On Sun, Feb 22, 2009 at 6:53 PM, Frederik Ramm <frede...@remote.org> wrote:
> Franc, > > Franc Carter wrote: > >> The talk-au list has been discussing the import and the current thoughts >> are here:- >> > > We did it almost the same way for a German administrative boundary import > in late 2008. Your Wiki table doesn't include the obvious relation tags > (type=multipolygon, boundary=administrative, admin_level=something) but I > guess these were implicit and so you didn't mention them? Also, for the > benefit of relation-unaware software, we usually tagged the member ways with > "boundary=administrative, admin_level=x", with x being the smallest > admin_level of all relations that used this way. We did not add any names or > "state:left=x, state:right=y" to the ways though. This is something I was unsure about - the talk page for this had lots of conflicting opinions ;-( > > We did not see a need to simplify the lines in our case and I cannot > imagine Aussie borders being so detailed that this should be necessary but > you'll know better. There's a lot 'over specification' in the data - 10's of points where one for each end would be fine. > > Multipolygons now support multiple outer rings (see the multipolygon > relation talk page in the Wiki - the "advanced multipolygons" secion really > needs to be moved to the main multipolygon page, it is actually in use and > not just "talk"), so that solves your problem of disjoint suburbs. Ok, I'll have a read and wrap my head around it > > When actually running the upload (we used the bulk_import.pl for this after > creating a suitable .osc file) you should make sure that you *not* attempt > to first upload all the nodes and then all the ways and then all the > relation (which is what we did), because unless the process is over in an > hour or so, you will have eager mappers removing your nodes for obvious lack > of use! In the end I had to download the hourly diffs while I was importing, > searching for "my" node IDs and checking who deleted some, then undelete > them with another script running in parallel, so that my import script would > find the nodes it expected to see. So the lesson learned here was, try and > upload a way as soon as you have the necessary nodes for it, and not at the > end. I thought of this ;-) The file has the data node/way specification just before it is first needed, so hopefully there will only be a short period where the data looks inconsistent cheers > > > Bye > Frederik > -- Franc
_______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk