[OSM-dev] Tiling the Planet and Missing Nodes
D Tucny
d at tucny.com
Sat Mar 28 14:50:36 GMT 2009
2009/3/28 Stefan de Konink <stefan at konink.de>
>
> > From my last run, way id 4043882 references node 365476284 that is not
> > present in the dump. This is with planet version 090325. Grep reveals
> > that 365476284 is only present in an "nd" tag as part of the ref=
> > statement. :-(
> >
> > My question is: what is the right interpretation of a missing node?
> > Should I simply pretend the node reference does not exist (e.g. delete
> > the vertex from the way)?
>
> The idea with that situation is for ways:
> - Download the way; and just reupload it, then it will be solved in the
> next run (because the way download hides invisible nodes)
>
This only fixes the problem of ways with nodes that shouldn't be there which
I don't believe is always the case... In most cases, nodes that should be in
a way have somehow become deleted, but, not from the way itself... The
typically accepted 'fix' for these is to modify the way in Potlatch, which,
as a byproduct of it's method of rewriting all related objects, restores the
nodes to their prior undeleted state... To blindly rewrite these damaged
ways removing the problem nodes would hide the problem but could leave the
way itself inaccurate... These ways should be manually inspected to work out
what the actual problem is and to repair it, or at the very least, a lot
more logic needs to be used in any automated process...
>
> For relations:
> - Download the relation; remove all nodes/ways/relations that you know
> that don't exist. Reupload it; will you get a 200; job well done ;)
>
Again, this isn't so simple... doing this would remove some errors (deleted
objects in a relation) and introduce new ones (relations that are missing
critical components or otherwise don't make sense)...
d
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/dev/attachments/20090328/37507db5/attachment.html>
More information about the dev
mailing list