wkt reader

[michaelkirk]

Fixes #19

I have a couple things I was hoping to get input on before submitting for actual review... questions inline. Update: OK! Ready for review!

Definitely some weirdness around handling POINT EMPTY, but I went with what felt to me like a reasonable compromise of trying to:

• handle POINT EMPTY where possible. Currently this is just Wkt->Wkt which isn't very likely to be used, but I suppose if someone wrote a custom processor, they might want to have the info.
• Ignore EMPTY in a MultiPoint.
• Else return Err when encountering a POINT EMPTY. I couldn't find a way to "be lazy" and ignore a POINT EMPTY for a top level geometry - we have to return something. And having Wkt("POINT EMPTY).to_geo() -> point!(x: 0, y: 0) seems likely to be a bad surprise.

[pka]

Looks like we can't avoid going into that rabbit hole. At least we're not alone (https://github.com/locationtech/jts/pull/567 https://trac.osgeo.org/geos/ticket/1005). Looks like your solution with an explicit empty_point is the cleanest one. But in the end we would probably need this for all geometry types?

Anyway let's merge this and improve handling of empty geometries seperately. Thanks a lot!

[michaelkirk]

in the end we would probably need this for all geometry types?

Luckily, I think POINT EMPTY is kind of an exception. It seems like we can more reasonably handle other empty geometries like MULTIPOINT EMPTY, LINESTRING EMPTY, etc. without exploding.

I've expounded on this here: https://github.com/georust/geozero/pull/21