[OSM-dev] TIGER import

[Dave Hansen]

On Thu, 2007-06-07 at 14:03 -0700, James Marca wrote:
> I understand from reading the mailing list archives (March, 2007, I
> think) that one problem is that roads that should be connected are not
> due to different coordinates of the end points (small errors). 

I just hacked up my copy of the ruby importer to generate XML files
instead of uploading back to the OSM server, and just dumped the xml
files back into JOSM.  The main modification I had to make was
generating my own ids for all of the objects, instead of getting them
from the OSM API.

Actually, the JOSM Validator plugin seems to detect these
close-but-different nodes pretty nicely.  It does seem to screw them up
when it tries to fix them automatically, though.  The only downside to
this approach (if that "Fix" is fixed) is that I still end up with a
bunch of connected, but single-segment ways.  Writing another validator
check and fix for this might be helpful to other people than just those
doing the TIGER import.  

One thing I just noted was that the TIGER data seems to be ~60m off for
a small street in my neighborhood.  anybody else see those kinds of
errors?

-- Dave