Waiting for PostGIS 3.1: ST_Subdivide and other function support with fixed ...

For a first glimpse, let's look at how fixed precision support affects ST_Subdivide function. In PostGIS 3.1, if you are running GEOS 3.9, you will have at your disposal an additional argument called gridsize

I'll give you a quick demonstration of how this works by chopping up my lovely state of Massachusetts which is part of Census state boundary file Census state boundaries

Original Geometry looks like this:

SELECT geom 
FROM states 
WHERE stusps = 'MA';

Double-precision subdivide

The old way, using double-precision coordinate system - this will yield 39 numbered rows:

SELECT f.ord, f.sd_geom
FROM states, 
  ST_Subdivide(states.geom, 300) WITH ORDINALITY AS f(sd_geom,ord) 
WHERE stusps = 'MA';

Fixed Precision Subdivide

Now let's try this using a fixed scale precision of 0.001 degrees. As with most things PostGIS the units of the gridsize is in the units of your spatial reference system. In this case we have NAD 83 long/lat, so we are setting the fixed precision to 0.001 degrees.

SELECT f.ord, f.sd_geom
FROM states ,
 ST_Subdivide(states.geom, 300,0.001) WITH ORDINALITY AS f(sd_geom,ord) 
WHERE stusps = 'MA';

and looks like this

The result is we end up with 10 rows instead of 39. You might be able to make out the edges are a bit smoother than the earlier image. This is because in a fixed precision, when the geometry is overlaid on the fixed grid the points smaller than a gridsize apart become one resulting in fewer points thus longer stretch of area before you hit a 300 subdivide point threshold and also a smoother image.

As you increase the grid size, you'll get a more pixelated result. Here we ramp up our grid size to 0.1 degrees and end up with PacMan fidelity. The result is 2 rows of very pixelated geometries.

SELECT f.ord, f.sd_geom
FROM states, 
 ST_Subdivide(states.geom, 300,0.1) WITH ORDINALITY AS f(sd_geom,ord) 
WHERE stusps = 'MA';

and looks like this