[postgis-users] newbie question about non closed rings

[Ben Madin]

G'day all,

a few cents more - this problem was a real blocker for my migration  
onto postgis - I had bad data sets, and they nearly all related to  
boundaries of tiny islands when I was working at a continental level -  
so I would have been quite happy to just loose the invalid geometries,  
rather than waste time trying to clean up unimportant data.

Maybe some development or instructions along the lines suggested by  
Simon (below) would help - getting data in to clean it (or discard it  
if deemed unworthy of repair) would be enormously beneficial.

Maybe an alternative way of handling it would be another  
geometrytype... essentially a GEOMETRY type but flagged or clearly  
'invalid' or at least 'un-validated' geometry. This might allow the  
use of 'cleaning' functions, but not analysis type functions so as not  
to affect overhead when under high query load, but allowing the tools  
in the database to repair invalid geometries.

cheers

Ben




On 19/04/2009, at 3:01 AM, postgis-users- 
request at postgis.refractions.net wrote:

> Before anything in PostGIS is changed to degrade this aspect of the  
> system perhaps someone ought to conduct a survey to ask if anything  
> should be changed and, if a respondant votes for a degrading of the  
> quality checking function, specifically ask what business  
> requirement demands that such a degrading takes place.
>
> Or perhaps my perspective on this is skewed by my view that a  
> database is about data management for businesses/departments and  
> less a tool for GIS practitioners.
>
> Surely, without changing anything, one can use PostGIS as a tool for  
> data cleaning. For example, why not load the data as WKT into a text  
> column and then update a geometry column after processing the text  
> data to ensure it is valid?

-- 

Ben Madin
REMOTE INFORMATION

t : +61 8 9192 5455
f : +61 8 9192 5535
m : 0448 887 220
Broome   WA   6725

ben at remoteinformation.com.au



							Out here, it pays to know...