I was doing some analysis on UNION against the same feature (on itself) to get
an understanding of what happens with slivers and overlaps. I started in ArcGIS
and was able to understand what happened with the combinations available there
Zone system used as input features composed of 463 zones including some
File with accidental overlaps:
Case 1 - self union, 1 ref to source, no gaps - Generates 525 - due to
lakes (3) and slivers (16) and overlaps (43)
-- overlaps times number of contributors i.e. two zones overlap, two overlap
features with same area but different source zone
-- slivers -1 as source (since there is none) and single feature.
Case 2 - self union, 1 gaps - 506 -
due to overlaps
Case 3 - self union, 2 refs to source, no gaps - 577
Case 4 - self union, 2 gaps - 558
Having done all this in Arc, I now wanted to check that QGIS produced the same
thing. I could not check for gaps / no gaps as there is no setting to select
regarding this. Also, the two tools I used via (1) menu Vector -->
Geoprocessing --> Union and (2) Toolbox Geoalgorithms --> Vector --> Polygons
--> Union do not allow for a single reference to the source feature set.
What puzzles me though is that instead of feature counts of the order listed
above, I get much more features and many, but not enough with nulls and 0 for
source ID fields etc. I cannot see an obvious way to get the resultant 4,139
features (with tool 2) down to approx 500 or so. Tool 1, returns 3,950
features on the same input feature set.
I also when selecting features in the area with overlaps return normal counts
of features i.e. 2, not multiple orders higher as I'd expect given the count of
features in the attribute table. Do some (lots) of them have no geometry? -
Actually having just added a field of area, I see this is not the case either
as too many have geometry on that basis.
Am I missing a key difference employed in the processing of QGIS union and a
way to get to the relevant features which approximate to the 500+ I expect?
Am using 2.12 on Windows if that has a bearing on this.
Qgis-user mailing list
List info: http://lists.osgeo.org/mailman/listinfo/qgis-user