[GRASS-user] generalize polygons including holes

[Markus Metz]

On Thu, Feb 5, 2015 at 4:56 PM, Moritz Lennert
<mlennert at club.worldonline.be> wrote:
> On 05/02/15 14:49, Markus Metz wrote:
>>
>> On Thu, Feb 5, 2015 at 11:50 AM, Moritz Lennert
>> <mlennert at club.worldonline.be> wrote:
>>>
>>> On 05/02/15 10:06, Markus Metz wrote:
>>>>
>>>>
>>>> On Thu, Feb 5, 2015 at 9:36 AM, Moritz Lennert
>>>> <mlennert at club.worldonline.be> wrote:
>>>>>
>>>>>
>>>>> [Please keep conversations on the list]
>>>>>
>>>>> On 04/02/15 16:12, Robert Nuske wrote:
>>>>>>
>>>>>>
>>>>>>
>>>>>> Hi Moritz
>>>>>>
>>>>>>>> is there a way to generalize polygons including their holes?
>>>>>>>> I am not too worried about topology, there are only polygons of one
>>>>>>>> class
>>>>>>>> in the dataset.
>>>>>>>>
>>>>>>>>
>>>>>>>> I tried the following
>>>>>>>> v.generalize in=m3 out=m5 type=area method=douglas threshold=5000
>>>>>>>>
>>>>>>>> which simplified the "outer boundaries" okay but did not touch the
>>>>>>>> "inner
>>>>>>>> boundaries"/ holes of the polygon.
>>>>>>>>
>>>>>>>>
>>>>>>>> Outer boundaries are on the left side in the picture and holes on
>>>>>>>> the
>>>>>>>> right
>>>>>>>> http://i.imgur.com/hHKr15m.png
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> It looks like the red boundaries are the generalized version of the
>>>>>>> original blue ones. Is that correct ? Are you sure that the original
>>>>>>> boundaries of the holes are identical to the new ones ? Could you
>>>>>>> display the result with polygon fill color set to transparent ?
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> That's right.
>>>>>> Original vector map is blue and simplified vector is red.
>>>>>> Both layers are 50% opaque in the following image
>>>>>> http://i.imgur.com/Rsx7Kh1.png
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> I imagine you are using GRASS 6 ? I can confirm the issue there with
>>>>> GRASS
>>>>> 6.4.4, but it has been solved in GRASS 7. See the attached image where
>>>>> you
>>>>> have two holes: both have been generalized with method=douglas (black =
>>>>> original, red=lower threshold, blue=higher threshold). You can clearly
>>>>> see
>>>>> that generalization is also happening for the hole boundaries.
>>>>
>>>>
>>>>
>>>> The right hole has not been generalized properly, the resultant
>>>> boundary intersects with itself. This has been fixed in GRASS 7
>>>> recently. Do you get the same result with recent GRASS 7.1?
>>>
>>>
>>>
>>> No. The example was with GRASS 7.0 from a week or two ago.
>>
>>
>> OK, the test was missing in 7.0, I have now backported it to 7.0 in r64471
>>
>>> In GRASS 7.1 no
>>> more intersection, but actually for that hole the generalization is less
>>> with a higher threshold (see attached image).
>>
>>
>> That hole is actually not generalized at all with thresh=500 because
>> it would damage topology.
>
>
> Maybe a little message would be helpful here to inform the user about this ?

This message exists already:

v.generalize test method=douglas thresh=500 out=test_gen_500
...
Generalization (douglas)...
 100%
WARNING: 1 boundaries were not modified because modification would damage
         topology
...

>>
>>> I've attached the vector file
>>> I used as a GRASS 7 vector pack for the NC location.
>>>
>>> The settings I used were:
>>>
>>> v.generalize test method=douglas thresh=100 out=test_gen_100
>>> v.generalize test method=douglas thresh=500 out=test_gen_500
>>>
>>>>
>>>> v.generalize does not distinguish between outer and inner rings, it
>>>> simply goes through all boundaries and generalizes each one.
>>>
>>>
>>>
>>> And this has always been the case or this is specific to GRASS7 ?
>>
>>
>> This has always been the case.
>
>
> Then why doesn't there seem to be any generalization happening for the hole
> boundaries of the test dataset with grass64release ?

With the test dataset you provided, GRASS 6.4 does generalize the hole
boundaries. I do not have the problem vector of Robert, so I can not
say if the hole boundaries in that vector are generalized or not.

Markus M