[Talk-us] MassGIS Building Import - process

Paul Norman penorman at mac.com
Sat Dec 15 21:45:19 GMT 2012


> From: andrzej zaborowski [mailto:balrogg at gmail.com]
> Subject: Re: [Talk-us] MassGIS Building Import - process
> 
> On 15 December 2012 20:09, Jeff Meyer <jeff at gwhat.org> wrote:
> > Paul - I've added a few comments and questions about changeset size
> > and revert policies on the Import Guidelines & Plan Outline wiki
> pages.
> >
> > Are there any recommended changeset size limits and/or revert plan
> > practices?
> 
> One good practice is not to revert data that is not known to be wrong.
>  If a big changeset fails halfway through it's possible to fix the
> remaining part to use the nodes that have been uploaded and continue,
> rather than delete the 1000s of nodes just to create new ones in the
> same places.
> 
> You can probably now do that in JOSM by downloading the changeset
> containing the orphaned nodes, opening in JOSM together with the data
> being uploaded and telling the validator to "fix all" duplicate nodes.

This hasn't come up for me on an import but I've tried it with normal
mapping. I don't believe you can do a fix all on duplicate nodes and instead
have to resolve them all individually


>  Myself I've been using the python scripts at
> http://wiki.openstreetmap.org/wiki/Upload.py in such situations,
> although the api is much more stable now than it was a couple years ago.

What's your workflow? Do changes in JOSM, save and then pass to the scripts?

> The other good practice, but possibly not usable with JOSM alone, is not
> to let the program upload the "naked" nodes in bulk and then the
> buildings in bulk.  You can sort the elements in such a way that every
> 50k element changeset contains say 45k nodes and 5k ways.  The scripts
> let you limit the number of elements in a chunk which the next chunk
> depends on to the minimum (optimally 0), this way there's no risk of a
> passer by spotting orphan nodes and deleting some causing you conflicts
> in your next chunk.

The best way to do this in JOSM alone is to only merge in 1-5k nodes+ways at
a time, review them, then upload. This also avoids most of the problems
above.

I would only ever do a 50k object changeset in very limited circumstances
where I am confident that it is safe to do so. Even then I'd try to keep it
under 25k normally. For "normal" imports I'd suggest 10k as a soft limit.




More information about the Talk-us mailing list