ST_Voronoi

[robe]

Once Voronoi is ported from JTS to GEOS ​http://trac.osgeo.org/geos/ticket/627 then this should be done to expose it in PostGIS

[m4farrel]

Hi,

Anyone working on this currently?

Would be very useful to have for manipulation of some bathymetric data I'm currently working with.

[strk]

Nobody working on this. I've recently tweaked the GEOS interface to fix the linear return and adding a bounding extent. I think the API is stable now, do you want to start with the PostGIS side of it ?

[m4farrel]

The task seems quite daunting to me, though I'd find contributing to PostGIS interesting.

I'm currently a student intern, so I've emailed my mentor to see if it would be appropriate to try to work on this. My software for the project I'm working on is written in Scala: I was also thinking about just using Geotools and the JTS Topology Suite to bake my data into the format that I desire.

Concerning the Voronoi builder implemented in GEOS, do you know if ST_Voronoi would be appropriate for use on sets containing ~2,000,000-10,000,000 point sites?

[strk]

I suspect that amount of points might require more RAM than available,but you can test that easily by providing such a big input in WKT or WKB format and feeding it to the voronoi diagram builder via the PHP binding, look here: ​https://github.com/libgeos/libgeos/blob/40c117aa088b084ee0f40b4333d63d4fd0acad56/php/test/test.php#L2046

[dbaston]

strk - I can tackle this one if you want.

[strk]

dbaston, sure, be my guest !

[robe]

dbaston,

You might want to talk to darkblueb (Brian Hamlin) to see how far he got with this. As I recall he had a patch raring to go but was too late for PostGIS 2.2 cut.

[robe]

ah found the relevant mailing list email

​https://lists.osgeo.org/pipermail/postgis-devel/2015-May/024944.html

[dbaston]

darkblueb - can you let me know if you're planning to work more on this? If not, I can pick up where you left off.

[robe]

For completeness, added dbaston's ST_Voronoi and darkblue_b

dbaston

​https://github.com/dbaston/postgis/tree/trac-2259

darkblue_b

​https://github.com/darkblue-b/postgis/commit/42e338462e8d1c717067a2a304e010caca0d3dd5

At a spot check dbaston's looks further along than darkblue_b and uses the new interator construct for 2.3. all missing in dbaston is the user facing documentation which I'm willing to do.

darkblue_b is missing c-unit tests, postgresql tests, make file changes and uses old structure.

Really sorry about this mixup. I think for future we need to make sure people note on tickets that they are working on something so there is no overlap of effort.

[dbaston]

PR at ​https://github.com/postgis/postgis/pull/73

FWIW, I don't think there's anything intrinsically good about using the iterator construct for this, it was just a convenient way to do the GEOS LineString hack, whose own merits could be debated.

[dbaston]

Committed to trunk at r14467.

[robe]

Dan,

As a rule, all new features should go in the NEWS file. Just remember for future reference.

committed to news file at r14469

[robe]

(In [14505]) Add examples for ST_Voronoi references #2259