Support heterogeneous features

[Sov-trotter]

I haven't removed the old meta implementation as planned due to some issues with MeshMeta. But for the new implemetation, tests are almost written.
We just need to fix the Mesh issue, add some other helpful constructors.

[Sov-trotter]

#71

[visr]

@SimonDanisch the basic reason for going down this path is #49 (comment). This PR is only a draft, but do you in general agree with the proposal of moving from PointMeta to Meta{Point}? Main downside is that Meta{Point} can no longer subtype AbstractPoint.

We are going with the name Feature here which is the GeoJSON concept it corresponds to, and to avoid the clash with Base.Meta, but the name is of course open for discussion. A Feature can be seen as a row in a table. In GeoJSON this table is a FeatureCollection. We could go with StructArray{Feature} here or wrap that in a FeatureCollection struct.

[SimonDanisch]

Main downside is that Meta{Point} can no longer subtype AbstractPoint.

Oh, yeah...that was 100% the reason, that I didn't go with it :D

I think having MetaPoint be an AbstractPoint is pretty important ;)
So one way of keeping that would be, to have Meta{X} a fallback, for types, that weren't explicitly created by the macro?

[visr]

I think that would work for us :)

[Sov-trotter]

closes #50 #49

[Sov-trotter]

I think this is ready to be merged?
Are there anymore methods that I should add?

[SimonDanisch]

@visr, can you give this a review as well? :)

[visr]

Yes, I'll have a look in a day or two.

[visr]

Thanks, left a few comments.

Since this will now only be an extension of the existing Meta types, I don't think Feature is a good name anymore. Meta is not an option due to Base.Meta. Alternatives include MetaT (like Triangle / TriangleP) or GeometryMeta / AnyMeta which goes more with PointMeta / PolygonMeta.

[Sov-trotter]

I think GeometryMeta is good to have?
Or even MetaT is good since it's synonymous/close to Meta{T}.
I think I should add the latter.

[visr]

Looks good to me!

[Sov-trotter]

I am extremely sorry for the scattered PR's(both here and in AbstractPlotting). But here's a sequence in which you might want to review/merge them:

1. Firstly this one
2. Give out points of geometries #73
3. Overload convert_arguments for LineString JuliaPlots/AbstractPlotting.jl#479 (Depends on the decompose() method in Give out points of geometries #73 )
4. Support GeometryBasics Polygon JuliaPlots/AbstractPlotting.jl#486 (Depends on the convert_arguments method for LineString in the above PR)

[Sov-trotter]

Hey @SimonDanisch if the changes are fine then this can be merged?

[SimonDanisch]

Sorry for being a bit slow! Thanks for you contribution!