Paul Ramsey

I think you may have nailed it. The whole core of that problem is not so much getting "wrong" results as getting "slightly different" results, and a stable answer would...

Just going to pop in here to note that your SQL solution is brilliant. 👍👍👍

I'm a little concerned about the CI failure, is this rebased to main? I don't really see how your API change could cause the failure, but neither do I want...

The thing I don't understand here is, isn't CascadedUnion already effectively doing this? The tree puts near-to-each-other things early in the union order. Is this on top of, or in...

Counter-intuitive. I'd have thought, working bottom-up, the disjoint parts would mostly come together near the end and get quickly no-opped with an envelope intersection test.

Very exciting. I think the preparedgeometryxy stuff is going to be required anyways, to really crank up point performance on the classic p-i-p case. A little surprised the gains were...

One thing I'd like to cogitate on is the extent to which this does/doesn't foreclose on a higher dimension strategy that splits the XY coordinates from the Z and M...

I don't think there's anything particularly magical about fixing the size of a CoordinateSequence. I look forward to seeing what happens when higher dimensionality comes into play. If one pre-supposes...

The basis for splitting off Z/M from X/Y is totally theoretical! Except it's also practically implemented in the Shape file. Also the OGR feature implementation. It packs the things we...

In the overlay code, a key problem is that CoordinateSequence spends some time on its own, after being found in the geometry and before being incorporated into a SegmentString, and...