GeoAPI

GeoAPI issues are now tracked on http://jira.codehaus.org/browse/GEO.

Added remaining interfaces for the "geometry" package. The "aggregate" package is now next on the list.

Martin, the "coordinate geometry" package (Section 6.4 in ISO/DIS 19107) seems to be divided between java packages
"org.opengis.spatialschema.geometry" and "org.opengis.spatialschema.geometry.geometry". Can we combine these two java packages into one?

Is there a rationale for the division?

The "org.opengis.spatialschema.geometry" package contained originaly the "root" classes in ISO 19107 terms. A few classes from section 6.4 were also put in this package because they seemed extensively used (DirectPosition, Envelope).

However, generally speaking, I'm unhappy with the package names in GeoAPI for the geometry part. I dislike especially "org.opengis.spatialschema.geometry.geometry" First, the repetition of "geometry" is not very nice. Second, I would have liked to drop the "spatialschema" part.

The mean problem is: are we allowed to change the package names, given that some implementations may already exists?

Well, I looked around and I couldn't find any 19107 implementations except that one GT inherited from Sys Technologies. The bigger problem is the "non-implementations", the ones which just implement "DirectPosition" and "Envelope". Those might cause some heartburn if moved. We could "deprecate, then delete" in two successive release cycles.

If we're going to do anything, though, we should do it sooner rather than later. I think people are starting to want 19107 implementations due to the fact that more standards are being released which use geometry as a building block.

I agree totally about both naming issues. Perhaps "...geometry.geometry" could be "...geometry.coordinate" and spatialschema could just disappear?

NOTE: The "...geometry.aggregate" package is incomplete. It's missing MultiSurface and MultiSolid. Should this be tracked under a different issue?

The discussion of relocating packages/classes seems germane to GEO-52, which is closed as "wontfix". If we do anything to the packages, maybe these others could be rearranged as well?

I would agree about dropping the "spatialschema" part and renaming the last "geometry" into something else (or move its content to the parent "geometry" package). I also believe that the two successive release cycles ("deprecate, then delete") would work at least in a J2SE 1.5 environment (because "covariant return type" would be of great help there). I'm not sure how we could do that in a J2SE 1.4 environment however.

We may propose that to OGC, and I agree that it would need to be done soon. We may target the OGC meeting in october 2006, since it is the only one I can attend in the next few months.

As you rightly point out, the "..geometry.aggregate" package is incomplete. The full topology package is also missing. I didn't had the time to finish them. They are actually the subject of this GEO-1 task. "Renaming geometry packages" should be a different JIRA task.

GEOT-52 was about moving 5 very specific interfaces; it was not a general discussion about package renaming. It was closed as "will not fix" simply because after more investigation, it appeared that those 5 interfaces were actually at the right place.

Typo: GEOT-52 above should have been GEO-52.

Martin can we close this bug now?

Do we have finished to create an interface for all ISO 19107 elements? It seems to me that we have not finished some packages like aggregates, and all topology interfaces?