[GRASS-user] v.overlay missing polygon

GRASS GRASS user
1

On 13/03/12 6:36 AM, grass-user-request@lists.osgeo.org wrote:

Dear Vincent,

thanks for your idea. I wasn't aware of the possibility to snap a vector
map to another one. Good thing to know!
However, it didn't solve my problem.
Actually, I checked the data with respect to my theory 2 again, and I
must say: Visually it seems, that in this case the two data sets I would
like to overlay match perfectly concerning the boundaries. The theory
might be wrong.

I did some further testing:

Even if I isolate the polygon that disappears in the overlay process

v.extract input=veg output=veg_137 list=137

and use it for the overlay

v.overlay ainput=plot atype=area binput=veg_137 btype=area operator=and
output=res_137_and

the result is erroneous. Nothing new, but still very strange.

Tricky. I am puzzled and need to get home;-)

Falko

Falko,

I agree - the same error reproduces for me, using your vectors.

Neither veg nor plot show any apparent errors in v.digit.

Anyone offer suggestions to try and debug this?

Richard

2

Falko,
if you try to make things match, looks like v.overlay returns a correct
result. Try the following command -- threshold 5 m is arbitrary, of
course :

v.edit map=plot type=boundary tool=move move=0,0 thresh=-1,5,0 bgmap=veg
snap=vertex bbox=3607160,5733860,3609000,57350

then
v.overlay --o ainput=veg atype=area binput=plot btype=area
output=res_overlay operator=and

(right now, I can't tell you what was wrong for the long western
polygon)

Vincent.

Le mardi 13 mars 2012 à 11:41 +1100, Richard Chirgwin a écrit :

On 13/03/12 6:36 AM, grass-user-request@lists.osgeo.org wrote:
> Dear Vincent,
>
> thanks for your idea. I wasn't aware of the possibility to snap a vector
> map to another one. Good thing to know!
> However, it didn't solve my problem.
> Actually, I checked the data with respect to my theory 2 again, and I
> must say: Visually it seems, that in this case the two data sets I would
> like to overlay match perfectly concerning the boundaries. The theory
> might be wrong.
>
>
> I did some further testing:
>
> Even if I isolate the polygon that disappears in the overlay process
>
> v.extract input=veg output=veg_137 list=137
>
> and use it for the overlay
>
> v.overlay ainput=plot atype=area binput=veg_137 btype=area operator=and
> output=res_137_and
>
> the result is erroneous. Nothing new, but still very strange.
>
>
> Tricky. I am puzzled and need to get home;-)
>
> Falko
Falko,

I agree - the same error reproduces for me, using your vectors.

Neither veg nor plot show any apparent errors in v.digit.

Anyone offer suggestions to try and debug this?

Richard

_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

3

Am 13.03.2012 07:53, schrieb Vincent Bain:

Falko,
if you try to make things match, looks like v.overlay returns a correct
result. Try the following command -- threshold 5 m is arbitrary, of
course :

v.edit map=plot type=boundary tool=move move=0,0 thresh=-1,5,0 bgmap=veg
snap=vertex bbox=3607160,5733860,3609000,57350

then
v.overlay --o ainput=veg atype=area binput=plot btype=area
output=res_overlay operator=and

(right now, I can't tell you what was wrong for the long western
polygon)

Vincent.

Le mardi 13 mars 2012 à 11:41 +1100, Richard Chirgwin a écrit :

On 13/03/12 6:36 AM, grass-user-request@lists.osgeo.org wrote:

Dear Vincent,

thanks for your idea. I wasn't aware of the possibility to snap a vector
map to another one. Good thing to know!
However, it didn't solve my problem.
Actually, I checked the data with respect to my theory 2 again, and I
must say: Visually it seems, that in this case the two data sets I would
like to overlay match perfectly concerning the boundaries. The theory
might be wrong.

I did some further testing:

Even if I isolate the polygon that disappears in the overlay process

v.extract input=veg output=veg_137 list=137

and use it for the overlay

v.overlay ainput=plot atype=area binput=veg_137 btype=area operator=and
output=res_137_and

the result is erroneous. Nothing new, but still very strange.

Tricky. I am puzzled and need to get home;-)

Falko

Falko,

I agree - the same error reproduces for me, using your vectors.

Neither veg nor plot show any apparent errors in v.digit.

Anyone offer suggestions to try and debug this?

Richard

_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

Dear Vincent,

I was able to reproduce your code. This is definitely an improvement! For the test data this solution works.
I applied the snapping code to a larger data set. There was only (or still) one case left where a polygon dissapeared. I guess the reason was, that there was a problematic overlap but no vertices within the snapping threshold. In this case it would be desirable that snapping could also be done with boundaries.

I think a snapping option in the manner of your suggestion should be an option in the v.overlay and v.patch tools. Disappearing polygons are really problematic for my work and this could be a solution.

Actually a friend told me that in PostGIS there is an option to snap all vector data to a grid. This would be possible in Grass if you would produce a vector grid (maybe from a raster file with an appropriate resolution) and use it as a bgmap. Did you ever try this?

Falko

4

Never tried this.
I am generally very cautious with snapping. Not knowing exactly how
grass handles it but one should be aware of possible topological
alterations, duplicate features, etc.

Vincent

Dear Vincent,

I was able to reproduce your code. This is definitely an improvement!
For the test data this solution works.
I applied the snapping code to a larger data set. There was only (or
still) one case left where a polygon dissapeared. I guess the reason
was, that there was a problematic overlap but no vertices within the
snapping threshold. In this case it would be desirable that snapping
could also be done with boundaries.

I think a snapping option in the manner of your suggestion should be an
option in the v.overlay and v.patch tools. Disappearing polygons are
really problematic for my work and this could be a solution.

Actually a friend told me that in PostGIS there is an option to snap all
vector data to a grid. This would be possible in Grass if you would
produce a vector grid (maybe from a raster file with an appropriate
resolution) and use it as a bgmap. Did you ever try this?

Falko
_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user