[OSM-talk] Way with only one single node

[Toby Murray]

On Tue, Mar 20, 2012 at 4:12 PM, Richard Fairhurst <richard at systemed.net> wrote:
> Stephan Knauss wrote:
>> Some editors create these ways. The most prominent one is Potlatch.
>> The Ticket is open since two years. xybot trys to correct some of
>> the problems until a bugfix is made in potlatch.
>
> xybot does not help its cause by having the same ticket for Potlatch 1 and
> Potlatch 2 (which are entirely different codebases and have different
> components in trac). When someone identifies reliable steps to reproduce in
> P2, we try and fix it and indeed have done in some cases IIRC.

There definitely seems to be a bug along these lines having to do with
ways that are members of relations.

I did some poking around in my pgsnapshot database. My first attempt
nearly killed my machine. Note to self: change postgres settings so
that it can't use up all 12 GB of memory in a single query. But then I
wrote a more scaled down query and got a few hits.

The first 10 single node ways I found had all been touched last by P2
as recently as March 18th. 7 of them are still members of a relation.
The other 3 are all parts of changesets that touched a lot of
relations. Behold:
http://www.openstreetmap.org/browse/way/4379724
http://www.openstreetmap.org/browse/way/4995040
http://www.openstreetmap.org/browse/way/4692950
http://www.openstreetmap.org/browse/way/4728054
http://www.openstreetmap.org/browse/way/4925460
http://www.openstreetmap.org/browse/way/5823610
http://www.openstreetmap.org/browse/way/6025527
http://www.openstreetmap.org/browse/way/7219350
http://www.openstreetmap.org/browse/way/7769909
http://www.openstreetmap.org/browse/way/7219309

These are all the single node ways with a way ID of less than 8
million. Considering we have over 100 million ways, there are probably
a couple hundred of these.

Hope that helps. Or did someone already know about this?

Toby