Implement ST_Boundary(Triangle)

[strk]

Seems to be a useful thing to have...

[nw]

The attached patch will implement this. A polygon would be pretty simple to do without going to GEOS also, though not trivial like the triangle.

[robe]

seems harmless enough.

[robe]

committed at r11358 . If you can get in a patch for TIN before end next week, I might be able to commit (and call this fully supporting tins in docs). Right now (I'd have to put in just a note that it supports triangle, which ain't bad since I think ST_Dump supports and they can do their own boundary by combining the 2).

[robe]

amended docs at r11359

[strk]

I'm seeing too many changes in lwgeom-geos connectors these days. May be worth some talk on the mailing list before it messes up too much ? Do I understand correctly that TRIANGLES are being threated like MULTILINESTRINGS when converted to GEOS ? Would it make more sense to convert to polygon ? Should we expose a Triangle geometry in GEOS C-API ?

[nw]

I don't think that triangles are converted to geos at all. A look through the LWGEOM2GEOS() function doesn't seem to have any entry for a TRIANGLETYPE in the switch, so I think you'd get a call to lwerror() if you somehow managed to pass in a triangle. Curved geometries will fail also. Perhaps something else is converting them before a call to LWGEOM2GEOS().

I don't think GEOS has a triangle type, so you'd have to convert to a polygon if you were to do this.

[strk]

ok then, I misread the patch, sorry for the noise.