GEOSClipByRect: preserve Z coordinate

[vadimcn]

For [multi]linestrings/linerings, Z coordinate is interpolated from that of the nearest existing vertices.

For [multi]polygons, Z coordinate of points created on the boundary is interpolated as above. For interior points, however, it is populated using the ElevationModel.

[vadimcn]

@dbaston does this look okay to merge now?

[dbaston]

Do you think it makes sense to modify the tests to remove the expected value and just check that the result matches GEOSIntersection ?

    void
    checkClipByRectIdentical(
        const char* wktInput,
        double xmin, double ymin,
        double xmax, double ymax)
    {
        input_ = fromWKT(wktInput);
        result_ = GEOSClipByRect(input_, xmin, ymin, xmax, ymax);


        GEOSGeometry* rect = GEOSGeom_createRectangle(xmin, ymin, xmax, ymax);
        expected_ = GEOSIntersection(input_, rect);
        GEOSGeom_destroy(rect);

        GEOSNormalize(result_);
        GEOSNormalize(expected_);

        bool equal =  GEOSEqualsIdentical(result_, expected_) == 1;
        if (!equal) {
            wkt_ = GEOSWKTWriter_write(wktw_, expected_);
            std::printf("EXP: %s\n", wkt_);
            GEOSFree(wkt_);
            wkt_ = GEOSWKTWriter_write(wktw_, result_);
            std::printf("OBT: %s\n", wkt_);
        }
        ensure(equal);
    }

[vadimcn]

I don't know enough about the GEOS project to have a strong opinion on this.
Do you postulate that the result of GEOSClipByRect should always be the same as GEOSIntersection with a box polygon? Or should it be the same after normalization, as you have above? Then maybe yes.

However, currently, several existing tests fail this comparison, see below. I'd say that fixing those is out of scope of this PR.

EXP: POINT EMPTY
OBT: GEOMETRYCOLLECTION EMPTY
[1=F]

EXP: POINT (15 10)
OBT: GEOMETRYCOLLECTION EMPTY
[3=F]

EXP: LINESTRING EMPTY
OBT: GEOMETRYCOLLECTION EMPTY
[4=F]

EXP: MULTILINESTRING ((10 10, 15 10), (10 10, 10 15))
OBT: GEOMETRYCOLLECTION EMPTY
[6=F]

EXP: MULTILINESTRING ((20 10, 20 20), (10 20, 20 20), (10 10, 20 10), (10 10, 10 20))
OBT: GEOMETRYCOLLECTION EMPTY
[10=F]

EXP: MULTILINESTRING ((20 10, 20 20), (10 20, 20 20), (10 10, 20 10), (10 10, 10 20))
OBT: GEOMETRYCOLLECTION EMPTY
[11=F]

[dbaston]

However, currently, several existing tests fail this comparison, see below.

Odd, they are passing for me after normalization. Either way, it can be beyond the scope of this PR.

[vadimcn]

Odd, they are passing for me after normalization. Either way, it can be beyond the scope of this PR.

I thought the plan was to update all tests, including those still using checkClipByRect. If you only changed checkClipByRectIdentical, I guess it'd work.

[dbaston]

Looking more closely at the examples you pointed out, I can see that the semantics are different from GEOSIntersection. I guess the tests should be left as-is.