[postgis-users] negative buffer

[Torsten Lange]

Again thank you, Vincent!
Now I got the idea!

Torsten

>
>>So 3 question remain:
>>What does st_buffer(geom,0) - just optimising the segments of 
>>the polygone?
>>    
>>
>
>It builds a buffer around a geometry with a radius of 0. Seems to be
>useless, but from the way the buffer is built, it is a method to
>transform a non-valid geometry into a valid one.
>
>  
>
>>Before I dropped the constraint it prevented the insert of 
>>the buffer geometry because it complained to be violated. 
>>Dropping it st_isvalid()however gives TRUE. This is a bit 
>>confusing to me. Do I mix something here? And are invalid 
>>geometries now likely to produced by postgis functions in 
>>some cases due to dropping the geometry check?
>>    
>>
>
>The geometry constraints check dimention, SRID and the type of geometry
>inserted, not its validity.
>If you had problems with the constraints, it is probably because the
>negative buffer yields a point or a linestring instead of a polygon. The
>geometry is probably valid, but it is not the right type, thus the
>constraint violation.
>You probably want to discard those geometries with a test on
>st_geometrytype, taking only polygons into account.
>
>Hope this helps.
>Vincent
>
>  
>
>>>>I have polygons with sometimes very peculiar shapes 
>>>>        
>>>>
>>representing the 
>>    
>>
>>>>distribution limits of specific types of sediments (e.g. 
>>>>        
>>>>
>>channel-like 
>>    
>>
>>>>branching, fluvial fans...).
>>>>When I try to create positive and negative buffers of that polygons 
>>>>using the combined (st_buffer(st_simply(geom,i)j)-method I succeed 
>>>>only in positive direction. In negative direction it very fastly 
>>>>violates enforce_geotype_geom-check. When I look on the 
>>>>        
>>>>
>>last results I 
>>    
>>
>>>>see, it is related to the channel-like branches.
>>>>Using the buffer-method in ArcGIS it works as expected. Is there a 
>>>>clever method getting it work in PostGIS too?
>>>>   
>>>>
>>>>        
>>>>
>>>As for getting negative buffers to be valid geometries, I 
>>>      
>>>
>>generally use 
>>    
>>
>>>the 0-buffer trick as in :
>>>st_buffer(st_buffer(geom, -i), 0). Not 100% sure it will work but it 
>>>generally fixes a lot of problems.
>>>
>>>Let us know of the results if you try that with your data.
>>>
>>>Vincent
>>>      
>>>