CRS not associated with geometries?

[asinghvi17]

When loading a Shapefile with defined CRS, the individual geometries always have GI.crs(geom) === nothing, even though GI.crs(table) isa GeoFormatTypes.ESRIWellKnownText. Would this need some form of refactor to store CRS in geometry as well?

[rafaqz]

Yes, there are no crs fields in the objects here. We could add them to most, but probably not to points.

And I guess strings are not actually copied/inlined for every polygon they are attached to? Its just a pointer to the same string underneath?

[asinghvi17]

I think so? Strings are heap allocated, so presumably yes, and Base.summarysize doesn't tell you whether there's a pointer there or not.

[asinghvi17]

Yes, I just checked and the pointer to the strings is the same for multiple geometries initialized with the same CRS object.

[asinghvi17]

Maybe it's a bad idea, but we could just wrap every Shapefile geometry as a GI geometry when they're returned from Handle....then the GI wrapper can hold crs.

[rafaqz]

Hmm maybe that will just work? But its a weird nested object in terms of default types. I think adding a field is better.

[asinghvi17]

I guess we would add a crs argument to all read functions in Shapefile.jl?

[rafaqz]

Ah right that's kind of annoying but yes probably

And pass the crs through