By: Jakub K

Jakub K — Wed, 10 Jun 2015 08:29:34 +0000

There’s another way..
Tesselate the polygon yourself. The relationship between how quickly the spatial query works (across all the rdbms’) is how much extra work the 2nd pass (after .Filter) has to do to get rid of the false positives.
By slicing your complex polygon into simpler, more rectangular shapes you remove the false positive ‘white space’ around your complex polygon. There’s some of my code and more info here – https://social.msdn.microsoft.com/Forums/sqlserver/en-US/78132bca-2a95-45c1-b93c-64c83827dc4f/sql-2k8r2-any-performance-hints-for-bulk-loading-data-from-spatialgis-db-into-a-data-warehouse?forum=sqlspatial
I’d be curious to see how the TVFs compare to the approaches you outlined above. Mine doesn’t scale too well; the overhead of the TVF recursively sliciing the polygons becomes obscene once you start throwing a dozen or so complex polygons at it.
I have a .netclr version of it as well that’ll eventually make it online. All coded and working, just the doco left… This one does scale very well

Comments on: Tuning slow spatial queries in SQL Server

By: Jakub K