[Talk-ca] Automated imports of Canvec?

[Tyler Gunn]

> With respect to the automated GeoBase import I did a few years back (of
> which that is a sample of)  I can provide .osm files with all the excluded
> data to anyone that asks. At one point I had them up on a free filesharing
> site but I think that went away. I don't think Potlach can read these files
> though. If someone wants to setup a server that can server them up as WMS
> I'd be happy to provide the files as well.

Hmm, I did a bit of playing around with Geobase NHN data (out of
curiosity), and I found the Geobase NHN import script is close to
being able to best CanVec in terms of converting lakes and rivers
over.  One thing I notice in the prairies is that the Canvec data
labels large rivers as "natural=water", and often omits the name.  The
same areas converted from Geobase (exact same geography) result in:
1) Properly tagged (okay, the script I'm using makes mistakes, but
they are fixable) rivers
2) More frequently named objects
3) Water bodies that are single large multi-polygons, rather than a
bunch of separate ways of relations describing the same area.
4) I get some rivers that become part of the multi-polygon for a lake,
making it a whole load of manual editing to split them up.  The source
Geobase NHN data specifies them separately.

I'm partly inclined to consider just fixing up the NHN Geobase import
scripts and importing the water bodies, rivers, streams, etc using
that.  Dunno.  I might be inclined to do a sample for an area in
Northern MB to see what I can figure out.

It does raise an interesting point in my mind; we don't really do much
about the fact that large water bodies get broken up across tiles into
separate relations, etc.  There really needs to be a tool to clean up
these kinds of things.   The same applies to an extent with large
wooded areas, but in my mind isn't as big of a deal as water bodies.

Tyler