Wrong explanation in FlatGeobuf Spatial Index

[mrmichaelf]

It' written:
"This Crunchy Data blog post has helpful examples for why sorting input is important."
The blog post says:
"Pre-ordering your rows randomly will result in a more balanced tree and measurable improvements in query performance."

[kylebarron]

FlatGeobuf's spatial index is always sorted on the geometries' hilbert value, so it doesn't matter what order the geometries are when passed into the FlatGeobuf writer.

[mrmichaelf]

The purpose of my comment is to shade a light about misleading documentation: One can understand "sorting input is important", while in reality "ordering your rows randomly will result... improvements in ... performance".

[kylebarron]

The implementations of the PostGIS R-Tree and FlatGeobuf's hilbert-packed R-Tree are different.

Maybe we should take out the reference to that blog post.