Setting precision on some but not all geometries causes intersection results to use the smallest nonzero gridSize

[Brendan Ward]

According to the comments in "geos_c.h.in" for GEOSGeom_setPrecision_r: "

• Note that operations will always be performed in the precision
• of the geometry with higher precision (smaller "gridSize").
• That same precision will be attached to the operation outputs.

"

Given inputs:

• "POLYGON ((0 0, 0 0.9, 0.9 0.9, 0.9 0, 0 0))"
• "POLYGON ((0.75 0, 0.75 0.75, 1.75 0.75, 1.75 0, 0.75 0))"

I then used GEOSGeom_setPrecision_r with a gridSize of 0.5 for the first geometry, which returns "POLYGON ((0 0, 0 1, 1 1, 1 0, 0 0))", and reports a gridSize of 0.5 via GEOSGeom_getPrecision_r.

Performing an intersection via GEOSIntersection_r, I get back "POLYGON ((1 0.75, 1 0, 0.75 0, 0.75 0.75, 1 0.75))" which reports a gridSize of 0.5 (second geometry reports a gridSize of 0).

If I then set a precision of 1 on the second geometry, which produces: "POLYGON ((1 0, 1 1, 2 1, 2 0, 1 0))" (second geometry then reports a gridSize of 1)

I get back "LINESTRING (1 1, 1 0)" from intersection, which reports a gridSize of 0.5

This may be expected behavior, but the comment in the code implies that the higher precision, in this case full precision as denoted by a gridSize of 0, would be used. If so, perhaps the comment could be clarified to note something like:

• if inputs both have nonzero gridSize, operations will be performed using the higher precision (smaller "gridSize").
• if one but not all of the inputs have a nonzero gridSize, and others have a gridSize of 0, the smallest gridSize will be used.

(only tested on GEOS 3.8 via C API)

[robe]

Retargeting in prep for GEOS 3.10.0 release