Geocoding Points Against a Grid of Cells/Tiles With/out Spatial Operators


First up, the solutions are very similiar.

The set up of the grid and points was shared.

This allowed a more direct, visual, comparison.

Where they are different is that one uses pure CPU to compute grid references, the other uses spatial index access.

Yes, in this article the traditional approach was faster, but I have not done any scalability or more detailed performance analysis.

The direct method may scale better where CPU is more available than database or disk.

Additionally, building a function that took the grid metadata but did not return the col/row references but rather the Morton key, that function could be used to determine partitions in a partitioned table that was organised on the Morton key.

The world is your oyster.
(Shakespeare’s The Merry Wives of Windsor?)