Sdo_intersection area and error message

Question

Hi
I wonder if anybody can offer any help/advice
I have been running a routine which loops through a spatial Layer A and for each polyon in the loop I pick up those polygons from a more detailed Layer B (over 5 million records) that interact then create a non spatial table output showing the area of intersection between the 2 layers.
This is used in the query to return the area of intersection:
ROUND(mdsys.sdo_geom.sdo_area(                              mdsys.sdo_geom.sdo_intersection(a.geom, b.'||l_spCol||', 0.0005),1,''unit=HECTARE'' ), 3)
The SDO_DIMINFO for both layers (UK, national grid) is:
((X, 0, 700000, 5E-10), (Y, 0, 1300000, 5E-10), , )
I have been running this successfully for a while now. However the 3rd party that supply layer B have done some major updates and I now have some errors when running the routine i.e. some of the polygons now cause errors.
The error I get is:
ORA-13347: the coordinates defining an arc are not distinct
It is a bit of a needle in a hay stack trying to find the ‘error’ polys and have found 3 so far. The way I have found the Layer B errors is by changing the tolerance in the above area calculation to 0.1 (in the polgon where Layer A fails), the query will run with the lower tolerance and a NULL record is returned for the error polys. I have taken these out of Spatial and loaded into ESRI, after geometry checks ESRI describes the error as “self intersections”. Once cleaned and reloaded the routine works ok. I did find some examples of these self intersections that are not causing a problem.
One thing that confused me was I got no error reported using this:
SELECT c.LAYERA_ID, SDO_GEOM.VALIDATE_GEOMETRY_WITH_CONTEXT(c.geom, 0.005)
FROM LAYER B c
I noticed the MDSYS.SDO_ELEM_INFO_ARRAY(1,1005…, was 1005 for the error polys rather than 1003, not sure about the significance of this!
I wondered whether there was a ‘better’ way of identifying the polygons upsetting the area calculation (and why VALIDATE_GEOMETRY_WITH_CONTEXT wasn’t picking these up).
And
How much accuracy would I lose in downgrading the tolerance of the area calculation to 0.1. Comparing results for 0.1 and 0.0005 they seem to be only different at 3 decimal places.
Thanks for any help/comments (I am fairly inexperienced on Oracle Spatial)….

Answer 1

Thanks again for the posts.
I have run my 5+million polygons through the has_arc procedure (thanks for that!) and this returned x4 polygons (3 of which I had identified as breaking my procedure). The additional one didn’t seem to be causing any problems – and ESRI didn’t find any geometry errors.
So I am inclined to think this issue has come about by the (recent) addition of these polygons with the ‘Circular Arcs’ to layer b (by a third party).
Moving forward when ever I get a cut of the latest version of the data (Layer b). I will run the has_arc function to see if any new polygons have gone into the layer – as it is likely these could be a problem. I have managed to now run the procedure with the addition of the x3 cleaned polygons. So hopefully when I start to re run the procedure on different Layer A's I will be ok i.e. Layer B wont break it (fingers crossed).
In summary to the problem I encountered:
I get Layer B from a third party (imported from their database), all in UK nat grid (metres) (81989). The SDO_DIMINFO setting coming from the source. It is based on detailed large scale mapping.
The procedure breaks on the x3 polygons when returning the area (SQL in original post) of intersection using 0.0005, I can by-pass the error using 0.1 and get a null intersection area returned.
I can get a spatial layer intersection output using the problem data.
I am still a bit confused as to what a circular arc actual means in terms of geometry (I haven’t done any digitising type work for a bout 15 years and my old topology type knowledge seems to have left me…) Once I understand it – I would like to raise it with the 3rd to ascertain why they are there. The polygons ‘look’ very simple!
And does anybody have any recommended reading re tolerances. The only GIS I have used in the last few years has been desktop GIS – where I haven’t really had to worry about setting appropriate tolerances. The stuff I have read on Oracle’s site has left me a bit puzzled.
Cheers