Add CRS constructor.

[evetion]

No description provided.

[visr]

Thanks for working on this. I added some comments.

[visr]

Do you know if this covers everything? What about PJ_TYPE_TEMPORAL_CRS or PJ_TYPE_COMPOUND_CRS? Could they also be geographic or projected? If we can't be certain, perhaps we should just stick to the PROJ API and add proj_get_type(::CRS) for slightly more convenience.

[evetion]

@rafaqz This PR includes a method to convert between GFT instances, like ArchGDAL.

[evetion]

Not sure why the tests are suddenly failing, like the PROJ version changed in between, but can't replicate it. @visr, any thoughts?

[visr]

That's your own show change from #73 ;)

Fixed in 29e3415

The new show is better, it shows explicitly it is a no-op (old then new):

Transformation
    source: WGS 84
    target: WGS 84
    direction: forward

Transformation noop
    description: Null geographic offset from WGS 84 to WGS 84
    definition: proj=noop ellps=GRS80
    direction: forward

[evetion]

LGTM

[visr]

It would be better to directly do this in the style of yeesian/ArchGDAL.jl#349 and JuliaGeo/LibGEOS.jl#149, since this adds a bunch of methods like proj_get_id_code(crs::CRS) = proj_get_id_code(crs.pj) that would soon be removed again.

[evetion]

The style being
Base.unsafe_convert(::Type{Ptr{Cvoid}}, x::CRS) = x.ptr
?

[visr]

Yes indeed, then we don't need to take out the pointer ourselves, which is less code and safer for GC.