On 10 February 2013 09:05, David Schneider <d.schnei...@tradermail.info> wrote: > Hi List! > > I searched an address in the eastern parts of Cape Town, and I was > suprised that nomatim didn't report it as residential. I checked the > data, and found that vast parts of Cape Town's residential areas are > made up from unclassified roads instead of residential! > From the metadata, this seems to be the result of an import some > years ago. >
Yes, the City of Cape Town (CoCT) municipality gave us road data a number of years ago. The data wasn't brilliant and did not make a distinction between residential roads and unclassified. We used the data to fill areas that hadn't been mapped. Mitchell's Plain, Rocklands, Khayelitsha, Eesterivier etc. The data was cleaned up (to a degree: over nodeing, connectivity, over slitting, etc) and added by myself (Firefishy), Adrian Frith and I think Ryan Peel. We were initially planning to use the http://osm.org/user/City%20of%20Cape%20Town%20Import > I feel this should be fixed. I don't know if this can be done auto- > matically, or a manual community effort is required. My proposal would > be to select "user=Firefishy maxspeed=60 highway=unclassified" in JOSM > and change all to residential. Now, a few roads that should be > unclassified will be residential, but the ratio seems to be 99% wrong > now vs. 1% wrong after this change. Of course, roads that have been > touched since the import will missed by this, but there will be an > improvement for sure. WAIT! Please do not select and mass change. Better to do the changes based on aerial imagery and local knowledge. It would be great to fix up the data. As detailed above I was not the only person to have add the CoCT data. > > Another thing that I noticed is that the roads are massively over-noded. > http://www.openstreetmap.org/browse/way/32216567 demonstrates both. It > should be residential, and if you compare to aerial images, the straight > section is represented with 6 nodes, 3 of them within 5m. I presume this > comes from calculating the centerline of a full shape, as the road widens > towards the east, where the many nodes are. > (I would map that road with 3 nodes, with the two on the eastern sides > interfacing between the NW-SE road, going away at 90° from there. I.e. > making a slight right turn when going east. The data is just the opposite, > going to the left, which gives an odd angle that is not there in reality > when you come from the south and turn east.) > JOSM has a built-in simply function, select way and press Shift-Y. Do not be too trigger happy with simplifying, especially if a mapper has spent time to create good natural curves. One of the annoying artifacts from the CoCT data is at T junctions there are often pointless nodes very close to the junctions, these should be removed or moved as they can disrupt mapping. Regards Grant (aka Firefishy) _______________________________________________ Talk-ZA mailing list Talk-ZA@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-za
