[GRASS5] GRASS5.1: shapefile import problems

[Radim Blazek]

On Monday 02 June 2003 23:36, Moritz Lennert wrote:
> Hello,
> 
> I'm trying to use / import a shapefile into 5.1.
> 
> When I use v.format, The map seems perfect and d.what.vect displays the
> correrct information. However, I seem to be limited to the dbf file as
> database since when I use v.db.connect to try to connect the file to a
> PostgreSQL database I'm told:
> 
> WARNING: Shapefile format cannot be updated.
> WARNING: Shapefile format cannot be updated.
> ERROR: Cannot open old vector cp at mlennert on level 0
> 
> When I use v.in.ogr + v.clean (tool=rmdupl,bpol) the file gets imported
> incorrectly, areas not being recognized as area and attributes being
> linked to the centroids, but not to the areas. I can link this map to the
> database, but that doesn't help me with wrong topology.
> 
> As info here's v.build output on the v.format shapefile:
> 
> v.build cp
> WARNING: Topology for shapefile areas is not fully supported.
> Registering areas: Topology was built.
> Number of nodes     :   2511
> Number of primitives:   2538
> Number of points    :   0
> Number of lines     :   0
> Number of boundaries:   1293
> Number of centroids :   1245
> Number of faces     :   0
> Number of kernels   :   0
> Number of areas     :   1245
> Number of isles     :   46
> Map is 3D           :   0
> 
> 
> And here v.build output on the v.in.ogr/v.clean tool=rmdupl,bpol file:
> 
> v.build cp_clean
> Registering lines:
> 10696 primitives registered
> Building areas:  100%
> 77 areas built
> 46 isles built
> Attaching islands:  100%
> Attaching centroids:  100%
> Topology was built.
> Number of nodes     :   4920
> Number of primitives:   10696
> Number of points    :   0
> Number of lines     :   0
> Number of boundaries:   9451
> Number of centroids :   1245
> Number of faces     :   0
> Number of kernels   :   0
> Number of areas     :   77
> Number of isles     :   46
> Map is 3D           :   0
> 
> The first one seems correct (I haven't verified for all areas), but the
> second is completely wrong.
> 
> So, my questions are:
> 
> 1) can I link a shapefile imported with v.format to a database table ?

No. Shapefile is linked by default to its dbf table. 

I think that to say "imported" is maybe a bit confusing, as it is not imported.
Vector created by v.format is something like link.

First version of shapefile support was using one column from dbf (defined in frmt file)
as category values, and user could link this vector to any table (including
original dbf). Problems were that:
1) dbf part of shapefile may be without unique id column
2) it was very difficult, to define this connection, for usual users
   (Yes, you have a shapefile so why do you want to link it to something else! :-)

> 2) is there a way I can ensure a more clean import with v.in.ogr

I would suggest tool=bpol,rmdupl, which means first break polygons
on intersections (boundary points shared by 3 and more areas
(or 2 areas + nothing)) and then remove duplicate boundaries.
tool=rmdupl,bpol must fail on shapefile, it is OK, because duplicate boundaries
were left in vector.

tool=bpol,rmdupl can work perfectly only on clean shapefile
(generated from coverage for example), if you have overlapping areas
in shapefile, you can try tool=bpol,break,rmdupl , but result
may never be perfect (in principle).

> 3) knowing that the shape format is very bad, and having access to
> ArcView, could I export the file to another format and import it into 5.1

I don't think that shapefile is bad, problem is that it is so different 
and allowes overlapping areas. For areas is better e00, but m.in.e00 
is not yet in 5.1. If you can get e00 for your file (impossible
from ArcView, I thing), you can import to 5.0, convert to 5.1 and 
link to dbf.

Radim