[postgis-users] PostGIS kills all connections to the PostgreSQLserver

[Paul Ramsey]

How would you suggest it be improved?

We could simply embed the ideas of the workaround into the function  
itself (if feature > magicsize and buffersize > magicsize then pre- 
simplifyslightly, or doublebuffer with simplification).  That might  
make cases better, but it might also screw up other cases which work  
perfectly fine right now.

Of course, you can't explain anything about ArcGIS, since the  
workings of it are invisible.  They could have a radically better  
approach to buffering, or they could have 200 case-dependent hacks to  
make their buffer work well across the cases they have compiled and  
know about over 20 years.  You don't know, I don't know, and that's  
the problem with trying to use it as a design exemplar.

P

On 20-Jul-06, at 7:47 AM, Michael de Armas (Micotan) wrote:

> Thanks for your solution. However, I believe the Buffer algorithm  
> should be improved to avoid this work around that only would  
> complicate an automation process and is still taking a while.
>
> I don't think there's any quality problem with the shape file, the  
> boundary is not simple because is following a river. If there was  
> such a quality problem, how would I explain that ArcGIS does the  
> same buffer (on that very shape) in 2 seconds maximum, and they are  
> even rendering.
>
> Bruce Rindahl wrote:
>> My guess as to why this is happening is the shapefile is really  
>> poor quality
>> with intersections (interior loops etc.)  Where did you get it?   
>> ESRI maps
>> and data CD?
>> The fix is to do multiple buffers.  I tested your query and it ran  
>> in 234
>> seconds (20000 buffer).
>> I then ran:
>> SELECT area(buffer(buffer(buffer(transform(GeomFromText('POLYGON 
>> ((....
>>   ......................
>> ......))', 4267), 32039), 1000), 9000), 10000);
>>
>> and got the same result in 25 seconds.  I think the first buffer  
>> in effect
>> cleans out the weird parts of the shape and then the rest are  
>> acting on a
>> cleaner shape.   A 60,000 buffer was done in 26 seconds.
>> Hope an order of magnitude helps!
>>
>> Bruce Rindahl
>>
>
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users