start using the preview button. One more try:


Comment from k1wi on 17 September 2015 at 15:39

That’s a great idea!

Comment from k1wi on 17 September 2015 at 15:42

Just one thing. Maybe indicate that the motorway_link is a way:

…is contained in the destination tag on the motorway_link way.

Comment from mvexel on 17 September 2015 at 16:25

Put the idea in its own blog post for better visibility.

Comment from mmd on 19 September 2015 at 09:37

Thanks for an interesting Overpass use case.

Please note that the Overpass API dev instance you used to pull out your stats only dates back to March 2015 (unlike the official Overpass API instance, which has all the object versions back to September 2012).

This might skew your figures a bit: let’s assume a node has been created in Jan 2014 and last touched in Feb 2015. The dev instance would could recognize it for Feb 2015. as the Jan 2015 version is not known (it was no longer part of the March 2015 planet file I used to initially set up the database).

This restriction is documented on the following wiki page: link

Comment from k1wi on 19 September 2015 at 10:40

Thanks mmd for the heads up. I used the dev instance because I was getting errors using the usual server. For reference, here’s the issue on github: I will check the data again when the problem is fixed.

Comment from k1wi on 19 September 2015 at 17:57

I was able to use the normal overpass turbo instance by using simpler queries that don’t end in error. The old data in the graph should be correct now.

Comment from Skybunny on 12 December 2015 at 09:18

It’s been a few months now since this graph was rendered…is this trend continuing in the U.S.? (Leading to, has destination surpassed exit_to at this point?)

Comment from k1wi on 12 December 2015 at 11:31

Read my new diary entry with the updated data.

Login to leave a comment