postgis/sfcgal/regress

These regress/sfcgal directory contains several tests closely based on 
current regress one. 

Usually they should behave really the same (i.e with both backends)
however some slightly differences beetween GEOS and SFCGAL backends 
are listed below:


==
sfcgal/ticket
==

2.OBT: #11|1.31708901596544e-9
2.EXP: #11|0

=> Numerical precision issue
   (We slightly modify the unit test to use SQL round)


18.OBT: ERROR:  LWGEOM2SFCGAL: Unknown geometry type !
18.EXP: #85|0

=> Curves are not (yet) handled by SFCGAL
   (To be done in future SFCGAL release)



132.OBT: #834|GEOMETRYCOLLECTION(POINT(0 0),LINESTRING(10 0,10 10))
132.EXP: #834|GEOMETRYCOLLECTION(POINT(0 0 5),LINESTRING(10 10 5,10 0 5))

=> For SFCGAL a 2D Intersection even with both 2D and 3D input data
   is assumed as 2D and so set output Z to 0



228.OBT: #1957|inf
228.EXP: #1957|1

=> Initial line is OGC invalid (single point),
   so SFCGAL return infinite distance,
   which .. also make sense


==
sfcgal/empty
==

29.OBT: ST_Intersection(geometry, empty) == geometry|010700000000000000
29.EXP: ST_Intersection(geometry, empty) == geometry|010300000000000000

30.OBT: ST_Intersection(empty, empty) == empty|010700000000000000
30.EXP: ST_Intersection(empty, empty) == empty|010300000000000000

=> SFCGAL don't consider keeping initial geometry type when intersection is null
   so a GEOMETRYCOLLECTION(EMPTY) is returned (for both tests)
   (maybe could be improved in SFCGAL)


33.OBT: ST_Distance(geometry, empty) == NULL|inf
33.EXP: ST_Distance(geometry, empty) == NULL|

=> SFCGAL returns an infinite distance to an empty geometry instead of null,
   which .. also make sense